# What does this PR do?
The goal of this PR is code base modernization.
Schema reflection code needed a minor adjustment to handle UnionTypes
and collections.abc.AsyncIterator. (Both are preferred for latest Python
releases.)
Note to reviewers: almost all changes here are automatically generated
by pyupgrade. Some additional unused imports were cleaned up. The only
change worth of note can be found under `docs/openapi_generator` and
`llama_stack/strong_typing/schema.py` where reflection code was updated
to deal with "newer" types.
Signed-off-by: Ihar Hrachyshka <ihar.hrachyshka@gmail.com>
# What does this PR do?
TLDR: Changes needed to get 100% passing tests for OpenAI API
verification tests when run against Llama Stack with the `together`,
`fireworks`, and `openai` providers. And `groq` is better than before,
at 88% passing.
This cleans up the OpenAI API support for image message types
(specifically `image_url` types) and handling of the `response_format`
chat completion parameter. Both of these required a few more Pydantic
model definitions in our Inference API, just to move from the
not-quite-right stubs I had in place to something fleshed out to match
the actual OpenAI API specs.
As part of testing this, I also found and fixed a bug in the litellm
implementation of openai_completion and openai_chat_completion, so the
providers based on those should actually be working now.
The method `prepare_openai_completion_params` in
`llama_stack/providers/utils/inference/openai_compat.py` was improved to
actually recursively clean up input parameters, including handling of
lists, dicts, and dumping of Pydantic models to dicts. These changes
were required to get to 100% passing tests on the OpenAI API
verification against the `openai` provider.
With the above, the together.ai provider was passing as well as it is
without Llama Stack. But, since we have Llama Stack in the middle, I
took the opportunity to clean up the together.ai provider so that it now
also passes the OpenAI API spec tests we have at 100%. That means
together.ai is now passing our verification test better when using an
OpenAI client talking to Llama Stack than it is when hitting together.ai
directly, without Llama Stack in the middle.
And, another round of work for Fireworks to improve translation of
incoming OpenAI chat completion requests to Llama Stack chat completion
requests gets the fireworks provider passing at 100%. The server-side
fireworks.ai tool calling support with OpenAI chat completions and Llama
4 models isn't great yet, but by pointing the OpenAI clients at Llama
Stack's API we can clean things up and get everything working as
expected for Llama 4 models.
## Test Plan
### OpenAI API Verification Tests
I ran the OpenAI API verification tests as below and 100% of the tests
passed.
First, start a Llama Stack server that runs the `openai` provider with
the `gpt-4o` and `gpt-4o-mini` models deployed. There's not a template
setup to do this out of the box, so I added a
`tests/verifications/openai-api-verification-run.yaml` to do this.
First, ensure you have the necessary API key environment variables set:
```
export TOGETHER_API_KEY="..."
export FIREWORKS_API_KEY="..."
export OPENAI_API_KEY="..."
```
Then, run a Llama Stack server that serves up all these providers:
```
llama stack run \
--image-type venv \
tests/verifications/openai-api-verification-run.yaml
```
Finally, generate a new verification report against all these providers,
both with and without the Llama Stack server in the middle.
```
python tests/verifications/generate_report.py \
--run-tests \
--provider \
together \
fireworks \
groq \
openai \
together-llama-stack \
fireworks-llama-stack \
groq-llama-stack \
openai-llama-stack
```
You'll see that most of the configurations with Llama Stack in the
middle now pass at 100%, even though some of them do not pass at 100%
when hitting the backend provider's API directly with an OpenAI client.
### OpenAI Completion Integration Tests with vLLM:
I also ran the smaller `test_openai_completion.py` test suite (that's
not yet merged with the verification tests) on multiple of the
providers, since I had to adjust the method signature of
openai_chat_completion a bit and thus had to touch lots of these
providers to match. Here's the tests I ran there, all passing:
```
VLLM_URL="http://localhost:8000/v1" INFERENCE_MODEL="meta-llama/Llama-3.2-3B-Instruct" llama stack build --template remote-vllm --image-type venv --run
```
in another terminal
```
LLAMA_STACK_CONFIG=http://localhost:8321 INFERENCE_MODEL="meta-llama/Llama-3.2-3B-Instruct" python -m pytest -v tests/integration/inference/test_openai_completion.py --text-model "meta-llama/Llama-3.2-3B-Instruct"
```
### OpenAI Completion Integration Tests with ollama
```
INFERENCE_MODEL="llama3.2:3b-instruct-q8_0" llama stack build --template ollama --image-type venv --run
```
in another terminal
```
LLAMA_STACK_CONFIG=http://localhost:8321 INFERENCE_MODEL="llama3.2:3b-instruct-q8_0" python -m pytest -v tests/integration/inference/test_openai_completion.py --text-model "llama3.2:3b-instruct-q8_0"
```
### OpenAI Completion Integration Tests with together.ai
```
INFERENCE_MODEL="meta-llama/Llama-3.2-3B-Instruct-Turbo" llama stack build --template together --image-type venv --run
```
in another terminal
```
LLAMA_STACK_CONFIG=http://localhost:8321 INFERENCE_MODEL="meta-llama/Llama-3.2-3B-Instruct-Turbo" python -m pytest -v tests/integration/inference/test_openai_completion.py --text-model "meta-llama/Llama-3.2-3B-Instruct-Turbo"
```
### OpenAI Completion Integration Tests with fireworks.ai
```
INFERENCE_MODEL="meta-llama/Llama-3.1-8B-Instruct" llama stack build --template fireworks --image-type venv --run
```
in another terminal
```
LLAMA_STACK_CONFIG=http://localhost:8321 INFERENCE_MODEL="meta-llama/Llama-3.1-8B-Instruct" python -m pytest -v tests/integration/inference/test_openai_completion.py --text-model "meta-llama/Llama-3.1-8B-Instruct"
---------
Signed-off-by: Ben Browning <bbrownin@redhat.com>
# What does this PR do?
This PR adds two methods to the Inference API:
- `batch_completion`
- `batch_chat_completion`
The motivation is for evaluations targeting a local inference engine
(like meta-reference or vllm) where batch APIs provide for a substantial
amount of acceleration.
Why did I not add this to `Api.batch_inference` though? That just
resulted in a _lot_ more book-keeping given the structure of Llama
Stack. Had I done that, I would have needed to create a notion of a
"batch model" resource, setup routing based on that, etc. This does not
sound ideal.
So what's the future of the batch inference API? I am not sure. Maybe we
can keep it for true _asynchronous_ execution. So you can submit
requests, and it can return a Job instance, etc.
## Test Plan
Run meta-reference-gpu using:
```bash
export INFERENCE_MODEL=meta-llama/Llama-4-Scout-17B-16E-Instruct
export INFERENCE_CHECKPOINT_DIR=../checkpoints/Llama-4-Scout-17B-16E-Instruct-20250331210000
export MODEL_PARALLEL_SIZE=4
export MAX_BATCH_SIZE=32
export MAX_SEQ_LEN=6144
LLAMA_MODELS_DEBUG=1 llama stack run meta-reference-gpu
```
Then run the batch inference test case.
# What does this PR do?
Move around bits. This makes the copies from llama-models _much_ easier
to maintain and ensures we don't entangle meta-reference specific
tidbits into llama-models code even by accident.
Also, kills the meta-reference-quantized-gpu distro and rolls
quantization deps into meta-reference-gpu.
## Test Plan
```
LLAMA_MODELS_DEBUG=1 \
with-proxy llama stack run meta-reference-gpu \
--env INFERENCE_MODEL=meta-llama/Llama-4-Scout-17B-16E-Instruct \
--env INFERENCE_CHECKPOINT_DIR=<DIR> \
--env MODEL_PARALLEL_SIZE=4 \
--env QUANTIZATION_TYPE=fp8_mixed
```
Start a server with and without quantization. Point integration tests to
it using:
```
pytest -s -v tests/integration/inference/test_text_inference.py \
--stack-config http://localhost:8321 --text-model meta-llama/Llama-4-Scout-17B-16E-Instruct
```
Running full Tool Calling required some updates to work e2e.
- Remove `python_start` and `python_end` tags
- Tool Call messages and Tool Resposne messages should end with
`<|eom|>`
- System prompt needed updates
```
You are a helpful assisant who can can answer general questions or invoke tools when necessary.
In addition to tool calls, you should also augment your responses by using the tool outputs.
```
### Test Plan
- Start server with meta-reference
```
LLAMA_STACK_DISABLE_VERSION_CHECK=1 LLAMA_MODELS_DEBUG=1 INFERENCE_MODEL=meta-llama/$MODEL llama stack run meta-reference-gpu
```
- Added **NEW** tests with 5 test cases for multi-turn tool calls
```
pytest -s -v --stack-config http://localhost:8321 tests/integration/inference/test_text_inference.py --text-model meta-llama/Llama-4-Scout-17B-16E-Instruct
```
- Also verified all vision and agent tests pass
# What does this PR do?
closes#1584
This should be a rather innocuous change.
## Test Plan
Verify that there's no more tool call parsing error for example in issue
<img width="1216" alt="image"
src="https://github.com/user-attachments/assets/a5a6f4e8-2093-4ca2-bc06-794b707a0429"
/>
LLAMA_STACK_CONFIG=fireworks pytest -s -v
tests/integration/agents/test_agents.py --safety-shield
meta-llama/Llama-Guard-3-8B --text-model
meta-llama/Llama-3.1-8B-Instruct
### What does this PR do?
Currently, `ToolCall.arguments` is a `Dict[str, RecursiveType]`.
However, on the client SDK side -- the `RecursiveType` gets deserialized
into a number ( both int and float get collapsed ) and hence when params
are `int` they get converted to float which might break client side
tools that might be doing type checking.
Closes: https://github.com/meta-llama/llama-stack/issues/1683
### Test Plan
Stainless changes --
https://github.com/meta-llama/llama-stack-client-python/pull/204
```
pytest -s -v --stack-config=fireworks tests/integration/agents/test_agents.py --text-model meta-llama/Llama-3.1-8B-Instruct
```
# What does this PR do?
Updated all instances of datetime.now() to use timezone.utc for
consistency in handling time across different systems. This ensures that
timestamps are always in Coordinated Universal Time (UTC), avoiding
issues with time zone discrepancies and promoting uniformity in
time-related data.
Signed-off-by: Sébastien Han <seb@redhat.com>
# Summary:
The current prompt doesn't work well and tend to overindex on tool
calling. This PR is not perfect, but should be an improvement over the
current prompt. We can keep iterating.
# Test Plan:
Ran on a (small) eval with 20 HotpotQA examples.
With current prompt:
https://gist.github.com/ehhuang/9f967e62751907165eb13781ea968f5c
{
│ 'basic::equality': {'accuracy': {'accuracy': 0.2, 'num_correct': 4.0,
'num_total': 20}},
│ 'F1ScoringFn': {
│ │ 'f1_average': 0.25333333333333335,
│ │ 'precision_average': 0.23301767676767676,
│ │ 'recall_average': 0.375
│ }
}
num_tool_calls=[5, 5, 5, 5, 5, 5, 2, 5, 5, 5, 5, 5, 2, 2, 1, 1, 2, 1, 2,
2]
num_examples_with_tool_call=20
num_examples_with_pythontag=0
#########################################################
With new prompt:
https://gist.github.com/ehhuang/6e4a8ecf54db68922c2be8700056f962
{
│ 'basic::equality': {'accuracy': {'accuracy': 0.25, 'num_correct': 5.0,
'num_total': 20}},
│ 'F1ScoringFn': {
│ │ 'f1_average': 0.35579260478321006,
│ │ 'precision_average': 0.32030238933180105,
│ │ 'recall_average': 0.6091666666666666
│ }
}
num_tool_calls=[2, 1, 1, 5, 5, 5, 1, 2, 1, 1, 2, 1, 1, 1, 1, 1, 2, 1, 3,
2]
num_examples_with_tool_call=20
num_examples_with_pythontag=0
The answers have higher recall, and make fewer tool calls. Note that
these were run with max_infer_iter=5, so the current prompt hits this
limit more often, and without the limit, someitmes goes into infinite
tool calling loop.
The data here is with 3.3-70B. Results are equally poor with either
prompt with 3.2-3B ~30 recall.
llama-models should have extremely minimal cruft. Its sole purpose
should be didactic -- show the simplest implementation of the llama
models and document the prompt formats, etc.
This PR is the complement to
https://github.com/meta-llama/llama-models/pull/279
## Test Plan
Ensure all `llama` CLI `model` sub-commands work:
```bash
llama model list
llama model download --model-id ...
llama model prompt-format -m ...
```
Ran tests:
```bash
cd tests/client-sdk
LLAMA_STACK_CONFIG=fireworks pytest -s -v inference/
LLAMA_STACK_CONFIG=fireworks pytest -s -v vector_io/
LLAMA_STACK_CONFIG=fireworks pytest -s -v agents/
```
Create a fresh venv `uv venv && source .venv/bin/activate` and run
`llama stack build --template fireworks --image-type venv` followed by
`llama stack run together --image-type venv` <-- the server runs
Also checked that the OpenAPI generator can run and there is no change
in the generated files as a result.
```bash
cd docs/openapi_generator
sh run_openapi_generator.sh
```
