phoenix/litellm-mirror
0
0
Fork 1
mirror of https://github.com/BerriAI/litellm.git synced 2025-04-24 18:24:20 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'litellm_contributors_apr_21' into feature/issue-8764-infinity

Browse source
This commit is contained in:
Ishaan Jaff 2025-04-21 17:23:02 -07:00 • committed by GitHub
commit f7eaba7716
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
336 changed files with 14257 additions and 4532 deletions
Download patch file Download diff file Expand all files Collapse all files

3
.gitignore vendored
View file

@ -86,4 +86,5 @@ litellm/proxy/db/migrations/0_init/migration.sql
litellm/proxy/db/migrations/*
litellm/proxy/migrations/*config.yaml
litellm/proxy/migrations/*
config.yaml
config.yaml
tests/litellm/litellm_core_utils/llm_cost_calc/log.txt

2
docs/my-website/docs/pass_through/cohere.md
View file

@ -4,7 +4,7 @@ Pass-through endpoints for Cohere - call provider-specific endpoint, in native f
| Feature | Supported | Notes |
|-------|-------|-------|
| Cost Tracking | ✅ | works across all integrations |
| Cost Tracking | ✅ | Supported for `/v1/chat`, and `/v2/chat` |
| Logging | ✅ | works across all integrations |
| End-user Tracking | ❌ | [Tell us if you need this](https://github.com/BerriAI/litellm/issues/new) |
| Streaming | ✅ | |

217
docs/my-website/docs/pass_through/mistral.md Normal file
View file

@ -0,0 +1,217 @@
# Mistral
Pass-through endpoints for Mistral - call provider-specific endpoint, in native format (no translation).
| Feature | Supported | Notes |
|-------|-------|-------|
| Cost Tracking | ❌ | Not supported |
| Logging | ✅ | works across all integrations |
| End-user Tracking | ❌ | [Tell us if you need this](https://github.com/BerriAI/litellm/issues/new) |
| Streaming | ✅ | |
Just replace `https://api.mistral.ai/v1` with `LITELLM_PROXY_BASE_URL/mistral` 🚀
#### **Example Usage**
```bash
curl -L -X POST 'http://0.0.0.0:4000/mistral/v1/ocr' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer sk-1234' \
-d '{
"model": "mistral-ocr-latest",
"document": {
"type": "image_url",
"image_url": "https://raw.githubusercontent.com/mistralai/cookbook/refs/heads/main/mistral/ocr/receipt.png"
}
}'
```
Supports **ALL** Mistral Endpoints (including streaming).
## Quick Start
Let's call the Mistral [`/chat/completions` endpoint](https://docs.mistral.ai/api/#tag/chat/operation/chat_completion_v1_chat_completions_post)
1. Add MISTRAL_API_KEY to your environment
```bash
export MISTRAL_API_KEY="sk-1234"
```
2. Start LiteLLM Proxy
```bash
litellm
# RUNNING on http://0.0.0.0:4000
```
3. Test it!
Let's call the Mistral `/ocr` endpoint
```bash
curl -L -X POST 'http://0.0.0.0:4000/mistral/v1/ocr' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer sk-1234' \
-d '{
"model": "mistral-ocr-latest",
"document": {
"type": "image_url",
"image_url": "https://raw.githubusercontent.com/mistralai/cookbook/refs/heads/main/mistral/ocr/receipt.png"
}
}'
```
## Examples
Anything after `http://0.0.0.0:4000/mistral` is treated as a provider-specific route, and handled accordingly.
Key Changes:
| **Original Endpoint** | **Replace With** |
|------------------------------------------------------|-----------------------------------|
| `https://api.mistral.ai/v1` | `http://0.0.0.0:4000/mistral` (LITELLM_PROXY_BASE_URL="http://0.0.0.0:4000") |
| `bearer $MISTRAL_API_KEY` | `bearer anything` (use `bearer LITELLM_VIRTUAL_KEY` if Virtual Keys are setup on proxy) |
### **Example 1: OCR endpoint**
#### LiteLLM Proxy Call
```bash
curl -L -X POST 'http://0.0.0.0:4000/mistral/v1/ocr' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer $LITELLM_API_KEY' \
-d '{
"model": "mistral-ocr-latest",
"document": {
"type": "image_url",
"image_url": "https://raw.githubusercontent.com/mistralai/cookbook/refs/heads/main/mistral/ocr/receipt.png"
}
}'
```
#### Direct Mistral API Call
```bash
curl https://api.mistral.ai/v1/ocr \
-H "Content-Type: application/json" \
-H "Authorization: Bearer ${MISTRAL_API_KEY}" \
-d '{
"model": "mistral-ocr-latest",
"document": {
"type": "document_url",
"document_url": "https://arxiv.org/pdf/2201.04234"
},
"include_image_base64": true
}'
```
### **Example 2: Chat API**
#### LiteLLM Proxy Call
```bash
curl -L -X POST 'http://0.0.0.0:4000/mistral/v1/chat/completions' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer $LITELLM_VIRTUAL_KEY' \
-d '{
"messages": [
{
"role": "user",
"content": "I am going to Paris, what should I see?"
}
],
"max_tokens": 2048,
"temperature": 0.8,
"top_p": 0.1,
"model": "mistral-large-latest",
}'
```
#### Direct Mistral API Call
```bash
curl -L -X POST 'https://api.mistral.ai/v1/chat/completions' \
-H 'Content-Type: application/json' \
-d '{
"messages": [
{
"role": "user",
"content": "I am going to Paris, what should I see?"
}
],
"max_tokens": 2048,
"temperature": 0.8,
"top_p": 0.1,
"model": "mistral-large-latest",
}'
```
## Advanced - Use with Virtual Keys
Pre-requisites
- [Setup proxy with DB](../proxy/virtual_keys.md#setup)
Use this, to avoid giving developers the raw Mistral API key, but still letting them use Mistral endpoints.
### Usage
1. Setup environment
```bash
export DATABASE_URL=""
export LITELLM_MASTER_KEY=""
export MISTRAL_API_BASE=""
```
```bash
litellm
# RUNNING on http://0.0.0.0:4000
```
2. Generate virtual key
```bash
curl -X POST 'http://0.0.0.0:4000/key/generate' \
-H 'Authorization: Bearer sk-1234' \
-H 'Content-Type: application/json' \
-d '{}'
```
Expected Response
```bash
{
...
"key": "sk-1234ewknldferwedojwojw"
}
```
3. Test it!
```bash
curl -L -X POST 'http://0.0.0.0:4000/mistral/v1/chat/completions' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer sk-1234ewknldferwedojwojw' \
--data '{
"messages": [
{
"role": "user",
"content": "I am going to Paris, what should I see?"
}
],
"max_tokens": 2048,
"temperature": 0.8,
"top_p": 0.1,
"model": "qwen2.5-7b-instruct",
}'
```

9
docs/my-website/docs/pass_through/vertex_ai.md
View file

@ -13,6 +13,15 @@ Pass-through endpoints for Vertex AI - call provider-specific endpoint, in nativ
| End-user Tracking | ❌ | [Tell us if you need this](https://github.com/BerriAI/litellm/issues/new) |
| Streaming | ✅ | |
## Supported Endpoints
LiteLLM supports 2 vertex ai passthrough routes:
1. `/vertex_ai` → routes to `https://{vertex_location}-aiplatform.googleapis.com/`
2. `/vertex_ai/discovery` → routes to [`https://discoveryengine.googleapis.com`](https://discoveryengine.googleapis.com/)
## How to use
Just replace `https://REGION-aiplatform.googleapis.com` with `LITELLM_PROXY_BASE_URL/vertex_ai`
LiteLLM supports 3 flows for calling Vertex AI endpoints via pass-through:

185
docs/my-website/docs/pass_through/vllm.md Normal file
View file

@ -0,0 +1,185 @@
# VLLM
Pass-through endpoints for VLLM - call provider-specific endpoint, in native format (no translation).
| Feature | Supported | Notes |
|-------|-------|-------|
| Cost Tracking | ❌ | Not supported |
| Logging | ✅ | works across all integrations |
| End-user Tracking | ❌ | [Tell us if you need this](https://github.com/BerriAI/litellm/issues/new) |
| Streaming | ✅ | |
Just replace `https://my-vllm-server.com` with `LITELLM_PROXY_BASE_URL/vllm` 🚀
#### **Example Usage**
```bash
curl -L -X GET 'http://0.0.0.0:4000/vllm/metrics' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer sk-1234' \
```
Supports **ALL** VLLM Endpoints (including streaming).
## Quick Start
Let's call the VLLM [`/metrics` endpoint](https://vllm.readthedocs.io/en/latest/api_reference/api_reference.html)
1. Add HOSTED VLLM API BASE to your environment
```bash
export HOSTED_VLLM_API_BASE="https://my-vllm-server.com"
```
2. Start LiteLLM Proxy
```bash
litellm
# RUNNING on http://0.0.0.0:4000
```
3. Test it!
Let's call the VLLM `/metrics` endpoint
```bash
curl -L -X GET 'http://0.0.0.0:4000/vllm/metrics' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer sk-1234' \
```
## Examples
Anything after `http://0.0.0.0:4000/vllm` is treated as a provider-specific route, and handled accordingly.
Key Changes:
| **Original Endpoint** | **Replace With** |
|------------------------------------------------------|-----------------------------------|
| `https://my-vllm-server.com` | `http://0.0.0.0:4000/vllm` (LITELLM_PROXY_BASE_URL="http://0.0.0.0:4000") |
| `bearer $VLLM_API_KEY` | `bearer anything` (use `bearer LITELLM_VIRTUAL_KEY` if Virtual Keys are setup on proxy) |
### **Example 1: Metrics endpoint**
#### LiteLLM Proxy Call
```bash
curl -L -X GET 'http://0.0.0.0:4000/vllm/metrics' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer $LITELLM_VIRTUAL_KEY' \
```
#### Direct VLLM API Call
```bash
curl -L -X GET 'https://my-vllm-server.com/metrics' \
-H 'Content-Type: application/json' \
```
### **Example 2: Chat API**
#### LiteLLM Proxy Call
```bash
curl -L -X POST 'http://0.0.0.0:4000/vllm/chat/completions' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer $LITELLM_VIRTUAL_KEY' \
-d '{
"messages": [
{
"role": "user",
"content": "I am going to Paris, what should I see?"
}
],
"max_tokens": 2048,
"temperature": 0.8,
"top_p": 0.1,
"model": "qwen2.5-7b-instruct",
}'
```
#### Direct VLLM API Call
```bash
curl -L -X POST 'https://my-vllm-server.com/chat/completions' \
-H 'Content-Type: application/json' \
-d '{
"messages": [
{
"role": "user",
"content": "I am going to Paris, what should I see?"
}
],
"max_tokens": 2048,
"temperature": 0.8,
"top_p": 0.1,
"model": "qwen2.5-7b-instruct",
}'
```
## Advanced - Use with Virtual Keys
Pre-requisites
- [Setup proxy with DB](../proxy/virtual_keys.md#setup)
Use this, to avoid giving developers the raw Cohere API key, but still letting them use Cohere endpoints.
### Usage
1. Setup environment
```bash
export DATABASE_URL=""
export LITELLM_MASTER_KEY=""
export HOSTED_VLLM_API_BASE=""
```
```bash
litellm
# RUNNING on http://0.0.0.0:4000
```
2. Generate virtual key
```bash
curl -X POST 'http://0.0.0.0:4000/key/generate' \
-H 'Authorization: Bearer sk-1234' \
-H 'Content-Type: application/json' \
-d '{}'
```
Expected Response
```bash
{
...
"key": "sk-1234ewknldferwedojwojw"
}
```
3. Test it!
```bash
curl -L -X POST 'http://0.0.0.0:4000/vllm/chat/completions' \
-H 'Content-Type: application/json' \
-H 'Authorization: Bearer sk-1234ewknldferwedojwojw' \
--data '{
"messages": [
{
"role": "user",
"content": "I am going to Paris, what should I see?"
}
],
"max_tokens": 2048,
"temperature": 0.8,
"top_p": 0.1,
"model": "qwen2.5-7b-instruct",
}'
```

117
docs/my-website/docs/providers/azure.md
View file

@ -1002,8 +1002,125 @@ Expected Response:
```
## **Azure Responses API**
| Property | Details |
|-------|-------|
| Description | Azure OpenAI Responses API |
| `custom_llm_provider` on LiteLLM | `azure/` |
| Supported Operations | `/v1/responses`|
| Azure OpenAI Responses API | [Azure OpenAI Responses API ↗](https://learn.microsoft.com/en-us/azure/ai-services/openai/how-to/responses?tabs=python-secure) |
| Cost Tracking, Logging Support | ✅ LiteLLM will log, track cost for Responses API Requests |
| Supported OpenAI Params | ✅ All OpenAI params are supported, [See here](https://github.com/BerriAI/litellm/blob/0717369ae6969882d149933da48eeb8ab0e691bd/litellm/llms/openai/responses/transformation.py#L23) |
## Usage
## Create a model response
<Tabs>
<TabItem value="litellm-sdk" label="LiteLLM SDK">
#### Non-streaming
```python showLineNumbers title="Azure Responses API"
import litellm
# Non-streaming response
response = litellm.responses(
model="azure/o1-pro",
input="Tell me a three sentence bedtime story about a unicorn.",
max_output_tokens=100,
api_key=os.getenv("AZURE_RESPONSES_OPENAI_API_KEY"),
api_base="https://litellm8397336933.openai.azure.com/",
api_version="2023-03-15-preview",
)
print(response)
```
#### Streaming
```python showLineNumbers title="Azure Responses API"
import litellm
# Streaming response
response = litellm.responses(
model="azure/o1-pro",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True,
api_key=os.getenv("AZURE_RESPONSES_OPENAI_API_KEY"),
api_base="https://litellm8397336933.openai.azure.com/",
api_version="2023-03-15-preview",
)
for event in response:
print(event)
```
</TabItem>
<TabItem value="proxy" label="OpenAI SDK with LiteLLM Proxy">
First, add this to your litellm proxy config.yaml:
```yaml showLineNumbers title="Azure Responses API"
model_list:
- model_name: o1-pro
litellm_params:
model: azure/o1-pro
api_key: os.environ/AZURE_RESPONSES_OPENAI_API_KEY
api_base: https://litellm8397336933.openai.azure.com/
api_version: 2023-03-15-preview
```
Start your LiteLLM proxy:
```bash
litellm --config /path/to/config.yaml
# RUNNING on http://0.0.0.0:4000
```
Then use the OpenAI SDK pointed to your proxy:
#### Non-streaming
```python showLineNumbers
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Non-streaming response
response = client.responses.create(
model="o1-pro",
input="Tell me a three sentence bedtime story about a unicorn."
)
print(response)
```
#### Streaming
```python showLineNumbers
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Streaming response
response = client.responses.create(
model="o1-pro",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
for event in response:
print(event)
```
</TabItem>
</Tabs>

152
docs/my-website/docs/providers/gemini.md
View file

@ -39,14 +39,164 @@ response = completion(
- temperature
- top_p
- max_tokens
- max_completion_tokens
- stream
- tools
- tool_choice
- functions
- response_format
- n
- stop
- logprobs
- frequency_penalty
- modalities
- reasoning_content
**Anthropic Params**
- thinking (used to set max budget tokens across anthropic/gemini models)
[**See Updated List**](https://github.com/BerriAI/litellm/blob/main/litellm/llms/gemini/chat/transformation.py#L70)
## Usage - Thinking / `reasoning_content`
LiteLLM translates OpenAI's `reasoning_effort` to Gemini's `thinking` parameter. [Code](https://github.com/BerriAI/litellm/blob/620664921902d7a9bfb29897a7b27c1a7ef4ddfb/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py#L362)
**Mapping**
| reasoning_effort | thinking |
| ---------------- | -------- |
| "low" | "budget_tokens": 1024 |
| "medium" | "budget_tokens": 2048 |
| "high" | "budget_tokens": 4096 |
<Tabs>
<TabItem value="sdk" label="SDK">
```python
from litellm import completion
resp = completion(
model="gemini/gemini-2.5-flash-preview-04-17",
messages=[{"role": "user", "content": "What is the capital of France?"}],
reasoning_effort="low",
)
```
</TabItem>
<TabItem value="proxy" label="PROXY">
1. Setup config.yaml
```yaml
- model_name: gemini-2.5-flash
litellm_params:
model: gemini/gemini-2.5-flash-preview-04-17
api_key: os.environ/GEMINI_API_KEY
```
2. Start proxy
```bash
litellm --config /path/to/config.yaml
```
3. Test it!
```bash
curl http://0.0.0.0:4000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer <YOUR-LITELLM-KEY>" \
-d '{
"model": "gemini-2.5-flash",
"messages": [{"role": "user", "content": "What is the capital of France?"}],
"reasoning_effort": "low"
}'
```
</TabItem>
</Tabs>
**Expected Response**
```python
ModelResponse(
id='chatcmpl-c542d76d-f675-4e87-8e5f-05855f5d0f5e',
created=1740470510,
model='claude-3-7-sonnet-20250219',
object='chat.completion',
system_fingerprint=None,
choices=[
Choices(
finish_reason='stop',
index=0,
message=Message(
content="The capital of France is Paris.",
role='assistant',
tool_calls=None,
function_call=None,
reasoning_content='The capital of France is Paris. This is a very straightforward factual question.'
),
)
],
usage=Usage(
completion_tokens=68,
prompt_tokens=42,
total_tokens=110,
completion_tokens_details=None,
prompt_tokens_details=PromptTokensDetailsWrapper(
audio_tokens=None,
cached_tokens=0,
text_tokens=None,
image_tokens=None
),
cache_creation_input_tokens=0,
cache_read_input_tokens=0
)
)
```
### Pass `thinking` to Gemini models
You can also pass the `thinking` parameter to Gemini models.
This is translated to Gemini's [`thinkingConfig` parameter](https://ai.google.dev/gemini-api/docs/thinking#set-budget).
<Tabs>
<TabItem value="sdk" label="SDK">
```python
response = litellm.completion(
model="gemini/gemini-2.5-flash-preview-04-17",
messages=[{"role": "user", "content": "What is the capital of France?"}],
thinking={"type": "enabled", "budget_tokens": 1024},
)
```
</TabItem>
<TabItem value="proxy" label="PROXY">
```bash
curl http://0.0.0.0:4000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $LITELLM_KEY" \
-d '{
"model": "gemini/gemini-2.5-flash-preview-04-17",
"messages": [{"role": "user", "content": "What is the capital of France?"}],
"thinking": {"type": "enabled", "budget_tokens": 1024}
}'
```
</TabItem>
</Tabs>
[**See Updated List**](https://github.com/BerriAI/litellm/blob/1c747f3ad372399c5b95cc5696b06a5fbe53186b/litellm/llms/vertex_httpx.py#L122)
## Passing Gemini Specific Params
### Response schema

6
docs/my-website/docs/providers/openai.md
View file

@ -163,6 +163,12 @@ os.environ["OPENAI_API_BASE"] = "openaiai-api-base" # OPTIONAL
| Model Name | Function Call |
|-----------------------|-----------------------------------------------------------------|
| gpt-4.1 | `response = completion(model="gpt-4.1", messages=messages)` |
| gpt-4.1-mini | `response = completion(model="gpt-4.1-mini", messages=messages)` |
| gpt-4.1-nano | `response = completion(model="gpt-4.1-nano", messages=messages)` |
| o4-mini | `response = completion(model="o4-mini", messages=messages)` |
| o3-mini | `response = completion(model="o3-mini", messages=messages)` |
| o3 | `response = completion(model="o3", messages=messages)` |
| o1-mini | `response = completion(model="o1-mini", messages=messages)` |
| o1-preview | `response = completion(model="o1-preview", messages=messages)` |
| gpt-4o-mini | `response = completion(model="gpt-4o-mini", messages=messages)` |

148
docs/my-website/docs/providers/vertex.md
View file

@ -542,6 +542,154 @@ print(resp)
```
### **Thinking / `reasoning_content`**
LiteLLM translates OpenAI's `reasoning_effort` to Gemini's `thinking` parameter. [Code](https://github.com/BerriAI/litellm/blob/620664921902d7a9bfb29897a7b27c1a7ef4ddfb/litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py#L362)
**Mapping**
| reasoning_effort | thinking |
| ---------------- | -------- |
| "low" | "budget_tokens": 1024 |
| "medium" | "budget_tokens": 2048 |
| "high" | "budget_tokens": 4096 |
<Tabs>
<TabItem value="sdk" label="SDK">
```python
from litellm import completion
# !gcloud auth application-default login - run this to add vertex credentials to your env
resp = completion(
model="vertex_ai/gemini-2.5-flash-preview-04-17",
messages=[{"role": "user", "content": "What is the capital of France?"}],
reasoning_effort="low",
vertex_project="project-id",
vertex_location="us-central1"
)
```
</TabItem>
<TabItem value="proxy" label="PROXY">
1. Setup config.yaml
```yaml
- model_name: gemini-2.5-flash
litellm_params:
model: vertex_ai/gemini-2.5-flash-preview-04-17
vertex_credentials: {"project_id": "project-id", "location": "us-central1", "project_key": "project-key"}
vertex_project: "project-id"
vertex_location: "us-central1"
```
2. Start proxy
```bash
litellm --config /path/to/config.yaml
```
3. Test it!
```bash
curl http://0.0.0.0:4000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer <YOUR-LITELLM-KEY>" \
-d '{
"model": "gemini-2.5-flash",
"messages": [{"role": "user", "content": "What is the capital of France?"}],
"reasoning_effort": "low"
}'
```
</TabItem>
</Tabs>
**Expected Response**
```python
ModelResponse(
id='chatcmpl-c542d76d-f675-4e87-8e5f-05855f5d0f5e',
created=1740470510,
model='claude-3-7-sonnet-20250219',
object='chat.completion',
system_fingerprint=None,
choices=[
Choices(
finish_reason='stop',
index=0,
message=Message(
content="The capital of France is Paris.",
role='assistant',
tool_calls=None,
function_call=None,
reasoning_content='The capital of France is Paris. This is a very straightforward factual question.'
),
)
],
usage=Usage(
completion_tokens=68,
prompt_tokens=42,
total_tokens=110,
completion_tokens_details=None,
prompt_tokens_details=PromptTokensDetailsWrapper(
audio_tokens=None,
cached_tokens=0,
text_tokens=None,
image_tokens=None
),
cache_creation_input_tokens=0,
cache_read_input_tokens=0
)
)
```
#### Pass `thinking` to Gemini models
You can also pass the `thinking` parameter to Gemini models.
This is translated to Gemini's [`thinkingConfig` parameter](https://ai.google.dev/gemini-api/docs/thinking#set-budget).
<Tabs>
<TabItem value="sdk" label="SDK">
```python
from litellm import completion
# !gcloud auth application-default login - run this to add vertex credentials to your env
response = litellm.completion(
model="vertex_ai/gemini-2.5-flash-preview-04-17",
messages=[{"role": "user", "content": "What is the capital of France?"}],
thinking={"type": "enabled", "budget_tokens": 1024},
vertex_project="project-id",
vertex_location="us-central1"
)
```
</TabItem>
<TabItem value="proxy" label="PROXY">
```bash
curl http://0.0.0.0:4000/v1/chat/completions \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $LITELLM_KEY" \
-d '{
"model": "vertex_ai/gemini-2.5-flash-preview-04-17",
"messages": [{"role": "user", "content": "What is the capital of France?"}],
"thinking": {"type": "enabled", "budget_tokens": 1024}
}'
```
</TabItem>
</Tabs>
### **Context Caching**
Use Vertex AI context caching is supported by calling provider api directly. (Unified Endpoint support comin soon.).

118
docs/my-website/docs/providers/vllm.md
View file

@ -161,6 +161,120 @@ curl -L -X POST 'http://0.0.0.0:4000/embeddings' \
Example Implementation from VLLM [here](https://github.com/vllm-project/vllm/pull/10020)
<Tabs>
<TabItem value="files_message" label="(Unified) Files Message">
Use this to send a video url to VLLM + Gemini in the same format, using OpenAI's `files` message type.
There are two ways to send a video url to VLLM:
1. Pass the video url directly
```
{"type": "file", "file": {"file_id": video_url}},
```
2. Pass the video data as base64
```
{"type": "file", "file": {"file_data": f"data:video/mp4;base64,{video_data_base64}"}}
```
<Tabs>
<TabItem value="sdk" label="SDK">
```python
from litellm import completion
messages=[
{
"role": "user",
"content": [
{
"type": "text",
"text": "Summarize the following video"
},
{
"type": "file",
"file": {
"file_id": "https://www.youtube.com/watch?v=dQw4w9WgXcQ"
}
}
]
}
]
# call vllm
os.environ["HOSTED_VLLM_API_BASE"] = "https://hosted-vllm-api.co"
os.environ["HOSTED_VLLM_API_KEY"] = "" # [optional], if your VLLM server requires an API key
response = completion(
model="hosted_vllm/qwen", # pass the vllm model name
messages=messages,
)
# call gemini
os.environ["GEMINI_API_KEY"] = "your-gemini-api-key"
response = completion(
model="gemini/gemini-1.5-flash", # pass the gemini model name
messages=messages,
)
print(response)
```
</TabItem>
<TabItem value="proxy" label="PROXY">
1. Setup config.yaml
```yaml
model_list:
- model_name: my-model
litellm_params:
model: hosted_vllm/qwen # add hosted_vllm/ prefix to route as OpenAI provider
api_base: https://hosted-vllm-api.co # add api base for OpenAI compatible provider
- model_name: my-gemini-model
litellm_params:
model: gemini/gemini-1.5-flash # add gemini/ prefix to route as Google AI Studio provider
api_key: os.environ/GEMINI_API_KEY
```
2. Start the proxy
```bash
$ litellm --config /path/to/config.yaml
# RUNNING on http://0.0.0.0:4000
```
3. Test it!
```bash
curl -X POST http://0.0.0.0:4000/chat/completions \
-H "Authorization: Bearer sk-1234" \
-H "Content-Type: application/json" \
-d '{
"model": "my-model",
"messages": [
{"role": "user", "content":
[
{"type": "text", "text": "Summarize the following video"},
{"type": "file", "file": {"file_id": "https://www.youtube.com/watch?v=dQw4w9WgXcQ"}}
]
}
]
}'
```
</TabItem>
</Tabs>
</TabItem>
<TabItem value="video_url" label="(VLLM-specific) Video Message">
Use this to send a video url to VLLM in it's native message format (`video_url`).
There are two ways to send a video url to VLLM:
1. Pass the video url directly
@ -249,6 +363,10 @@ curl -X POST http://0.0.0.0:4000/chat/completions \
</Tabs>
</TabItem>
</Tabs>
## (Deprecated) for `vllm pip package`
### Using - `litellm.completion`

1
docs/my-website/docs/proxy/config_settings.md
View file

@ -334,7 +334,6 @@ router_settings:
| AZURE_STORAGE_TENANT_ID | The Application Tenant ID to use for Authentication to Azure Blob Storage logging
| AZURE_STORAGE_CLIENT_ID | The Application Client ID to use for Authentication to Azure Blob Storage logging
| AZURE_STORAGE_CLIENT_SECRET | The Application Client Secret to use for Authentication to Azure Blob Storage logging
| BERRISPEND_ACCOUNT_ID | Account ID for BerriSpend service
| BRAINTRUST_API_KEY | API key for Braintrust integration
| CIRCLE_OIDC_TOKEN | OpenID Connect token for CircleCI

4
docs/my-website/docs/proxy/logging.md
View file

@ -862,7 +862,7 @@ Add the following to your env
```shell
OTEL_EXPORTER="otlp_http"
OTEL_ENDPOINT="http:/0.0.0.0:4317"
OTEL_ENDPOINT="http://0.0.0.0:4317"
OTEL_HEADERS="x-honeycomb-team=<your-api-key>" # Optional
```
@ -2501,4 +2501,4 @@ litellm_settings:
:::info
`thresholds` are not required by default, but you can tune the values to your needs.
Default values is `4` for all categories
::: -->
::: -->

108
docs/my-website/docs/proxy/model_discovery.md Normal file
View file

@ -0,0 +1,108 @@
# Model Discovery
Use this to give users an accurate list of models available behind provider endpoint, when calling `/v1/models` for wildcard models.
## Supported Models
- Fireworks AI
- OpenAI
- Gemini
- LiteLLM Proxy
- Topaz
- Anthropic
- XAI
- VLLM
- Vertex AI
### Usage
**1. Setup config.yaml**
```yaml
model_list:
- model_name: xai/*
litellm_params:
model: xai/*
api_key: os.environ/XAI_API_KEY
litellm_settings:
check_provider_endpoint: true # 👈 Enable checking provider endpoint for wildcard models
```
**2. Start proxy**
```bash
litellm --config /path/to/config.yaml
# RUNNING on http://0.0.0.0:4000
```
**3. Call `/v1/models`**
```bash
curl -X GET "http://localhost:4000/v1/models" -H "Authorization: Bearer $LITELLM_KEY"
```
Expected response
```json
{
"data": [
{
"id": "xai/grok-2-1212",
"object": "model",
"created": 1677610602,
"owned_by": "openai"
},
{
"id": "xai/grok-2-vision-1212",
"object": "model",
"created": 1677610602,
"owned_by": "openai"
},
{
"id": "xai/grok-3-beta",
"object": "model",
"created": 1677610602,
"owned_by": "openai"
},
{
"id": "xai/grok-3-fast-beta",
"object": "model",
"created": 1677610602,
"owned_by": "openai"
},
{
"id": "xai/grok-3-mini-beta",
"object": "model",
"created": 1677610602,
"owned_by": "openai"
},
{
"id": "xai/grok-3-mini-fast-beta",
"object": "model",
"created": 1677610602,
"owned_by": "openai"
},
{
"id": "xai/grok-beta",
"object": "model",
"created": 1677610602,
"owned_by": "openai"
},
{
"id": "xai/grok-vision-beta",
"object": "model",
"created": 1677610602,
"owned_by": "openai"
},
{
"id": "xai/grok-2-image-1212",
"object": "model",
"created": 1677610602,
"owned_by": "openai"
}
],
"object": "list"
}
```

4
docs/my-website/docs/reasoning_content.md
View file

@ -16,6 +16,8 @@ Supported Providers:
- Vertex AI (Anthropic) (`vertexai/`)
- OpenRouter (`openrouter/`)
- XAI (`xai/`)
- Google AI Studio (`google/`)
- Vertex AI (`vertex_ai/`)
LiteLLM will standardize the `reasoning_content` in the response and `thinking_blocks` in the assistant message.
@ -23,7 +25,7 @@ LiteLLM will standardize the `reasoning_content` in the response and `thinking_b
"message": {
...
"reasoning_content": "The capital of France is Paris.",
"thinking_blocks": [
"thinking_blocks": [ # only returned for Anthropic models
{
"type": "thinking",
"thinking": "The capital of France is Paris.",

493
docs/my-website/docs/response_api.md
View file

@ -14,22 +14,22 @@ LiteLLM provides a BETA endpoint in the spec of [OpenAI's `/responses` API](http
| Fallbacks | ✅ | Works between supported models |
| Loadbalancing | ✅ | Works between supported models |
| Supported LiteLLM Versions | 1.63.8+ | |
| Supported LLM providers | `openai` | |
| Supported LLM providers | **All LiteLLM supported providers** | `openai`, `anthropic`, `bedrock`, `vertex_ai`, `gemini`, `azure`, `azure_ai` etc. |
## Usage
## Create a model response
### LiteLLM Python SDK
<Tabs>
<TabItem value="litellm-sdk" label="LiteLLM SDK">
<TabItem value="openai" label="OpenAI">
#### Non-streaming
```python
```python showLineNumbers title="OpenAI Non-streaming Response"
import litellm
# Non-streaming response
response = litellm.responses(
model="o1-pro",
model="openai/o1-pro",
input="Tell me a three sentence bedtime story about a unicorn.",
max_output_tokens=100
)
@ -38,12 +38,12 @@ print(response)
```
#### Streaming
```python
```python showLineNumbers title="OpenAI Streaming Response"
import litellm
# Streaming response
response = litellm.responses(
model="o1-pro",
model="openai/o1-pro",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
@ -53,58 +53,169 @@ for event in response:
```
</TabItem>
<TabItem value="proxy" label="OpenAI SDK with LiteLLM Proxy">
First, add this to your litellm proxy config.yaml:
```yaml
model_list:
- model_name: o1-pro
litellm_params:
model: openai/o1-pro
api_key: os.environ/OPENAI_API_KEY
```
Start your LiteLLM proxy:
```bash
litellm --config /path/to/config.yaml
# RUNNING on http://0.0.0.0:4000
```
Then use the OpenAI SDK pointed to your proxy:
<TabItem value="anthropic" label="Anthropic">
#### Non-streaming
```python
from openai import OpenAI
```python showLineNumbers title="Anthropic Non-streaming Response"
import litellm
import os
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Set API key
os.environ["ANTHROPIC_API_KEY"] = "your-anthropic-api-key"
# Non-streaming response
response = client.responses.create(
model="o1-pro",
input="Tell me a three sentence bedtime story about a unicorn."
response = litellm.responses(
model="anthropic/claude-3-5-sonnet-20240620",
input="Tell me a three sentence bedtime story about a unicorn.",
max_output_tokens=100
)
print(response)
```
#### Streaming
```python
from openai import OpenAI
```python showLineNumbers title="Anthropic Streaming Response"
import litellm
import os
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Set API key
os.environ["ANTHROPIC_API_KEY"] = "your-anthropic-api-key"
# Streaming response
response = client.responses.create(
model="o1-pro",
response = litellm.responses(
model="anthropic/claude-3-5-sonnet-20240620",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
for event in response:
print(event)
```
</TabItem>
<TabItem value="vertex" label="Vertex AI">
#### Non-streaming
```python showLineNumbers title="Vertex AI Non-streaming Response"
import litellm
import os
# Set credentials - Vertex AI uses application default credentials
# Run 'gcloud auth application-default login' to authenticate
os.environ["VERTEXAI_PROJECT"] = "your-gcp-project-id"
os.environ["VERTEXAI_LOCATION"] = "us-central1"
# Non-streaming response
response = litellm.responses(
model="vertex_ai/gemini-1.5-pro",
input="Tell me a three sentence bedtime story about a unicorn.",
max_output_tokens=100
)
print(response)
```
#### Streaming
```python showLineNumbers title="Vertex AI Streaming Response"
import litellm
import os
# Set credentials - Vertex AI uses application default credentials
# Run 'gcloud auth application-default login' to authenticate
os.environ["VERTEXAI_PROJECT"] = "your-gcp-project-id"
os.environ["VERTEXAI_LOCATION"] = "us-central1"
# Streaming response
response = litellm.responses(
model="vertex_ai/gemini-1.5-pro",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
for event in response:
print(event)
```
</TabItem>
<TabItem value="bedrock" label="AWS Bedrock">
#### Non-streaming
```python showLineNumbers title="AWS Bedrock Non-streaming Response"
import litellm
import os
# Set AWS credentials
os.environ["AWS_ACCESS_KEY_ID"] = "your-access-key-id"
os.environ["AWS_SECRET_ACCESS_KEY"] = "your-secret-access-key"
os.environ["AWS_REGION_NAME"] = "us-west-2" # or your AWS region
# Non-streaming response
response = litellm.responses(
model="bedrock/anthropic.claude-3-sonnet-20240229-v1:0",
input="Tell me a three sentence bedtime story about a unicorn.",
max_output_tokens=100
)
print(response)
```
#### Streaming
```python showLineNumbers title="AWS Bedrock Streaming Response"
import litellm
import os
# Set AWS credentials
os.environ["AWS_ACCESS_KEY_ID"] = "your-access-key-id"
os.environ["AWS_SECRET_ACCESS_KEY"] = "your-secret-access-key"
os.environ["AWS_REGION_NAME"] = "us-west-2" # or your AWS region
# Streaming response
response = litellm.responses(
model="bedrock/anthropic.claude-3-sonnet-20240229-v1:0",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
for event in response:
print(event)
```
</TabItem>
<TabItem value="gemini" label="Google AI Studio">
#### Non-streaming
```python showLineNumbers title="Google AI Studio Non-streaming Response"
import litellm
import os
# Set API key for Google AI Studio
os.environ["GEMINI_API_KEY"] = "your-gemini-api-key"
# Non-streaming response
response = litellm.responses(
model="gemini/gemini-1.5-flash",
input="Tell me a three sentence bedtime story about a unicorn.",
max_output_tokens=100
)
print(response)
```
#### Streaming
```python showLineNumbers title="Google AI Studio Streaming Response"
import litellm
import os
# Set API key for Google AI Studio
os.environ["GEMINI_API_KEY"] = "your-gemini-api-key"
# Streaming response
response = litellm.responses(
model="gemini/gemini-1.5-flash",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
@ -115,3 +226,297 @@ for event in response:
</TabItem>
</Tabs>
### LiteLLM Proxy with OpenAI SDK
First, set up and start your LiteLLM proxy server.
```bash title="Start LiteLLM Proxy Server"
litellm --config /path/to/config.yaml
# RUNNING on http://0.0.0.0:4000
```
<Tabs>
<TabItem value="openai" label="OpenAI">
First, add this to your litellm proxy config.yaml:
```yaml showLineNumbers title="OpenAI Proxy Configuration"
model_list:
- model_name: openai/o1-pro
litellm_params:
model: openai/o1-pro
api_key: os.environ/OPENAI_API_KEY
```
#### Non-streaming
```python showLineNumbers title="OpenAI Proxy Non-streaming Response"
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Non-streaming response
response = client.responses.create(
model="openai/o1-pro",
input="Tell me a three sentence bedtime story about a unicorn."
)
print(response)
```
#### Streaming
```python showLineNumbers title="OpenAI Proxy Streaming Response"
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Streaming response
response = client.responses.create(
model="openai/o1-pro",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
for event in response:
print(event)
```
</TabItem>
<TabItem value="anthropic" label="Anthropic">
First, add this to your litellm proxy config.yaml:
```yaml showLineNumbers title="Anthropic Proxy Configuration"
model_list:
- model_name: anthropic/claude-3-5-sonnet-20240620
litellm_params:
model: anthropic/claude-3-5-sonnet-20240620
api_key: os.environ/ANTHROPIC_API_KEY
```
#### Non-streaming
```python showLineNumbers title="Anthropic Proxy Non-streaming Response"
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Non-streaming response
response = client.responses.create(
model="anthropic/claude-3-5-sonnet-20240620",
input="Tell me a three sentence bedtime story about a unicorn."
)
print(response)
```
#### Streaming
```python showLineNumbers title="Anthropic Proxy Streaming Response"
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Streaming response
response = client.responses.create(
model="anthropic/claude-3-5-sonnet-20240620",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
for event in response:
print(event)
```
</TabItem>
<TabItem value="vertex" label="Vertex AI">
First, add this to your litellm proxy config.yaml:
```yaml showLineNumbers title="Vertex AI Proxy Configuration"
model_list:
- model_name: vertex_ai/gemini-1.5-pro
litellm_params:
model: vertex_ai/gemini-1.5-pro
vertex_project: your-gcp-project-id
vertex_location: us-central1
```
#### Non-streaming
```python showLineNumbers title="Vertex AI Proxy Non-streaming Response"
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Non-streaming response
response = client.responses.create(
model="vertex_ai/gemini-1.5-pro",
input="Tell me a three sentence bedtime story about a unicorn."
)
print(response)
```
#### Streaming
```python showLineNumbers title="Vertex AI Proxy Streaming Response"
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Streaming response
response = client.responses.create(
model="vertex_ai/gemini-1.5-pro",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
for event in response:
print(event)
```
</TabItem>
<TabItem value="bedrock" label="AWS Bedrock">
First, add this to your litellm proxy config.yaml:
```yaml showLineNumbers title="AWS Bedrock Proxy Configuration"
model_list:
- model_name: bedrock/anthropic.claude-3-sonnet-20240229-v1:0
litellm_params:
model: bedrock/anthropic.claude-3-sonnet-20240229-v1:0
aws_access_key_id: os.environ/AWS_ACCESS_KEY_ID
aws_secret_access_key: os.environ/AWS_SECRET_ACCESS_KEY
aws_region_name: us-west-2
```
#### Non-streaming
```python showLineNumbers title="AWS Bedrock Proxy Non-streaming Response"
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Non-streaming response
response = client.responses.create(
model="bedrock/anthropic.claude-3-sonnet-20240229-v1:0",
input="Tell me a three sentence bedtime story about a unicorn."
)
print(response)
```
#### Streaming
```python showLineNumbers title="AWS Bedrock Proxy Streaming Response"
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Streaming response
response = client.responses.create(
model="bedrock/anthropic.claude-3-sonnet-20240229-v1:0",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
for event in response:
print(event)
```
</TabItem>
<TabItem value="gemini" label="Google AI Studio">
First, add this to your litellm proxy config.yaml:
```yaml showLineNumbers title="Google AI Studio Proxy Configuration"
model_list:
- model_name: gemini/gemini-1.5-flash
litellm_params:
model: gemini/gemini-1.5-flash
api_key: os.environ/GEMINI_API_KEY
```
#### Non-streaming
```python showLineNumbers title="Google AI Studio Proxy Non-streaming Response"
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Non-streaming response
response = client.responses.create(
model="gemini/gemini-1.5-flash",
input="Tell me a three sentence bedtime story about a unicorn."
)
print(response)
```
#### Streaming
```python showLineNumbers title="Google AI Studio Proxy Streaming Response"
from openai import OpenAI
# Initialize client with your proxy URL
client = OpenAI(
base_url="http://localhost:4000", # Your proxy URL
api_key="your-api-key" # Your proxy API key
)
# Streaming response
response = client.responses.create(
model="gemini/gemini-1.5-flash",
input="Tell me a three sentence bedtime story about a unicorn.",
stream=True
)
for event in response:
print(event)
```
</TabItem>
</Tabs>
## Supported Responses API Parameters
| Provider | Supported Parameters |
|----------|---------------------|
| `openai` | [All Responses API parameters are supported](https://github.com/BerriAI/litellm/blob/7c3df984da8e4dff9201e4c5353fdc7a2b441831/litellm/llms/openai/responses/transformation.py#L23) |
| `azure` | [All Responses API parameters are supported](https://github.com/BerriAI/litellm/blob/7c3df984da8e4dff9201e4c5353fdc7a2b441831/litellm/llms/openai/responses/transformation.py#L23) |
| `anthropic` | [See supported parameters here](https://github.com/BerriAI/litellm/blob/f39d9178868662746f159d5ef642c7f34f9bfe5f/litellm/responses/litellm_completion_transformation/transformation.py#L57) |
| `bedrock` | [See supported parameters here](https://github.com/BerriAI/litellm/blob/f39d9178868662746f159d5ef642c7f34f9bfe5f/litellm/responses/litellm_completion_transformation/transformation.py#L57) |
| `gemini` | [See supported parameters here](https://github.com/BerriAI/litellm/blob/f39d9178868662746f159d5ef642c7f34f9bfe5f/litellm/responses/litellm_completion_transformation/transformation.py#L57) |
| `vertex_ai` | [See supported parameters here](https://github.com/BerriAI/litellm/blob/f39d9178868662746f159d5ef642c7f34f9bfe5f/litellm/responses/litellm_completion_transformation/transformation.py#L57) |
| `azure_ai` | [See supported parameters here](https://github.com/BerriAI/litellm/blob/f39d9178868662746f159d5ef642c7f34f9bfe5f/litellm/responses/litellm_completion_transformation/transformation.py#L57) |
| All other llm api providers | [See supported parameters here](https://github.com/BerriAI/litellm/blob/f39d9178868662746f159d5ef642c7f34f9bfe5f/litellm/responses/litellm_completion_transformation/transformation.py#L57) |

146
docs/my-website/docs/tutorials/openai_codex.md Normal file
View file

@ -0,0 +1,146 @@
import Image from '@theme/IdealImage';
import Tabs from '@theme/Tabs';
import TabItem from '@theme/TabItem';
# Using LiteLLM with OpenAI Codex
This guide walks you through connecting OpenAI Codex to LiteLLM. Using LiteLLM with Codex allows teams to:
- Access 100+ LLMs through the Codex interface
- Use powerful models like Gemini through a familiar interface
- Track spend and usage with LiteLLM's built-in analytics
- Control model access with virtual keys
<Image img={require('../../img/litellm_codex.gif')} />
## Quickstart
:::info
Requires LiteLLM v1.66.3.dev5 and higher
:::
Make sure to set up LiteLLM with the [LiteLLM Getting Started Guide](../proxy/docker_quick_start.md).
## 1. Install OpenAI Codex
Install the OpenAI Codex CLI tool globally using npm:
<Tabs>
<TabItem value="npm" label="npm">
```bash showLineNumbers
npm i -g @openai/codex
```
</TabItem>
<TabItem value="yarn" label="yarn">
```bash showLineNumbers
yarn global add @openai/codex
```
</TabItem>
</Tabs>
## 2. Start LiteLLM Proxy
<Tabs>
<TabItem value="docker" label="Docker">
```bash showLineNumbers
docker run \
-v $(pwd)/litellm_config.yaml:/app/config.yaml \
-p 4000:4000 \
ghcr.io/berriai/litellm:main-latest \
--config /app/config.yaml
```
</TabItem>
<TabItem value="pip" label="LiteLLM CLI">
```bash showLineNumbers
litellm --config /path/to/config.yaml
```
</TabItem>
</Tabs>
LiteLLM should now be running on [http://localhost:4000](http://localhost:4000)
## 3. Configure LiteLLM for Model Routing
Ensure your LiteLLM Proxy is properly configured to route to your desired models. Create a `litellm_config.yaml` file with the following content:
```yaml showLineNumbers
model_list:
- model_name: o3-mini
litellm_params:
model: openai/o3-mini
api_key: os.environ/OPENAI_API_KEY
- model_name: claude-3-7-sonnet-latest
litellm_params:
model: anthropic/claude-3-7-sonnet-latest
api_key: os.environ/ANTHROPIC_API_KEY
- model_name: gemini-2.0-flash
litellm_params:
model: gemini/gemini-2.0-flash
api_key: os.environ/GEMINI_API_KEY
litellm_settings:
drop_params: true
```
This configuration enables routing to specific OpenAI, Anthropic, and Gemini models with explicit names.
## 4. Configure Codex to Use LiteLLM Proxy
Set the required environment variables to point Codex to your LiteLLM Proxy:
```bash
# Point to your LiteLLM Proxy server
export OPENAI_BASE_URL=http://0.0.0.0:4000
# Use your LiteLLM API key (if you've set up authentication)
export OPENAI_API_KEY="sk-1234"
```
## 5. Run Codex with Gemini
With everything configured, you can now run Codex with Gemini:
```bash showLineNumbers
codex --model gemini-2.0-flash --full-auto
```
<Image img={require('../../img/litellm_codex.gif')} />
The `--full-auto` flag allows Codex to automatically generate code without additional prompting.
## 6. Advanced Options
### Using Different Models
You can use any model configured in your LiteLLM proxy:
```bash
# Use Claude models
codex --model claude-3-7-sonnet-latest
# Use Google AI Studio Gemini models
codex --model gemini/gemini-2.0-flash
```
## Troubleshooting
- If you encounter connection issues, ensure your LiteLLM Proxy is running and accessible at the specified URL
- Verify your LiteLLM API key is valid if you're using authentication
- Check that your model routing configuration is correct
- For model-specific errors, ensure the model is properly configured in your LiteLLM setup
## Additional Resources
- [LiteLLM Docker Quick Start Guide](../proxy/docker_quick_start.md)
- [OpenAI Codex GitHub Repository](https://github.com/openai/codex)
- [LiteLLM Virtual Keys and Authentication](../proxy/virtual_keys.md)

128
docs/my-website/docs/tutorials/prompt_caching.md Normal file
View file

@ -0,0 +1,128 @@
import Image from '@theme/IdealImage';
import Tabs from '@theme/Tabs';
import TabItem from '@theme/TabItem';
# Auto-Inject Prompt Caching Checkpoints
Reduce costs by up to 90% by using LiteLLM to auto-inject prompt caching checkpoints.
<Image img={require('../../img/auto_prompt_caching.png')} style={{ width: '800px', height: 'auto' }} />
## How it works
LiteLLM can automatically inject prompt caching checkpoints into your requests to LLM providers. This allows:
- **Cost Reduction**: Long, static parts of your prompts can be cached to avoid repeated processing
- **No need to modify your application code**: You can configure the auto-caching behavior in the LiteLLM UI or in the `litellm config.yaml` file.
## Configuration
You need to specify `cache_control_injection_points` in your model configuration. This tells LiteLLM:
1. Where to add the caching directive (`location`)
2. Which message to target (`role`)
LiteLLM will then automatically add a `cache_control` directive to the specified messages in your requests:
```json
"cache_control": {
"type": "ephemeral"
}
```
## Usage Example
In this example, we'll configure caching for system messages by adding the directive to all messages with `role: system`.
<Tabs>
<TabItem value="litellm config.yaml" label="litellm config.yaml">
```yaml showLineNumbers title="litellm config.yaml"
model_list:
- model_name: anthropic-auto-inject-cache-system-message
litellm_params:
model: anthropic/claude-3-5-sonnet-20240620
api_key: os.environ/ANTHROPIC_API_KEY
cache_control_injection_points:
- location: message
role: system
```
</TabItem>
<TabItem value="UI" label="LiteLLM UI">
On the LiteLLM UI, you can specify the `cache_control_injection_points` in the `Advanced Settings` tab when adding a model.
<Image img={require('../../img/ui_auto_prompt_caching.png')}/>
</TabItem>
</Tabs>
## Detailed Example
### 1. Original Request to LiteLLM
In this example, we have a very long, static system message and a varying user message. It's efficient to cache the system message since it rarely changes.
```json
{
"messages": [
{
"role": "system",
"content": [
{
"type": "text",
"text": "You are a helpful assistant. This is a set of very long instructions that you will follow. Here is a legal document that you will use to answer the user's question."
}
]
},
{
"role": "user",
"content": [
{
"type": "text",
"text": "What is the main topic of this legal document?"
}
]
}
]
}
```
### 2. LiteLLM's Modified Request
LiteLLM auto-injects the caching directive into the system message based on our configuration:
```json
{
"messages": [
{
"role": "system",
"content": [
{
"type": "text",
"text": "You are a helpful assistant. This is a set of very long instructions that you will follow. Here is a legal document that you will use to answer the user's question.",
"cache_control": {"type": "ephemeral"}
}
]
},
{
"role": "user",
"content": [
{
"type": "text",
"text": "What is the main topic of this legal document?"
}
]
}
]
}
```
When the model provider processes this request, it will recognize the caching directive and only process the system message once, caching it for subsequent requests.

74
docs/my-website/docs/tutorials/scim_litellm.md Normal file
View file

@ -0,0 +1,74 @@
import Image from '@theme/IdealImage';
# SCIM with LiteLLM
Enables identity providers (Okta, Azure AD, OneLogin, etc.) to automate user and team (group) provisioning, updates, and deprovisioning on LiteLLM.
This tutorial will walk you through the steps to connect your IDP to LiteLLM SCIM Endpoints.
### Supported SSO Providers for SCIM
Below is a list of supported SSO providers for connecting to LiteLLM SCIM Endpoints.
- Microsoft Entra ID (Azure AD)
- Okta
- Google Workspace
- OneLogin
- Keycloak
- Auth0
## 1. Get your SCIM Tenant URL and Bearer Token
On LiteLLM, navigate to the Settings > Admin Settings > SCIM. On this page you will create a SCIM Token, this allows your IDP to authenticate to litellm `/scim` endpoints.
<Image img={require('../../img/scim_2.png')} style={{ width: '800px', height: 'auto' }} />
## 2. Connect your IDP to LiteLLM SCIM Endpoints
On your IDP provider, navigate to your SSO application and select `Provisioning` > `New provisioning configuration`.
On this page, paste in your litellm scim tenant url and bearer token.
Once this is pasted in, click on `Test Connection` to ensure your IDP can authenticate to the LiteLLM SCIM endpoints.
<Image img={require('../../img/scim_4.png')} style={{ width: '800px', height: 'auto' }} />
## 3. Test SCIM Connection
### 3.1 Assign the group to your LiteLLM Enterprise App
On your IDP Portal, navigate to `Enterprise Applications` > Select your litellm app
<Image img={require('../../img/msft_enterprise_app.png')} style={{ width: '800px', height: 'auto' }} />
<br />
<br />
Once you've selected your litellm app, click on `Users and Groups` > `Add user/group`
<Image img={require('../../img/msft_enterprise_assign_group.png')} style={{ width: '800px', height: 'auto' }} />
<br />
Now select the group you created in step 1.1. And add it to the LiteLLM Enterprise App. At this point we have added `Production LLM Evals Group` to the LiteLLM Enterprise App. The next step is having LiteLLM automatically create the `Production LLM Evals Group` on the LiteLLM DB when a new user signs in.
<Image img={require('../../img/msft_enterprise_select_group.png')} style={{ width: '800px', height: 'auto' }} />
### 3.2 Sign in to LiteLLM UI via SSO
Sign into the LiteLLM UI via SSO. You should be redirected to the Entra ID SSO page. This SSO sign in flow will trigger LiteLLM to fetch the latest Groups and Members from Azure Entra ID.
<Image img={require('../../img/msft_sso_sign_in.png')} style={{ width: '800px', height: 'auto' }} />
### 3.3 Check the new team on LiteLLM UI
On the LiteLLM UI, Navigate to `Teams`, You should see the new team `Production LLM Evals Group` auto-created on LiteLLM.
<Image img={require('../../img/msft_auto_team.png')} style={{ width: '900px', height: 'auto' }} />

BIN
docs/my-website/img/auto_prompt_caching.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 1.8 MiB

BIN
docs/my-website/img/litellm_codex.gif Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 12 MiB

BIN
docs/my-website/img/release_notes/new_tag_usage.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 207 KiB

BIN
docs/my-website/img/release_notes/new_team_usage.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 268 KiB

BIN
docs/my-website/img/release_notes/new_team_usage_highlight.jpg Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 999 KiB

BIN
docs/my-website/img/release_notes/unified_responses_api_rn.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 244 KiB

BIN
docs/my-website/img/scim_0.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 380 KiB

BIN
docs/my-website/img/scim_1.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 231 KiB

BIN
docs/my-website/img/scim_2.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 261 KiB

BIN
docs/my-website/img/scim_3.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 413 KiB

BIN
docs/my-website/img/scim_4.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 274 KiB

BIN
docs/my-website/img/scim_integration.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 31 KiB

BIN
docs/my-website/img/ui_auto_prompt_caching.png Normal file
View file

Binary file not shown.
Side by side

After

Width:  |  Height:  |  Size: 103 KiB

7
docs/my-website/package-lock.json generated
View file

@ -12455,9 +12455,10 @@
}
},
"node_modules/http-proxy-middleware": {
"version": "2.0.7",
"resolved": "https://registry.npmjs.org/http-proxy-middleware/-/http-proxy-middleware-2.0.7.tgz",
"integrity": "sha512-fgVY8AV7qU7z/MmXJ/rxwbrtQH4jBQ9m7kp3llF0liB7glmFeVZFBepQb32T3y8n8k2+AEYuMPCpinYW+/CuRA==",
"version": "2.0.9",
"resolved": "https://registry.npmjs.org/http-proxy-middleware/-/http-proxy-middleware-2.0.9.tgz",
"integrity": "sha512-c1IyJYLYppU574+YI7R4QyX2ystMtVXZwIdzazUIPIJsHuWNd+mho2j+bKoHftndicGj9yh+xjd+l0yj7VeT1Q==",
"license": "MIT",
"dependencies": {
"@types/http-proxy": "^1.17.8",
"http-proxy": "^1.18.1",

2
docs/my-website/release_notes/v1.57.8-stable/index.md
View file

@ -38,7 +38,7 @@ hide_table_of_contents: false
2. OpenAI Moderations - `omni-moderation-latest` support. [Start Here](https://docs.litellm.ai/docs/moderation)
3. Azure O1 - fake streaming support. This ensures if a `stream=true` is passed, the response is streamed. [Start Here](https://docs.litellm.ai/docs/providers/azure)
4. Anthropic - non-whitespace char stop sequence handling - [PR](https://github.com/BerriAI/litellm/pull/7484)
5. Azure OpenAI - support Entra id username + password based auth. [Start Here](https://docs.litellm.ai/docs/providers/azure#entrata-id---use-tenant_id-client_id-client_secret)
5. Azure OpenAI - support Entra ID username + password based auth. [Start Here](https://docs.litellm.ai/docs/providers/azure#entra-id---use-tenant_id-client_id-client_secret)
6. LM Studio - embedding route support. [Start Here](https://docs.litellm.ai/docs/providers/lm-studio)
7. WatsonX - ZenAPIKeyAuth support. [Start Here](https://docs.litellm.ai/docs/providers/watsonx)

153
docs/my-website/release_notes/v1.67.0-stable/index.md Normal file
View file

@ -0,0 +1,153 @@
---
title: v1.67.0-stable - SCIM Integration
slug: v1.67.0-stable
date: 2025-04-19T10:00:00
authors:
- name: Krrish Dholakia
title: CEO, LiteLLM
url: https://www.linkedin.com/in/krish-d/
image_url: https://media.licdn.com/dms/image/v2/D4D03AQGrlsJ3aqpHmQ/profile-displayphoto-shrink_400_400/B4DZSAzgP7HYAg-/0/1737327772964?e=1749686400&v=beta&t=Hkl3U8Ps0VtvNxX0BNNq24b4dtX5wQaPFp6oiKCIHD8
- name: Ishaan Jaffer
title: CTO, LiteLLM
url: https://www.linkedin.com/in/reffajnaahsi/
image_url: https://pbs.twimg.com/profile_images/1613813310264340481/lz54oEiB_400x400.jpg
tags: ["sso", "unified_file_id", "cost_tracking", "security"]
hide_table_of_contents: false
---
import Image from '@theme/IdealImage';
import Tabs from '@theme/Tabs';
import TabItem from '@theme/TabItem';
## Key Highlights
- **SCIM Integration**: Enables identity providers (Okta, Azure AD, OneLogin, etc.) to automate user and team (group) provisioning, updates, and deprovisioning
- **Team and Tag based usage tracking**: You can now see usage and spend by team and tag at 1M+ spend logs.
- **Unified Responses API**: Support for calling Anthropic, Gemini, Groq, etc. via OpenAI's new Responses API.
Let's dive in.
## SCIM Integration
<Image img={require('../../img/scim_integration.png')}/>
This release adds SCIM support to LiteLLM. This allows your SSO provider (Okta, Azure AD, etc) to automatically create/delete users, teams, and memberships on LiteLLM. This means that when you remove a team on your SSO provider, your SSO provider will automatically delete the corresponding team on LiteLLM.
[Read more](../../docs/tutorials/scim_litellm)
## Team and Tag based usage tracking
<Image img={require('../../img/release_notes/new_team_usage_highlight.jpg')}/>
This release improves team and tag based usage tracking at 1m+ spend logs, making it easy to monitor your LLM API Spend in production. This covers:
- View **daily spend** by teams + tags
- View **usage / spend by key**, within teams
- View **spend by multiple tags**
- Allow **internal users** to view spend of teams they're a member of
[Read more](#management-endpoints--ui)
## Unified Responses API
This release allows you to call Azure OpenAI, Anthropic, AWS Bedrock, and Google Vertex AI models via the POST /v1/responses endpoint on LiteLLM. This means you can now use popular tools like [OpenAI Codex](https://docs.litellm.ai/docs/tutorials/openai_codex) with your own models.
<Image img={require('../../img/release_notes/unified_responses_api_rn.png')}/>
[Read more](https://docs.litellm.ai/docs/response_api)
## New Models / Updated Models
- **OpenAI**
1. gpt-4.1, gpt-4.1-mini, gpt-4.1-nano, o3, o3-mini, o4-mini pricing - [Get Started](../../docs/providers/openai#usage), [PR](https://github.com/BerriAI/litellm/pull/9990)
2. o4 - correctly map o4 to openai o_series model
- **Azure AI**
1. Phi-4 output cost per token fix - [PR](https://github.com/BerriAI/litellm/pull/9880)
2. Responses API support [Get Started](../../docs/providers/azure#azure-responses-api),[PR](https://github.com/BerriAI/litellm/pull/10116)
- **Anthropic**
1. redacted message thinking support - [Get Started](../../docs/providers/anthropic#usage---thinking--reasoning_content),[PR](https://github.com/BerriAI/litellm/pull/10129)
- **Cohere**
1. `/v2/chat` Passthrough endpoint support w/ cost tracking - [Get Started](../../docs/pass_through/cohere), [PR](https://github.com/BerriAI/litellm/pull/9997)
- **Azure**
1. Support azure tenant_id/client_id env vars - [Get Started](../../docs/providers/azure#entra-id---use-tenant_id-client_id-client_secret), [PR](https://github.com/BerriAI/litellm/pull/9993)
2. Fix response_format check for 2025+ api versions - [PR](https://github.com/BerriAI/litellm/pull/9993)
3. Add gpt-4.1, gpt-4.1-mini, gpt-4.1-nano, o3, o3-mini, o4-mini pricing
- **VLLM**
1. Files - Support 'file' message type for VLLM video url's - [Get Started](../../docs/providers/vllm#send-video-url-to-vllm), [PR](https://github.com/BerriAI/litellm/pull/10129)
2. Passthrough - new `/vllm/` passthrough endpoint support [Get Started](../../docs/pass_through/vllm), [PR](https://github.com/BerriAI/litellm/pull/10002)
- **Mistral**
1. new `/mistral` passthrough endpoint support [Get Started](../../docs/pass_through/mistral), [PR](https://github.com/BerriAI/litellm/pull/10002)
- **AWS**
1. New mapped bedrock regions - [PR](https://github.com/BerriAI/litellm/pull/9430)
- **VertexAI / Google AI Studio**
1. Gemini - Response format - Retain schema field ordering for google gemini and vertex by specifying propertyOrdering - [Get Started](../../docs/providers/vertex#json-schema), [PR](https://github.com/BerriAI/litellm/pull/9828)
2. Gemini-2.5-flash - return reasoning content [Google AI Studio](../../docs/providers/gemini#usage---thinking--reasoning_content), [Vertex AI](../../docs/providers/vertex#thinking--reasoning_content)
3. Gemini-2.5-flash - pricing + model information [PR](https://github.com/BerriAI/litellm/pull/10125)
4. Passthrough - new `/vertex_ai/discovery` route - enables calling AgentBuilder API routes [Get Started](../../docs/pass_through/vertex_ai#supported-api-endpoints), [PR](https://github.com/BerriAI/litellm/pull/10084)
- **Fireworks AI**
1. return tool calling responses in `tool_calls` field (fireworks incorrectly returns this as a json str in content) [PR](https://github.com/BerriAI/litellm/pull/10130)
- **Triton**
1. Remove fixed remove bad_words / stop words from `/generate` call - [Get Started](../../docs/providers/triton-inference-server#triton-generate---chat-completion), [PR](https://github.com/BerriAI/litellm/pull/10163)
- **Other**
1. Support for all litellm providers on Responses API (works with Codex) - [Get Started](../../docs/tutorials/openai_codex), [PR](https://github.com/BerriAI/litellm/pull/10132)
2. Fix combining multiple tool calls in streaming response - [Get Started](../../docs/completion/stream#helper-function), [PR](https://github.com/BerriAI/litellm/pull/10040)
## Spend Tracking Improvements
- **Cost Control** - inject cache control points in prompt for cost reduction [Get Started](../../docs/tutorials/prompt_caching), [PR](https://github.com/BerriAI/litellm/pull/10000)
- **Spend Tags** - spend tags in headers - support x-litellm-tags even if tag based routing not enabled [Get Started](../../docs/proxy/request_headers#litellm-headers), [PR](https://github.com/BerriAI/litellm/pull/10000)
- **Gemini-2.5-flash** - support cost calculation for reasoning tokens [PR](https://github.com/BerriAI/litellm/pull/10141)
## Management Endpoints / UI
- **Users**
1. Show created_at and updated_at on users page - [PR](https://github.com/BerriAI/litellm/pull/10033)
- **Virtual Keys**
1. Filter by key alias - https://github.com/BerriAI/litellm/pull/10085
- **Usage Tab**
1. Team based usage
- New `LiteLLM_DailyTeamSpend` Table for aggregate team based usage logging - [PR](https://github.com/BerriAI/litellm/pull/10039)
- New Team based usage dashboard + new `/team/daily/activity` API - [PR](https://github.com/BerriAI/litellm/pull/10081)
- Return team alias on /team/daily/activity API - [PR](https://github.com/BerriAI/litellm/pull/10157)
- allow internal user view spend for teams they belong to - [PR](https://github.com/BerriAI/litellm/pull/10157)
- allow viewing top keys by team - [PR](https://github.com/BerriAI/litellm/pull/10157)
<Image img={require('../../img/release_notes/new_team_usage.png')}/>
2. Tag Based Usage
- New `LiteLLM_DailyTagSpend` Table for aggregate tag based usage logging - [PR](https://github.com/BerriAI/litellm/pull/10071)
- Restrict to only Proxy Admins - [PR](https://github.com/BerriAI/litellm/pull/10157)
- allow viewing top keys by tag
- Return tags passed in request (i.e. dynamic tags) on `/tag/list` API - [PR](https://github.com/BerriAI/litellm/pull/10157)
<Image img={require('../../img/release_notes/new_tag_usage.png')}/>
3. Track prompt caching metrics in daily user, team, tag tables - [PR](https://github.com/BerriAI/litellm/pull/10029)
4. Show usage by key (on all up, team, and tag usage dashboards) - [PR](https://github.com/BerriAI/litellm/pull/10157)
5. swap old usage with new usage tab
- **Models**
1. Make columns resizable/hideable - [PR](https://github.com/BerriAI/litellm/pull/10119)
- **API Playground**
1. Allow internal user to call api playground - [PR](https://github.com/BerriAI/litellm/pull/10157)
- **SCIM**
1. Add LiteLLM SCIM Integration for Team and User management - [Get Started](../../docs/tutorials/scim_litellm), [PR](https://github.com/BerriAI/litellm/pull/10072)
## Logging / Guardrail Integrations
- **GCS**
1. Fix gcs pub sub logging with env var GCS_PROJECT_ID - [Get Started](../../docs/observability/gcs_bucket_integration#usage), [PR](https://github.com/BerriAI/litellm/pull/10042)
- **AIM**
1. Add litellm call id passing to Aim guardrails on pre and post-hooks calls - [Get Started](../../docs/proxy/guardrails/aim_security), [PR](https://github.com/BerriAI/litellm/pull/10021)
- **Azure blob storage**
1. Ensure logging works in high throughput scenarios - [Get Started](../../docs/proxy/logging#azure-blob-storage), [PR](https://github.com/BerriAI/litellm/pull/9962)
## General Proxy Improvements
- **Support setting `litellm.modify_params` via env var** [PR](https://github.com/BerriAI/litellm/pull/9964)
- **Model Discovery** - Check provider’s `/models` endpoints when calling proxy’s `/v1/models` endpoint - [Get Started](../../docs/proxy/model_discovery), [PR](https://github.com/BerriAI/litellm/pull/9958)
- **`/utils/token_counter`** - fix retrieving custom tokenizer for db models - [Get Started](../../docs/proxy/configs#set-custom-tokenizer), [PR](https://github.com/BerriAI/litellm/pull/10047)
- **Prisma migrate** - handle existing columns in db table - [PR](https://github.com/BerriAI/litellm/pull/10138)

6
docs/my-website/sidebars.js
View file

@ -69,6 +69,7 @@ const sidebars = {
"proxy/clientside_auth",
"proxy/request_headers",
"proxy/response_headers",
"proxy/model_discovery",
],
},
{
@ -101,6 +102,7 @@ const sidebars = {
"proxy/admin_ui_sso",
"proxy/self_serve",
"proxy/public_teams",
"tutorials/scim_litellm",
"proxy/custom_sso",
"proxy/ui_credentials",
"proxy/ui_logs"
@ -330,6 +332,8 @@ const sidebars = {
"pass_through/vertex_ai",
"pass_through/google_ai_studio",
"pass_through/cohere",
"pass_through/vllm",
"pass_through/mistral",
"pass_through/openai_passthrough",
"pass_through/anthropic_completion",
"pass_through/bedrock",
@ -443,7 +447,9 @@ const sidebars = {
label: "Tutorials",
items: [
"tutorials/openweb_ui",
"tutorials/openai_codex",
"tutorials/msft_sso",
"tutorials/prompt_caching",
"tutorials/tag_management",
'tutorials/litellm_proxy_aporia',
{

BIN
litellm-proxy-extras/dist/litellm_proxy_extras-0.1.8-py3-none-any.whl vendored Normal file
View file

Binary file not shown.

BIN
litellm-proxy-extras/dist/litellm_proxy_extras-0.1.8.tar.gz vendored Normal file
View file

Binary file not shown.

4
litellm-proxy-extras/litellm_proxy_extras/migrations/20250415151647_add_cache_read_write_tokens_daily_spend_transactions/migration.sql Normal file
View file

@ -0,0 +1,4 @@
-- AlterTable
ALTER TABLE "LiteLLM_DailyUserSpend" ADD COLUMN "cache_creation_input_tokens" INTEGER NOT NULL DEFAULT 0,
ADD COLUMN "cache_read_input_tokens" INTEGER NOT NULL DEFAULT 0;

36
litellm-proxy-extras/litellm_proxy_extras/migrations/20250415191926_add_daily_team_table/migration.sql Normal file
View file

@ -0,0 +1,36 @@
-- CreateTable
CREATE TABLE "LiteLLM_DailyTeamSpend" (
"id" TEXT NOT NULL,
"team_id" TEXT NOT NULL,
"date" TEXT NOT NULL,
"api_key" TEXT NOT NULL,
"model" TEXT NOT NULL,
"model_group" TEXT,
"custom_llm_provider" TEXT,
"prompt_tokens" INTEGER NOT NULL DEFAULT 0,
"completion_tokens" INTEGER NOT NULL DEFAULT 0,
"spend" DOUBLE PRECISION NOT NULL DEFAULT 0.0,
"api_requests" INTEGER NOT NULL DEFAULT 0,
"successful_requests" INTEGER NOT NULL DEFAULT 0,
"failed_requests" INTEGER NOT NULL DEFAULT 0,
"created_at" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
"updated_at" TIMESTAMP(3) NOT NULL,
CONSTRAINT "LiteLLM_DailyTeamSpend_pkey" PRIMARY KEY ("id")
);
-- CreateIndex
CREATE INDEX "LiteLLM_DailyTeamSpend_date_idx" ON "LiteLLM_DailyTeamSpend"("date");
-- CreateIndex
CREATE INDEX "LiteLLM_DailyTeamSpend_team_id_idx" ON "LiteLLM_DailyTeamSpend"("team_id");
-- CreateIndex
CREATE INDEX "LiteLLM_DailyTeamSpend_api_key_idx" ON "LiteLLM_DailyTeamSpend"("api_key");
-- CreateIndex
CREATE INDEX "LiteLLM_DailyTeamSpend_model_idx" ON "LiteLLM_DailyTeamSpend"("model");
-- CreateIndex
CREATE UNIQUE INDEX "LiteLLM_DailyTeamSpend_team_id_date_api_key_model_custom_ll_key" ON "LiteLLM_DailyTeamSpend"("team_id", "date", "api_key", "model", "custom_llm_provider");

45
litellm-proxy-extras/litellm_proxy_extras/migrations/20250416115320_add_tag_table_to_db/migration.sql Normal file
View file

@ -0,0 +1,45 @@
-- AlterTable
ALTER TABLE "LiteLLM_DailyTeamSpend" ADD COLUMN "cache_creation_input_tokens" INTEGER NOT NULL DEFAULT 0,
ADD COLUMN "cache_read_input_tokens" INTEGER NOT NULL DEFAULT 0;
-- CreateTable
CREATE TABLE "LiteLLM_DailyTagSpend" (
"id" TEXT NOT NULL,
"tag" TEXT NOT NULL,
"date" TEXT NOT NULL,
"api_key" TEXT NOT NULL,
"model" TEXT NOT NULL,
"model_group" TEXT,
"custom_llm_provider" TEXT,
"prompt_tokens" INTEGER NOT NULL DEFAULT 0,
"completion_tokens" INTEGER NOT NULL DEFAULT 0,
"cache_read_input_tokens" INTEGER NOT NULL DEFAULT 0,
"cache_creation_input_tokens" INTEGER NOT NULL DEFAULT 0,
"spend" DOUBLE PRECISION NOT NULL DEFAULT 0.0,
"api_requests" INTEGER NOT NULL DEFAULT 0,
"successful_requests" INTEGER NOT NULL DEFAULT 0,
"failed_requests" INTEGER NOT NULL DEFAULT 0,
"created_at" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
"updated_at" TIMESTAMP(3) NOT NULL,
CONSTRAINT "LiteLLM_DailyTagSpend_pkey" PRIMARY KEY ("id")
);
-- CreateIndex
CREATE UNIQUE INDEX "LiteLLM_DailyTagSpend_tag_key" ON "LiteLLM_DailyTagSpend"("tag");
-- CreateIndex
CREATE INDEX "LiteLLM_DailyTagSpend_date_idx" ON "LiteLLM_DailyTagSpend"("date");
-- CreateIndex
CREATE INDEX "LiteLLM_DailyTagSpend_tag_idx" ON "LiteLLM_DailyTagSpend"("tag");
-- CreateIndex
CREATE INDEX "LiteLLM_DailyTagSpend_api_key_idx" ON "LiteLLM_DailyTagSpend"("api_key");
-- CreateIndex
CREATE INDEX "LiteLLM_DailyTagSpend_model_idx" ON "LiteLLM_DailyTagSpend"("model");
-- CreateIndex
CREATE UNIQUE INDEX "LiteLLM_DailyTagSpend_tag_date_api_key_model_custom_llm_pro_key" ON "LiteLLM_DailyTagSpend"("tag", "date", "api_key", "model", "custom_llm_provider");

3
litellm-proxy-extras/litellm_proxy_extras/migrations/20250416151339_drop_tag_uniqueness_requirement/migration.sql Normal file
View file

@ -0,0 +1,3 @@
-- DropIndex
DROP INDEX "LiteLLM_DailyTagSpend_tag_key";

3
litellm-proxy-extras/litellm_proxy_extras/migrations/20250416185146_add_allowed_routes_litellm_verification_token/migration.sql Normal file
View file

@ -0,0 +1,3 @@
-- AlterTable
ALTER TABLE "LiteLLM_VerificationToken" ADD COLUMN "allowed_routes" TEXT[] DEFAULT ARRAY[]::TEXT[];

57
litellm-proxy-extras/litellm_proxy_extras/schema.prisma
View file

@ -169,6 +169,7 @@ model LiteLLM_VerificationToken {
budget_duration String?
budget_reset_at DateTime?
allowed_cache_controls String[] @default([])
allowed_routes String[] @default([])
model_spend Json @default("{}")
model_max_budget Json @default("{}")
budget_id String?
@ -326,6 +327,8 @@ model LiteLLM_DailyUserSpend {
custom_llm_provider String?
prompt_tokens Int @default(0)
completion_tokens Int @default(0)
cache_read_input_tokens Int @default(0)
cache_creation_input_tokens Int @default(0)
spend Float @default(0.0)
api_requests Int @default(0)
successful_requests Int @default(0)
@ -340,6 +343,60 @@ model LiteLLM_DailyUserSpend {
@@index([model])
}
// Track daily team spend metrics per model and key
model LiteLLM_DailyTeamSpend {
id String @id @default(uuid())
team_id String
date String
api_key String
model String
model_group String?
custom_llm_provider String?
prompt_tokens Int @default(0)
completion_tokens Int @default(0)
cache_read_input_tokens Int @default(0)
cache_creation_input_tokens Int @default(0)
spend Float @default(0.0)
api_requests Int @default(0)
successful_requests Int @default(0)
failed_requests Int @default(0)
created_at DateTime @default(now())
updated_at DateTime @updatedAt
@@unique([team_id, date, api_key, model, custom_llm_provider])
@@index([date])
@@index([team_id])
@@index([api_key])
@@index([model])
}
// Track daily team spend metrics per model and key
model LiteLLM_DailyTagSpend {
id String @id @default(uuid())
tag String
date String
api_key String
model String
model_group String?
custom_llm_provider String?
prompt_tokens Int @default(0)
completion_tokens Int @default(0)
cache_read_input_tokens Int @default(0)
cache_creation_input_tokens Int @default(0)
spend Float @default(0.0)
api_requests Int @default(0)
successful_requests Int @default(0)
failed_requests Int @default(0)
created_at DateTime @default(now())
updated_at DateTime @updatedAt
@@unique([tag, date, api_key, model, custom_llm_provider])
@@index([date])
@@index([tag])
@@index([api_key])
@@index([model])
}
// Track the status of cron jobs running. Only allow one pod to run the job at a time
model LiteLLM_CronJob {

73
litellm-proxy-extras/litellm_proxy_extras/utils.py
View file

@ -1,6 +1,7 @@
import glob
import os
import random
import re
import subprocess
import time
from pathlib import Path
@ -82,6 +83,26 @@ class ProxyExtrasDBManager:
logger.info(f"Found {len(migration_paths)} migrations at {migrations_dir}")
return [Path(p).parent.name for p in migration_paths]
@staticmethod
def _roll_back_migration(migration_name: str):
"""Mark a specific migration as rolled back"""
subprocess.run(
["prisma", "migrate", "resolve", "--rolled-back", migration_name],
timeout=60,
check=True,
capture_output=True,
)
@staticmethod
def _resolve_specific_migration(migration_name: str):
"""Mark a specific migration as applied"""
subprocess.run(
["prisma", "migrate", "resolve", "--applied", migration_name],
timeout=60,
check=True,
capture_output=True,
)
@staticmethod
def _resolve_all_migrations(migrations_dir: str):
"""Mark all existing migrations as applied"""
@ -141,7 +162,34 @@ class ProxyExtrasDBManager:
return True
except subprocess.CalledProcessError as e:
logger.info(f"prisma db error: {e.stderr}, e: {e.stdout}")
if (
if "P3009" in e.stderr:
# Extract the failed migration name from the error message
migration_match = re.search(
r"`(\d+_.*)` migration", e.stderr
)
if migration_match:
failed_migration = migration_match.group(1)
logger.info(
f"Found failed migration: {failed_migration}, marking as rolled back"
)
# Mark the failed migration as rolled back
subprocess.run(
[
"prisma",
"migrate",
"resolve",
"--rolled-back",
failed_migration,
],
timeout=60,
check=True,
capture_output=True,
text=True,
)
logger.info(
f"✅ Migration {failed_migration} marked as rolled back... retrying"
)
elif (
"P3005" in e.stderr
and "database schema is not empty" in e.stderr
):
@ -155,6 +203,29 @@ class ProxyExtrasDBManager:
ProxyExtrasDBManager._resolve_all_migrations(migrations_dir)
logger.info("✅ All migrations resolved.")
return True
elif (
"P3018" in e.stderr
): # PostgreSQL error code for duplicate column
logger.info(
"Migration already exists, resolving specific migration"
)
# Extract the migration name from the error message
migration_match = re.search(
r"Migration name: (\d+_.*)", e.stderr
)
if migration_match:
migration_name = migration_match.group(1)
logger.info(f"Rolling back migration {migration_name}")
ProxyExtrasDBManager._roll_back_migration(
migration_name
)
logger.info(
f"Resolving migration {migration_name} that failed due to existing columns"
)
ProxyExtrasDBManager._resolve_specific_migration(
migration_name
)
logger.info("✅ Migration resolved.")
else:
# Use prisma db push with increased timeout
subprocess.run(

4
litellm-proxy-extras/poetry.lock generated
View file

@ -1,7 +1,7 @@
# This file is automatically @generated by Poetry 1.8.3 and should not be changed by hand.
# This file is automatically @generated by Poetry 2.1.2 and should not be changed by hand.
package = []
[metadata]
lock-version = "2.0"
lock-version = "2.1"
python-versions = ">=3.8.1,<4.0, !=3.9.7"
content-hash = "2cf39473e67ff0615f0a61c9d2ac9f02b38cc08cbb1bdb893d89bee002646623"

4
litellm-proxy-extras/pyproject.toml
View file

@ -1,6 +1,6 @@
[tool.poetry]
name = "litellm-proxy-extras"
version = "0.1.7"
version = "0.1.11"
description = "Additional files for the LiteLLM Proxy. Reduces the size of the main litellm package."
authors = ["BerriAI"]
readme = "README.md"
@ -22,7 +22,7 @@ requires = ["poetry-core"]
build-backend = "poetry.core.masonry.api"
[tool.commitizen]
version = "0.1.7"
version = "0.1.11"
version_files = [
"pyproject.toml:version",
"../requirements.txt:litellm-proxy-extras==",

99
litellm/__init__.py
View file

@ -128,19 +128,19 @@ prometheus_initialize_budget_metrics: Optional[bool] = False
require_auth_for_metrics_endpoint: Optional[bool] = False
argilla_batch_size: Optional[int] = None
datadog_use_v1: Optional[bool] = False # if you want to use v1 datadog logged payload
gcs_pub_sub_use_v1: Optional[
bool
] = False # if you want to use v1 gcs pubsub logged payload
gcs_pub_sub_use_v1: Optional[bool] = (
False # if you want to use v1 gcs pubsub logged payload
)
argilla_transformation_object: Optional[Dict[str, Any]] = None
_async_input_callback: List[
Union[str, Callable, CustomLogger]
] = [] # internal variable - async custom callbacks are routed here.
_async_success_callback: List[
Union[str, Callable, CustomLogger]
] = [] # internal variable - async custom callbacks are routed here.
_async_failure_callback: List[
Union[str, Callable, CustomLogger]
] = [] # internal variable - async custom callbacks are routed here.
_async_input_callback: List[Union[str, Callable, CustomLogger]] = (
[]
) # internal variable - async custom callbacks are routed here.
_async_success_callback: List[Union[str, Callable, CustomLogger]] = (
[]
) # internal variable - async custom callbacks are routed here.
_async_failure_callback: List[Union[str, Callable, CustomLogger]] = (
[]
) # internal variable - async custom callbacks are routed here.
pre_call_rules: List[Callable] = []
post_call_rules: List[Callable] = []
turn_off_message_logging: Optional[bool] = False
@ -148,18 +148,18 @@ log_raw_request_response: bool = False
redact_messages_in_exceptions: Optional[bool] = False
redact_user_api_key_info: Optional[bool] = False
filter_invalid_headers: Optional[bool] = False
add_user_information_to_llm_headers: Optional[
bool
] = None # adds user_id, team_id, token hash (params from StandardLoggingMetadata) to request headers
add_user_information_to_llm_headers: Optional[bool] = (
None # adds user_id, team_id, token hash (params from StandardLoggingMetadata) to request headers
)
store_audit_logs = False # Enterprise feature, allow users to see audit logs
### end of callbacks #############
email: Optional[
str
] = None # Not used anymore, will be removed in next MAJOR release - https://github.com/BerriAI/litellm/discussions/648
token: Optional[
str
] = None # Not used anymore, will be removed in next MAJOR release - https://github.com/BerriAI/litellm/discussions/648
email: Optional[str] = (
None # Not used anymore, will be removed in next MAJOR release - https://github.com/BerriAI/litellm/discussions/648
)
token: Optional[str] = (
None # Not used anymore, will be removed in next MAJOR release - https://github.com/BerriAI/litellm/discussions/648
)
telemetry = True
max_tokens: int = DEFAULT_MAX_TOKENS # OpenAI Defaults
drop_params = bool(os.getenv("LITELLM_DROP_PARAMS", False))
@ -235,20 +235,24 @@ enable_loadbalancing_on_batch_endpoints: Optional[bool] = None
enable_caching_on_provider_specific_optional_params: bool = (
False # feature-flag for caching on optional params - e.g. 'top_k'
)
caching: bool = False # Not used anymore, will be removed in next MAJOR release - https://github.com/BerriAI/litellm/discussions/648
caching_with_models: bool = False # # Not used anymore, will be removed in next MAJOR release - https://github.com/BerriAI/litellm/discussions/648
cache: Optional[
Cache
] = None # cache object <- use this - https://docs.litellm.ai/docs/caching
caching: bool = (
False # Not used anymore, will be removed in next MAJOR release - https://github.com/BerriAI/litellm/discussions/648
)
caching_with_models: bool = (
False # # Not used anymore, will be removed in next MAJOR release - https://github.com/BerriAI/litellm/discussions/648
)
cache: Optional[Cache] = (
None # cache object <- use this - https://docs.litellm.ai/docs/caching
)
default_in_memory_ttl: Optional[float] = None
default_redis_ttl: Optional[float] = None
default_redis_batch_cache_expiry: Optional[float] = None
model_alias_map: Dict[str, str] = {}
model_group_alias_map: Dict[str, str] = {}
max_budget: float = 0.0 # set the max budget across all providers
budget_duration: Optional[
str
] = None # proxy only - resets budget after fixed duration. You can set duration as seconds ("30s"), minutes ("30m"), hours ("30h"), days ("30d").
budget_duration: Optional[str] = (
None # proxy only - resets budget after fixed duration. You can set duration as seconds ("30s"), minutes ("30m"), hours ("30h"), days ("30d").
)
default_soft_budget: float = (
DEFAULT_SOFT_BUDGET # by default all litellm proxy keys have a soft budget of 50.0
)
@ -257,11 +261,15 @@ forward_traceparent_to_llm_provider: bool = False
_current_cost = 0.0 # private variable, used if max budget is set
error_logs: Dict = {}
add_function_to_prompt: bool = False # if function calling not supported by api, append function call details to system prompt
add_function_to_prompt: bool = (
False # if function calling not supported by api, append function call details to system prompt
)
client_session: Optional[httpx.Client] = None
aclient_session: Optional[httpx.AsyncClient] = None
model_fallbacks: Optional[List] = None # Deprecated for 'litellm.fallbacks'
model_cost_map_url: str = "https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
model_cost_map_url: str = (
"https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json"
)
suppress_debug_info = False
dynamodb_table_name: Optional[str] = None
s3_callback_params: Optional[Dict] = None
@ -284,7 +292,9 @@ disable_end_user_cost_tracking_prometheus_only: Optional[bool] = None
custom_prometheus_metadata_labels: List[str] = []
#### REQUEST PRIORITIZATION ####
priority_reservation: Optional[Dict[str, float]] = None
force_ipv4: bool = False # when True, litellm will force ipv4 for all LLM requests. Some users have seen httpx ConnectionError when using ipv6.
force_ipv4: bool = (
False # when True, litellm will force ipv4 for all LLM requests. Some users have seen httpx ConnectionError when using ipv6.
)
module_level_aclient = AsyncHTTPHandler(
timeout=request_timeout, client_alias="module level aclient"
)
@ -298,13 +308,13 @@ fallbacks: Optional[List] = None
context_window_fallbacks: Optional[List] = None
content_policy_fallbacks: Optional[List] = None
allowed_fails: int = 3
num_retries_per_request: Optional[
int
] = None # for the request overall (incl. fallbacks + model retries)
num_retries_per_request: Optional[int] = (
None # for the request overall (incl. fallbacks + model retries)
)
####### SECRET MANAGERS #####################
secret_manager_client: Optional[
Any
] = None # list of instantiated key management clients - e.g. azure kv, infisical, etc.
secret_manager_client: Optional[Any] = (
None # list of instantiated key management clients - e.g. azure kv, infisical, etc.
)
_google_kms_resource_name: Optional[str] = None
_key_management_system: Optional[KeyManagementSystem] = None
_key_management_settings: KeyManagementSettings = KeyManagementSettings()
@ -945,6 +955,7 @@ from .llms.infinity.embedding.transformation import InfinityEmbeddingConfig
from .llms.azure_ai.chat.transformation import AzureAIStudioConfig
from .llms.mistral.mistral_chat_transformation import MistralConfig
from .llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from .llms.azure.responses.transformation import AzureOpenAIResponsesAPIConfig
from .llms.openai.chat.o_series_transformation import (
OpenAIOSeriesConfig as OpenAIO1Config, # maintain backwards compatibility
OpenAIOSeriesConfig,
@ -1061,10 +1072,10 @@ from .types.llms.custom_llm import CustomLLMItem
from .types.utils import GenericStreamingChunk
custom_provider_map: List[CustomLLMItem] = []
_custom_providers: List[
str
] = [] # internal helper util, used to track names of custom providers
disable_hf_tokenizer_download: Optional[
bool
] = None # disable huggingface tokenizer download. Defaults to openai clk100
_custom_providers: List[str] = (
[]
) # internal helper util, used to track names of custom providers
disable_hf_tokenizer_download: Optional[bool] = (
None # disable huggingface tokenizer download. Defaults to openai clk100
)
global_disable_no_log_param: bool = False

5
litellm/assistants/main.py
View file

@ -304,6 +304,11 @@ def create_assistants(
"response_format": response_format,
}
# only send params that are not None
create_assistant_data = {
k: v for k, v in create_assistant_data.items() if v is not None
}
response: Optional[Union[Coroutine[Any, Any, Assistant], Assistant]] = None
if custom_llm_provider == "openai":
api_base = (

9
litellm/constants.py
View file

@ -21,9 +21,18 @@ DEFAULT_MAX_TOKENS = 256 # used when providers need a default
MAX_SIZE_PER_ITEM_IN_MEMORY_CACHE_IN_KB = 1024 # 1MB = 1024KB
SINGLE_DEPLOYMENT_TRAFFIC_FAILURE_THRESHOLD = 1000 # Minimum number of requests to consider "reasonable traffic". Used for single-deployment cooldown logic.
DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET = 1024
DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET = 2048
DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET = 4096
########## Networking constants ##############################################################
_DEFAULT_TTL_FOR_HTTPX_CLIENTS = 3600 # 1 hour, re-use the same httpx client for 1 hour
########### v2 Architecture constants for managing writing updates to the database ###########
REDIS_UPDATE_BUFFER_KEY = "litellm_spend_update_buffer"
REDIS_DAILY_SPEND_UPDATE_BUFFER_KEY = "litellm_daily_spend_update_buffer"
REDIS_DAILY_TEAM_SPEND_UPDATE_BUFFER_KEY = "litellm_daily_team_spend_update_buffer"
REDIS_DAILY_TAG_SPEND_UPDATE_BUFFER_KEY = "litellm_daily_tag_spend_update_buffer"
MAX_REDIS_BUFFER_DEQUEUE_COUNT = 100
MAX_SIZE_IN_MEMORY_QUEUE = 10000
MAX_IN_MEMORY_QUEUE_FLUSH_COUNT = 1000

48
litellm/integrations/azure_storage/azure_storage.py
View file

@ -1,14 +1,15 @@
import asyncio
import json
import os
import time
import uuid
from datetime import datetime, timedelta
from typing import List, Optional
from litellm._logging import verbose_logger
from litellm.constants import AZURE_STORAGE_MSFT_VERSION
from litellm.constants import _DEFAULT_TTL_FOR_HTTPX_CLIENTS, AZURE_STORAGE_MSFT_VERSION
from litellm.integrations.custom_batch_logger import CustomBatchLogger
from litellm.llms.azure.common_utils import get_azure_ad_token_from_entrata_id
from litellm.llms.azure.common_utils import get_azure_ad_token_from_entra_id
from litellm.llms.custom_httpx.http_handler import (
AsyncHTTPHandler,
get_async_httpx_client,
@ -48,14 +49,17 @@ class AzureBlobStorageLogger(CustomBatchLogger):
"Missing required environment variable: AZURE_STORAGE_FILE_SYSTEM"
)
self.azure_storage_file_system: str = _azure_storage_file_system
self._service_client = None
# Time that the azure service client expires, in order to reset the connection pool and keep it fresh
self._service_client_timeout: Optional[float] = None
# Internal variables used for Token based authentication
self.azure_auth_token: Optional[
str
] = None # the Azure AD token to use for Azure Storage API requests
self.token_expiry: Optional[
datetime
] = None # the expiry time of the currentAzure AD token
self.azure_auth_token: Optional[str] = (
None # the Azure AD token to use for Azure Storage API requests
)
self.token_expiry: Optional[datetime] = (
None # the expiry time of the currentAzure AD token
)
asyncio.create_task(self.periodic_flush())
self.flush_lock = asyncio.Lock()
@ -291,7 +295,7 @@ class AzureBlobStorageLogger(CustomBatchLogger):
"Missing required environment variable: AZURE_STORAGE_CLIENT_SECRET"
)
token_provider = get_azure_ad_token_from_entrata_id(
token_provider = get_azure_ad_token_from_entra_id(
tenant_id=tenant_id,
client_id=client_id,
client_secret=client_secret,
@ -324,6 +328,25 @@ class AzureBlobStorageLogger(CustomBatchLogger):
f"AzureBlobStorageLogger is only available for premium users. {CommonProxyErrors.not_premium_user}"
)
async def get_service_client(self):
from azure.storage.filedatalake.aio import DataLakeServiceClient
# expire old clients to recover from connection issues
if (
self._service_client_timeout
and self._service_client
and self._service_client_timeout > time.time()
):
await self._service_client.close()
self._service_client = None
if not self._service_client:
self._service_client = DataLakeServiceClient(
account_url=f"https://{self.azure_storage_account_name}.dfs.core.windows.net",
credential=self.azure_storage_account_key,
)
self._service_client_timeout = time.time() + _DEFAULT_TTL_FOR_HTTPX_CLIENTS
return self._service_client
async def upload_to_azure_data_lake_with_azure_account_key(
self, payload: StandardLoggingPayload
):
@ -332,13 +355,10 @@ class AzureBlobStorageLogger(CustomBatchLogger):
This is used when Azure Storage Account Key is set - Azure Storage Account Key does not work directly with Azure Rest API
"""
from azure.storage.filedatalake.aio import DataLakeServiceClient
# Create an async service client
service_client = DataLakeServiceClient(
account_url=f"https://{self.azure_storage_account_name}.dfs.core.windows.net",
credential=self.azure_storage_account_key,
)
service_client = await self.get_service_client()
# Get file system client
file_system_client = service_client.get_file_system_client(
file_system=self.azure_storage_file_system

2
litellm/integrations/gcs_pubsub/pub_sub.py
View file

@ -75,7 +75,7 @@ class GcsPubSubLogger(CustomBatchLogger):
vertex_project,
) = await vertex_chat_completion._ensure_access_token_async(
credentials=self.path_service_account_json,
project_id=None,
project_id=self.project_id,
custom_llm_provider="vertex_ai",
)

41
litellm/litellm_core_utils/llm_cost_calc/utils.py
View file

@ -265,8 +265,10 @@ def generic_cost_per_token(
)
## CALCULATE OUTPUT COST
text_tokens = usage.completion_tokens
text_tokens = 0
audio_tokens = 0
reasoning_tokens = 0
is_text_tokens_total = False
if usage.completion_tokens_details is not None:
audio_tokens = (
cast(
@ -280,9 +282,20 @@ def generic_cost_per_token(
Optional[int],
getattr(usage.completion_tokens_details, "text_tokens", None),
)
or usage.completion_tokens # default to completion tokens, if this field is not set
or 0 # default to completion tokens, if this field is not set
)
reasoning_tokens = (
cast(
Optional[int],
getattr(usage.completion_tokens_details, "reasoning_tokens", 0),
)
or 0
)
if text_tokens == 0:
text_tokens = usage.completion_tokens
if text_tokens == usage.completion_tokens:
is_text_tokens_total = True
## TEXT COST
completion_cost = float(text_tokens) * completion_base_cost
@ -290,12 +303,26 @@ def generic_cost_per_token(
"output_cost_per_audio_token"
)
_output_cost_per_reasoning_token: Optional[float] = model_info.get(
"output_cost_per_reasoning_token"
)
## AUDIO COST
if (
_output_cost_per_audio_token is not None
and audio_tokens is not None
and audio_tokens > 0
):
if not is_text_tokens_total and audio_tokens is not None and audio_tokens > 0:
_output_cost_per_audio_token = (
_output_cost_per_audio_token
if _output_cost_per_audio_token is not None
else completion_base_cost
)
completion_cost += float(audio_tokens) * _output_cost_per_audio_token
## REASONING COST
if not is_text_tokens_total and reasoning_tokens and reasoning_tokens > 0:
_output_cost_per_reasoning_token = (
_output_cost_per_reasoning_token
if _output_cost_per_reasoning_token is not None
else completion_base_cost
)
completion_cost += float(reasoning_tokens) * _output_cost_per_reasoning_token
return prompt_cost, completion_cost

10
litellm/litellm_core_utils/llm_response_utils/convert_dict_to_response.py
View file

@ -14,6 +14,7 @@ from litellm.types.llms.openai import ChatCompletionThinkingBlock
from litellm.types.utils import (
ChatCompletionDeltaToolCall,
ChatCompletionMessageToolCall,
ChatCompletionRedactedThinkingBlock,
Choices,
Delta,
EmbeddingResponse,
@ -486,7 +487,14 @@ def convert_to_model_response_object( # noqa: PLR0915
)
# Handle thinking models that display `thinking_blocks` within `content`
thinking_blocks: Optional[List[ChatCompletionThinkingBlock]] = None
thinking_blocks: Optional[
List[
Union[
ChatCompletionThinkingBlock,
ChatCompletionRedactedThinkingBlock,
]
]
] = None
if "thinking_blocks" in choice["message"]:
thinking_blocks = choice["message"]["thinking_blocks"]
provider_specific_fields["thinking_blocks"] = thinking_blocks

56
litellm/litellm_core_utils/prompt_templates/common_utils.py
View file

@ -471,3 +471,59 @@ def unpack_defs(schema, defs):
unpack_defs(ref, defs)
value["items"] = ref
continue
def _get_image_mime_type_from_url(url: str) -> Optional[str]:
"""
Get mime type for common image URLs
See gemini mime types: https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/image-understanding#image-requirements
Supported by Gemini:
application/pdf
audio/mpeg
audio/mp3
audio/wav
image/png
image/jpeg
image/webp
text/plain
video/mov
video/mpeg
video/mp4
video/mpg
video/avi
video/wmv
video/mpegps
video/flv
"""
url = url.lower()
# Map file extensions to mime types
mime_types = {
# Images
(".jpg", ".jpeg"): "image/jpeg",
(".png",): "image/png",
(".webp",): "image/webp",
# Videos
(".mp4",): "video/mp4",
(".mov",): "video/mov",
(".mpeg", ".mpg"): "video/mpeg",
(".avi",): "video/avi",
(".wmv",): "video/wmv",
(".mpegps",): "video/mpegps",
(".flv",): "video/flv",
# Audio
(".mp3",): "audio/mp3",
(".wav",): "audio/wav",
(".mpeg",): "audio/mpeg",
# Documents
(".pdf",): "application/pdf",
(".txt",): "text/plain",
}
# Check each extension group against the URL
for extensions, mime_type in mime_types.items():
if any(url.endswith(ext) for ext in extensions):
return mime_type
return None

8
litellm/litellm_core_utils/prompt_templates/factory.py
View file

@ -2258,6 +2258,14 @@ def _parse_content_type(content_type: str) -> str:
return m.get_content_type()
def _parse_mime_type(base64_data: str) -> Optional[str]:
mime_type_match = re.match(r"data:(.*?);base64", base64_data)
if mime_type_match:
return mime_type_match.group(1)
else:
return None
class BedrockImageProcessor:
"""Handles both sync and async image processing for Bedrock conversations."""

101
litellm/litellm_core_utils/streaming_chunk_builder_utils.py
View file

@ -106,74 +106,63 @@ class ChunkProcessor:
def get_combined_tool_content(
self, tool_call_chunks: List[Dict[str, Any]]
) -> List[ChatCompletionMessageToolCall]:
argument_list: List[str] = []
delta = tool_call_chunks[0]["choices"][0]["delta"]
id = None
name = None
type = None
tool_calls_list: List[ChatCompletionMessageToolCall] = []
prev_index = None
prev_name = None
prev_id = None
curr_id = None
curr_index = 0
tool_call_map: Dict[
int, Dict[str, Any]
] = {} # Map to store tool calls by index
for chunk in tool_call_chunks:
choices = chunk["choices"]
for choice in choices:
delta = choice.get("delta", {})
tool_calls = delta.get("tool_calls", "")
# Check if a tool call is present
if tool_calls and tool_calls[0].function is not None:
if tool_calls[0].id:
id = tool_calls[0].id
curr_id = id
if prev_id is None:
prev_id = curr_id
if tool_calls[0].index:
curr_index = tool_calls[0].index
if tool_calls[0].function.arguments:
# Now, tool_calls is expected to be a dictionary
arguments = tool_calls[0].function.arguments
argument_list.append(arguments)
if tool_calls[0].function.name:
name = tool_calls[0].function.name
if tool_calls[0].type:
type = tool_calls[0].type
if prev_index is None:
prev_index = curr_index
if prev_name is None:
prev_name = name
if curr_index != prev_index: # new tool call
combined_arguments = "".join(argument_list)
tool_calls = delta.get("tool_calls", [])
for tool_call in tool_calls:
if not tool_call or not hasattr(tool_call, "function"):
continue
index = getattr(tool_call, "index", 0)
if index not in tool_call_map:
tool_call_map[index] = {
"id": None,
"name": None,
"type": None,
"arguments": [],
}
if hasattr(tool_call, "id") and tool_call.id:
tool_call_map[index]["id"] = tool_call.id
if hasattr(tool_call, "type") and tool_call.type:
tool_call_map[index]["type"] = tool_call.type
if hasattr(tool_call, "function"):
if (
hasattr(tool_call.function, "name")
and tool_call.function.name
):
tool_call_map[index]["name"] = tool_call.function.name
if (
hasattr(tool_call.function, "arguments")
and tool_call.function.arguments
):
tool_call_map[index]["arguments"].append(
tool_call.function.arguments
)
# Convert the map to a list of tool calls
for index in sorted(tool_call_map.keys()):
tool_call_data = tool_call_map[index]
if tool_call_data["id"] and tool_call_data["name"]:
combined_arguments = "".join(tool_call_data["arguments"]) or "{}"
tool_calls_list.append(
ChatCompletionMessageToolCall(
id=prev_id,
id=tool_call_data["id"],
function=Function(
arguments=combined_arguments,
name=prev_name,
name=tool_call_data["name"],
),
type=type,
type=tool_call_data["type"] or "function",
)
)
argument_list = [] # reset
prev_index = curr_index
prev_id = curr_id
prev_name = name
combined_arguments = (
"".join(argument_list) or "{}"
) # base case, return empty dict
tool_calls_list.append(
ChatCompletionMessageToolCall(
id=id,
type="function",
function=Function(
arguments=combined_arguments,
name=name,
),
)
)
return tool_calls_list

36
litellm/llms/anthropic/chat/handler.py
View file

@ -29,6 +29,7 @@ from litellm.types.llms.anthropic import (
UsageDelta,
)
from litellm.types.llms.openai import (
ChatCompletionRedactedThinkingBlock,
ChatCompletionThinkingBlock,
ChatCompletionToolCallChunk,
)
@ -501,18 +502,19 @@ class ModelResponseIterator:
) -> Tuple[
str,
Optional[ChatCompletionToolCallChunk],
List[ChatCompletionThinkingBlock],
List[Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]],
Dict[str, Any],
]:
"""
Helper function to handle the content block delta
"""
text = ""
tool_use: Optional[ChatCompletionToolCallChunk] = None
provider_specific_fields = {}
content_block = ContentBlockDelta(**chunk) # type: ignore
thinking_blocks: List[ChatCompletionThinkingBlock] = []
thinking_blocks: List[
Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]
] = []
self.content_blocks.append(content_block)
if "text" in content_block["delta"]:
@ -541,20 +543,25 @@ class ModelResponseIterator:
)
]
provider_specific_fields["thinking_blocks"] = thinking_blocks
return text, tool_use, thinking_blocks, provider_specific_fields
def _handle_reasoning_content(
self, thinking_blocks: List[ChatCompletionThinkingBlock]
self,
thinking_blocks: List[
Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]
],
) -> Optional[str]:
"""
Handle the reasoning content
"""
reasoning_content = None
for block in thinking_blocks:
thinking_content = cast(Optional[str], block.get("thinking"))
if reasoning_content is None:
reasoning_content = ""
if "thinking" in block:
reasoning_content += block["thinking"]
if thinking_content is not None:
reasoning_content += thinking_content
return reasoning_content
def chunk_parser(self, chunk: dict) -> ModelResponseStream:
@ -567,7 +574,13 @@ class ModelResponseIterator:
usage: Optional[Usage] = None
provider_specific_fields: Dict[str, Any] = {}
reasoning_content: Optional[str] = None
thinking_blocks: Optional[List[ChatCompletionThinkingBlock]] = None
thinking_blocks: Optional[
List[
Union[
ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock
]
]
] = None
index = int(chunk.get("index", 0))
if type_chunk == "content_block_delta":
@ -605,6 +618,15 @@ class ModelResponseIterator:
},
"index": self.tool_index,
}
elif (
content_block_start["content_block"]["type"] == "redacted_thinking"
):
thinking_blocks = [
ChatCompletionRedactedThinkingBlock(
type="redacted_thinking",
data=content_block_start["content_block"]["data"],
)
]
elif type_chunk == "content_block_stop":
ContentBlockStop(**chunk) # type: ignore
# check if tool call content block

65
litellm/llms/anthropic/chat/transformation.py
View file

@ -7,6 +7,9 @@ import httpx
import litellm
from litellm.constants import (
DEFAULT_ANTHROPIC_CHAT_MAX_TOKENS,
DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
RESPONSE_FORMAT_TOOL_NAME,
)
from litellm.litellm_core_utils.core_helpers import map_finish_reason
@ -27,6 +30,7 @@ from litellm.types.llms.openai import (
REASONING_EFFORT,
AllMessageValues,
ChatCompletionCachedContent,
ChatCompletionRedactedThinkingBlock,
ChatCompletionSystemMessage,
ChatCompletionThinkingBlock,
ChatCompletionToolCallChunk,
@ -276,11 +280,20 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig):
if reasoning_effort is None:
return None
elif reasoning_effort == "low":
return AnthropicThinkingParam(type="enabled", budget_tokens=1024)
return AnthropicThinkingParam(
type="enabled",
budget_tokens=DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
)
elif reasoning_effort == "medium":
return AnthropicThinkingParam(type="enabled", budget_tokens=2048)
return AnthropicThinkingParam(
type="enabled",
budget_tokens=DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
)
elif reasoning_effort == "high":
return AnthropicThinkingParam(type="enabled", budget_tokens=4096)
return AnthropicThinkingParam(
type="enabled",
budget_tokens=DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
)
else:
raise ValueError(f"Unmapped reasoning effort: {reasoning_effort}")
@ -563,13 +576,21 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig):
) -> Tuple[
str,
Optional[List[Any]],
Optional[List[ChatCompletionThinkingBlock]],
Optional[
List[
Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]
]
],
Optional[str],
List[ChatCompletionToolCallChunk],
]:
text_content = ""
citations: Optional[List[Any]] = None
thinking_blocks: Optional[List[ChatCompletionThinkingBlock]] = None
thinking_blocks: Optional[
List[
Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]
]
] = None
reasoning_content: Optional[str] = None
tool_calls: List[ChatCompletionToolCallChunk] = []
for idx, content in enumerate(completion_response["content"]):
@ -588,20 +609,30 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig):
index=idx,
)
)
## CITATIONS
if content.get("citations", None) is not None:
if citations is None:
citations = []
citations.append(content["citations"])
if content.get("thinking", None) is not None:
elif content.get("thinking", None) is not None:
if thinking_blocks is None:
thinking_blocks = []
thinking_blocks.append(cast(ChatCompletionThinkingBlock, content))
elif content["type"] == "redacted_thinking":
if thinking_blocks is None:
thinking_blocks = []
thinking_blocks.append(
cast(ChatCompletionRedactedThinkingBlock, content)
)
## CITATIONS
if content.get("citations") is not None:
if citations is None:
citations = []
citations.append(content["citations"])
if thinking_blocks is not None:
reasoning_content = ""
for block in thinking_blocks:
if "thinking" in block:
reasoning_content += block["thinking"]
thinking_content = cast(Optional[str], block.get("thinking"))
if thinking_content is not None:
reasoning_content += thinking_content
return text_content, citations, thinking_blocks, reasoning_content, tool_calls
def calculate_usage(
@ -691,7 +722,13 @@ class AnthropicConfig(AnthropicModelInfo, BaseConfig):
else:
text_content = ""
citations: Optional[List[Any]] = None
thinking_blocks: Optional[List[ChatCompletionThinkingBlock]] = None
thinking_blocks: Optional[
List[
Union[
ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock
]
]
] = None
reasoning_content: Optional[str] = None
tool_calls: List[ChatCompletionToolCallChunk] = []

1
litellm/llms/azure/assistants.py
View file

@ -288,6 +288,7 @@ class AzureAssistantsAPI(BaseAzureLLM):
timeout=timeout,
max_retries=max_retries,
client=client,
litellm_params=litellm_params,
)
thread_message: OpenAIMessage = openai_client.beta.threads.messages.create( # type: ignore

2
litellm/llms/azure/chat/o_series_transformation.py
View file

@ -79,7 +79,7 @@ class AzureOpenAIO1Config(OpenAIOSeriesConfig):
return True
def is_o_series_model(self, model: str) -> bool:
return "o1" in model or "o3" in model or "o_series/" in model
return "o1" in model or "o3" in model or "o4" in model or "o_series/" in model
def transform_request(
self,

8
litellm/llms/azure/common_utils.py
View file

@ -61,7 +61,7 @@ def process_azure_headers(headers: Union[httpx.Headers, dict]) -> dict:
return {**llm_response_headers, **openai_headers}
def get_azure_ad_token_from_entrata_id(
def get_azure_ad_token_from_entra_id(
tenant_id: str,
client_id: str,
client_secret: str,
@ -81,7 +81,7 @@ def get_azure_ad_token_from_entrata_id(
"""
from azure.identity import ClientSecretCredential, get_bearer_token_provider
verbose_logger.debug("Getting Azure AD Token from Entrata ID")
verbose_logger.debug("Getting Azure AD Token from Entra ID")
if tenant_id.startswith("os.environ/"):
_tenant_id = get_secret_str(tenant_id)
@ -324,9 +324,9 @@ class BaseAzureLLM(BaseOpenAILLM):
timeout = litellm_params.get("timeout")
if not api_key and tenant_id and client_id and client_secret:
verbose_logger.debug(
"Using Azure AD Token Provider from Entrata ID for Azure Auth"
"Using Azure AD Token Provider from Entra ID for Azure Auth"
)
azure_ad_token_provider = get_azure_ad_token_from_entrata_id(
azure_ad_token_provider = get_azure_ad_token_from_entra_id(
tenant_id=tenant_id,
client_id=client_id,
client_secret=client_secret,

94
litellm/llms/azure/responses/transformation.py Normal file
View file

@ -0,0 +1,94 @@
from typing import TYPE_CHECKING, Any, Optional, cast
import httpx
import litellm
from litellm.llms.openai.responses.transformation import OpenAIResponsesAPIConfig
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import *
from litellm.utils import _add_path_to_api_base
if TYPE_CHECKING:
from litellm.litellm_core_utils.litellm_logging import Logging as _LiteLLMLoggingObj
LiteLLMLoggingObj = _LiteLLMLoggingObj
else:
LiteLLMLoggingObj = Any
class AzureOpenAIResponsesAPIConfig(OpenAIResponsesAPIConfig):
def validate_environment(
self,
headers: dict,
model: str,
api_key: Optional[str] = None,
) -> dict:
api_key = (
api_key
or litellm.api_key
or litellm.azure_key
or get_secret_str("AZURE_OPENAI_API_KEY")
or get_secret_str("AZURE_API_KEY")
)
headers.update(
{
"Authorization": f"Bearer {api_key}",
}
)
return headers
def get_complete_url(
self,
api_base: Optional[str],
api_key: Optional[str],
model: str,
optional_params: dict,
litellm_params: dict,
stream: Optional[bool] = None,
) -> str:
"""
Constructs a complete URL for the API request.
Args:
- api_base: Base URL, e.g.,
"https://litellm8397336933.openai.azure.com"
OR
"https://litellm8397336933.openai.azure.com/openai/responses?api-version=2024-05-01-preview"
- model: Model name.
- optional_params: Additional query parameters, including "api_version".
- stream: If streaming is required (optional).
Returns:
- A complete URL string, e.g.,
"https://litellm8397336933.openai.azure.com/openai/responses?api-version=2024-05-01-preview"
"""
api_base = api_base or litellm.api_base or get_secret_str("AZURE_API_BASE")
if api_base is None:
raise ValueError(
f"api_base is required for Azure AI Studio. Please set the api_base parameter. Passed `api_base={api_base}`"
)
original_url = httpx.URL(api_base)
# Extract api_version or use default
api_version = cast(Optional[str], litellm_params.get("api_version"))
# Create a new dictionary with existing params
query_params = dict(original_url.params)
# Add api_version if needed
if "api-version" not in query_params and api_version:
query_params["api-version"] = api_version
# Add the path to the base URL
if "/openai/responses" not in api_base:
new_url = _add_path_to_api_base(
api_base=api_base, ending_path="/openai/responses"
)
else:
new_url = api_base
# Use the new query_params dictionary
final_url = httpx.URL(new_url).copy_with(params=query_params)
return str(final_url)

52
litellm/llms/azure_ai/chat/transformation.py
View file

@ -1,3 +1,4 @@
import enum
from typing import Any, List, Optional, Tuple, cast
from urllib.parse import urlparse
@ -19,6 +20,10 @@ from litellm.types.utils import ModelResponse, ProviderField
from litellm.utils import _add_path_to_api_base, supports_tool_choice
class AzureFoundryErrorStrings(str, enum.Enum):
SET_EXTRA_PARAMETERS_TO_PASS_THROUGH = "Set extra-parameters to 'pass-through'"
class AzureAIStudioConfig(OpenAIConfig):
def get_supported_openai_params(self, model: str) -> List:
model_supports_tool_choice = True # azure ai supports this by default
@ -240,12 +245,18 @@ class AzureAIStudioConfig(OpenAIConfig):
) -> bool:
should_drop_params = litellm_params.get("drop_params") or litellm.drop_params
error_text = e.response.text
if should_drop_params and "Extra inputs are not permitted" in error_text:
return True
elif (
"unknown field: parameter index is not a valid field" in error_text
): # remove index from tool calls
return True
elif (
AzureFoundryErrorStrings.SET_EXTRA_PARAMETERS_TO_PASS_THROUGH.value
in error_text
): # remove extra-parameters from tool calls
return True
return super().should_retry_llm_api_inside_llm_translation_on_http_error(
e=e, litellm_params=litellm_params
)
@ -265,5 +276,46 @@ class AzureAIStudioConfig(OpenAIConfig):
litellm.remove_index_from_tool_calls(
messages=_messages,
)
elif (
AzureFoundryErrorStrings.SET_EXTRA_PARAMETERS_TO_PASS_THROUGH.value
in e.response.text
):
request_data = self._drop_extra_params_from_request_data(
request_data, e.response.text
)
data = drop_params_from_unprocessable_entity_error(e=e, data=request_data)
return data
def _drop_extra_params_from_request_data(
self, request_data: dict, error_text: str
) -> dict:
params_to_drop = self._extract_params_to_drop_from_error_text(error_text)
if params_to_drop:
for param in params_to_drop:
if param in request_data:
request_data.pop(param, None)
return request_data
def _extract_params_to_drop_from_error_text(
self, error_text: str
) -> Optional[List[str]]:
"""
Error text looks like this"
"Extra parameters ['stream_options', 'extra-parameters'] are not allowed when extra-parameters is not set or set to be 'error'.
"""
import re
# Extract parameters within square brackets
match = re.search(r"\[(.*?)\]", error_text)
if not match:
return []
# Parse the extracted string into a list of parameter names
params_str = match.group(1)
params = []
for param in params_str.split(","):
# Clean up the parameter name (remove quotes, spaces)
clean_param = param.strip().strip("'").strip('"')
if clean_param:
params.append(clean_param)
return params

64
litellm/llms/base_llm/base_model_iterator.py
View file

@ -1,9 +1,16 @@
import json
from abc import abstractmethod
from typing import Optional, Union
from typing import List, Optional, Union, cast
import litellm
from litellm.types.utils import GenericStreamingChunk, ModelResponseStream
from litellm.types.utils import (
Choices,
Delta,
GenericStreamingChunk,
ModelResponse,
ModelResponseStream,
StreamingChoices,
)
class BaseModelResponseIterator:
@ -121,6 +128,59 @@ class BaseModelResponseIterator:
raise RuntimeError(f"Error parsing chunk: {e},\nReceived chunk: {chunk}")
class MockResponseIterator: # for returning ai21 streaming responses
def __init__(
self, model_response: ModelResponse, json_mode: Optional[bool] = False
):
self.model_response = model_response
self.json_mode = json_mode
self.is_done = False
# Sync iterator
def __iter__(self):
return self
def _chunk_parser(self, chunk_data: ModelResponse) -> ModelResponseStream:
try:
streaming_choices: List[StreamingChoices] = []
for choice in chunk_data.choices:
streaming_choices.append(
StreamingChoices(
index=choice.index,
delta=Delta(
**cast(Choices, choice).message.model_dump(),
),
finish_reason=choice.finish_reason,
)
)
processed_chunk = ModelResponseStream(
id=chunk_data.id,
object="chat.completion",
created=chunk_data.created,
model=chunk_data.model,
choices=streaming_choices,
)
return processed_chunk
except Exception as e:
raise ValueError(f"Failed to decode chunk: {chunk_data}. Error: {e}")
def __next__(self):
if self.is_done:
raise StopIteration
self.is_done = True
return self._chunk_parser(self.model_response)
# Async iterator
def __aiter__(self):
return self
async def __anext__(self):
if self.is_done:
raise StopAsyncIteration
self.is_done = True
return self._chunk_parser(self.model_response)
class FakeStreamResponseIterator:
def __init__(self, model_response, json_mode: Optional[bool] = False):
self.model_response = model_response

3
litellm/llms/base_llm/responses/transformation.py
View file

@ -73,7 +73,10 @@ class BaseResponsesAPIConfig(ABC):
def get_complete_url(
self,
api_base: Optional[str],
api_key: Optional[str],
model: str,
optional_params: dict,
litellm_params: dict,
stream: Optional[bool] = None,
) -> str:
"""

40
litellm/llms/bedrock/chat/converse_transformation.py
View file

@ -22,6 +22,7 @@ from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMExcepti
from litellm.types.llms.bedrock import *
from litellm.types.llms.openai import (
AllMessageValues,
ChatCompletionRedactedThinkingBlock,
ChatCompletionResponseMessage,
ChatCompletionSystemMessage,
ChatCompletionThinkingBlock,
@ -375,25 +376,27 @@ class AmazonConverseConfig(BaseConfig):
system_content_blocks: List[SystemContentBlock] = []
for idx, message in enumerate(messages):
if message["role"] == "system":
_system_content_block: Optional[SystemContentBlock] = None
_cache_point_block: Optional[SystemContentBlock] = None
if isinstance(message["content"], str) and len(message["content"]) > 0:
_system_content_block = SystemContentBlock(text=message["content"])
_cache_point_block = self._get_cache_point_block(
system_prompt_indices.append(idx)
if isinstance(message["content"], str) and message["content"]:
system_content_blocks.append(
SystemContentBlock(text=message["content"])
)
cache_block = self._get_cache_point_block(
message, block_type="system"
)
if cache_block:
system_content_blocks.append(cache_block)
elif isinstance(message["content"], list):
for m in message["content"]:
if m.get("type", "") == "text" and len(m["text"]) > 0:
_system_content_block = SystemContentBlock(text=m["text"])
_cache_point_block = self._get_cache_point_block(
if m.get("type") == "text" and m.get("text"):
system_content_blocks.append(
SystemContentBlock(text=m["text"])
)
cache_block = self._get_cache_point_block(
m, block_type="system"
)
if _system_content_block is not None:
system_content_blocks.append(_system_content_block)
if _cache_point_block is not None:
system_content_blocks.append(_cache_point_block)
system_prompt_indices.append(idx)
if cache_block:
system_content_blocks.append(cache_block)
if len(system_prompt_indices) > 0:
for idx in reversed(system_prompt_indices):
messages.pop(idx)
@ -627,9 +630,11 @@ class AmazonConverseConfig(BaseConfig):
def _transform_thinking_blocks(
self, thinking_blocks: List[BedrockConverseReasoningContentBlock]
) -> List[ChatCompletionThinkingBlock]:
) -> List[Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]]:
"""Return a consistent format for thinking blocks between Anthropic and Bedrock."""
thinking_blocks_list: List[ChatCompletionThinkingBlock] = []
thinking_blocks_list: List[
Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]
] = []
for block in thinking_blocks:
if "reasoningText" in block:
_thinking_block = ChatCompletionThinkingBlock(type="thinking")
@ -640,6 +645,11 @@ class AmazonConverseConfig(BaseConfig):
if _signature is not None:
_thinking_block["signature"] = _signature
thinking_blocks_list.append(_thinking_block)
elif "redactedContent" in block:
_redacted_block = ChatCompletionRedactedThinkingBlock(
type="redacted_thinking", data=block["redactedContent"]
)
thinking_blocks_list.append(_redacted_block)
return thinking_blocks_list
def _transform_usage(self, usage: ConverseTokenUsageBlock) -> Usage:

76
litellm/llms/bedrock/chat/invoke_handler.py
View file

@ -50,6 +50,7 @@ from litellm.llms.custom_httpx.http_handler import (
)
from litellm.types.llms.bedrock import *
from litellm.types.llms.openai import (
ChatCompletionRedactedThinkingBlock,
ChatCompletionThinkingBlock,
ChatCompletionToolCallChunk,
ChatCompletionToolCallFunctionChunk,
@ -1255,19 +1256,33 @@ class AWSEventStreamDecoder:
def translate_thinking_blocks(
self, thinking_block: BedrockConverseReasoningContentBlockDelta
) -> Optional[List[ChatCompletionThinkingBlock]]:
) -> Optional[
List[Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]]
]:
"""
Translate the thinking blocks to a string
"""
thinking_blocks_list: List[ChatCompletionThinkingBlock] = []
_thinking_block = ChatCompletionThinkingBlock(type="thinking")
thinking_blocks_list: List[
Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]
] = []
_thinking_block: Optional[
Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]
] = None
if "text" in thinking_block:
_thinking_block = ChatCompletionThinkingBlock(type="thinking")
_thinking_block["thinking"] = thinking_block["text"]
elif "signature" in thinking_block:
_thinking_block = ChatCompletionThinkingBlock(type="thinking")
_thinking_block["signature"] = thinking_block["signature"]
_thinking_block["thinking"] = "" # consistent with anthropic response
thinking_blocks_list.append(_thinking_block)
elif "redactedContent" in thinking_block:
_thinking_block = ChatCompletionRedactedThinkingBlock(
type="redacted_thinking", data=thinking_block["redactedContent"]
)
if _thinking_block is not None:
thinking_blocks_list.append(_thinking_block)
return thinking_blocks_list
def converse_chunk_parser(self, chunk_data: dict) -> ModelResponseStream:
@ -1279,31 +1294,44 @@ class AWSEventStreamDecoder:
usage: Optional[Usage] = None
provider_specific_fields: dict = {}
reasoning_content: Optional[str] = None
thinking_blocks: Optional[List[ChatCompletionThinkingBlock]] = None
thinking_blocks: Optional[
List[
Union[
ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock
]
]
] = None
index = int(chunk_data.get("contentBlockIndex", 0))
if "start" in chunk_data:
start_obj = ContentBlockStartEvent(**chunk_data["start"])
self.content_blocks = [] # reset
if (
start_obj is not None
and "toolUse" in start_obj
and start_obj["toolUse"] is not None
):
## check tool name was formatted by litellm
_response_tool_name = start_obj["toolUse"]["name"]
response_tool_name = get_bedrock_tool_name(
response_tool_name=_response_tool_name
)
tool_use = {
"id": start_obj["toolUse"]["toolUseId"],
"type": "function",
"function": {
"name": response_tool_name,
"arguments": "",
},
"index": index,
}
if start_obj is not None:
if "toolUse" in start_obj and start_obj["toolUse"] is not None:
## check tool name was formatted by litellm
_response_tool_name = start_obj["toolUse"]["name"]
response_tool_name = get_bedrock_tool_name(
response_tool_name=_response_tool_name
)
tool_use = {
"id": start_obj["toolUse"]["toolUseId"],
"type": "function",
"function": {
"name": response_tool_name,
"arguments": "",
},
"index": index,
}
elif (
"reasoningContent" in start_obj
and start_obj["reasoningContent"] is not None
): # redacted thinking can be in start object
thinking_blocks = self.translate_thinking_blocks(
start_obj["reasoningContent"]
)
provider_specific_fields = {
"reasoningContent": start_obj["reasoningContent"],
}
elif "delta" in chunk_data:
delta_obj = ContentBlockDeltaEvent(**chunk_data["delta"])
self.content_blocks.append(delta_obj)

36
litellm/llms/bedrock/common_utils.py
View file

@ -44,7 +44,18 @@ class AmazonBedrockGlobalConfig:
)
def get_ap_regions(self) -> List[str]:
return ["ap-northeast-1", "ap-northeast-2", "ap-northeast-3", "ap-south-1"]
"""
Source: https://www.aws-services.info/bedrock.html
"""
return [
"ap-northeast-1", # Asia Pacific (Tokyo)
"ap-northeast-2", # Asia Pacific (Seoul)
"ap-northeast-3", # Asia Pacific (Osaka)
"ap-south-1", # Asia Pacific (Mumbai)
"ap-south-2", # Asia Pacific (Hyderabad)
"ap-southeast-1", # Asia Pacific (Singapore)
"ap-southeast-2", # Asia Pacific (Sydney)
]
def get_sa_regions(self) -> List[str]:
return ["sa-east-1"]
@ -54,10 +65,14 @@ class AmazonBedrockGlobalConfig:
Source: https://www.aws-services.info/bedrock.html
"""
return [
"eu-west-1",
"eu-west-2",
"eu-west-3",
"eu-central-1",
"eu-west-1", # Europe (Ireland)
"eu-west-2", # Europe (London)
"eu-west-3", # Europe (Paris)
"eu-central-1", # Europe (Frankfurt)
"eu-central-2", # Europe (Zurich)
"eu-south-1", # Europe (Milan)
"eu-south-2", # Europe (Spain)
"eu-north-1", # Europe (Stockholm)
]
def get_ca_regions(self) -> List[str]:
@ -68,11 +83,12 @@ class AmazonBedrockGlobalConfig:
Source: https://www.aws-services.info/bedrock.html
"""
return [
"us-east-2",
"us-east-1",
"us-west-1",
"us-west-2",
"us-gov-west-1",
"us-east-1", # US East (N. Virginia)
"us-east-2", # US East (Ohio)
"us-west-1", # US West (N. California)
"us-west-2", # US West (Oregon)
"us-gov-east-1", # AWS GovCloud (US-East)
"us-gov-west-1", # AWS GovCloud (US-West)
]

2
litellm/llms/custom_httpx/http_handler.py
View file

@ -8,6 +8,7 @@ import httpx
from httpx import USE_CLIENT_DEFAULT, AsyncHTTPTransport, HTTPTransport
import litellm
from litellm.constants import _DEFAULT_TTL_FOR_HTTPX_CLIENTS
from litellm.litellm_core_utils.logging_utils import track_llm_api_timing
from litellm.types.llms.custom_http import *
@ -31,7 +32,6 @@ headers = {
# https://www.python-httpx.org/advanced/timeouts
_DEFAULT_TIMEOUT = httpx.Timeout(timeout=5.0, connect=5.0)
_DEFAULT_TTL_FOR_HTTPX_CLIENTS = 3600 # 1 hour, re-use the same httpx client for 1 hour
def mask_sensitive_info(error_message):

83
litellm/llms/custom_httpx/llm_http_handler.py
View file

@ -11,6 +11,7 @@ from litellm._logging import verbose_logger
from litellm.llms.base_llm.audio_transcription.transformation import (
BaseAudioTranscriptionConfig,
)
from litellm.llms.base_llm.base_model_iterator import MockResponseIterator
from litellm.llms.base_llm.chat.transformation import BaseConfig
from litellm.llms.base_llm.embedding.transformation import BaseEmbeddingConfig
from litellm.llms.base_llm.files.transformation import BaseFilesConfig
@ -228,11 +229,17 @@ class BaseLLMHTTPHandler:
api_key: Optional[str] = None,
headers: Optional[dict] = {},
client: Optional[Union[HTTPHandler, AsyncHTTPHandler]] = None,
provider_config: Optional[BaseConfig] = None,
):
json_mode: bool = optional_params.pop("json_mode", False)
extra_body: Optional[dict] = optional_params.pop("extra_body", None)
fake_stream = fake_stream or optional_params.pop("fake_stream", False)
provider_config = ProviderConfigManager.get_provider_chat_config(
model=model, provider=litellm.LlmProviders(custom_llm_provider)
provider_config = (
provider_config
or ProviderConfigManager.get_provider_chat_config(
model=model, provider=litellm.LlmProviders(custom_llm_provider)
)
)
if provider_config is None:
raise ValueError(
@ -267,6 +274,9 @@ class BaseLLMHTTPHandler:
headers=headers,
)
if extra_body is not None:
data = {**data, **extra_body}
headers = provider_config.sign_request(
headers=headers,
optional_params=optional_params,
@ -313,6 +323,7 @@ class BaseLLMHTTPHandler:
),
litellm_params=litellm_params,
json_mode=json_mode,
optional_params=optional_params,
)
else:
@ -374,6 +385,7 @@ class BaseLLMHTTPHandler:
),
litellm_params=litellm_params,
json_mode=json_mode,
optional_params=optional_params,
)
return CustomStreamWrapper(
completion_stream=completion_stream,
@ -422,6 +434,7 @@ class BaseLLMHTTPHandler:
model: str,
messages: list,
logging_obj,
optional_params: dict,
litellm_params: dict,
timeout: Union[float, httpx.Timeout],
fake_stream: bool = False,
@ -453,11 +466,22 @@ class BaseLLMHTTPHandler:
)
if fake_stream is True:
completion_stream = provider_config.get_model_response_iterator(
streaming_response=response.json(),
sync_stream=True,
model_response: ModelResponse = provider_config.transform_response(
model=model,
raw_response=response,
model_response=litellm.ModelResponse(),
logging_obj=logging_obj,
request_data=data,
messages=messages,
optional_params=optional_params,
litellm_params=litellm_params,
encoding=None,
json_mode=json_mode,
)
completion_stream: Any = MockResponseIterator(
model_response=model_response, json_mode=json_mode
)
else:
completion_stream = provider_config.get_model_response_iterator(
streaming_response=response.iter_lines(),
@ -487,6 +511,7 @@ class BaseLLMHTTPHandler:
logging_obj: LiteLLMLoggingObj,
data: dict,
litellm_params: dict,
optional_params: dict,
fake_stream: bool = False,
client: Optional[AsyncHTTPHandler] = None,
json_mode: Optional[bool] = None,
@ -505,6 +530,7 @@ class BaseLLMHTTPHandler:
)
completion_stream, _response_headers = await self.make_async_call_stream_helper(
model=model,
custom_llm_provider=custom_llm_provider,
provider_config=provider_config,
api_base=api_base,
@ -516,6 +542,8 @@ class BaseLLMHTTPHandler:
fake_stream=fake_stream,
client=client,
litellm_params=litellm_params,
optional_params=optional_params,
json_mode=json_mode,
)
streamwrapper = CustomStreamWrapper(
completion_stream=completion_stream,
@ -527,6 +555,7 @@ class BaseLLMHTTPHandler:
async def make_async_call_stream_helper(
self,
model: str,
custom_llm_provider: str,
provider_config: BaseConfig,
api_base: str,
@ -536,8 +565,10 @@ class BaseLLMHTTPHandler:
logging_obj: LiteLLMLoggingObj,
timeout: Union[float, httpx.Timeout],
litellm_params: dict,
optional_params: dict,
fake_stream: bool = False,
client: Optional[AsyncHTTPHandler] = None,
json_mode: Optional[bool] = None,
) -> Tuple[Any, httpx.Headers]:
"""
Helper function for making an async call with stream.
@ -568,8 +599,21 @@ class BaseLLMHTTPHandler:
)
if fake_stream is True:
completion_stream = provider_config.get_model_response_iterator(
streaming_response=response.json(), sync_stream=False
model_response: ModelResponse = provider_config.transform_response(
model=model,
raw_response=response,
model_response=litellm.ModelResponse(),
logging_obj=logging_obj,
request_data=data,
messages=messages,
optional_params=optional_params,
litellm_params=litellm_params,
encoding=None,
json_mode=json_mode,
)
completion_stream: Any = MockResponseIterator(
model_response=model_response, json_mode=json_mode
)
else:
completion_stream = provider_config.get_model_response_iterator(
@ -594,8 +638,12 @@ class BaseLLMHTTPHandler:
"""
Some providers like Bedrock invoke do not support the stream parameter in the request body, we only pass `stream` in the request body the provider supports it.
"""
if fake_stream is True:
return data
# remove 'stream' from data
new_data = data.copy()
new_data.pop("stream", None)
return new_data
if provider_config.supports_stream_param_in_request_body is True:
data["stream"] = True
return data
@ -1011,9 +1059,16 @@ class BaseLLMHTTPHandler:
if extra_headers:
headers.update(extra_headers)
# Check if streaming is requested
stream = response_api_optional_request_params.get("stream", False)
api_base = responses_api_provider_config.get_complete_url(
api_base=litellm_params.api_base,
api_key=litellm_params.api_key,
model=model,
optional_params=response_api_optional_request_params,
litellm_params=dict(litellm_params),
stream=stream,
)
data = responses_api_provider_config.transform_responses_api_request(
@ -1035,9 +1090,6 @@ class BaseLLMHTTPHandler:
},
)
# Check if streaming is requested
stream = response_api_optional_request_params.get("stream", False)
try:
if stream:
# For streaming, use stream=True in the request
@ -1126,9 +1178,16 @@ class BaseLLMHTTPHandler:
if extra_headers:
headers.update(extra_headers)
# Check if streaming is requested
stream = response_api_optional_request_params.get("stream", False)
api_base = responses_api_provider_config.get_complete_url(
api_base=litellm_params.api_base,
api_key=litellm_params.api_key,
model=model,
optional_params=response_api_optional_request_params,
litellm_params=dict(litellm_params),
stream=stream,
)
data = responses_api_provider_config.transform_responses_api_request(
@ -1149,8 +1208,6 @@ class BaseLLMHTTPHandler:
"headers": headers,
},
)
# Check if streaming is requested
stream = response_api_optional_request_params.get("stream", False)
try:
if stream:

16
litellm/llms/databricks/chat/transformation.py
View file

@ -37,6 +37,7 @@ from litellm.types.llms.databricks import (
)
from litellm.types.llms.openai import (
AllMessageValues,
ChatCompletionRedactedThinkingBlock,
ChatCompletionThinkingBlock,
ChatCompletionToolChoiceFunctionParam,
ChatCompletionToolChoiceObjectParam,
@ -314,13 +315,24 @@ class DatabricksConfig(DatabricksBase, OpenAILikeChatConfig, AnthropicConfig):
@staticmethod
def extract_reasoning_content(
content: Optional[AllDatabricksContentValues],
) -> Tuple[Optional[str], Optional[List[ChatCompletionThinkingBlock]]]:
) -> Tuple[
Optional[str],
Optional[
List[
Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]
]
],
]:
"""
Extract and return the reasoning content and thinking blocks
"""
if content is None:
return None, None
thinking_blocks: Optional[List[ChatCompletionThinkingBlock]] = None
thinking_blocks: Optional[
List[
Union[ChatCompletionThinkingBlock, ChatCompletionRedactedThinkingBlock]
]
] = None
reasoning_content: Optional[str] = None
if isinstance(content, list):
for item in content:

110
litellm/llms/fireworks_ai/chat/transformation.py
View file

@ -1,15 +1,33 @@
from typing import List, Literal, Optional, Tuple, Union, cast
import json
import uuid
from typing import Any, List, Literal, Optional, Tuple, Union, cast
import httpx
import litellm
from litellm.constants import RESPONSE_FORMAT_TOOL_NAME
from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj
from litellm.litellm_core_utils.llm_response_utils.get_headers import (
get_response_headers,
)
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import (
AllMessageValues,
ChatCompletionImageObject,
ChatCompletionToolParam,
OpenAIChatCompletionToolParam,
)
from litellm.types.utils import ProviderSpecificModelInfo
from litellm.types.utils import (
ChatCompletionMessageToolCall,
Choices,
Function,
Message,
ModelResponse,
ProviderSpecificModelInfo,
)
from ...openai.chat.gpt_transformation import OpenAIGPTConfig
from ..common_utils import FireworksAIException
class FireworksAIConfig(OpenAIGPTConfig):
@ -219,6 +237,94 @@ class FireworksAIConfig(OpenAIGPTConfig):
headers=headers,
)
def _handle_message_content_with_tool_calls(
self,
message: Message,
tool_calls: Optional[List[ChatCompletionToolParam]],
) -> Message:
"""
Fireworks AI sends tool calls in the content field instead of tool_calls
Relevant Issue: https://github.com/BerriAI/litellm/issues/7209#issuecomment-2813208780
"""
if (
tool_calls is not None
and message.content is not None
and message.tool_calls is None
):
try:
function = Function(**json.loads(message.content))
if function.name != RESPONSE_FORMAT_TOOL_NAME and function.name in [
tool["function"]["name"] for tool in tool_calls
]:
tool_call = ChatCompletionMessageToolCall(
function=function, id=str(uuid.uuid4()), type="function"
)
message.tool_calls = [tool_call]
message.content = None
except Exception:
pass
return message
def transform_response(
self,
model: str,
raw_response: httpx.Response,
model_response: ModelResponse,
logging_obj: LiteLLMLoggingObj,
request_data: dict,
messages: List[AllMessageValues],
optional_params: dict,
litellm_params: dict,
encoding: Any,
api_key: Optional[str] = None,
json_mode: Optional[bool] = None,
) -> ModelResponse:
## LOGGING
logging_obj.post_call(
input=messages,
api_key=api_key,
original_response=raw_response.text,
additional_args={"complete_input_dict": request_data},
)
## RESPONSE OBJECT
try:
completion_response = raw_response.json()
except Exception as e:
response_headers = getattr(raw_response, "headers", None)
raise FireworksAIException(
message="Unable to get json response - {}, Original Response: {}".format(
str(e), raw_response.text
),
status_code=raw_response.status_code,
headers=response_headers,
)
raw_response_headers = dict(raw_response.headers)
additional_headers = get_response_headers(raw_response_headers)
response = ModelResponse(**completion_response)
if response.model is not None:
response.model = "fireworks_ai/" + response.model
## FIREWORKS AI sends tool calls in the content field instead of tool_calls
for choice in response.choices:
cast(
Choices, choice
).message = self._handle_message_content_with_tool_calls(
message=cast(Choices, choice).message,
tool_calls=optional_params.get("tools", None),
)
response._hidden_params = {"additional_headers": additional_headers}
return response
def _get_openai_compatible_provider_info(
self, api_base: Optional[str], api_key: Optional[str]
) -> Tuple[Optional[str], Optional[str]]:

7
litellm/llms/gemini/chat/transformation.py
View file

@ -7,6 +7,7 @@ from litellm.litellm_core_utils.prompt_templates.factory import (
)
from litellm.types.llms.openai import AllMessageValues
from litellm.types.llms.vertex_ai import ContentType, PartType
from litellm.utils import supports_reasoning
from ...vertex_ai.gemini.transformation import _gemini_convert_messages_with_history
from ...vertex_ai.gemini.vertex_and_google_ai_studio_gemini import VertexGeminiConfig
@ -67,7 +68,7 @@ class GoogleAIStudioGeminiConfig(VertexGeminiConfig):
return super().get_config()
def get_supported_openai_params(self, model: str) -> List[str]:
return [
supported_params = [
"temperature",
"top_p",
"max_tokens",
@ -83,6 +84,10 @@ class GoogleAIStudioGeminiConfig(VertexGeminiConfig):
"frequency_penalty",
"modalities",
]
if supports_reasoning(model):
supported_params.append("reasoning_effort")
supported_params.append("thinking")
return supported_params
def map_openai_params(
self,

19
litellm/llms/groq/chat/transformation.py
View file

@ -14,10 +14,10 @@ from litellm.types.llms.openai import (
ChatCompletionToolParamFunctionChunk,
)
from ...openai.chat.gpt_transformation import OpenAIGPTConfig
from ...openai_like.chat.transformation import OpenAILikeChatConfig
class GroqChatConfig(OpenAIGPTConfig):
class GroqChatConfig(OpenAILikeChatConfig):
frequency_penalty: Optional[int] = None
function_call: Optional[Union[str, dict]] = None
functions: Optional[list] = None
@ -57,6 +57,14 @@ class GroqChatConfig(OpenAIGPTConfig):
def get_config(cls):
return super().get_config()
def get_supported_openai_params(self, model: str) -> list:
base_params = super().get_supported_openai_params(model)
try:
base_params.remove("max_retries")
except ValueError:
pass
return base_params
def _transform_messages(self, messages: List[AllMessageValues], model: str) -> List:
for idx, message in enumerate(messages):
"""
@ -124,8 +132,11 @@ class GroqChatConfig(OpenAIGPTConfig):
optional_params: dict,
model: str,
drop_params: bool = False,
replace_max_completion_tokens_with_max_tokens: bool = False, # groq supports max_completion_tokens
) -> dict:
_response_format = non_default_params.get("response_format")
if self._should_fake_stream(non_default_params):
optional_params["fake_stream"] = True
if _response_format is not None and isinstance(_response_format, dict):
json_schema: Optional[dict] = None
if "response_schema" in _response_format:
@ -152,6 +163,8 @@ class GroqChatConfig(OpenAIGPTConfig):
non_default_params.pop(
"response_format", None
) # only remove if it's a json_schema - handled via using groq's tool calling params.
return super().map_openai_params(
optional_params = super().map_openai_params(
non_default_params, optional_params, model, drop_params
)
return optional_params

80
litellm/llms/hosted_vllm/chat/transformation.py
View file

@ -2,9 +2,19 @@
Translate from OpenAI's `/v1/chat/completions` to VLLM's `/v1/chat/completions`
"""
from typing import Optional, Tuple
from typing import List, Optional, Tuple, cast
from litellm.litellm_core_utils.prompt_templates.common_utils import (
_get_image_mime_type_from_url,
)
from litellm.litellm_core_utils.prompt_templates.factory import _parse_mime_type
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import (
AllMessageValues,
ChatCompletionFileObject,
ChatCompletionVideoObject,
ChatCompletionVideoUrlObject,
)
from ....utils import _remove_additional_properties, _remove_strict_from_schema
from ...openai.chat.gpt_transformation import OpenAIGPTConfig
@ -38,3 +48,71 @@ class HostedVLLMChatConfig(OpenAIGPTConfig):
api_key or get_secret_str("HOSTED_VLLM_API_KEY") or "fake-api-key"
) # vllm does not require an api key
return api_base, dynamic_api_key
def _is_video_file(self, content_item: ChatCompletionFileObject) -> bool:
"""
Check if the file is a video
- format: video/<extension>
- file_data: base64 encoded video data
- file_id: infer mp4 from extension
"""
file = content_item.get("file", {})
format = file.get("format")
file_data = file.get("file_data")
file_id = file.get("file_id")
if content_item.get("type") != "file":
return False
if format and format.startswith("video/"):
return True
elif file_data:
mime_type = _parse_mime_type(file_data)
if mime_type and mime_type.startswith("video/"):
return True
elif file_id:
mime_type = _get_image_mime_type_from_url(file_id)
if mime_type and mime_type.startswith("video/"):
return True
return False
def _convert_file_to_video_url(
self, content_item: ChatCompletionFileObject
) -> ChatCompletionVideoObject:
file = content_item.get("file", {})
file_id = file.get("file_id")
file_data = file.get("file_data")
if file_id:
return ChatCompletionVideoObject(
type="video_url", video_url=ChatCompletionVideoUrlObject(url=file_id)
)
elif file_data:
return ChatCompletionVideoObject(
type="video_url", video_url=ChatCompletionVideoUrlObject(url=file_data)
)
raise ValueError("file_id or file_data is required")
def _transform_messages(
self, messages: List[AllMessageValues], model: str
) -> List[AllMessageValues]:
"""
Support translating video files from file_id or file_data to video_url
"""
for message in messages:
if message["role"] == "user":
message_content = message.get("content")
if message_content and isinstance(message_content, list):
replaced_content_items: List[
Tuple[int, ChatCompletionFileObject]
] = []
for idx, content_item in enumerate(message_content):
if content_item.get("type") == "file":
content_item = cast(ChatCompletionFileObject, content_item)
if self._is_video_file(content_item):
replaced_content_items.append((idx, content_item))
for idx, content_item in replaced_content_items:
message_content[idx] = self._convert_file_to_video_url(
content_item
)
transformed_messages = super()._transform_messages(messages, model)
return transformed_messages

1
litellm/llms/litellm_proxy/chat/transformation.py
View file

@ -13,6 +13,7 @@ class LiteLLMProxyChatConfig(OpenAIGPTConfig):
def get_supported_openai_params(self, model: str) -> List:
list = super().get_supported_openai_params(model)
list.append("thinking")
list.append("reasoning_effort")
return list
def _map_openai_params(

5
litellm/llms/openai/chat/o_series_transformation.py
View file

@ -131,7 +131,10 @@ class OpenAIOSeriesConfig(OpenAIGPTConfig):
def is_model_o_series_model(self, model: str) -> bool:
if model in litellm.open_ai_chat_completion_models and (
"o1" in model or "o3" in model
"o1" in model
or "o3" in model
or "o4"
in model # [TODO] make this a more generic check (e.g. using `openai-o-series` as provider like gemini)
):
return True
return False

3
litellm/llms/openai/responses/transformation.py
View file

@ -110,7 +110,10 @@ class OpenAIResponsesAPIConfig(BaseResponsesAPIConfig):
def get_complete_url(
self,
api_base: Optional[str],
api_key: Optional[str],
model: str,
optional_params: dict,
litellm_params: dict,
stream: Optional[bool] = None,
) -> str:
"""

45
litellm/llms/openai_like/chat/transformation.py
View file

@ -7,7 +7,7 @@ from typing import TYPE_CHECKING, Any, List, Optional, Tuple, Union
import httpx
from litellm.secret_managers.main import get_secret_str
from litellm.types.llms.openai import ChatCompletionAssistantMessage
from litellm.types.llms.openai import AllMessageValues, ChatCompletionAssistantMessage
from litellm.types.utils import ModelResponse
from ...openai.chat.gpt_transformation import OpenAIGPTConfig
@ -25,7 +25,6 @@ class OpenAILikeChatConfig(OpenAIGPTConfig):
self,
api_base: Optional[str],
api_key: Optional[str],
model: Optional[str] = None,
) -> Tuple[Optional[str], Optional[str]]:
api_base = api_base or get_secret_str("OPENAI_LIKE_API_BASE") # type: ignore
dynamic_api_key = (
@ -74,8 +73,8 @@ class OpenAILikeChatConfig(OpenAIGPTConfig):
messages: List,
print_verbose,
encoding,
json_mode: bool,
custom_llm_provider: str,
json_mode: Optional[bool],
custom_llm_provider: Optional[str],
base_model: Optional[str],
) -> ModelResponse:
response_json = response.json()
@ -97,14 +96,46 @@ class OpenAILikeChatConfig(OpenAIGPTConfig):
returned_response = ModelResponse(**response_json)
returned_response.model = (
custom_llm_provider + "/" + (returned_response.model or "")
)
if custom_llm_provider is not None:
returned_response.model = (
custom_llm_provider + "/" + (returned_response.model or "")
)
if base_model is not None:
returned_response._hidden_params["model"] = base_model
return returned_response
def transform_response(
self,
model: str,
raw_response: httpx.Response,
model_response: ModelResponse,
logging_obj: LiteLLMLoggingObj,
request_data: dict,
messages: List[AllMessageValues],
optional_params: dict,
litellm_params: dict,
encoding: Any,
api_key: Optional[str] = None,
json_mode: Optional[bool] = None,
) -> ModelResponse:
return OpenAILikeChatConfig._transform_response(
model=model,
response=raw_response,
model_response=model_response,
stream=optional_params.get("stream", False),
logging_obj=logging_obj,
optional_params=optional_params,
api_key=api_key,
data=request_data,
messages=messages,
print_verbose=None,
encoding=None,
json_mode=json_mode,
custom_llm_provider=None,
base_model=None,
)
def map_openai_params(
self,
non_default_params: dict,

2
litellm/llms/triton/completion/transformation.py
View file

@ -201,8 +201,6 @@ class TritonGenerateConfig(TritonConfig):
"max_tokens": int(
optional_params.get("max_tokens", DEFAULT_MAX_TOKENS_FOR_TRITON)
),
"bad_words": [""],
"stop_words": [""],
},
"stream": bool(stream),
}

50
litellm/llms/vertex_ai/common_utils.py
View file

@ -1,5 +1,5 @@
from typing import Any, Dict, List, Literal, Optional, Set, Tuple, Union, get_type_hints
import re
from typing import Any, Dict, List, Literal, Optional, Set, Tuple, Union, get_type_hints
import httpx
@ -165,9 +165,18 @@ def _check_text_in_content(parts: List[PartType]) -> bool:
return has_text_param
def _build_vertex_schema(parameters: dict):
def _build_vertex_schema(parameters: dict, add_property_ordering: bool = False):
"""
This is a modified version of https://github.com/google-gemini/generative-ai-python/blob/8f77cc6ac99937cd3a81299ecf79608b91b06bbb/google/generativeai/types/content_types.py#L419
Updates the input parameters, removing extraneous fields, adjusting types, unwinding $defs, and adding propertyOrdering if specified, returning the updated parameters.
Parameters:
parameters: dict - the json schema to build from
add_property_ordering: bool - whether to add propertyOrdering to the schema. This is only applicable to schemas for structured outputs. See
set_schema_property_ordering for more details.
Returns:
parameters: dict - the input parameters, modified in place
"""
# Get valid fields from Schema TypedDict
valid_schema_fields = set(get_type_hints(Schema).keys())
@ -186,8 +195,40 @@ def _build_vertex_schema(parameters: dict):
add_object_type(parameters)
# Postprocessing
# Filter out fields that don't exist in Schema
filtered_parameters = filter_schema_fields(parameters, valid_schema_fields)
return filtered_parameters
parameters = filter_schema_fields(parameters, valid_schema_fields)
if add_property_ordering:
set_schema_property_ordering(parameters)
return parameters
def set_schema_property_ordering(
schema: Dict[str, Any], depth: int = 0
) -> Dict[str, Any]:
"""
vertex ai and generativeai apis order output of fields alphabetically, unless you specify the order.
python dicts retain order, so we just use that. Note that this field only applies to structured outputs, and not tools.
Function tools are not afflicted by the same alphabetical ordering issue, (the order of keys returned seems to be arbitrary, up to the model)
https://cloud.google.com/vertex-ai/docs/reference/rest/v1/projects.locations.cachedContents#Schema.FIELDS.property_ordering
Args:
schema: The schema dictionary to process
depth: Current recursion depth to prevent infinite loops
"""
if depth > DEFAULT_MAX_RECURSE_DEPTH:
raise ValueError(
f"Max depth of {DEFAULT_MAX_RECURSE_DEPTH} exceeded while processing schema. Please check the schema for excessive nesting."
)
if "properties" in schema and isinstance(schema["properties"], dict):
# retain propertyOrdering as an escape hatch if user already specifies it
if "propertyOrdering" not in schema:
schema["propertyOrdering"] = [k for k, v in schema["properties"].items()]
for k, v in schema["properties"].items():
set_schema_property_ordering(v, depth + 1)
if "items" in schema:
set_schema_property_ordering(schema["items"], depth + 1)
return schema
def filter_schema_fields(
@ -360,6 +401,7 @@ def construct_target_url(
Constructed Url:
POST https://LOCATION-aiplatform.googleapis.com/{version}/projects/PROJECT_ID/locations/LOCATION/cachedContents
"""
new_base_url = httpx.URL(base_url)
if "locations" in requested_route: # contains the target project id + location
if vertex_project and vertex_location:

65
litellm/llms/vertex_ai/gemini/transformation.py
View file

@ -12,6 +12,9 @@ from pydantic import BaseModel
import litellm
from litellm._logging import verbose_logger
from litellm.litellm_core_utils.prompt_templates.common_utils import (
_get_image_mime_type_from_url,
)
from litellm.litellm_core_utils.prompt_templates.factory import (
convert_to_anthropic_image_obj,
convert_to_gemini_tool_call_invoke,
@ -99,62 +102,6 @@ def _process_gemini_image(image_url: str, format: Optional[str] = None) -> PartT
raise e
def _get_image_mime_type_from_url(url: str) -> Optional[str]:
"""
Get mime type for common image URLs
See gemini mime types: https://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/image-understanding#image-requirements
Supported by Gemini:
application/pdf
audio/mpeg
audio/mp3
audio/wav
image/png
image/jpeg
image/webp
text/plain
video/mov
video/mpeg
video/mp4
video/mpg
video/avi
video/wmv
video/mpegps
video/flv
"""
url = url.lower()
# Map file extensions to mime types
mime_types = {
# Images
(".jpg", ".jpeg"): "image/jpeg",
(".png",): "image/png",
(".webp",): "image/webp",
# Videos
(".mp4",): "video/mp4",
(".mov",): "video/mov",
(".mpeg", ".mpg"): "video/mpeg",
(".avi",): "video/avi",
(".wmv",): "video/wmv",
(".mpegps",): "video/mpegps",
(".flv",): "video/flv",
# Audio
(".mp3",): "audio/mp3",
(".wav",): "audio/wav",
(".mpeg",): "audio/mpeg",
# Documents
(".pdf",): "application/pdf",
(".txt",): "text/plain",
}
# Check each extension group against the URL
for extensions, mime_type in mime_types.items():
if any(url.endswith(ext) for ext in extensions):
return mime_type
return None
def _gemini_convert_messages_with_history( # noqa: PLR0915
messages: List[AllMessageValues],
) -> List[ContentType]:
@ -269,6 +216,11 @@ def _gemini_convert_messages_with_history( # noqa: PLR0915
msg_dict = messages[msg_i] # type: ignore
assistant_msg = ChatCompletionAssistantMessage(**msg_dict) # type: ignore
_message_content = assistant_msg.get("content", None)
reasoning_content = assistant_msg.get("reasoning_content", None)
if reasoning_content is not None:
assistant_content.append(
PartType(thought=True, text=reasoning_content)
)
if _message_content is not None and isinstance(_message_content, list):
_parts = []
for element in _message_content:
@ -276,6 +228,7 @@ def _gemini_convert_messages_with_history( # noqa: PLR0915
if element["type"] == "text":
_part = PartType(text=element["text"])
_parts.append(_part)
assistant_content.extend(_parts)
elif (
_message_content is not None

110
litellm/llms/vertex_ai/gemini/vertex_and_google_ai_studio_gemini.py
View file

@ -24,6 +24,11 @@ import litellm
import litellm.litellm_core_utils
import litellm.litellm_core_utils.litellm_logging
from litellm import verbose_logger
from litellm.constants import (
DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
)
from litellm.litellm_core_utils.core_helpers import map_finish_reason
from litellm.llms.base_llm.chat.transformation import BaseConfig, BaseLLMException
from litellm.llms.custom_httpx.http_handler import (
@ -31,6 +36,7 @@ from litellm.llms.custom_httpx.http_handler import (
HTTPHandler,
get_async_httpx_client,
)
from litellm.types.llms.anthropic import AnthropicThinkingParam
from litellm.types.llms.openai import (
AllMessageValues,
ChatCompletionResponseMessage,
@ -45,6 +51,7 @@ from litellm.types.llms.vertex_ai import (
ContentType,
FunctionCallingConfig,
FunctionDeclaration,
GeminiThinkingConfig,
GenerateContentResponseBody,
HttpxPartType,
LogprobsResult,
@ -59,7 +66,7 @@ from litellm.types.utils import (
TopLogprob,
Usage,
)
from litellm.utils import CustomStreamWrapper, ModelResponse
from litellm.utils import CustomStreamWrapper, ModelResponse, supports_reasoning
from ....utils import _remove_additional_properties, _remove_strict_from_schema
from ..common_utils import VertexAIError, _build_vertex_schema
@ -190,7 +197,7 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
return super().get_config()
def get_supported_openai_params(self, model: str) -> List[str]:
return [
supported_params = [
"temperature",
"top_p",
"max_tokens",
@ -207,9 +214,13 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
"extra_headers",
"seed",
"logprobs",
"top_logprobs", # Added this to list of supported openAI params
"top_logprobs",
"modalities",
]
if supports_reasoning(model):
supported_params.append("reasoning_effort")
supported_params.append("thinking")
return supported_params
def map_tool_choice_values(
self, model: str, tool_choice: Union[str, dict]
@ -313,9 +324,14 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
if isinstance(old_schema, list):
for item in old_schema:
if isinstance(item, dict):
item = _build_vertex_schema(parameters=item)
item = _build_vertex_schema(
parameters=item, add_property_ordering=True
)
elif isinstance(old_schema, dict):
old_schema = _build_vertex_schema(parameters=old_schema)
old_schema = _build_vertex_schema(
parameters=old_schema, add_property_ordering=True
)
return old_schema
def apply_response_schema_transformation(self, value: dict, optional_params: dict):
@ -342,6 +358,43 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
value=optional_params["response_schema"]
)
@staticmethod
def _map_reasoning_effort_to_thinking_budget(
reasoning_effort: str,
) -> GeminiThinkingConfig:
if reasoning_effort == "low":
return {
"thinkingBudget": DEFAULT_REASONING_EFFORT_LOW_THINKING_BUDGET,
"includeThoughts": True,
}
elif reasoning_effort == "medium":
return {
"thinkingBudget": DEFAULT_REASONING_EFFORT_MEDIUM_THINKING_BUDGET,
"includeThoughts": True,
}
elif reasoning_effort == "high":
return {
"thinkingBudget": DEFAULT_REASONING_EFFORT_HIGH_THINKING_BUDGET,
"includeThoughts": True,
}
else:
raise ValueError(f"Invalid reasoning effort: {reasoning_effort}")
@staticmethod
def _map_thinking_param(
thinking_param: AnthropicThinkingParam,
) -> GeminiThinkingConfig:
thinking_enabled = thinking_param.get("type") == "enabled"
thinking_budget = thinking_param.get("budget_tokens")
params: GeminiThinkingConfig = {}
if thinking_enabled:
params["includeThoughts"] = True
if thinking_budget:
params["thinkingBudget"] = thinking_budget
return params
def map_openai_params(
self,
non_default_params: Dict,
@ -398,6 +451,16 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
optional_params["tool_choice"] = _tool_choice_value
elif param == "seed":
optional_params["seed"] = value
elif param == "reasoning_effort" and isinstance(value, str):
optional_params[
"thinkingConfig"
] = VertexGeminiConfig._map_reasoning_effort_to_thinking_budget(value)
elif param == "thinking":
optional_params[
"thinkingConfig"
] = VertexGeminiConfig._map_thinking_param(
cast(AnthropicThinkingParam, value)
)
elif param == "modalities" and isinstance(value, list):
response_modalities = []
for modality in value:
@ -513,19 +576,28 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
def get_assistant_content_message(
self, parts: List[HttpxPartType]
) -> Optional[str]:
_content_str = ""
) -> Tuple[Optional[str], Optional[str]]:
content_str: Optional[str] = None
reasoning_content_str: Optional[str] = None
for part in parts:
_content_str = ""
if "text" in part:
_content_str += part["text"]
elif "inlineData" in part: # base64 encoded image
_content_str += "data:{};base64,{}".format(
part["inlineData"]["mimeType"], part["inlineData"]["data"]
)
if len(_content_str) > 0:
if part.get("thought") is True:
if reasoning_content_str is None:
reasoning_content_str = ""
reasoning_content_str += _content_str
else:
if content_str is None:
content_str = ""
content_str += _content_str
if _content_str:
return _content_str
return None
return content_str, reasoning_content_str
def _transform_parts(
self,
@ -676,6 +748,7 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
audio_tokens: Optional[int] = None
text_tokens: Optional[int] = None
prompt_tokens_details: Optional[PromptTokensDetailsWrapper] = None
reasoning_tokens: Optional[int] = None
if "cachedContentTokenCount" in completion_response["usageMetadata"]:
cached_tokens = completion_response["usageMetadata"][
"cachedContentTokenCount"
@ -686,7 +759,10 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
audio_tokens = detail["tokenCount"]
elif detail["modality"] == "TEXT":
text_tokens = detail["tokenCount"]
if "thoughtsTokenCount" in completion_response["usageMetadata"]:
reasoning_tokens = completion_response["usageMetadata"][
"thoughtsTokenCount"
]
prompt_tokens_details = PromptTokensDetailsWrapper(
cached_tokens=cached_tokens,
audio_tokens=audio_tokens,
@ -702,6 +778,7 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
),
total_tokens=completion_response["usageMetadata"].get("totalTokenCount", 0),
prompt_tokens_details=prompt_tokens_details,
reasoning_tokens=reasoning_tokens,
)
return usage
@ -730,11 +807,16 @@ class VertexGeminiConfig(VertexAIBaseConfig, BaseConfig):
citation_metadata.append(candidate["citationMetadata"])
if "parts" in candidate["content"]:
chat_completion_message[
"content"
] = VertexGeminiConfig().get_assistant_content_message(
(
content,
reasoning_content,
) = VertexGeminiConfig().get_assistant_content_message(
parts=candidate["content"]["parts"]
)
if content is not None:
chat_completion_message["content"] = content
if reasoning_content is not None:
chat_completion_message["reasoning_content"] = reasoning_content
functions, tools = self._transform_parts(
parts=candidate["content"]["parts"],

2
litellm/llms/watsonx/common_utils.py
View file

@ -38,7 +38,7 @@ def generate_iam_token(api_key=None, **params) -> str:
headers = {}
headers["Content-Type"] = "application/x-www-form-urlencoded"
if api_key is None:
api_key = get_secret_str("WX_API_KEY") or get_secret_str("WATSONX_API_KEY")
api_key = get_secret_str("WX_API_KEY") or get_secret_str("WATSONX_API_KEY") or get_secret_str("WATSONX_APIKEY")
if api_key is None:
raise ValueError("API key is required")
headers["Accept"] = "application/json"

76
litellm/main.py
View file

@ -1435,6 +1435,7 @@ def completion( # type: ignore # noqa: PLR0915
custom_llm_provider=custom_llm_provider,
encoding=encoding,
stream=stream,
provider_config=provider_config,
)
except Exception as e:
## LOGGING - log the original exception returned
@ -1596,6 +1597,37 @@ def completion( # type: ignore # noqa: PLR0915
additional_args={"headers": headers},
)
response = _response
elif custom_llm_provider == "fireworks_ai":
## COMPLETION CALL
try:
response = base_llm_http_handler.completion(
model=model,
messages=messages,
headers=headers,
model_response=model_response,
api_key=api_key,
api_base=api_base,
acompletion=acompletion,
logging_obj=logging,
optional_params=optional_params,
litellm_params=litellm_params,
timeout=timeout, # type: ignore
client=client,
custom_llm_provider=custom_llm_provider,
encoding=encoding,
stream=stream,
provider_config=provider_config,
)
except Exception as e:
## LOGGING - log the original exception returned
logging.post_call(
input=messages,
api_key=api_key,
original_response=str(e),
additional_args={"headers": headers},
)
raise e
elif custom_llm_provider == "groq":
api_base = (
api_base # for deepinfra/perplexity/anyscale/groq/friendliai we check in get_llm_provider and pass in the api base from there
@ -1622,24 +1654,22 @@ def completion( # type: ignore # noqa: PLR0915
): # completion(top_k=3) > openai_config(top_k=3) <- allows for dynamic variables to be passed in
optional_params[k] = v
response = groq_chat_completions.completion(
response = base_llm_http_handler.completion(
model=model,
stream=stream,
messages=messages,
headers=headers,
model_response=model_response,
print_verbose=print_verbose,
api_key=api_key,
api_base=api_base,
acompletion=acompletion,
logging_obj=logging,
api_base=api_base,
model_response=model_response,
optional_params=optional_params,
litellm_params=litellm_params,
logger_fn=logger_fn,
timeout=timeout, # type: ignore
custom_prompt_dict=custom_prompt_dict,
client=client, # pass AsyncOpenAI, OpenAI client
custom_llm_provider=custom_llm_provider,
timeout=timeout,
headers=headers,
encoding=encoding,
api_key=api_key,
logging_obj=logging, # model call logging done inside the class as we make need to modify I/O to fit aleph alpha's requirements
client=client,
)
elif custom_llm_provider == "aiohttp_openai":
# NEW aiohttp provider for 10-100x higher RPS
@ -2658,9 +2688,9 @@ def completion( # type: ignore # noqa: PLR0915
"aws_region_name" not in optional_params
or optional_params["aws_region_name"] is None
):
optional_params["aws_region_name"] = (
aws_bedrock_client.meta.region_name
)
optional_params[
"aws_region_name"
] = aws_bedrock_client.meta.region_name
bedrock_route = BedrockModelInfo.get_bedrock_route(model)
if bedrock_route == "converse":
@ -4382,9 +4412,9 @@ def adapter_completion(
new_kwargs = translation_obj.translate_completion_input_params(kwargs=kwargs)
response: Union[ModelResponse, CustomStreamWrapper] = completion(**new_kwargs) # type: ignore
translated_response: Optional[Union[BaseModel, AdapterCompletionStreamWrapper]] = (
None
)
translated_response: Optional[
Union[BaseModel, AdapterCompletionStreamWrapper]
] = None
if isinstance(response, ModelResponse):
translated_response = translation_obj.translate_completion_output_params(
response=response
@ -5804,9 +5834,9 @@ def stream_chunk_builder( # noqa: PLR0915
]
if len(content_chunks) > 0:
response["choices"][0]["message"]["content"] = (
processor.get_combined_content(content_chunks)
)
response["choices"][0]["message"][
"content"
] = processor.get_combined_content(content_chunks)
reasoning_chunks = [
chunk
@ -5817,9 +5847,9 @@ def stream_chunk_builder( # noqa: PLR0915
]
if len(reasoning_chunks) > 0:
response["choices"][0]["message"]["reasoning_content"] = (
processor.get_combined_reasoning_content(reasoning_chunks)
)
response["choices"][0]["message"][
"reasoning_content"
] = processor.get_combined_reasoning_content(reasoning_chunks)
audio_chunks = [
chunk

462
litellm/model_prices_and_context_window_backup.json
View file

@ -5,6 +5,7 @@
"max_output_tokens": "max output tokens, if the provider specifies it. if not default to max_tokens",
"input_cost_per_token": 0.0000,
"output_cost_per_token": 0.000,
"output_cost_per_reasoning_token": 0.000,
"litellm_provider": "one of https://docs.litellm.ai/docs/providers",
"mode": "one of: chat, embedding, completion, image_generation, audio_transcription, audio_speech, image_generation, moderation, rerank",
"supports_function_calling": true,
@ -600,6 +601,40 @@
"supports_vision": true,
"supports_prompt_caching": true
},
"o3": {
"max_tokens": 100000,
"max_input_tokens": 200000,
"max_output_tokens": 100000,
"input_cost_per_token": 1e-5,
"output_cost_per_token": 4e-5,
"cache_read_input_token_cost": 2.5e-6,
"litellm_provider": "openai",
"mode": "chat",
"supports_function_calling": true,
"supports_parallel_function_calling": false,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_response_schema": true,
"supports_reasoning": true,
"supports_tool_choice": true
},
"o3-2025-04-16": {
"max_tokens": 100000,
"max_input_tokens": 200000,
"max_output_tokens": 100000,
"input_cost_per_token": 1e-5,
"output_cost_per_token": 4e-5,
"cache_read_input_token_cost": 2.5e-6,
"litellm_provider": "openai",
"mode": "chat",
"supports_function_calling": true,
"supports_parallel_function_calling": false,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_response_schema": true,
"supports_reasoning": true,
"supports_tool_choice": true
},
"o3-mini": {
"max_tokens": 100000,
"max_input_tokens": 200000,
@ -634,6 +669,40 @@
"supports_reasoning": true,
"supports_tool_choice": true
},
"o4-mini": {
"max_tokens": 100000,
"max_input_tokens": 200000,
"max_output_tokens": 100000,
"input_cost_per_token": 1.1e-6,
"output_cost_per_token": 4.4e-6,
"cache_read_input_token_cost": 2.75e-7,
"litellm_provider": "openai",
"mode": "chat",
"supports_function_calling": true,
"supports_parallel_function_calling": false,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_response_schema": true,
"supports_reasoning": true,
"supports_tool_choice": true
},
"o4-mini-2025-04-16": {
"max_tokens": 100000,
"max_input_tokens": 200000,
"max_output_tokens": 100000,
"input_cost_per_token": 1.1e-6,
"output_cost_per_token": 4.4e-6,
"cache_read_input_token_cost": 2.75e-7,
"litellm_provider": "openai",
"mode": "chat",
"supports_function_calling": true,
"supports_parallel_function_calling": false,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_response_schema": true,
"supports_reasoning": true,
"supports_tool_choice": true
},
"o1-mini-2024-09-12": {
"max_tokens": 65536,
"max_input_tokens": 128000,
@ -1403,6 +1472,295 @@
"litellm_provider": "openai",
"supported_endpoints": ["/v1/audio/speech"]
},
"azure/computer-use-preview": {
"max_tokens": 1024,
"max_input_tokens": 8192,
"max_output_tokens": 1024,
"input_cost_per_token": 0.000003,
"output_cost_per_token": 0.000012,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/responses"],
"supported_modalities": ["text", "image"],
"supported_output_modalities": ["text"],
"supports_function_calling": true,
"supports_parallel_function_calling": true,
"supports_response_schema": true,
"supports_vision": true,
"supports_prompt_caching": false,
"supports_system_messages": true,
"supports_tool_choice": true,
"supports_native_streaming": false,
"supports_reasoning": true
},
"azure/gpt-4o-audio-preview-2024-12-17": {
"max_tokens": 16384,
"max_input_tokens": 128000,
"max_output_tokens": 16384,
"input_cost_per_token": 0.0000025,
"input_cost_per_audio_token": 0.00004,
"output_cost_per_token": 0.00001,
"output_cost_per_audio_token": 0.00008,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions"],
"supported_modalities": ["text", "audio"],
"supported_output_modalities": ["text", "audio"],
"supports_function_calling": true,
"supports_parallel_function_calling": true,
"supports_response_schema": false,
"supports_vision": false,
"supports_prompt_caching": false,
"supports_system_messages": true,
"supports_tool_choice": true,
"supports_native_streaming": true,
"supports_reasoning": false
},
"azure/gpt-4o-mini-audio-preview-2024-12-17": {
"max_tokens": 16384,
"max_input_tokens": 128000,
"max_output_tokens": 16384,
"input_cost_per_token": 0.0000025,
"input_cost_per_audio_token": 0.00004,
"output_cost_per_token": 0.00001,
"output_cost_per_audio_token": 0.00008,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions"],
"supported_modalities": ["text", "audio"],
"supported_output_modalities": ["text", "audio"],
"supports_function_calling": true,
"supports_parallel_function_calling": true,
"supports_response_schema": false,
"supports_vision": false,
"supports_prompt_caching": false,
"supports_system_messages": true,
"supports_tool_choice": true,
"supports_native_streaming": true,
"supports_reasoning": false
},
"azure/gpt-4.1": {
"max_tokens": 32768,
"max_input_tokens": 1047576,
"max_output_tokens": 32768,
"input_cost_per_token": 2e-6,
"output_cost_per_token": 8e-6,
"input_cost_per_token_batches": 1e-6,
"output_cost_per_token_batches": 4e-6,
"cache_read_input_token_cost": 0.5e-6,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions", "/v1/batch", "/v1/responses"],
"supported_modalities": ["text", "image"],
"supported_output_modalities": ["text"],
"supports_function_calling": true,
"supports_parallel_function_calling": true,
"supports_response_schema": true,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_system_messages": true,
"supports_tool_choice": true,
"supports_native_streaming": true,
"supports_web_search": true,
"search_context_cost_per_query": {
"search_context_size_low": 30e-3,
"search_context_size_medium": 35e-3,
"search_context_size_high": 50e-3
}
},
"azure/gpt-4.1-2025-04-14": {
"max_tokens": 32768,
"max_input_tokens": 1047576,
"max_output_tokens": 32768,
"input_cost_per_token": 2e-6,
"output_cost_per_token": 8e-6,
"input_cost_per_token_batches": 1e-6,
"output_cost_per_token_batches": 4e-6,
"cache_read_input_token_cost": 0.5e-6,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions", "/v1/batch", "/v1/responses"],
"supported_modalities": ["text", "image"],
"supported_output_modalities": ["text"],
"supports_function_calling": true,
"supports_parallel_function_calling": true,
"supports_response_schema": true,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_system_messages": true,
"supports_tool_choice": true,
"supports_native_streaming": true,
"supports_web_search": true,
"search_context_cost_per_query": {
"search_context_size_low": 30e-3,
"search_context_size_medium": 35e-3,
"search_context_size_high": 50e-3
}
},
"azure/gpt-4.1-mini": {
"max_tokens": 32768,
"max_input_tokens": 1047576,
"max_output_tokens": 32768,
"input_cost_per_token": 0.4e-6,
"output_cost_per_token": 1.6e-6,
"input_cost_per_token_batches": 0.2e-6,
"output_cost_per_token_batches": 0.8e-6,
"cache_read_input_token_cost": 0.1e-6,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions", "/v1/batch", "/v1/responses"],
"supported_modalities": ["text", "image"],
"supported_output_modalities": ["text"],
"supports_function_calling": true,
"supports_parallel_function_calling": true,
"supports_response_schema": true,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_system_messages": true,
"supports_tool_choice": true,
"supports_native_streaming": true,
"supports_web_search": true,
"search_context_cost_per_query": {
"search_context_size_low": 25e-3,
"search_context_size_medium": 27.5e-3,
"search_context_size_high": 30e-3
}
},
"azure/gpt-4.1-mini-2025-04-14": {
"max_tokens": 32768,
"max_input_tokens": 1047576,
"max_output_tokens": 32768,
"input_cost_per_token": 0.4e-6,
"output_cost_per_token": 1.6e-6,
"input_cost_per_token_batches": 0.2e-6,
"output_cost_per_token_batches": 0.8e-6,
"cache_read_input_token_cost": 0.1e-6,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions", "/v1/batch", "/v1/responses"],
"supported_modalities": ["text", "image"],
"supported_output_modalities": ["text"],
"supports_function_calling": true,
"supports_parallel_function_calling": true,
"supports_response_schema": true,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_system_messages": true,
"supports_tool_choice": true,
"supports_native_streaming": true,
"supports_web_search": true,
"search_context_cost_per_query": {
"search_context_size_low": 25e-3,
"search_context_size_medium": 27.5e-3,
"search_context_size_high": 30e-3
}
},
"azure/gpt-4.1-nano": {
"max_tokens": 32768,
"max_input_tokens": 1047576,
"max_output_tokens": 32768,
"input_cost_per_token": 0.1e-6,
"output_cost_per_token": 0.4e-6,
"input_cost_per_token_batches": 0.05e-6,
"output_cost_per_token_batches": 0.2e-6,
"cache_read_input_token_cost": 0.025e-6,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions", "/v1/batch", "/v1/responses"],
"supported_modalities": ["text", "image"],
"supported_output_modalities": ["text"],
"supports_function_calling": true,
"supports_parallel_function_calling": true,
"supports_response_schema": true,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_system_messages": true,
"supports_tool_choice": true,
"supports_native_streaming": true
},
"azure/gpt-4.1-nano-2025-04-14": {
"max_tokens": 32768,
"max_input_tokens": 1047576,
"max_output_tokens": 32768,
"input_cost_per_token": 0.1e-6,
"output_cost_per_token": 0.4e-6,
"input_cost_per_token_batches": 0.05e-6,
"output_cost_per_token_batches": 0.2e-6,
"cache_read_input_token_cost": 0.025e-6,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions", "/v1/batch", "/v1/responses"],
"supported_modalities": ["text", "image"],
"supported_output_modalities": ["text"],
"supports_function_calling": true,
"supports_parallel_function_calling": true,
"supports_response_schema": true,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_system_messages": true,
"supports_tool_choice": true,
"supports_native_streaming": true
},
"azure/o3": {
"max_tokens": 100000,
"max_input_tokens": 200000,
"max_output_tokens": 100000,
"input_cost_per_token": 1e-5,
"output_cost_per_token": 4e-5,
"cache_read_input_token_cost": 2.5e-6,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions", "/v1/batch", "/v1/responses"],
"supported_modalities": ["text", "image"],
"supported_output_modalities": ["text"],
"supports_function_calling": true,
"supports_parallel_function_calling": false,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_response_schema": true,
"supports_reasoning": true,
"supports_tool_choice": true
},
"azure/o3-2025-04-16": {
"max_tokens": 100000,
"max_input_tokens": 200000,
"max_output_tokens": 100000,
"input_cost_per_token": 1e-5,
"output_cost_per_token": 4e-5,
"cache_read_input_token_cost": 2.5e-6,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions", "/v1/batch", "/v1/responses"],
"supported_modalities": ["text", "image"],
"supported_output_modalities": ["text"],
"supports_function_calling": true,
"supports_parallel_function_calling": false,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_response_schema": true,
"supports_reasoning": true,
"supports_tool_choice": true
},
"azure/o4-mini": {
"max_tokens": 100000,
"max_input_tokens": 200000,
"max_output_tokens": 100000,
"input_cost_per_token": 1.1e-6,
"output_cost_per_token": 4.4e-6,
"cache_read_input_token_cost": 2.75e-7,
"litellm_provider": "azure",
"mode": "chat",
"supported_endpoints": ["/v1/chat/completions", "/v1/batch", "/v1/responses"],
"supported_modalities": ["text", "image"],
"supported_output_modalities": ["text"],
"supports_function_calling": true,
"supports_parallel_function_calling": false,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_response_schema": true,
"supports_reasoning": true,
"supports_tool_choice": true
},
"azure/gpt-4o-mini-realtime-preview-2024-12-17": {
"max_tokens": 4096,
"max_input_tokens": 128000,
@ -1579,6 +1937,23 @@
"supports_system_messages": true,
"supports_tool_choice": true
},
"azure/o4-mini-2025-04-16": {
"max_tokens": 100000,
"max_input_tokens": 200000,
"max_output_tokens": 100000,
"input_cost_per_token": 1.1e-6,
"output_cost_per_token": 4.4e-6,
"cache_read_input_token_cost": 2.75e-7,
"litellm_provider": "azure",
"mode": "chat",
"supports_function_calling": true,
"supports_parallel_function_calling": false,
"supports_vision": true,
"supports_prompt_caching": true,
"supports_response_schema": true,
"supports_reasoning": true,
"supports_tool_choice": true
},
"azure/o3-mini-2025-01-31": {
"max_tokens": 100000,
"max_input_tokens": 200000,
@ -5025,6 +5400,64 @@
"source": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#gemini-2.0-flash",
"supports_tool_choice": true
},
"gemini/gemini-2.5-flash-preview-04-17": {
"max_tokens": 65536,
"max_input_tokens": 1048576,
"max_output_tokens": 65536,
"max_images_per_prompt": 3000,
"max_videos_per_prompt": 10,
"max_video_length": 1,
"max_audio_length_hours": 8.4,
"max_audio_per_prompt": 1,
"max_pdf_size_mb": 30,
"input_cost_per_audio_token": 1e-6,
"input_cost_per_token": 0.15e-6,
"output_cost_per_token": 0.6e-6,
"output_cost_per_reasoning_token": 3.5e-6,
"litellm_provider": "gemini",
"mode": "chat",
"rpm": 10,
"tpm": 250000,
"supports_system_messages": true,
"supports_function_calling": true,
"supports_vision": true,
"supports_reasoning": true,
"supports_response_schema": true,
"supports_audio_output": false,
"supports_tool_choice": true,
"supported_endpoints": ["/v1/chat/completions", "/v1/completions"],
"supported_modalities": ["text", "image", "audio", "video"],
"supported_output_modalities": ["text"],
"source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview"
},
"gemini-2.5-flash-preview-04-17": {
"max_tokens": 65536,
"max_input_tokens": 1048576,
"max_output_tokens": 65536,
"max_images_per_prompt": 3000,
"max_videos_per_prompt": 10,
"max_video_length": 1,
"max_audio_length_hours": 8.4,
"max_audio_per_prompt": 1,
"max_pdf_size_mb": 30,
"input_cost_per_audio_token": 1e-6,
"input_cost_per_token": 0.15e-6,
"output_cost_per_token": 0.6e-6,
"output_cost_per_reasoning_token": 3.5e-6,
"litellm_provider": "vertex_ai-language-models",
"mode": "chat",
"supports_reasoning": true,
"supports_system_messages": true,
"supports_function_calling": true,
"supports_vision": true,
"supports_response_schema": true,
"supports_audio_output": false,
"supports_tool_choice": true,
"supported_endpoints": ["/v1/chat/completions", "/v1/completions", "/v1/batch"],
"supported_modalities": ["text", "image", "audio", "video"],
"supported_output_modalities": ["text"],
"source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview"
},
"gemini-2.0-flash": {
"max_tokens": 8192,
"max_input_tokens": 1048576,
@ -5099,6 +5532,35 @@
"source": "https://cloud.google.com/vertex-ai/generative-ai/docs/learn/models#gemini-2.0-flash",
"supports_tool_choice": true
},
"gemini-2.5-pro-preview-03-25": {
"max_tokens": 65536,
"max_input_tokens": 1048576,
"max_output_tokens": 65536,
"max_images_per_prompt": 3000,
"max_videos_per_prompt": 10,
"max_video_length": 1,
"max_audio_length_hours": 8.4,
"max_audio_per_prompt": 1,
"max_pdf_size_mb": 30,
"input_cost_per_audio_token": 0.00000125,
"input_cost_per_token": 0.00000125,
"input_cost_per_token_above_200k_tokens": 0.0000025,
"output_cost_per_token": 0.00001,
"output_cost_per_token_above_200k_tokens": 0.000015,
"litellm_provider": "vertex_ai-language-models",
"mode": "chat",
"supports_reasoning": true,
"supports_system_messages": true,
"supports_function_calling": true,
"supports_vision": true,
"supports_response_schema": true,
"supports_audio_output": false,
"supports_tool_choice": true,
"supported_endpoints": ["/v1/chat/completions", "/v1/completions", "/v1/batch"],
"supported_modalities": ["text", "image", "audio", "video"],
"supported_output_modalities": ["text"],
"source": "https://ai.google.dev/gemini-api/docs/models#gemini-2.5-flash-preview"
},
"gemini/gemini-2.0-pro-exp-02-05": {
"max_tokens": 8192,
"max_input_tokens": 2097152,

1
litellm/openai-responses-starter-app Submodule

@ -0,0 +1 @@
Subproject commit bf0485467c343957ba5c217db777f407b2e65453

0
litellm/proxy/_experimental/out/_next/static/cxKWUnb1loD53gKRUqiFm/_buildManifest.js → litellm/proxy/_experimental/out/_next/static/FPIQgzUY81b7nl8zNun4_/_buildManifest.js
View file

0
litellm/proxy/_experimental/out/_next/static/cxKWUnb1loD53gKRUqiFm/_ssgManifest.js → litellm/proxy/_experimental/out/_next/static/FPIQgzUY81b7nl8zNun4_/_ssgManifest.js
View file

1
litellm/proxy/_experimental/out/_next/static/chunks/250-a927a558002d8fb9.js Normal file
View file

File diff suppressed because one or more lines are too long

1
litellm/proxy/_experimental/out/_next/static/chunks/250-ef859ae76fcc7788.js
View file

File diff suppressed because one or more lines are too long

2
litellm/proxy/_experimental/out/_next/static/chunks/261-d4b99bc9f53d4ef3.js → litellm/proxy/_experimental/out/_next/static/chunks/261-57d48f76eec1e568.js
View file

File diff suppressed because one or more lines are too long

2
litellm/proxy/_experimental/out/_next/static/chunks/42-6810261f4d6c8bbf.js → litellm/proxy/_experimental/out/_next/static/chunks/42-59f99bfbf676f282.js
View file

File diff suppressed because one or more lines are too long

0
litellm/proxy/_experimental/out/_next/static/chunks/699-b82671c2a3a100c7.js → litellm/proxy/_experimental/out/_next/static/chunks/699-99a8a36b70ac90c1.js
View file

11
litellm/proxy/_experimental/out/_next/static/chunks/817-c8c94d51d28af935.js
View file

File diff suppressed because one or more lines are too long

11
litellm/proxy/_experimental/out/_next/static/chunks/860-c1d8f124df444312.js Normal file
View file

File diff suppressed because one or more lines are too long

0
litellm/proxy/_experimental/out/_next/static/chunks/899-8d832fe7c09b2afe.js → litellm/proxy/_experimental/out/_next/static/chunks/899-9af4feaf6f21839c.js
View file

1
litellm/proxy/_experimental/out/_next/static/chunks/app/layout-429ad74a94df7643.js Normal file
View file

@ -0,0 +1 @@
(self.webpackChunk_N_E=self.webpackChunk_N_E||[]).push([[185],{96443:function(n,e,t){Promise.resolve().then(t.t.bind(t,39974,23)),Promise.resolve().then(t.t.bind(t,2778,23))},2778:function(){},39974:function(n){n.exports={style:{fontFamily:"'__Inter_cf7686', '__Inter_Fallback_cf7686'",fontStyle:"normal"},className:"__className_cf7686"}}},function(n){n.O(0,[919,986,971,117,744],function(){return n(n.s=96443)}),_N_E=n.O()}]);

1
litellm/proxy/_experimental/out/_next/static/chunks/app/layout-af8319e6c59a08da.js
View file

@ -1 +0,0 @@
(self.webpackChunk_N_E=self.webpackChunk_N_E||[]).push([[185],{6580:function(n,e,t){Promise.resolve().then(t.t.bind(t,39974,23)),Promise.resolve().then(t.t.bind(t,2778,23))},2778:function(){},39974:function(n){n.exports={style:{fontFamily:"'__Inter_cf7686', '__Inter_Fallback_cf7686'",fontStyle:"normal"},className:"__className_cf7686"}}},function(n){n.O(0,[919,986,971,117,744],function(){return n(n.s=6580)}),_N_E=n.O()}]);

2
litellm/proxy/_experimental/out/_next/static/chunks/app/model_hub/page-068a441595bd0fc3.js → litellm/proxy/_experimental/out/_next/static/chunks/app/model_hub/page-cde2fb783e81a6c1.js
View file

@ -1 +1 @@
(self.webpackChunk_N_E=self.webpackChunk_N_E||[]).push([[418],{11790:function(e,n,u){Promise.resolve().then(u.bind(u,52829))},52829:function(e,n,u){"use strict";u.r(n),u.d(n,{default:function(){return f}});var t=u(57437),s=u(2265),r=u(99376),c=u(92699);function f(){let e=(0,r.useSearchParams)().get("key"),[n,u]=(0,s.useState)(null);return(0,s.useEffect)(()=>{e&&u(e)},[e]),(0,t.jsx)(c.Z,{accessToken:n,publicPage:!0,premiumUser:!1})}}},function(e){e.O(0,[42,261,250,699,971,117,744],function(){return e(e.s=11790)}),_N_E=e.O()}]);
(self.webpackChunk_N_E=self.webpackChunk_N_E||[]).push([[418],{21024:function(e,n,u){Promise.resolve().then(u.bind(u,52829))},52829:function(e,n,u){"use strict";u.r(n),u.d(n,{default:function(){return f}});var t=u(57437),s=u(2265),r=u(99376),c=u(92699);function f(){let e=(0,r.useSearchParams)().get("key"),[n,u]=(0,s.useState)(null);return(0,s.useEffect)(()=>{e&&u(e)},[e]),(0,t.jsx)(c.Z,{accessToken:n,publicPage:!0,premiumUser:!1})}}},function(e){e.O(0,[42,261,250,699,971,117,744],function(){return e(e.s=21024)}),_N_E=e.O()}]);

Some files were not shown because too many files have changed in this diff Show more