Merge b82fe58f6c into b82af5b826

docs/my-website/docs/completion/video.md

@@ -0,0 +1,149 @@
+import Tabs from '@theme/Tabs';
+import TabItem from '@theme/TabItem';
+
+# Using Video Models
+
+## Quick Start
+Example passing videos to a model 
+
+
+## Proxy Config setup
+
+Admins can configure file storage settings and retention policies in their LiteLLM config:
+
+```yaml
+model_list: 
+  - model_name: vertex_ai/*
+    litellm_params:
+      model: vertex_ai/*
+  - model_name: bedrock/*
+    litellm_params:
+      model: bedrock/*
+
+
+files_settings:
+  # Configure storage providers and retention  
+    - custom_llm_provider: azure
+      api_base: https://exampleopenaiendpoint-production.up.railway.app
+      api_key: fake-key
+      api_version: "2023-03-15-preview"
+    - custom_llm_provider: openai
+      api_key: os.environ/OPENAI_API_KEY
+    - custom_llm_provider: bedrock
+      api_key: os.environ/BEDROCK_API_KEY
+      api_base: https://bedrock.us-east-1.amazonaws.com
+      retention_period: 7
+    - custom_llm_provider: vertex_ai
+      bucket_name: my-vertex_ai-bucket
+      retention_period: 7
+```
+
+This configuration:
+- Sets up storage providers (GCS and S3) with their retention policies
+- Configures provider-specific file endpoints for Azure and OpenAI
+
+## 1. Local File and process on Vertex + Bedrock 
+
+When uploading a local file, the process follows these steps:
+
+1. Client makes a POST request to `/files` endpoint with the local file
+2. Client MUST specify required storage custom_llm_provider(s) based on intended model usage
+3. LiteLLM uploads the file to specified custom_llm_provider(s)
+4. A file ID is returned to the client
+5. Files have a 7-day retention policy (configured by admin)
+
+```bash
+# custom_llm_provider is required
+curl https://api.litellm.ai/v1/files \
+  -H "Authorization: Bearer sk-1234" \
+  -F purpose="fine-tune" \
+  -F file="@mydata.jsonl" \
+  -F custom_llm_provider='["vertex_ai"]'  # Required: specify ["vertex_ai"], ["bedrock"], or ["vertex_ai", "bedrock"]
+```
+
+
+After uploading, you can use the file_id in a /chat/completions request:
+
+```bash
+curl "https://api.litellm.ai/v1/chat/completions" \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer sk-1234" \
+  -d '{
+    "model": "bedrock/amazon.nova-pro-v1:0",
+    "input": [
+      {
+        "role": "user",
+        "content": [
+          {
+            "type": "input_file",
+            "file_id": "file-6F2ksmvXxt4VdoqmHRw6kL"
+          },
+          {
+            "type": "input_text",
+            "text": "What is happening in this video?"
+          }
+        ]
+      }
+    ]
+  }'
+```
+
+## 2. Existing file in S3 and process on Vertex + Bedrock 
+
+For files already in S3, you can avoid downloading and re-uploading by using a presigned URL:
+
+```bash
+# custom_llm_provider is required
+curl -X POST "https://api.litellm.ai/files" \
+  -H "Authorization: Bearer sk-1234" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "source": "https://your-bucket.s3.amazonaws.com/path/to/mydata.jsonl?X-Amz-Algorithm=AWS4-HMAC-SHA256&X-Amz-Credential=...",
+    "custom_llm_provider": ["vertex_ai"]  # Required: specify target custom_llm_provider
+  }'
+```
+
+This performs a direct copy from the S3 presigned URL to the specified custom_llm_provider(s), which is more efficient than downloading and re-uploading.
+
+After the S3 file is copied, you can use the returned file_id in your /chat/completions request:
+
+```bash
+curl "https://api.litellm.ai/v1/chat/completions" \
+  -H "Content-Type: application/json" \
+  -H "Authorization: Bearer sk-1234" \
+  -d '{
+    "model": "vertex_ai/gemini-pro-vision",
+    "input": [
+      {
+        "role": "user",
+        "content": [
+          {
+            "type": "input_file",
+            "file_id": "file-8H3jsmwYyt6WepqnKRw9mN"
+          },
+          {
+            "type": "input_text",
+            "text": "Describe the main events in this video"
+          }
+        ]
+      }
+    ]
+  }'
+```
+
+
+## Best Practices
+
+**Storage Location Selection**: 
+   - You MUST specify custom_llm_provider based on your model requirements
+   - For Vertex AI models, specify `["vertex_ai"]`
+   - For Bedrock models, specify `["bedrock"]`
+   - For fallback scenarios, specify both `["vertex_ai", "bedrock"]`
+   - Avoid unnecessary copies by selecting only required destinations
+## Model Compatibility
+
+This file handling approach is compatible with various model backends:
+- Vertex AI
+- AWS Bedrock
+- vLLM (for video processing, using their multimodal input format)
+

docs/my-website/sidebars.js

@@ -260,6 +260,7 @@ const sidebars = {
         "completion/web_search",
         "completion/document_understanding",
         "completion/vision",
+        "completion/video",
         "completion/json_mode",
         "reasoning_content",
         "completion/prompt_caching",