forked from phoenix-oss/llama-stack-mirror
8de8eb03c8
## What does this PR do?
This is a long-pending change and particularly important to get done
now.
Specifically:
- we cannot "localize" (aka download) any URLs from media attachments
anywhere near our modeling code. it must be done within llama-stack.
- `PIL.Image` is infesting all our APIs via `ImageMedia ->
InterleavedTextMedia` and that cannot be right at all. Anything in the
API surface must be "naturally serializable". We need a standard `{
type: "image", image_url: "<...>" }` which is more extensible
- `UserMessage`, `SystemMessage`, etc. are moved completely to
llama-stack from the llama-models repository.
See https://github.com/meta-llama/llama-models/pull/244 for the
corresponding PR in llama-models.
## Test Plan
```bash
cd llama_stack/providers/tests
pytest -s -v -k "fireworks or ollama or together" inference/test_vision_inference.py
pytest -s -v -k "(fireworks or ollama or together) and llama_3b" inference/test_text_inference.py
pytest -s -v -k chroma memory/test_memory.py \
--env EMBEDDING_DIMENSION=384 --env CHROMA_DB_PATH=/tmp/foobar
pytest -s -v -k fireworks agents/test_agents.py \
--safety-shield=meta-llama/Llama-Guard-3-8B \
--inference-model=meta-llama/Llama-3.1-8B-Instruct
```
Updated the client sdk (see PR ...), installed the SDK in the same
environment and then ran the SDK tests:
```bash
cd tests/client-sdk
LLAMA_STACK_CONFIG=together pytest -s -v agents/test_agents.py
LLAMA_STACK_CONFIG=ollama pytest -s -v memory/test_memory.py
# this one needed a bit of hacking in the run.yaml to ensure I could register the vision model correctly
INFERENCE_MODEL=llama3.2-vision:latest LLAMA_STACK_CONFIG=ollama pytest -s -v inference/test_inference.py
```
111 lines
4.3 KiB
Python
111 lines
4.3 KiB
Python
# Copyright (c) Meta Platforms, Inc. and affiliates.
|
|
# All rights reserved.
|
|
#
|
|
# This source code is licensed under the terms described in the LICENSE file in
|
|
# the root directory of this source tree.
|
|
|
|
from typing import List
|
|
|
|
from llama_stack.distribution.datatypes import * # noqa: F403
|
|
|
|
|
|
EMBEDDING_DEPS = [
|
|
"blobfile",
|
|
"chardet",
|
|
"pypdf",
|
|
"tqdm",
|
|
"numpy",
|
|
"scikit-learn",
|
|
"scipy",
|
|
"nltk",
|
|
"sentencepiece",
|
|
"transformers",
|
|
# this happens to work because special dependencies are always installed last
|
|
# so if there was a regular torch installed first, this would be ignored
|
|
# we need a better way to do this to identify potential conflicts, etc.
|
|
# for now, this lets us significantly reduce the size of the container which
|
|
# does not have any "local" inference code (and hence does not need GPU-enabled torch)
|
|
"torch --index-url https://download.pytorch.org/whl/cpu",
|
|
"sentence-transformers --no-deps",
|
|
]
|
|
|
|
|
|
def available_providers() -> List[ProviderSpec]:
|
|
return [
|
|
InlineProviderSpec(
|
|
api=Api.memory,
|
|
provider_type="inline::meta-reference",
|
|
pip_packages=EMBEDDING_DEPS + ["faiss-cpu"],
|
|
module="llama_stack.providers.inline.memory.faiss",
|
|
config_class="llama_stack.providers.inline.memory.faiss.FaissImplConfig",
|
|
deprecation_warning="Please use the `inline::faiss` provider instead.",
|
|
api_dependencies=[Api.inference],
|
|
),
|
|
InlineProviderSpec(
|
|
api=Api.memory,
|
|
provider_type="inline::faiss",
|
|
pip_packages=EMBEDDING_DEPS + ["faiss-cpu"],
|
|
module="llama_stack.providers.inline.memory.faiss",
|
|
config_class="llama_stack.providers.inline.memory.faiss.FaissImplConfig",
|
|
api_dependencies=[Api.inference],
|
|
),
|
|
remote_provider_spec(
|
|
Api.memory,
|
|
AdapterSpec(
|
|
adapter_type="chromadb",
|
|
pip_packages=EMBEDDING_DEPS + ["chromadb-client"],
|
|
module="llama_stack.providers.remote.memory.chroma",
|
|
config_class="llama_stack.providers.remote.memory.chroma.ChromaRemoteImplConfig",
|
|
),
|
|
api_dependencies=[Api.inference],
|
|
),
|
|
InlineProviderSpec(
|
|
api=Api.memory,
|
|
provider_type="inline::chromadb",
|
|
pip_packages=EMBEDDING_DEPS + ["chromadb"],
|
|
module="llama_stack.providers.inline.memory.chroma",
|
|
config_class="llama_stack.providers.inline.memory.chroma.ChromaInlineImplConfig",
|
|
api_dependencies=[Api.inference],
|
|
),
|
|
remote_provider_spec(
|
|
Api.memory,
|
|
AdapterSpec(
|
|
adapter_type="pgvector",
|
|
pip_packages=EMBEDDING_DEPS + ["psycopg2-binary"],
|
|
module="llama_stack.providers.remote.memory.pgvector",
|
|
config_class="llama_stack.providers.remote.memory.pgvector.PGVectorConfig",
|
|
),
|
|
api_dependencies=[Api.inference],
|
|
),
|
|
remote_provider_spec(
|
|
Api.memory,
|
|
AdapterSpec(
|
|
adapter_type="weaviate",
|
|
pip_packages=EMBEDDING_DEPS + ["weaviate-client"],
|
|
module="llama_stack.providers.remote.memory.weaviate",
|
|
config_class="llama_stack.providers.remote.memory.weaviate.WeaviateConfig",
|
|
provider_data_validator="llama_stack.providers.remote.memory.weaviate.WeaviateRequestProviderData",
|
|
),
|
|
api_dependencies=[Api.inference],
|
|
),
|
|
remote_provider_spec(
|
|
api=Api.memory,
|
|
adapter=AdapterSpec(
|
|
adapter_type="sample",
|
|
pip_packages=[],
|
|
module="llama_stack.providers.remote.memory.sample",
|
|
config_class="llama_stack.providers.remote.memory.sample.SampleConfig",
|
|
),
|
|
api_dependencies=[],
|
|
),
|
|
remote_provider_spec(
|
|
Api.memory,
|
|
AdapterSpec(
|
|
adapter_type="qdrant",
|
|
pip_packages=EMBEDDING_DEPS + ["qdrant-client"],
|
|
module="llama_stack.providers.remote.memory.qdrant",
|
|
config_class="llama_stack.providers.remote.memory.qdrant.QdrantConfig",
|
|
),
|
|
api_dependencies=[Api.inference],
|
|
),
|
|
]
|