phoenix-oss/llama-stack-mirror
0
0
Fork 1
mirror of https://github.com/meta-llama/llama-stack.git synced 2025-12-22 22:39:41 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Merge branch 'main' into make-provider-model-id-non-optional

Browse source
This commit is contained in:
Matthew Farrellee 2025-07-16 11:59:36 -04:00
commit bd9bad5f84
176 changed files with 4121 additions and 2735 deletions
Download patch file Download diff file Expand all files Collapse all files

2
.github/CODEOWNERS vendored
View file

@ -2,4 +2,4 @@
# These owners will be the default owners for everything in
# the repo. Unless a later match takes precedence,
* @ashwinb @yanxi0830 @hardikjshah @raghotham @ehhuang @terrytangyuan @leseb @bbrowning @reluctantfuturist
* @ashwinb @yanxi0830 @hardikjshah @raghotham @ehhuang @terrytangyuan @leseb @bbrowning @reluctantfuturist @mattf

30
.github/ISSUE_TEMPLATE/tech-debt.yml vendored Normal file
View file

@ -0,0 +1,30 @@
name: 🔧 Tech Debt
description: Something that is functional but should be improved or optimizied
labels: ["tech-debt"]
body:
- type: textarea
id: tech-debt-explanation
attributes:
label: 🤔 What is the technical debt you think should be addressed?
description: >
A clear and concise description of _what_ needs to be addressed - ensure you are describing
constitutes [technical debt](https://en.wikipedia.org/wiki/Technical_debt) and is not a bug
or feature request.
validations:
required: true
- type: textarea
id: tech-debt-motivation
attributes:
label: 💡 What is the benefit of addressing this technical debt?
description: >
A clear and concise description of _why_ this work is needed.
validations:
required: true
- type: textarea
id: other-thoughts
attributes:
label: Other thoughts
description: >
Any thoughts about how this may result in complexity in the codebase, or other trade-offs.

2
.github/actions/setup-ollama/action.yml vendored
View file

@ -7,3 +7,5 @@ runs:
shell: bash
run: |
docker run -d --name ollama -p 11434:11434 docker.io/leseb/ollama-with-models
echo "Verifying Ollama status..."
timeout 30 bash -c 'while ! curl -s -L http://127.0.0.1:11434; do sleep 1 && echo "."; done'

19
.github/actions/setup-runner/action.yml vendored
View file

@ -5,6 +5,10 @@ inputs:
description: The Python version to use
required: false
default: "3.12"
client-version:
description: The llama-stack-client-python version to test against (latest or published)
required: false
default: "latest"
runs:
using: "composite"
steps:
@ -20,8 +24,17 @@ runs:
run: |
uv sync --all-groups
uv pip install ollama faiss-cpu
# always test against the latest version of the client
# TODO: this is not necessarily a good idea. we need to test against both published and latest
# to find out backwards compatibility issues.
# Install llama-stack-client-python based on the client-version input
if [ "${{ inputs.client-version }}" = "latest" ]; then
echo "Installing latest llama-stack-client-python from main branch"
uv pip install git+https://github.com/meta-llama/llama-stack-client-python.git@main
elif [ "${{ inputs.client-version }}" = "published" ]; then
echo "Installing published llama-stack-client-python from PyPI"
uv pip install llama-stack-client
else
echo "Invalid client-version: ${{ inputs.client-version }}"
exit 1
fi
uv pip install -e .

355
.github/workflows/gha_workflow_llama_stack_tests.yml vendored
View file

@ -1,355 +0,0 @@
name: "Run Llama-stack Tests"
on:
#### Temporarily disable PR runs until tests run as intended within mainline.
#TODO Add this back.
#pull_request_target:
# types: ["opened"]
# branches:
# - 'main'
# paths:
# - 'llama_stack/**/*.py'
# - 'tests/**/*.py'
workflow_dispatch:
inputs:
runner:
description: 'GHA Runner Scale Set label to run workflow on.'
required: true
default: "llama-stack-gha-runner-gpu"
checkout_reference:
description: "The branch, tag, or SHA to checkout"
required: true
default: "main"
debug:
description: 'Run debugging steps?'
required: false
default: "true"
sleep_time:
description: '[DEBUG] sleep time for debugging'
required: true
default: "0"
provider_id:
description: 'ID of your provider'
required: true
default: "meta_reference"
model_id:
description: 'Shorthand name for target model ID (llama_3b or llama_8b)'
required: true
default: "llama_3b"
model_override_3b:
description: 'Specify shorthand model for <llama_3b> '
required: false
default: "Llama3.2-3B-Instruct"
model_override_8b:
description: 'Specify shorthand model for <llama_8b> '
required: false
default: "Llama3.1-8B-Instruct"
env:
# ID used for each test's provider config
PROVIDER_ID: "${{ inputs.provider_id || 'meta_reference' }}"
# Path to model checkpoints within EFS volume
MODEL_CHECKPOINT_DIR: "/data/llama"
# Path to directory to run tests from
TESTS_PATH: "${{ github.workspace }}/llama_stack/providers/tests"
# Keep track of a list of model IDs that are valid to use within pytest fixture marks
AVAILABLE_MODEL_IDs: "llama_3b llama_8b"
# Shorthand name for model ID, used in pytest fixture marks
MODEL_ID: "${{ inputs.model_id || 'llama_3b' }}"
# Override the `llama_3b` / `llama_8b' models, else use the default.
LLAMA_3B_OVERRIDE: "${{ inputs.model_override_3b || 'Llama3.2-3B-Instruct' }}"
LLAMA_8B_OVERRIDE: "${{ inputs.model_override_8b || 'Llama3.1-8B-Instruct' }}"
# Defines which directories in TESTS_PATH to exclude from the test loop
EXCLUDED_DIRS: "__pycache__"
# Defines the output xml reports generated after a test is run
REPORTS_GEN: ""
jobs:
execute_workflow:
name: Execute workload on Self-Hosted GPU k8s runner
permissions:
pull-requests: write
defaults:
run:
shell: bash
runs-on: ${{ inputs.runner != '' && inputs.runner || 'llama-stack-gha-runner-gpu' }}
if: always()
steps:
##############################
#### INITIAL DEBUG CHECKS ####
##############################
- name: "[DEBUG] Check content of the EFS mount"
id: debug_efs_volume
continue-on-error: true
if: inputs.debug == 'true'
run: |
echo "========= Content of the EFS mount ============="
ls -la ${{ env.MODEL_CHECKPOINT_DIR }}
- name: "[DEBUG] Get runner container OS information"
id: debug_os_info
if: ${{ inputs.debug == 'true' }}
run: |
cat /etc/os-release
- name: "[DEBUG] Print environment variables"
id: debug_env_vars
if: ${{ inputs.debug == 'true' }}
run: |
echo "PROVIDER_ID = ${PROVIDER_ID}"
echo "MODEL_CHECKPOINT_DIR = ${MODEL_CHECKPOINT_DIR}"
echo "AVAILABLE_MODEL_IDs = ${AVAILABLE_MODEL_IDs}"
echo "MODEL_ID = ${MODEL_ID}"
echo "LLAMA_3B_OVERRIDE = ${LLAMA_3B_OVERRIDE}"
echo "LLAMA_8B_OVERRIDE = ${LLAMA_8B_OVERRIDE}"
echo "EXCLUDED_DIRS = ${EXCLUDED_DIRS}"
echo "REPORTS_GEN = ${REPORTS_GEN}"
############################
#### MODEL INPUT CHECKS ####
############################
- name: "Check if env.model_id is valid"
id: check_model_id
run: |
if [[ " ${AVAILABLE_MODEL_IDs[@]} " =~ " ${MODEL_ID} " ]]; then
echo "Model ID '${MODEL_ID}' is valid."
else
echo "Model ID '${MODEL_ID}' is invalid. Terminating workflow."
exit 1
fi
#######################
#### CODE CHECKOUT ####
#######################
- name: "Checkout 'meta-llama/llama-stack' repository"
id: checkout_repo
uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
with:
ref: ${{ inputs.branch }}
- name: "[DEBUG] Content of the repository after checkout"
id: debug_content_after_checkout
if: ${{ inputs.debug == 'true' }}
run: |
ls -la ${GITHUB_WORKSPACE}
##########################################################
#### OPTIONAL SLEEP DEBUG ####
# #
# Use to "exec" into the test k8s POD and run tests #
# manually to identify what dependencies are being used. #
# #
##########################################################
- name: "[DEBUG] sleep"
id: debug_sleep
if: ${{ inputs.debug == 'true' && inputs.sleep_time != '' }}
run: |
sleep ${{ inputs.sleep_time }}
############################
#### UPDATE SYSTEM PATH ####
############################
- name: "Update path: execute"
id: path_update_exec
run: |
# .local/bin is needed for certain libraries installed below to be recognized
# when calling their executable to install sub-dependencies
mkdir -p ${HOME}/.local/bin
echo "${HOME}/.local/bin" >> "$GITHUB_PATH"
#####################################
#### UPDATE CHECKPOINT DIRECTORY ####
#####################################
- name: "Update checkpoint directory"
id: checkpoint_update
run: |
echo "Checkpoint directory: ${MODEL_CHECKPOINT_DIR}/$LLAMA_3B_OVERRIDE"
if [ "${MODEL_ID}" = "llama_3b" ] && [ -d "${MODEL_CHECKPOINT_DIR}/${LLAMA_3B_OVERRIDE}" ]; then
echo "MODEL_CHECKPOINT_DIR=${MODEL_CHECKPOINT_DIR}/${LLAMA_3B_OVERRIDE}" >> "$GITHUB_ENV"
elif [ "${MODEL_ID}" = "llama_8b" ] && [ -d "${MODEL_CHECKPOINT_DIR}/${LLAMA_8B_OVERRIDE}" ]; then
echo "MODEL_CHECKPOINT_DIR=${MODEL_CHECKPOINT_DIR}/${LLAMA_8B_OVERRIDE}" >> "$GITHUB_ENV"
else
echo "MODEL_ID & LLAMA_*B_OVERRIDE are not a valid pairing. Terminating workflow."
exit 1
fi
- name: "[DEBUG] Checkpoint update check"
id: debug_checkpoint_update
if: ${{ inputs.debug == 'true' }}
run: |
echo "MODEL_CHECKPOINT_DIR (after update) = ${MODEL_CHECKPOINT_DIR}"
##################################
#### DEPENDENCY INSTALLATIONS ####
##################################
- name: "Installing 'apt' required packages"
id: install_apt
run: |
echo "[STEP] Installing 'apt' required packages"
sudo apt update -y
sudo apt install -y python3 python3-pip npm wget
- name: "Installing packages with 'curl'"
id: install_curl
run: |
curl -fsSL https://ollama.com/install.sh | sh
- name: "Installing packages with 'wget'"
id: install_wget
run: |
wget https://repo.continuum.io/miniconda/Miniconda3-latest-Linux-x86_64.sh
chmod +x Miniconda3-latest-Linux-x86_64.sh
./Miniconda3-latest-Linux-x86_64.sh -b install -c pytorch -c nvidia faiss-gpu=1.9.0
# Add miniconda3 bin to system path
echo "${HOME}/miniconda3/bin" >> "$GITHUB_PATH"
- name: "Installing packages with 'npm'"
id: install_npm_generic
run: |
sudo npm install -g junit-merge
- name: "Installing pip dependencies"
id: install_pip_generic
run: |
echo "[STEP] Installing 'llama-stack' models"
pip install -U pip setuptools
pip install -r requirements.txt
pip install -e .
pip install -U \
torch torchvision \
pytest pytest_asyncio \
fairscale lm-format-enforcer \
zmq chardet pypdf \
pandas sentence_transformers together \
aiosqlite
- name: "Installing packages with conda"
id: install_conda_generic
run: |
conda install -q -c pytorch -c nvidia faiss-gpu=1.9.0
#############################################################
#### TESTING TO BE DONE FOR BOTH PRS AND MANUAL DISPATCH ####
#############################################################
- name: "Run Tests: Loop"
id: run_tests_loop
working-directory: "${{ github.workspace }}"
run: |
pattern=""
for dir in llama_stack/providers/tests/*; do
if [ -d "$dir" ]; then
dir_name=$(basename "$dir")
if [[ ! " $EXCLUDED_DIRS " =~ " $dir_name " ]]; then
for file in "$dir"/test_*.py; do
test_name=$(basename "$file")
new_file="result-${dir_name}-${test_name}.xml"
if torchrun $(which pytest) -s -v ${TESTS_PATH}/${dir_name}/${test_name} -m "${PROVIDER_ID} and ${MODEL_ID}" \
--junitxml="${{ github.workspace }}/${new_file}"; then
echo "Ran test: ${test_name}"
else
echo "Did NOT run test: ${test_name}"
fi
pattern+="${new_file} "
done
fi
fi
done
echo "REPORTS_GEN=$pattern" >> "$GITHUB_ENV"
- name: "Test Summary: Merge"
id: test_summary_merge
working-directory: "${{ github.workspace }}"
run: |
echo "Merging the following test result files: ${REPORTS_GEN}"
# Defaults to merging them into 'merged-test-results.xml'
junit-merge ${{ env.REPORTS_GEN }}
############################################
#### AUTOMATIC TESTING ON PULL REQUESTS ####
############################################
#### Run tests ####
- name: "PR - Run Tests"
id: pr_run_tests
working-directory: "${{ github.workspace }}"
if: github.event_name == 'pull_request_target'
run: |
echo "[STEP] Running PyTest tests at 'GITHUB_WORKSPACE' path: ${GITHUB_WORKSPACE} | path: ${{ github.workspace }}"
# (Optional) Add more tests here.
# Merge test results with 'merged-test-results.xml' from above.
# junit-merge <new-test-results> merged-test-results.xml
#### Create test summary ####
- name: "PR - Test Summary"
id: pr_test_summary_create
if: github.event_name == 'pull_request_target'
uses: test-summary/action@31493c76ec9e7aa675f1585d3ed6f1da69269a86 # v2.4
with:
paths: "${{ github.workspace }}/merged-test-results.xml"
output: test-summary.md
- name: "PR - Upload Test Summary"
id: pr_test_summary_upload
if: github.event_name == 'pull_request_target'
uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
with:
name: test-summary
path: test-summary.md
#### Update PR request ####
- name: "PR - Update comment"
id: pr_update_comment
if: github.event_name == 'pull_request_target'
uses: thollander/actions-comment-pull-request@24bffb9b452ba05a4f3f77933840a6a841d1b32b # v3.0.1
with:
filePath: test-summary.md
########################
#### MANUAL TESTING ####
########################
#### Run tests ####
- name: "Manual - Run Tests: Prep"
id: manual_run_tests
working-directory: "${{ github.workspace }}"
if: github.event_name == 'workflow_dispatch'
run: |
echo "[STEP] Running PyTest tests at 'GITHUB_WORKSPACE' path: ${{ github.workspace }}"
#TODO Use this when collection errors are resolved
# pytest -s -v -m "${PROVIDER_ID} and ${MODEL_ID}" --junitxml="${{ github.workspace }}/merged-test-results.xml"
# (Optional) Add more tests here.
# Merge test results with 'merged-test-results.xml' from above.
# junit-merge <new-test-results> merged-test-results.xml
#### Create test summary ####
- name: "Manual - Test Summary"
id: manual_test_summary
if: always() && github.event_name == 'workflow_dispatch'
uses: test-summary/action@31493c76ec9e7aa675f1585d3ed6f1da69269a86 # v2.4
with:
paths: "${{ github.workspace }}/merged-test-results.xml"

8
.github/workflows/install-script-ci.yml vendored
View file

@ -3,10 +3,10 @@ name: Installer CI
on:
pull_request:
paths:
- 'install.sh'
- 'scripts/install.sh'
push:
paths:
- 'install.sh'
- 'scripts/install.sh'
schedule:
- cron: '0 2 * * *' # every day at 02:00 UTC
@ -16,11 +16,11 @@ jobs:
steps:
- uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # 4.2.2
- name: Run ShellCheck on install.sh
run: shellcheck install.sh
run: shellcheck scripts/install.sh
smoke-test:
needs: lint
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # 4.2.2
- name: Run installer end-to-end
run: ./install.sh
run: ./scripts/install.sh

2
.github/workflows/integration-auth-tests.yml vendored
View file

@ -35,7 +35,7 @@ jobs:
- name: Install minikube
if: ${{ matrix.auth-provider == 'kubernetes' }}
uses: medyagh/setup-minikube@cea33675329b799adccc9526aa5daccc26cd5052 # v0.0.19
uses: medyagh/setup-minikube@e3c7f79eb1e997eabccc536a6cf318a2b0fe19d9 # v0.0.20
- name: Start minikube
if: ${{ matrix.auth-provider == 'oauth2_token' }}

47
.github/workflows/integration-tests.yml vendored
View file

@ -12,22 +12,49 @@ on:
- 'pyproject.toml'
- 'requirements.txt'
- '.github/workflows/integration-tests.yml' # This workflow
- '.github/actions/setup-ollama/action.yml'
schedule:
- cron: '0 0 * * *' # Daily at 12 AM UTC
workflow_dispatch:
inputs:
test-all-client-versions:
description: 'Test against both the latest and published versions'
type: boolean
default: false
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
jobs:
test-matrix:
discover-tests:
runs-on: ubuntu-latest
outputs:
test-type: ${{ steps.generate-matrix.outputs.test-type }}
steps:
- name: Checkout repository
uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
- name: Generate test matrix
id: generate-matrix
run: |
# Get test directories dynamically, excluding non-test directories
TEST_TYPES=$(find tests/integration -maxdepth 1 -mindepth 1 -type d -printf "%f\n" |
grep -Ev "^(__pycache__|fixtures|test_cases)$" |
sort | jq -R -s -c 'split("\n")[:-1]')
echo "test-type=$TEST_TYPES" >> $GITHUB_OUTPUT
test-matrix:
needs: discover-tests
runs-on: ubuntu-latest
strategy:
fail-fast: false
matrix:
# Listing tests manually since some of them currently fail
# TODO: generate matrix list from tests/integration when fixed
test-type: [agents, inference, datasets, inspect, scoring, post_training, providers, tool_runtime, vector_io]
test-type: ${{ fromJson(needs.discover-tests.outputs.test-type) }}
client-type: [library, server]
python-version: ["3.12", "3.13"]
fail-fast: false # we want to run all tests regardless of failure
client-version: ${{ (github.event_name == 'schedule' || github.event.inputs.test-all-client-versions == 'true') && fromJSON('["published", "latest"]') || fromJSON('["latest"]') }}
steps:
- name: Checkout repository
@ -37,6 +64,7 @@ jobs:
uses: ./.github/actions/setup-runner
with:
python-version: ${{ matrix.python-version }}
client-version: ${{ matrix.client-version }}
- name: Setup ollama
uses: ./.github/actions/setup-ollama
@ -53,9 +81,11 @@ jobs:
- name: Run Integration Tests
env:
OLLAMA_INFERENCE_MODEL: "meta-llama/Llama-3.2-3B-Instruct" # for server tests
OLLAMA_INFERENCE_MODEL: "llama3.2:3b-instruct-fp16" # for server tests
ENABLE_OLLAMA: "ollama" # for server tests
OLLAMA_URL: "http://0.0.0.0:11434"
SAFETY_MODEL: "llama-guard3:1b"
LLAMA_STACK_CLIENT_TIMEOUT: "300" # Increased timeout for eval operations
# Use 'shell' to get pipefail behavior
# https://docs.github.com/en/actions/reference/workflow-syntax-for-github-actions#exit-codes-and-error-action-preference
# TODO: write a precommit hook to detect if a test contains a pipe but does not use 'shell: bash'
@ -68,8 +98,9 @@ jobs:
fi
uv run pytest -s -v tests/integration/${{ matrix.test-type }} --stack-config=${stack_config} \
-k "not(builtin_tool or safety_with_image or code_interpreter or test_rag)" \
--text-model="ollama/meta-llama/Llama-3.2-3B-Instruct" \
--text-model="ollama/llama3.2:3b-instruct-fp16" \
--embedding-model=all-MiniLM-L6-v2 \
--safety-shield=$SAFETY_MODEL \
--color=yes \
--capture=tee-sys | tee pytest-${{ matrix.test-type }}.log
@ -88,7 +119,7 @@ jobs:
if: ${{ always() }}
uses: actions/upload-artifact@ea165f8d65b6e75b540449e92b4886f43607fa02 # v4.6.2
with:
name: logs-${{ github.run_id }}-${{ github.run_attempt }}-${{ matrix.client-type }}-${{ matrix.test-type }}-${{ matrix.python-version }}
name: logs-${{ github.run_id }}-${{ github.run_attempt }}-${{ matrix.client-type }}-${{ matrix.test-type }}-${{ matrix.python-version }}-${{ matrix.client-version }}
path: |
*.log
retention-days: 1

69
.github/workflows/tests.yml vendored
View file

@ -1,69 +0,0 @@
name: auto-tests
on:
# pull_request:
workflow_dispatch:
inputs:
commit_sha:
description: 'Specific Commit SHA to trigger on'
required: false
default: $GITHUB_SHA # default to the last commit of $GITHUB_REF branch
jobs:
test-llama-stack-as-library:
runs-on: ubuntu-latest
env:
TOGETHER_API_KEY: ${{ secrets.TOGETHER_API_KEY }}
FIREWORKS_API_KEY: ${{ secrets.FIREWORKS_API_KEY }}
TAVILY_SEARCH_API_KEY: ${{ secrets.TAVILY_SEARCH_API_KEY }}
strategy:
matrix:
provider: [fireworks, together]
steps:
- uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
with:
ref: ${{ github.event.inputs.commit_sha }}
- name: Echo commit SHA
run: |
echo "Triggered on commit SHA: ${{ github.event.inputs.commit_sha }}"
git rev-parse HEAD
- name: Install dependencies
run: |
python -m pip install --upgrade pip
pip install -r requirements.txt pytest
pip install -e .
- name: Build providers
run: |
llama stack build --template ${{ matrix.provider }} --image-type venv
- name: Install the latest llama-stack-client & llama-models packages
run: |
pip install -e git+https://github.com/meta-llama/llama-stack-client-python.git#egg=llama-stack-client
pip install -e git+https://github.com/meta-llama/llama-models.git#egg=llama-models
- name: Run client-sdk test
working-directory: "${{ github.workspace }}"
env:
REPORT_OUTPUT: md_report.md
shell: bash
run: |
pip install --upgrade pytest-md-report
echo "REPORT_FILE=${REPORT_OUTPUT}" >> "$GITHUB_ENV"
export INFERENCE_MODEL=meta-llama/Llama-3.1-8B-Instruct
LLAMA_STACK_CONFIG=./llama_stack/templates/${{ matrix.provider }}/run.yaml pytest --md-report --md-report-verbose=1 ./tests/client-sdk/inference/ --md-report-output "$REPORT_OUTPUT"
- name: Output reports to the job summary
if: always()
shell: bash
run: |
if [ -f "$REPORT_FILE" ]; then
echo "<details><summary> Test Report for ${{ matrix.provider }} </summary>" >> $GITHUB_STEP_SUMMARY
echo "" >> $GITHUB_STEP_SUMMARY
cat "$REPORT_FILE" >> $GITHUB_STEP_SUMMARY
echo "" >> $GITHUB_STEP_SUMMARY
echo "</details>" >> $GITHUB_STEP_SUMMARY
fi

15
.pre-commit-config.yaml
View file

@ -29,7 +29,7 @@ repos:
- id: check-toml
- repo: https://github.com/Lucas-C/pre-commit-hooks
rev: v1.5.4
rev: v1.5.5
hooks:
- id: insert-license
files: \.py$|\.sh$
@ -38,7 +38,7 @@ repos:
- docs/license_header.txt
- repo: https://github.com/astral-sh/ruff-pre-commit
rev: v0.9.4
rev: v0.12.2
hooks:
- id: ruff
args: [ --fix ]
@ -46,14 +46,14 @@ repos:
- id: ruff-format
- repo: https://github.com/adamchainz/blacken-docs
rev: 1.19.0
rev: 1.19.1
hooks:
- id: blacken-docs
additional_dependencies:
- black==24.3.0
- repo: https://github.com/astral-sh/uv-pre-commit
rev: 0.7.8
rev: 0.7.20
hooks:
- id: uv-lock
- id: uv-export
@ -66,7 +66,7 @@ repos:
]
- repo: https://github.com/pre-commit/mirrors-mypy
rev: v1.15.0
rev: v1.16.1
hooks:
- id: mypy
additional_dependencies:
@ -133,3 +133,8 @@ repos:
ci:
autofix_commit_msg: 🎨 [pre-commit.ci] Auto format from pre-commit.com hooks
autoupdate_commit_msg: ⬆ [pre-commit.ci] pre-commit autoupdate
autofix_prs: true
autoupdate_branch: ''
autoupdate_schedule: weekly
skip: []
submodules: false

6
CONTRIBUTING.md
View file

@ -66,7 +66,7 @@ You can install the dependencies by running:
```bash
cd llama-stack
uv sync --extra dev
uv sync --group dev
uv pip install -e .
source .venv/bin/activate
```
@ -112,7 +112,7 @@ uv run pre-commit run --all-files
## Running tests
You can find the Llama Stack testing documentation here [here](tests/README.md).
You can find the Llama Stack testing documentation [here](https://github.com/meta-llama/llama-stack/blob/main/tests/README.md).
## Adding a new dependency to the project
@ -168,7 +168,7 @@ manually as they are auto-generated.
### Updating the provider documentation
If you have made changes to a provider's configuration, you should run `./scripts/distro_codegen.py`
If you have made changes to a provider's configuration, you should run `./scripts/provider_codegen.py`
to re-generate the documentation. You should not change `docs/source/.../providers/` files manually
as they are auto-generated.
Note that the provider "description" field will be used to generate the provider documentation.

2
README.md
View file

@ -77,7 +77,7 @@ As more providers start supporting Llama 4, you can use them in Llama Stack as w
To try Llama Stack locally, run:
```bash
curl -LsSf https://github.com/meta-llama/llama-stack/raw/main/install.sh | bash
curl -LsSf https://github.com/meta-llama/llama-stack/raw/main/scripts/install.sh | bash
```
### Overview

24
docs/_static/llama-stack-spec.html vendored
View file

@ -11340,6 +11340,9 @@
},
"embedding_dimension": {
"type": "integer"
},
"vector_db_name": {
"type": "string"
}
},
"additionalProperties": false,
@ -13590,10 +13593,6 @@
"provider_id": {
"type": "string",
"description": "The ID of the provider to use for this vector store."
},
"provider_vector_db_id": {
"type": "string",
"description": "The provider-specific vector database ID."
}
},
"additionalProperties": false,
@ -14796,7 +14795,8 @@
"description": "Template for formatting each retrieved chunk in the context. Available placeholders: {index} (1-based chunk ordinal), {chunk.content} (chunk content string), {metadata} (chunk metadata dict). Default: \"Result {index}\\nContent: {chunk.content}\\nMetadata: {metadata}\\n\""
},
"mode": {
"type": "string",
"$ref": "#/components/schemas/RAGSearchMode",
"default": "vector",
"description": "Search mode for retrieval—either \"vector\", \"keyword\", or \"hybrid\". Default \"vector\"."
},
"ranker": {
@ -14831,6 +14831,16 @@
}
}
},
"RAGSearchMode": {
"type": "string",
"enum": [
"vector",
"keyword",
"hybrid"
],
"title": "RAGSearchMode",
"description": "Search modes for RAG query retrieval: - VECTOR: Uses vector similarity search for semantic matching - KEYWORD: Uses keyword-based search for exact matching - HYBRID: Combines both vector and keyword search for better results"
},
"RRFRanker": {
"type": "object",
"properties": {
@ -15623,6 +15633,10 @@
"type": "string",
"description": "The identifier of the provider."
},
"vector_db_name": {
"type": "string",
"description": "The name of the vector database."
},
"provider_vector_db_id": {
"type": "string",
"description": "The identifier of the vector database in the provider."

23
docs/_static/llama-stack-spec.yaml vendored
View file

@ -7984,6 +7984,8 @@ components:
type: string
embedding_dimension:
type: integer
vector_db_name:
type: string
additionalProperties: false
required:
- identifier
@ -9494,10 +9496,6 @@ components:
type: string
description: >-
The ID of the provider to use for this vector store.
provider_vector_db_id:
type: string
description: >-
The provider-specific vector database ID.
additionalProperties: false
required:
- name
@ -10346,7 +10344,8 @@ components:
content string), {metadata} (chunk metadata dict). Default: "Result {index}\nContent:
{chunk.content}\nMetadata: {metadata}\n"
mode:
type: string
$ref: '#/components/schemas/RAGSearchMode'
default: vector
description: >-
Search mode for retrieval—either "vector", "keyword", or "hybrid". Default
"vector".
@ -10373,6 +10372,17 @@ components:
mapping:
default: '#/components/schemas/DefaultRAGQueryGeneratorConfig'
llm: '#/components/schemas/LLMRAGQueryGeneratorConfig'
RAGSearchMode:
type: string
enum:
- vector
- keyword
- hybrid
title: RAGSearchMode
description: >-
Search modes for RAG query retrieval: - VECTOR: Uses vector similarity search
for semantic matching - KEYWORD: Uses keyword-based search for exact matching
- HYBRID: Combines both vector and keyword search for better results
RRFRanker:
type: object
properties:
@ -10933,6 +10943,9 @@ components:
provider_id:
type: string
description: The identifier of the provider.
vector_db_name:
type: string
description: The name of the vector database.
provider_vector_db_id:
type: string
description: >-

10
rfcs/RFC-0001-llama-stack.md → docs/original_rfc.md
View file

@ -1,5 +1,7 @@
# The Llama Stack API
*Originally authored Jul 23, 2024*
**Authors:**
* Meta: @raghotham, @ashwinb, @hjshah, @jspisak
@ -24,7 +26,7 @@ Meta releases weights of both the pretrained and instruction fine-tuned Llama mo
### Model Lifecycle
![Figure 1: Model Life Cycle](../docs/resources/model-lifecycle.png)
![Figure 1: Model Life Cycle](resources/model-lifecycle.png)
For each of the operations that need to be performed (e.g. fine tuning, inference, evals etc) during the model life cycle, we identified the capabilities as toolchain APIs that are needed. Some of these capabilities are primitive operations like inference while other capabilities like synthetic data generation are composed of other capabilities. The list of APIs we have identified to support the lifecycle of Llama models is below:
@ -37,7 +39,7 @@ For each of the operations that need to be performed (e.g. fine tuning, inferenc
### Agentic System
![Figure 2: Agentic System](../docs/resources/agentic-system.png)
![Figure 2: Agentic System](resources/agentic-system.png)
In addition to the model lifecycle, we considered the different components involved in an agentic system. Specifically around tool calling and shields. Since the model may decide to call tools, a single model inference call is not enough. What’s needed is an agentic loop consisting of tool calls and inference. The model provides separate tokens representing end-of-message and end-of-turn. A message represents a possible stopping point for execution where the model can inform the execution environment that a tool call needs to be made. The execution environment, upon execution, adds back the result to the context window and makes another inference call. This process can get repeated until an end-of-turn token is generated.
Note that as of today, in the OSS world, such a “loop” is often coded explicitly via elaborate prompt engineering using a ReAct pattern (typically) or preconstructed execution graph. Llama 3.1 (and future Llamas) attempts to absorb this multi-step reasoning loop inside the main model itself.
@ -63,9 +65,9 @@ The sequence diagram that details the steps is [here](https://github.com/meta-ll
We define the Llama Stack as a layer cake shown below.
![Figure 3: Llama Stack](../docs/resources/llama-stack.png)
![Figure 3: Llama Stack](resources/llama-stack.png)
The API is defined in the [YAML](../docs/_static/llama-stack-spec.yaml) and [HTML](../docs/_static/llama-stack-spec.html) files.
The API is defined in the [YAML](_static/llama-stack-spec.yaml) and [HTML](_static/llama-stack-spec.html) files.
## Sample implementations

6
docs/source/advanced_apis/eval/index.md Normal file
View file

@ -0,0 +1,6 @@
# Eval Providers
This section contains documentation for all available providers for the **eval** API.
- [inline::meta-reference](inline_meta-reference.md)
- [remote::nvidia](remote_nvidia.md)

21
docs/source/advanced_apis/eval/inline_meta-reference.md Normal file
View file

@ -0,0 +1,21 @@
# inline::meta-reference
## Description
Meta's reference implementation of evaluation tasks with support for multiple languages and evaluation metrics.
## Configuration
| Field | Type | Required | Default | Description |
|-------|------|----------|---------|-------------|
| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | |
## Sample Configuration
```yaml
kvstore:
type: sqlite
db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/meta_reference_eval.db
```

19
docs/source/advanced_apis/eval/remote_nvidia.md Normal file
View file

@ -0,0 +1,19 @@
# remote::nvidia
## Description
NVIDIA's evaluation provider for running evaluation tasks on NVIDIA's platform.
## Configuration
| Field | Type | Required | Default | Description |
|-------|------|----------|---------|-------------|
| `evaluator_url` | `<class 'str'>` | No | http://0.0.0.0:7331 | The url for accessing the evaluator service |
## Sample Configuration
```yaml
evaluator_url: ${env.NVIDIA_EVALUATOR_URL:=http://localhost:7331}
```

0
docs/source/concepts/evaluation_concepts.md → docs/source/advanced_apis/evaluation_concepts.md
View file

33
docs/source/advanced_apis/index.md Normal file
View file

@ -0,0 +1,33 @@
# Advanced APIs
## Post-training
Fine-tunes a model.
```{toctree}
:maxdepth: 1
post_training/index
```
## Eval
Generates outputs (via Inference or Agents) and perform scoring.
```{toctree}
:maxdepth: 1
eval/index
```
```{include} evaluation_concepts.md
:start-after: ## Evaluation Concepts
```
## Scoring
Evaluates the outputs of the system.
```{toctree}
:maxdepth: 1
scoring/index
```

0
docs/source/providers/post_training/huggingface.md → docs/source/advanced_apis/post_training/huggingface.md
View file

7
docs/source/advanced_apis/post_training/index.md Normal file
View file

@ -0,0 +1,7 @@
# Post_Training Providers
This section contains documentation for all available providers for the **post_training** API.
- [inline::huggingface](inline_huggingface.md)
- [inline::torchtune](inline_torchtune.md)
- [remote::nvidia](remote_nvidia.md)

33
docs/source/advanced_apis/post_training/inline_huggingface.md Normal file
View file

@ -0,0 +1,33 @@
# inline::huggingface
## Description
HuggingFace-based post-training provider for fine-tuning models using the HuggingFace ecosystem.
## Configuration
| Field | Type | Required | Default | Description |
|-------|------|----------|---------|-------------|
| `device` | `<class 'str'>` | No | cuda | |
| `distributed_backend` | `Literal['fsdp', 'deepspeed'` | No | | |
| `checkpoint_format` | `Literal['full_state', 'huggingface'` | No | huggingface | |
| `chat_template` | `<class 'str'>` | No | |
| `model_specific_config` | `<class 'dict'>` | No | {'trust_remote_code': True, 'attn_implementation': 'sdpa'} | |
| `max_seq_length` | `<class 'int'>` | No | 2048 | |
| `gradient_checkpointing` | `<class 'bool'>` | No | False | |
| `save_total_limit` | `<class 'int'>` | No | 3 | |
| `logging_steps` | `<class 'int'>` | No | 10 | |
| `warmup_ratio` | `<class 'float'>` | No | 0.1 | |
| `weight_decay` | `<class 'float'>` | No | 0.01 | |
| `dataloader_num_workers` | `<class 'int'>` | No | 4 | |
| `dataloader_pin_memory` | `<class 'bool'>` | No | True | |
## Sample Configuration
```yaml
checkpoint_format: huggingface
distributed_backend: null
device: cpu
```

20
docs/source/advanced_apis/post_training/inline_torchtune.md Normal file
View file

@ -0,0 +1,20 @@
# inline::torchtune
## Description
TorchTune-based post-training provider for fine-tuning and optimizing models using Meta's TorchTune framework.
## Configuration
| Field | Type | Required | Default | Description |
|-------|------|----------|---------|-------------|
| `torch_seed` | `int \| None` | No | | |
| `checkpoint_format` | `Literal['meta', 'huggingface'` | No | meta | |
## Sample Configuration
```yaml
checkpoint_format: meta
```

0
docs/source/providers/post_training/nvidia_nemo.md → docs/source/advanced_apis/post_training/nvidia_nemo.md
View file

28
docs/source/advanced_apis/post_training/remote_nvidia.md Normal file
View file

@ -0,0 +1,28 @@
# remote::nvidia
## Description
NVIDIA's post-training provider for fine-tuning models on NVIDIA's platform.
## Configuration
| Field | Type | Required | Default | Description |
|-------|------|----------|---------|-------------|
| `api_key` | `str \| None` | No | | The NVIDIA API key. |
| `dataset_namespace` | `str \| None` | No | default | The NVIDIA dataset namespace. |
| `project_id` | `str \| None` | No | test-example-model@v1 | The NVIDIA project ID. |
| `customizer_url` | `str \| None` | No | | Base URL for the NeMo Customizer API |
| `timeout` | `<class 'int'>` | No | 300 | Timeout for the NVIDIA Post Training API |
| `max_retries` | `<class 'int'>` | No | 3 | Maximum number of retries for the NVIDIA Post Training API |
| `output_model_dir` | `<class 'str'>` | No | test-example-model@v1 | Directory to save the output model |
## Sample Configuration
```yaml
api_key: ${env.NVIDIA_API_KEY:=}
dataset_namespace: ${env.NVIDIA_DATASET_NAMESPACE:=default}
project_id: ${env.NVIDIA_PROJECT_ID:=test-project}
customizer_url: ${env.NVIDIA_CUSTOMIZER_URL:=http://nemo.test}
```

0
docs/source/providers/post_training/torchtune.md → docs/source/advanced_apis/post_training/torchtune.md
View file

7
docs/source/advanced_apis/scoring/index.md Normal file
View file

@ -0,0 +1,7 @@
# Scoring Providers
This section contains documentation for all available providers for the **scoring** API.
- [inline::basic](inline_basic.md)
- [inline::braintrust](inline_braintrust.md)
- [inline::llm-as-judge](inline_llm-as-judge.md)

13
docs/source/advanced_apis/scoring/inline_basic.md Normal file
View file

@ -0,0 +1,13 @@
# inline::basic
## Description
Basic scoring provider for simple evaluation metrics and scoring functions.
## Sample Configuration
```yaml
{}
```

19
docs/source/advanced_apis/scoring/inline_braintrust.md Normal file
View file

@ -0,0 +1,19 @@
# inline::braintrust
## Description
Braintrust scoring provider for evaluation and scoring using the Braintrust platform.
## Configuration
| Field | Type | Required | Default | Description |
|-------|------|----------|---------|-------------|
| `openai_api_key` | `str \| None` | No | | The OpenAI API Key |
## Sample Configuration
```yaml
openai_api_key: ${env.OPENAI_API_KEY:=}
```

13
docs/source/advanced_apis/scoring/inline_llm-as-judge.md Normal file
View file

@ -0,0 +1,13 @@
# inline::llm-as-judge
## Description
LLM-as-judge scoring provider that uses language models to evaluate and score responses.
## Sample Configuration
```yaml
{}
```

3
docs/source/building_applications/index.md
View file

@ -1,4 +1,4 @@
# Building AI Applications (Examples)
# AI Application Examples
Llama Stack provides all the building blocks needed to create sophisticated AI applications.
@ -27,4 +27,5 @@ tools
evals
telemetry
safety
playground/index
```

6
docs/source/playground/index.md → docs/source/building_applications/playground/index.md
View file

@ -1,4 +1,4 @@
# Llama Stack Playground
## Llama Stack Playground
```{note}
The Llama Stack Playground is currently experimental and subject to change. We welcome feedback and contributions to help improve it.
@ -9,7 +9,7 @@ The Llama Stack Playground is an simple interface which aims to:
- Demo **end-to-end** application code to help users get started to build their own applications
- Provide an **UI** to help users inspect and understand Llama Stack API providers and resources
## Key Features
### Key Features
#### Playground
Interactive pages for users to play with and explore Llama Stack API capabilities.
@ -90,7 +90,7 @@ Interactive pages for users to play with and explore Llama Stack API capabilitie
- Under the hood, it uses Llama Stack's `/<resources>/list` API to get information about each resources.
- Please visit [Core Concepts](https://llama-stack.readthedocs.io/en/latest/concepts/index.html) for more details about the resources.
## Starting the Llama Stack Playground
### Starting the Llama Stack Playground
To start the Llama Stack Playground, run the following commands:

50
docs/source/introduction/index.md → docs/source/concepts/architecture.md
View file

@ -1,31 +1,39 @@
# Why Llama Stack?
## Llama Stack architecture
Building production AI applications today requires solving multiple challenges:
**Infrastructure Complexity**
- Running large language models efficiently requires specialized infrastructure.
- Different deployment scenarios (local development, cloud, edge) need different solutions.
- Moving from development to production often requires significant rework.
**Essential Capabilities**
- Safety guardrails and content filtering are necessary in an enterprise setting.
- Just model inference is not enough - Knowledge retrieval and RAG capabilities are required.
- Nearly any application needs composable multi-step workflows.
- Finally, without monitoring, observability and evaluation, you end up operating in the dark.
**Lack of Flexibility and Choice**
- Directly integrating with multiple providers creates tight coupling.
- Different providers have different APIs and abstractions.
- Changing providers requires significant code changes.
### Our Solution: A Universal Stack
Llama Stack allows you to build different layers of distributions for your AI workloads using various SDKs and API providers.
```{image} ../../_static/llama-stack.png
:alt: Llama Stack
:width: 400px
```
### Benefits of Llama stack
#### Current challenges in custom AI applications
Building production AI applications today requires solving multiple challenges:
Infrastructure Complexity
- Running large language models efficiently requires specialized infrastructure.
- Different deployment scenarios (local development, cloud, edge) need different solutions.
- Moving from development to production often requires significant rework.
**Essential Capabilities**
- Safety guardrails and content filtering are necessary in an enterprise setting.
- Just model inference is not enough - Knowledge retrieval and RAG capabilities are required.
- Nearly any application needs composable multi-step workflows.
- Without monitoring, observability and evaluation, you end up operating in the dark.
**Lack of Flexibility and Choice**
- Directly integrating with multiple providers creates tight coupling.
- Different providers have different APIs and abstractions.
- Changing providers requires significant code changes.
#### Our Solution: A Universal Stack
Llama Stack addresses these challenges through a service-oriented, API-first approach:
**Develop Anywhere, Deploy Everywhere**

12
docs/source/concepts/index.md
View file

@ -2,6 +2,10 @@
Given Llama Stack's service-oriented philosophy, a few concepts and workflows arise which may not feel completely natural in the LLM landscape, especially if you are coming with a background in other frameworks.
```{include} architecture.md
:start-after: ## Llama Stack architecture
```
```{include} apis.md
:start-after: ## APIs
```
@ -10,14 +14,10 @@ Given Llama Stack's service-oriented philosophy, a few concepts and workflows ar
:start-after: ## API Providers
```
```{include} resources.md
:start-after: ## Resources
```
```{include} distributions.md
:start-after: ## Distributions
```
```{include} evaluation_concepts.md
:start-after: ## Evaluation Concepts
```{include} resources.md
:start-after: ## Resources
```

11
docs/source/conf.py
View file

@ -52,7 +52,18 @@ extensions = [
"sphinxcontrib.redoc",
"sphinxcontrib.mermaid",
"sphinxcontrib.video",
"sphinx_reredirects"
]
redirects = {
"providers/post_training/index": "../../advanced_apis/post_training/index.html",
"providers/eval/index": "../../advanced_apis/eval/index.html",
"providers/scoring/index": "../../advanced_apis/scoring/index.html",
"playground/index": "../../building_applications/playground/index.html",
"openai/index": "../../providers/index.html#openai-api-compatibility",
"introduction/index": "../concepts/index.html#llama-stack-architecture"
}
myst_enable_extensions = ["colon_fence"]
html_theme = "sphinx_rtd_theme"

4
docs/source/deploying/index.md Normal file
View file

@ -0,0 +1,4 @@
# Deployment Examples
```{include} kubernetes_deployment.md
```

2
docs/source/distributions/kubernetes_deployment.md → docs/source/deploying/kubernetes_deployment.md
View file

@ -1,4 +1,4 @@
# Kubernetes Deployment Guide
## Kubernetes Deployment Guide
Instead of starting the Llama Stack and vLLM servers locally. We can deploy them in a Kubernetes cluster.

4
docs/source/distributions/building_distro.md
View file

@ -145,6 +145,10 @@ $ llama stack build --template starter
...
You can now edit ~/.llama/distributions/llamastack-starter/starter-run.yaml and run `llama stack run ~/.llama/distributions/llamastack-starter/starter-run.yaml`
```
```{tip}
The generated `run.yaml` file is a starting point for your configuration. For comprehensive guidance on customizing it for your specific needs, infrastructure, and deployment scenarios, see [Customizing Your run.yaml Configuration](customizing_run_yaml.md).
```
:::
:::{tab-item} Building from Scratch

4
docs/source/distributions/configuration.md
View file

@ -2,6 +2,10 @@
The Llama Stack runtime configuration is specified as a YAML file. Here is a simplified version of an example configuration file for the Ollama distribution:
```{note}
The default `run.yaml` files generated by templates are starting points for your configuration. For guidance on customizing these files for your specific needs, see [Customizing Your run.yaml Configuration](customizing_run_yaml.md).
```
```{dropdown} 👋 Click here for a Sample Configuration File
```yaml

40
docs/source/distributions/customizing_run_yaml.md Normal file
View file

@ -0,0 +1,40 @@
# Customizing run.yaml Files
The `run.yaml` files generated by Llama Stack templates are **starting points** designed to be customized for your specific needs. They are not meant to be used as-is in production environments.
## Key Points
- **Templates are starting points**: Generated `run.yaml` files contain defaults for development/testing
- **Customization expected**: Update URLs, credentials, models, and settings for your environment
- **Version control separately**: Keep customized configs in your own repository
- **Environment-specific**: Create different configurations for dev, staging, production
## What You Can Customize
You can customize:
- **Provider endpoints**: Change `http://localhost:8000` to your actual servers
- **Swap providers**: Replace default providers (e.g., swap Tavily with Brave for search)
- **Storage paths**: Move from `/tmp/` to production directories
- **Authentication**: Add API keys, SSL, timeouts
- **Models**: Different model sizes for dev vs prod
- **Database settings**: Switch from SQLite to PostgreSQL
- **Tool configurations**: Add custom tools and integrations
## Best Practices
- Use environment variables for secrets and environment-specific values
- Create separate `run.yaml` files for different environments (dev, staging, prod)
- Document your changes with comments
- Test configurations before deployment
- Keep your customized configs in version control
Example structure:
```
your-project/
├── configs/
│ ├── dev-run.yaml
│ ├── prod-run.yaml
└── README.md
```
The goal is to take the generated template and adapt it to your specific infrastructure and operational needs.

10
docs/source/distributions/index.md
View file

@ -6,13 +6,9 @@ This section provides an overview of the distributions available in Llama Stack.
```{toctree}
:maxdepth: 3
list_of_distributions
building_distro
customizing_run_yaml
importing_as_library
configuration
list_of_distributions
kubernetes_deployment
building_distro
on_device_distro
remote_hosted_distro
self_hosted_distro
```

10
docs/source/distributions/k8s/apply.sh
View file

@ -6,12 +6,12 @@
# This source code is licensed under the terms described in the LICENSE file in
# the root directory of this source tree.
export POSTGRES_USER=${POSTGRES_USER:-llamastack}
export POSTGRES_DB=${POSTGRES_DB:-llamastack}
export POSTGRES_PASSWORD=${POSTGRES_PASSWORD:-llamastack}
export POSTGRES_USER=llamastack
export POSTGRES_DB=llamastack
export POSTGRES_PASSWORD=llamastack
export INFERENCE_MODEL=${INFERENCE_MODEL:-meta-llama/Llama-3.2-3B-Instruct}
export SAFETY_MODEL=${SAFETY_MODEL:-meta-llama/Llama-Guard-3-1B}
export INFERENCE_MODEL=meta-llama/Llama-3.2-3B-Instruct
export SAFETY_MODEL=meta-llama/Llama-Guard-3-1B
# HF_TOKEN should be set by the user; base64 encode it for the secret
if [ -n "${HF_TOKEN:-}" ]; then

2
docs/source/distributions/k8s/vllm-k8s.yaml.template
View file

@ -32,7 +32,7 @@ spec:
image: vllm/vllm-openai:latest
command: ["/bin/sh", "-c"]
args:
- "vllm serve ${INFERENCE_MODEL} --dtype float16 --enforce-eager --max-model-len 4096 --gpu-memory-utilization 0.6"
- "vllm serve ${INFERENCE_MODEL} --dtype float16 --enforce-eager --max-model-len 4096 --gpu-memory-utilization 0.6 --enable-auto-tool-choice --tool-call-parser llama4_pythonic"
env:
- name: INFERENCE_MODEL
value: "${INFERENCE_MODEL}"

6
docs/source/distributions/ondevice_distro/android_sdk.md
View file

@ -13,7 +13,7 @@ Latest Release Notes: [link](https://github.com/meta-llama/llama-stack-client-ko
*Tagged releases are stable versions of the project. While we strive to maintain a stable main branch, it's not guaranteed to be free of bugs or issues.*
## Android Demo App
Check out our demo app to see how to integrate Llama Stack into your Android app: [Android Demo App](https://github.com/meta-llama/llama-stack-client-kotlin/tree/examples/android_app)
Check out our demo app to see how to integrate Llama Stack into your Android app: [Android Demo App](https://github.com/meta-llama/llama-stack-client-kotlin/tree/latest-release/examples/android_app)
The key files in the app are `ExampleLlamaStackLocalInference.kt`, `ExampleLlamaStackRemoteInference.kts`, and `MainActivity.java`. With encompassed business logic, the app shows how to use Llama Stack for both the environments.
@ -68,7 +68,7 @@ Ensure the Llama Stack server version is the same as the Kotlin SDK Library for
Other inference providers: [Table](https://llama-stack.readthedocs.io/en/latest/index.html#supported-llama-stack-implementations)
How to set remote localhost in Demo App: [Settings](https://github.com/meta-llama/llama-stack-apps/tree/main/examples/android_app#settings)
How to set remote localhost in Demo App: [Settings](https://github.com/meta-llama/llama-stack-client-kotlin/tree/latest-release/examples/android_app#settings)
### Initialize the Client
A client serves as the primary interface for interacting with a specific inference type and its associated parameters. Only after client is initialized then you can configure and start inferences.
@ -135,7 +135,7 @@ val result = client!!.inference().chatCompletionStreaming(
### Setup Custom Tool Calling
Android demo app for more details: [Custom Tool Calling](https://github.com/meta-llama/llama-stack-apps/tree/main/examples/android_app#tool-calling)
Android demo app for more details: [Custom Tool Calling](https://github.com/meta-llama/llama-stack-client-kotlin/tree/latest-release/examples/android_app#tool-calling)
## Advanced Users

1
docs/source/distributions/starting_llama_stack_server.md
View file

@ -28,5 +28,4 @@ If you have built a container image and want to deploy it in a Kubernetes cluste
importing_as_library
configuration
kubernetes_deployment
```

24
docs/source/getting_started/detailed_tutorial.md
View file

@ -1,4 +1,4 @@
# Detailed Tutorial
## Detailed Tutorial
In this guide, we'll walk through how you can use the Llama Stack (server and client SDK) to test a simple agent.
A Llama Stack agent is a simple integrated system that can perform tasks by combining a Llama model for reasoning with
@ -10,7 +10,7 @@ Llama Stack is a stateful service with REST APIs to support seamless transition
In this guide, we'll walk through how to build a RAG agent locally using Llama Stack with [Ollama](https://ollama.com/)
as the inference [provider](../providers/index.md#inference) for a Llama Model.
## Step 1: Installation and Setup
### Step 1: Installation and Setup
Install Ollama by following the instructions on the [Ollama website](https://ollama.com/download), then
download Llama 3.2 3B model, and then start the Ollama service.
@ -45,7 +45,7 @@ Setup your virtual environment.
uv sync --python 3.12
source .venv/bin/activate
```
## Step 2: Run Llama Stack
### Step 2: Run Llama Stack
Llama Stack is a server that exposes multiple APIs, you connect with it using the Llama Stack client SDK.
::::{tab-set}
@ -54,7 +54,7 @@ Llama Stack is a server that exposes multiple APIs, you connect with it using th
You can use Python to build and run the Llama Stack server, which is useful for testing and development.
Llama Stack uses a [YAML configuration file](../distributions/configuration.md) to specify the stack setup,
which defines the providers and their settings.
which defines the providers and their settings. The generated configuration serves as a starting point that you can [customize for your specific needs](../distributions/customizing_run_yaml.md).
Now let's build and run the Llama Stack config for Ollama.
We use `starter` as template. By default all providers are disabled, this requires enable ollama by passing environment variables.
@ -77,7 +77,7 @@ ENABLE_OLLAMA=ollama INFERENCE_MODEL="llama3.2:3b" llama stack build --template
You can use a container image to run the Llama Stack server. We provide several container images for the server
component that works with different inference providers out of the box. For this guide, we will use
`llamastack/distribution-starter` as the container image. If you'd like to build your own image or customize the
configurations, please check out [this guide](../references/index.md).
configurations, please check out [this guide](../distributions/building_distro.md).
First lets setup some environment variables and create a local directory to mount into the container’s file system.
```bash
export INFERENCE_MODEL="llama3.2:3b"
@ -132,7 +132,7 @@ Now you can use the Llama Stack client to run inference and build agents!
You can reuse the server setup or use the [Llama Stack Client](https://github.com/meta-llama/llama-stack-client-python/).
Note that the client package is already included in the `llama-stack` package.
## Step 3: Run Client CLI
### Step 3: Run Client CLI
Open a new terminal and navigate to the same directory you started the server from. Then set up a new or activate your
existing server virtual environment.
@ -232,7 +232,7 @@ OpenAIChatCompletion(
)
```
## Step 4: Run the Demos
### Step 4: Run the Demos
Note that these demos show the [Python Client SDK](../references/python_sdk_reference/index.md).
Other SDKs are also available, please refer to the [Client SDK](../index.md#client-sdks) list for the complete options.
@ -242,7 +242,7 @@ Other SDKs are also available, please refer to the [Client SDK](../index.md#clie
:::{tab-item} Basic Inference
Now you can run inference using the Llama Stack client SDK.
### i. Create the Script
#### i. Create the Script
Create a file `inference.py` and add the following code:
```python
@ -269,7 +269,7 @@ response = client.chat.completions.create(
print(response)
```
### ii. Run the Script
#### ii. Run the Script
Let's run the script using `uv`
```bash
uv run python inference.py
@ -283,7 +283,7 @@ OpenAIChatCompletion(id='chatcmpl-30cd0f28-a2ad-4b6d-934b-13707fc60ebf', choices
:::{tab-item} Build a Simple Agent
Next we can move beyond simple inference and build an agent that can perform tasks using the Llama Stack server.
### i. Create the Script
#### i. Create the Script
Create a file `agent.py` and add the following code:
```python
@ -455,7 +455,7 @@ uv run python agent.py
For our last demo, we can build a RAG agent that can answer questions about the Torchtune project using the documents
in a vector database.
### i. Create the Script
#### i. Create the Script
Create a file `rag_agent.py` and add the following code:
```python
@ -533,7 +533,7 @@ for t in turns:
for event in AgentEventLogger().log(stream):
event.print()
```
### ii. Run the Script
#### ii. Run the Script
Let's run the script using `uv`
```bash
uv run python rag_agent.py

126
docs/source/getting_started/index.md
View file

@ -1,123 +1,13 @@
# Quickstart
# Getting Started
Get started with Llama Stack in minutes!
Llama Stack is a stateful service with REST APIs to support the seamless transition of AI applications across different
environments. You can build and test using a local server first and deploy to a hosted endpoint for production.
In this guide, we'll walk through how to build a RAG application locally using Llama Stack with [Ollama](https://ollama.com/)
as the inference [provider](../providers/inference/index) for a Llama Model.
**💡 Notebook Version:** You can also follow this quickstart guide in a Jupyter notebook format: [quick_start.ipynb](https://github.com/meta-llama/llama-stack/blob/main/docs/quick_start.ipynb)
#### Step 1: Install and setup
1. Install [uv](https://docs.astral.sh/uv/)
2. Run inference on a Llama model with [Ollama](https://ollama.com/download)
```bash
ollama run llama3.2:3b --keepalive 60m
```{include} quickstart.md
:start-after: ## Quickstart
```
#### Step 2: Run the Llama Stack server
We will use `uv` to run the Llama Stack server.
```bash
INFERENCE_MODEL=llama3.2:3b uv run --with llama-stack llama stack build --template starter --image-type venv --run
```{include} libraries.md
:start-after: ## Libraries (SDKs)
```
#### Step 3: Run the demo
Now open up a new terminal and copy the following script into a file named `demo_script.py`.
```python
from llama_stack_client import Agent, AgentEventLogger, RAGDocument, LlamaStackClient
vector_db_id = "my_demo_vector_db"
client = LlamaStackClient(base_url="http://localhost:8321")
models = client.models.list()
# Select the first LLM and first embedding models
model_id = next(m for m in models if m.model_type == "llm").identifier
embedding_model_id = (
em := next(m for m in models if m.model_type == "embedding")
).identifier
embedding_dimension = em.metadata["embedding_dimension"]
_ = client.vector_dbs.register(
vector_db_id=vector_db_id,
embedding_model=embedding_model_id,
embedding_dimension=embedding_dimension,
provider_id="faiss",
)
source = "https://www.paulgraham.com/greatwork.html"
print("rag_tool> Ingesting document:", source)
document = RAGDocument(
document_id="document_1",
content=source,
mime_type="text/html",
metadata={},
)
client.tool_runtime.rag_tool.insert(
documents=[document],
vector_db_id=vector_db_id,
chunk_size_in_tokens=50,
)
agent = Agent(
client,
model=model_id,
instructions="You are a helpful assistant",
tools=[
{
"name": "builtin::rag/knowledge_search",
"args": {"vector_db_ids": [vector_db_id]},
}
],
)
prompt = "How do you do great work?"
print("prompt>", prompt)
response = agent.create_turn(
messages=[{"role": "user", "content": prompt}],
session_id=agent.create_session("rag_session"),
stream=True,
)
for log in AgentEventLogger().log(response):
log.print()
```{include} detailed_tutorial.md
:start-after: ## Detailed Tutorial
```
We will use `uv` to run the script
```
uv run --with llama-stack-client,fire,requests demo_script.py
```
And you should see output like below.
```
rag_tool> Ingesting document: https://www.paulgraham.com/greatwork.html
prompt> How do you do great work?
inference> [knowledge_search(query="What is the key to doing great work")]
tool_execution> Tool:knowledge_search Args:{'query': 'What is the key to doing great work'}
tool_execution> Tool:knowledge_search Response:[TextContentItem(text='knowledge_search tool found 5 chunks:\nBEGIN of knowledge_search tool results.\n', type='text'), TextContentItem(text="Result 1:\nDocument_id:docum\nContent: work. Doing great work means doing something important\nso well that you expand people's ideas of what's possible. But\nthere's no threshold for importance. It's a matter of degree, and\noften hard to judge at the time anyway.\n", type='text'), TextContentItem(text="Result 2:\nDocument_id:docum\nContent: work. Doing great work means doing something important\nso well that you expand people's ideas of what's possible. But\nthere's no threshold for importance. It's a matter of degree, and\noften hard to judge at the time anyway.\n", type='text'), TextContentItem(text="Result 3:\nDocument_id:docum\nContent: work. Doing great work means doing something important\nso well that you expand people's ideas of what's possible. But\nthere's no threshold for importance. It's a matter of degree, and\noften hard to judge at the time anyway.\n", type='text'), TextContentItem(text="Result 4:\nDocument_id:docum\nContent: work. Doing great work means doing something important\nso well that you expand people's ideas of what's possible. But\nthere's no threshold for importance. It's a matter of degree, and\noften hard to judge at the time anyway.\n", type='text'), TextContentItem(text="Result 5:\nDocument_id:docum\nContent: work. Doing great work means doing something important\nso well that you expand people's ideas of what's possible. But\nthere's no threshold for importance. It's a matter of degree, and\noften hard to judge at the time anyway.\n", type='text'), TextContentItem(text='END of knowledge_search tool results.\n', type='text')]
inference> Based on the search results, it seems that doing great work means doing something important so well that you expand people's ideas of what's possible. However, there is no clear threshold for importance, and it can be difficult to judge at the time.
To further clarify, I would suggest that doing great work involves:
* Completing tasks with high quality and attention to detail
* Expanding on existing knowledge or ideas
* Making a positive impact on others through your work
* Striving for excellence and continuous improvement
Ultimately, great work is about making a meaningful contribution and leaving a lasting impression.
```
Congratulations! You've successfully built your first RAG application using Llama Stack! 🎉🥳
## Next Steps
Now you're ready to dive deeper into Llama Stack!
- Explore the [Detailed Tutorial](./detailed_tutorial.md).
- Try the [Getting Started Notebook](https://github.com/meta-llama/llama-stack/blob/main/docs/getting_started.ipynb).
- Browse more [Notebooks on GitHub](https://github.com/meta-llama/llama-stack/tree/main/docs/notebooks).
- Learn about Llama Stack [Concepts](../concepts/index.md).
- Discover how to [Build Llama Stacks](../distributions/index.md).
- Refer to our [References](../references/index.md) for details on the Llama CLI and Python SDK.
- Check out the [llama-stack-apps](https://github.com/meta-llama/llama-stack-apps/tree/main/examples) repository for example applications and tutorials.

10
docs/source/getting_started/libraries.md Normal file
View file

@ -0,0 +1,10 @@
## Libraries (SDKs)
We have a number of client-side SDKs available for different languages.
| **Language** | **Client SDK** | **Package** |
| :----: | :----: | :----: |
| Python | [llama-stack-client-python](https://github.com/meta-llama/llama-stack-client-python) | [![PyPI version](https://img.shields.io/pypi/v/llama_stack_client.svg)](https://pypi.org/project/llama_stack_client/)
| Swift | [llama-stack-client-swift](https://github.com/meta-llama/llama-stack-client-swift/tree/latest-release) | [![Swift Package Index](https://img.shields.io/endpoint?url=https%3A%2F%2Fswiftpackageindex.com%2Fapi%2Fpackages%2Fmeta-llama%2Fllama-stack-client-swift%2Fbadge%3Ftype%3Dswift-versions)](https://swiftpackageindex.com/meta-llama/llama-stack-client-swift)
| Node | [llama-stack-client-node](https://github.com/meta-llama/llama-stack-client-node) | [![NPM version](https://img.shields.io/npm/v/llama-stack-client.svg)](https://npmjs.org/package/llama-stack-client)
| Kotlin | [llama-stack-client-kotlin](https://github.com/meta-llama/llama-stack-client-kotlin/tree/latest-release) | [![Maven version](https://img.shields.io/maven-central/v/com.llama.llamastack/llama-stack-client-kotlin)](https://central.sonatype.com/artifact/com.llama.llamastack/llama-stack-client-kotlin)

123
docs/source/getting_started/quickstart.md Normal file
View file

@ -0,0 +1,123 @@
## Quickstart
Get started with Llama Stack in minutes!
Llama Stack is a stateful service with REST APIs to support the seamless transition of AI applications across different
environments. You can build and test using a local server first and deploy to a hosted endpoint for production.
In this guide, we'll walk through how to build a RAG application locally using Llama Stack with [Ollama](https://ollama.com/)
as the inference [provider](../providers/inference/index) for a Llama Model.
**💡 Notebook Version:** You can also follow this quickstart guide in a Jupyter notebook format: [quick_start.ipynb](https://github.com/meta-llama/llama-stack/blob/main/docs/quick_start.ipynb)
#### Step 1: Install and setup
1. Install [uv](https://docs.astral.sh/uv/)
2. Run inference on a Llama model with [Ollama](https://ollama.com/download)
```bash
ollama run llama3.2:3b --keepalive 60m
```
#### Step 2: Run the Llama Stack server
We will use `uv` to run the Llama Stack server.
```bash
INFERENCE_MODEL=llama3.2:3b uv run --with llama-stack llama stack build --template starter --image-type venv --run
```
#### Step 3: Run the demo
Now open up a new terminal and copy the following script into a file named `demo_script.py`.
```python
from llama_stack_client import Agent, AgentEventLogger, RAGDocument, LlamaStackClient
vector_db_id = "my_demo_vector_db"
client = LlamaStackClient(base_url="http://localhost:8321")
models = client.models.list()
# Select the first LLM and first embedding models
model_id = next(m for m in models if m.model_type == "llm").identifier
embedding_model_id = (
em := next(m for m in models if m.model_type == "embedding")
).identifier
embedding_dimension = em.metadata["embedding_dimension"]
_ = client.vector_dbs.register(
vector_db_id=vector_db_id,
embedding_model=embedding_model_id,
embedding_dimension=embedding_dimension,
provider_id="faiss",
)
source = "https://www.paulgraham.com/greatwork.html"
print("rag_tool> Ingesting document:", source)
document = RAGDocument(
document_id="document_1",
content=source,
mime_type="text/html",
metadata={},
)
client.tool_runtime.rag_tool.insert(
documents=[document],
vector_db_id=vector_db_id,
chunk_size_in_tokens=50,
)
agent = Agent(
client,
model=model_id,
instructions="You are a helpful assistant",
tools=[
{
"name": "builtin::rag/knowledge_search",
"args": {"vector_db_ids": [vector_db_id]},
}
],
)
prompt = "How do you do great work?"
print("prompt>", prompt)
response = agent.create_turn(
messages=[{"role": "user", "content": prompt}],
session_id=agent.create_session("rag_session"),
stream=True,
)
for log in AgentEventLogger().log(response):
log.print()
```
We will use `uv` to run the script
```
uv run --with llama-stack-client,fire,requests demo_script.py
```
And you should see output like below.
```
rag_tool> Ingesting document: https://www.paulgraham.com/greatwork.html
prompt> How do you do great work?
inference> [knowledge_search(query="What is the key to doing great work")]
tool_execution> Tool:knowledge_search Args:{'query': 'What is the key to doing great work'}
tool_execution> Tool:knowledge_search Response:[TextContentItem(text='knowledge_search tool found 5 chunks:\nBEGIN of knowledge_search tool results.\n', type='text'), TextContentItem(text="Result 1:\nDocument_id:docum\nContent: work. Doing great work means doing something important\nso well that you expand people's ideas of what's possible. But\nthere's no threshold for importance. It's a matter of degree, and\noften hard to judge at the time anyway.\n", type='text'), TextContentItem(text="Result 2:\nDocument_id:docum\nContent: work. Doing great work means doing something important\nso well that you expand people's ideas of what's possible. But\nthere's no threshold for importance. It's a matter of degree, and\noften hard to judge at the time anyway.\n", type='text'), TextContentItem(text="Result 3:\nDocument_id:docum\nContent: work. Doing great work means doing something important\nso well that you expand people's ideas of what's possible. But\nthere's no threshold for importance. It's a matter of degree, and\noften hard to judge at the time anyway.\n", type='text'), TextContentItem(text="Result 4:\nDocument_id:docum\nContent: work. Doing great work means doing something important\nso well that you expand people's ideas of what's possible. But\nthere's no threshold for importance. It's a matter of degree, and\noften hard to judge at the time anyway.\n", type='text'), TextContentItem(text="Result 5:\nDocument_id:docum\nContent: work. Doing great work means doing something important\nso well that you expand people's ideas of what's possible. But\nthere's no threshold for importance. It's a matter of degree, and\noften hard to judge at the time anyway.\n", type='text'), TextContentItem(text='END of knowledge_search tool results.\n', type='text')]
inference> Based on the search results, it seems that doing great work means doing something important so well that you expand people's ideas of what's possible. However, there is no clear threshold for importance, and it can be difficult to judge at the time.
To further clarify, I would suggest that doing great work involves:
* Completing tasks with high quality and attention to detail
* Expanding on existing knowledge or ideas
* Making a positive impact on others through your work
* Striving for excellence and continuous improvement
Ultimately, great work is about making a meaningful contribution and leaving a lasting impression.
```
Congratulations! You've successfully built your first RAG application using Llama Stack! 🎉🥳
### Next Steps
Now you're ready to dive deeper into Llama Stack!
- Explore the [Detailed Tutorial](./detailed_tutorial.md).
- Try the [Getting Started Notebook](https://github.com/meta-llama/llama-stack/blob/main/docs/getting_started.ipynb).
- Browse more [Notebooks on GitHub](https://github.com/meta-llama/llama-stack/tree/main/docs/notebooks).
- Learn about Llama Stack [Concepts](../concepts/index.md).
- Discover how to [Build Llama Stacks](../distributions/index.md).
- Refer to our [References](../references/index.md) for details on the Llama CLI and Python SDK.
- Check out the [llama-stack-apps](https://github.com/meta-llama/llama-stack-apps/tree/main/examples) repository for example applications and tutorials.

17
docs/source/index.md
View file

@ -40,17 +40,6 @@ Kotlin.
- Ready to build? Check out the [Quick Start](getting_started/index) to get started.
- Want to contribute? See the [Contributing](contributing/index) guide.
## Client SDKs
We have a number of client-side SDKs available for different languages.
| **Language** | **Client SDK** | **Package** |
| :----: | :----: | :----: |
| Python | [llama-stack-client-python](https://github.com/meta-llama/llama-stack-client-python) | [![PyPI version](https://img.shields.io/pypi/v/llama_stack_client.svg)](https://pypi.org/project/llama_stack_client/)
| Swift | [llama-stack-client-swift](https://github.com/meta-llama/llama-stack-client-swift/tree/latest-release) | [![Swift Package Index](https://img.shields.io/endpoint?url=https%3A%2F%2Fswiftpackageindex.com%2Fapi%2Fpackages%2Fmeta-llama%2Fllama-stack-client-swift%2Fbadge%3Ftype%3Dswift-versions)](https://swiftpackageindex.com/meta-llama/llama-stack-client-swift)
| Node | [llama-stack-client-node](https://github.com/meta-llama/llama-stack-client-node) | [![NPM version](https://img.shields.io/npm/v/llama-stack-client.svg)](https://npmjs.org/package/llama-stack-client)
| Kotlin | [llama-stack-client-kotlin](https://github.com/meta-llama/llama-stack-client-kotlin/tree/latest-release) | [![Maven version](https://img.shields.io/maven-central/v/com.llama.llamastack/llama-stack-client-kotlin)](https://central.sonatype.com/artifact/com.llama.llamastack/llama-stack-client-kotlin)
## Supported Llama Stack Implementations
A number of "adapters" are available for some popular Inference and Vector Store providers. For other APIs (particularly Safety and Agents), we provide *reference implementations* you can use to get started. We expect this list to grow over time. We are slowly onboarding more providers to the ecosystem as we get more confidence in the APIs.
@ -133,14 +122,12 @@ A number of "adapters" are available for some popular Inference and Vector Store
self
getting_started/index
getting_started/detailed_tutorial
introduction/index
concepts/index
openai/index
providers/index
distributions/index
advanced_apis/index
building_applications/index
playground/index
deploying/index
contributing/index
references/index
```

72
docs/source/providers/index.md
View file

@ -1,4 +1,4 @@
# Providers Overview
# API Providers Overview
The goal of Llama Stack is to build an ecosystem where users can easily swap out different implementations for the same API. Examples for these include:
- LLM inference providers (e.g., Meta Reference, Ollama, Fireworks, Together, AWS Bedrock, Groq, Cerebras, SambaNova, vLLM, OpenAI, Anthropic, Gemini, WatsonX, etc.),
@ -13,13 +13,25 @@ Providers come in two flavors:
Importantly, Llama Stack always strives to provide at least one fully inline provider for each API so you can iterate on a fully featured environment locally.
## External Providers
Llama Stack supports external providers that live outside of the main codebase. This allows you to create and maintain your own providers independently.
```{toctree}
:maxdepth: 1
external
external.md
```
```{include} openai.md
:start-after: ## OpenAI API Compatibility
```
## Inference
Runs inference with an LLM.
```{toctree}
:maxdepth: 1
inference/index
```
## Agents
@ -40,33 +52,6 @@ Interfaces with datasets and data loaders.
datasetio/index
```
## Eval
Generates outputs (via Inference or Agents) and perform scoring.
```{toctree}
:maxdepth: 1
eval/index
```
## Inference
Runs inference with an LLM.
```{toctree}
:maxdepth: 1
inference/index
```
## Post Training
Fine-tunes a model.
```{toctree}
:maxdepth: 1
post_training/index
```
## Safety
Applies safety policies to the output at a Systems (not only model) level.
@ -76,15 +61,6 @@ Applies safety policies to the output at a Systems (not only model) level.
safety/index
```
## Scoring
Evaluates the outputs of the system.
```{toctree}
:maxdepth: 1
scoring/index
```
## Telemetry
Collects telemetry data from the system.
@ -94,15 +70,6 @@ Collects telemetry data from the system.
telemetry/index
```
## Tool Runtime
Is associated with the ToolGroup resouces.
```{toctree}
:maxdepth: 1
tool_runtime/index
```
## Vector IO
Vector IO refers to operations on vector databases, such as adding documents, searching, and deleting documents.
@ -114,3 +81,12 @@ io and database are used to store and retrieve documents for retrieval.
vector_io/index
```
## Tool Runtime
Is associated with the ToolGroup resources.
```{toctree}
:maxdepth: 1
tool_runtime/index
```

30
docs/source/openai/index.md → docs/source/providers/openai.md
View file

@ -1,14 +1,14 @@
# OpenAI API Compatibility
## OpenAI API Compatibility
## Server path
### Server path
Llama Stack exposes an OpenAI-compatible API endpoint at `/v1/openai/v1`. So, for a Llama Stack server running locally on port `8321`, the full url to the OpenAI-compatible API endpoint is `http://localhost:8321/v1/openai/v1`.
## Clients
### Clients
You should be able to use any client that speaks OpenAI APIs with Llama Stack. We regularly test with the official Llama Stack clients as well as OpenAI's official Python client.
### Llama Stack Client
#### Llama Stack Client
When using the Llama Stack client, set the `base_url` to the root of your Llama Stack server. It will automatically route OpenAI-compatible requests to the right server endpoint for you.
@ -18,7 +18,7 @@ from llama_stack_client import LlamaStackClient
client = LlamaStackClient(base_url="http://localhost:8321")
```
### OpenAI Client
#### OpenAI Client
When using an OpenAI client, set the `base_url` to the `/v1/openai/v1` path on your Llama Stack server.
@ -30,9 +30,9 @@ client = OpenAI(base_url="http://localhost:8321/v1/openai/v1", api_key="none")
Regardless of the client you choose, the following code examples should all work the same.
## APIs implemented
### APIs implemented
### Models
#### Models
Many of the APIs require you to pass in a model parameter. To see the list of models available in your Llama Stack server:
@ -40,13 +40,13 @@ Many of the APIs require you to pass in a model parameter. To see the list of mo
models = client.models.list()
```
### Responses
#### Responses
:::{note}
The Responses API implementation is still in active development. While it is quite usable, there are still unimplemented parts of the API. We'd love feedback on any use-cases you try that do not work to help prioritize the pieces left to implement. Please open issues in the [meta-llama/llama-stack](https://github.com/meta-llama/llama-stack) GitHub repository with details of anything that does not work.
:::
#### Simple inference
##### Simple inference
Request:
@ -66,7 +66,7 @@ Syntax whispers secrets sweet
Code's gentle silence
```
#### Structured Output
##### Structured Output
Request:
@ -106,9 +106,9 @@ Example output:
{ "participants": ["Alice", "Bob"] }
```
### Chat Completions
#### Chat Completions
#### Simple inference
##### Simple inference
Request:
@ -129,7 +129,7 @@ Logic flows like a river
Code's gentle beauty
```
#### Structured Output
##### Structured Output
Request:
@ -170,9 +170,9 @@ Example output:
{ "participants": ["Alice", "Bob"] }
```
### Completions
#### Completions
#### Simple inference
##### Simple inference
Request:

2
docs/source/providers/vector_io/inline_milvus.md
View file

@ -11,7 +11,7 @@ Please refer to the remote provider documentation.
| Field | Type | Required | Default | Description |
|-------|------|----------|---------|-------------|
| `db_path` | `<class 'str'>` | No | PydanticUndefined | |
| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | |
| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend (SQLite only for now) |
| `consistency_level` | `<class 'str'>` | No | Strong | The consistency level of the Milvus server |
## Sample Configuration

6
docs/source/providers/vector_io/inline_sqlite-vec.md
View file

@ -205,12 +205,16 @@ See [sqlite-vec's GitHub repo](https://github.com/asg017/sqlite-vec/tree/main) f
| Field | Type | Required | Default | Description |
|-------|------|----------|---------|-------------|
| `db_path` | `<class 'str'>` | No | PydanticUndefined | |
| `db_path` | `<class 'str'>` | No | PydanticUndefined | Path to the SQLite database file |
| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend (SQLite only for now) |
## Sample Configuration
```yaml
db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/sqlite_vec.db
kvstore:
type: sqlite
db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/sqlite_vec_registry.db
```

6
docs/source/providers/vector_io/inline_sqlite_vec.md
View file

@ -10,12 +10,16 @@ Please refer to the sqlite-vec provider documentation.
| Field | Type | Required | Default | Description |
|-------|------|----------|---------|-------------|
| `db_path` | `<class 'str'>` | No | PydanticUndefined | |
| `db_path` | `<class 'str'>` | No | PydanticUndefined | Path to the SQLite database file |
| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend (SQLite only for now) |
## Sample Configuration
```yaml
db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/sqlite_vec.db
kvstore:
type: sqlite
db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/sqlite_vec_registry.db
```

5
docs/source/providers/vector_io/remote_milvus.md
View file

@ -114,7 +114,7 @@ For more details on TLS configuration, refer to the [TLS setup guide](https://mi
| `uri` | `<class 'str'>` | No | PydanticUndefined | The URI of the Milvus server |
| `token` | `str \| None` | No | PydanticUndefined | The token of the Milvus server |
| `consistency_level` | `<class 'str'>` | No | Strong | The consistency level of the Milvus server |
| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig, annotation=NoneType, required=False, default='sqlite', discriminator='type'` | No | | Config for KV store backend (SQLite only for now) |
| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig` | No | sqlite | Config for KV store backend |
| `config` | `dict` | No | {} | This configuration allows additional fields to be passed through to the underlying Milvus client. See the [Milvus](https://milvus.io/docs/install-overview.md) documentation for more details about Milvus in general. |
> **Note**: This configuration class accepts additional fields beyond those listed above. You can pass any additional configuration options that will be forwarded to the underlying provider.
@ -124,6 +124,9 @@ For more details on TLS configuration, refer to the [TLS setup guide](https://mi
```yaml
uri: ${env.MILVUS_ENDPOINT}
token: ${env.MILVUS_TOKEN}
kvstore:
type: sqlite
db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/milvus_remote_registry.db
```

4
docs/source/providers/vector_io/remote_pgvector.md
View file

@ -40,6 +40,7 @@ See [PGVector's documentation](https://github.com/pgvector/pgvector) for more de
| `db` | `str \| None` | No | postgres | |
| `user` | `str \| None` | No | postgres | |
| `password` | `str \| None` | No | mysecretpassword | |
| `kvstore` | `utils.kvstore.config.RedisKVStoreConfig \| utils.kvstore.config.SqliteKVStoreConfig \| utils.kvstore.config.PostgresKVStoreConfig \| utils.kvstore.config.MongoDBKVStoreConfig, annotation=NoneType, required=False, default='sqlite', discriminator='type'` | No | | Config for KV store backend (SQLite only for now) |
## Sample Configuration
@ -49,6 +50,9 @@ port: ${env.PGVECTOR_PORT:=5432}
db: ${env.PGVECTOR_DB}
user: ${env.PGVECTOR_USER}
password: ${env.PGVECTOR_PASSWORD}
kvstore:
type: sqlite
db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/pgvector_registry.db
```

4
docs/source/providers/vector_io/remote_weaviate.md
View file

@ -36,7 +36,9 @@ See [Weaviate's documentation](https://weaviate.io/developers/weaviate) for more
## Sample Configuration
```yaml
{}
kvstore:
type: sqlite
db_path: ${env.SQLITE_STORE_DIR:=~/.llama/dummy}/weaviate_registry.db
```

401
docs/source/references/llama_stack_client_cli_reference.md
View file

@ -9,7 +9,8 @@ The `llama-stack-client` CLI allows you to query information about the distribut
llama-stack-client
Usage: llama-stack-client [OPTIONS] COMMAND [ARGS]...
Welcome to the LlamaStackClient CLI
Welcome to the llama-stack-client CLI - a command-line interface for
interacting with Llama Stack
Options:
--version Show the version and exit.
@ -35,6 +36,7 @@ Commands:
```
### `llama-stack-client configure`
Configure Llama Stack Client CLI.
```bash
llama-stack-client configure
> Enter the host name of the Llama Stack distribution server: localhost
@ -42,7 +44,24 @@ llama-stack-client configure
Done! You can now use the Llama Stack Client CLI with endpoint http://localhost:8321
```
Optional arguments:
- `--endpoint`: Llama Stack distribution endpoint
- `--api-key`: Llama Stack distribution API key
## `llama-stack-client inspect version`
Inspect server configuration.
```bash
llama-stack-client inspect version
```
```bash
VersionInfo(version='0.2.14')
```
### `llama-stack-client providers list`
Show available providers on distribution endpoint
```bash
llama-stack-client providers list
```
@ -66,9 +85,74 @@ llama-stack-client providers list
+-----------+----------------+-----------------+
```
### `llama-stack-client providers inspect`
Show specific provider configuration on distribution endpoint
```bash
llama-stack-client providers inspect <provider_id>
```
## Inference
Inference (chat).
### `llama-stack-client inference chat-completion`
Show available inference chat completion endpoints on distribution endpoint
```bash
llama-stack-client inference chat-completion --message <message> [--stream] [--session] [--model-id]
```
```bash
OpenAIChatCompletion(
id='chatcmpl-aacd11f3-8899-4ec5-ac5b-e655132f6891',
choices=[
OpenAIChatCompletionChoice(
finish_reason='stop',
index=0,
message=OpenAIChatCompletionChoiceMessageOpenAIAssistantMessageParam(
role='assistant',
content='The captain of the whaleship Pequod in Nathaniel Hawthorne\'s novel "Moby-Dick" is Captain
Ahab. He\'s a vengeful and obsessive old sailor who\'s determined to hunt down and kill the white sperm whale
Moby-Dick, whom he\'s lost his leg to in a previous encounter.',
name=None,
tool_calls=None,
refusal=None,
annotations=None,
audio=None,
function_call=None
),
logprobs=None
)
],
created=1752578797,
model='llama3.2:3b-instruct-fp16',
object='chat.completion',
service_tier=None,
system_fingerprint='fp_ollama',
usage={
'completion_tokens': 67,
'prompt_tokens': 33,
'total_tokens': 100,
'completion_tokens_details': None,
'prompt_tokens_details': None
}
)
```
Required arguments:
**Note:** At least one of these parameters is required for chat completion
- `--message`: Message
- `--session`: Start a Chat Session
Optional arguments:
- `--stream`: Stream
- `--model-id`: Model ID
## Model Management
Manage GenAI models.
### `llama-stack-client models list`
Show available llama models at distribution endpoint
```bash
llama-stack-client models list
```
@ -85,6 +169,7 @@ Total models: 1
```
### `llama-stack-client models get`
Show details of a specific model at the distribution endpoint
```bash
llama-stack-client models get Llama3.1-8B-Instruct
```
@ -105,69 +190,92 @@ Model RandomModel is not found at distribution endpoint host:port. Please ensure
```
### `llama-stack-client models register`
Register a new model at distribution endpoint
```bash
llama-stack-client models register <model_id> [--provider-id <provider_id>] [--provider-model-id <provider_model_id>] [--metadata <metadata>]
llama-stack-client models register <model_id> [--provider-id <provider_id>] [--provider-model-id <provider_model_id>] [--metadata <metadata>] [--model-type <model_type>]
```
### `llama-stack-client models update`
Required arguments:
- `MODEL_ID`: Model ID
- `--provider-id`: Provider ID for the model
Optional arguments:
- `--provider-model-id`: Provider's model ID
- `--metadata`: JSON metadata for the model
- `--model-type`: Model type: `llm`, `embedding`
### `llama-stack-client models unregister`
Unregister a model from distribution endpoint
```bash
llama-stack-client models update <model_id> [--provider-id <provider_id>] [--provider-model-id <provider_model_id>] [--metadata <metadata>]
```
### `llama-stack-client models delete`
```bash
llama-stack-client models delete <model_id>
llama-stack-client models unregister <model_id>
```
## Vector DB Management
Manage vector databases.
### `llama-stack-client vector_dbs list`
Show available vector dbs on distribution endpoint
```bash
llama-stack-client vector_dbs list
```
```
+--------------+----------------+---------------------+---------------+------------------------+
| identifier | provider_id | provider_resource_id| vector_db_type| params |
+==============+================+=====================+===============+========================+
| test_bank | meta-reference | test_bank | vector | embedding_model: all-MiniLM-L6-v2
embedding_dimension: 384|
+--------------+----------------+---------------------+---------------+------------------------+
┏━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ identifier ┃ provider_id ┃ provider_resource_id ┃ vector_db_type ┃ params ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩
│ my_demo_vector_db │ faiss │ my_demo_vector_db │ │ embedding_dimension: 384 │
│ │ │ │ │ embedding_model: all-MiniLM-L6-v2 │
│ │ │ │ │ type: vector_db │
│ │ │ │ │ │
└──────────────────────────┴─────────────┴──────────────────────────┴────────────────┴───────────────────────────────────┘
```
### `llama-stack-client vector_dbs register`
Create a new vector db
```bash
llama-stack-client vector_dbs register <vector-db-id> [--provider-id <provider-id>] [--provider-vector-db-id <provider-vector-db-id>] [--embedding-model <embedding-model>] [--embedding-dimension <embedding-dimension>]
```
Required arguments:
- `VECTOR_DB_ID`: Vector DB ID
Optional arguments:
- `--provider-id`: Provider ID for the vector db
- `--provider-vector-db-id`: Provider's vector db ID
- `--embedding-model`: Embedding model to use. Default: "all-MiniLM-L6-v2"
- `--embedding-model`: Embedding model to use. Default: `all-MiniLM-L6-v2`
- `--embedding-dimension`: Dimension of embeddings. Default: 384
### `llama-stack-client vector_dbs unregister`
Delete a vector db
```bash
llama-stack-client vector_dbs unregister <vector-db-id>
```
Required arguments:
- `VECTOR_DB_ID`: Vector DB ID
## Shield Management
Manage safety shield services.
### `llama-stack-client shields list`
Show available safety shields on distribution endpoint
```bash
llama-stack-client shields list
```
```
+--------------+----------+----------------+-------------+
| identifier | params | provider_id | type |
+==============+==========+================+=============+
| llama_guard | {} | meta-reference | llama_guard |
+--------------+----------+----------------+-------------+
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┓
┃ identifier ┃ provider_alias ┃ params ┃ provider_id ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┩
│ ollama │ ollama/llama-guard3:1b │ │ llama-guard │
└──────────────────────────────────┴───────────────────────────────────────────────────────────────────────┴───────────────────────┴────────────────────────────────────┘
```
### `llama-stack-client shields register`
Register a new safety shield
```bash
llama-stack-client shields register --shield-id <shield-id> [--provider-id <provider-id>] [--provider-shield-id <provider-shield-id>] [--params <params>]
```
@ -180,41 +288,29 @@ Optional arguments:
- `--provider-shield-id`: Provider's shield ID
- `--params`: JSON configuration parameters for the shield
## Eval Task Management
### `llama-stack-client benchmarks list`
```bash
llama-stack-client benchmarks list
```
### `llama-stack-client benchmarks register`
```bash
llama-stack-client benchmarks register --eval-task-id <eval-task-id> --dataset-id <dataset-id> --scoring-functions <function1> [<function2> ...] [--provider-id <provider-id>] [--provider-eval-task-id <provider-eval-task-id>] [--metadata <metadata>]
```
Required arguments:
- `--eval-task-id`: ID of the eval task
- `--dataset-id`: ID of the dataset to evaluate
- `--scoring-functions`: One or more scoring functions to use for evaluation
Optional arguments:
- `--provider-id`: Provider ID for the eval task
- `--provider-eval-task-id`: Provider's eval task ID
- `--metadata`: Metadata for the eval task in JSON format
## Eval execution
Run evaluation tasks.
### `llama-stack-client eval run-benchmark`
Run a evaluation benchmark task
```bash
llama-stack-client eval run-benchmark <eval-task-id1> [<eval-task-id2> ...] --eval-task-config <config-file> --output-dir <output-dir> [--num-examples <num>] [--visualize]
llama-stack-client eval run-benchmark <eval-task-id1> [<eval-task-id2> ...] --eval-task-config <config-file> --output-dir <output-dir> --model-id <model-id> [--num-examples <num>] [--visualize] [--repeat-penalty <repeat-penalty>] [--top-p <top-p>] [--max-tokens <max-tokens>]
```
Required arguments:
- `--eval-task-config`: Path to the eval task config file in JSON format
- `--output-dir`: Path to the directory where evaluation results will be saved
- `--model-id`: model id to run the benchmark eval on
Optional arguments:
- `--num-examples`: Number of examples to evaluate (useful for debugging)
- `--visualize`: If set, visualizes evaluation results after completion
- `--repeat-penalty`: repeat-penalty in the sampling params to run generation
- `--top-p`: top-p in the sampling params to run generation
- `--max-tokens`: max-tokens in the sampling params to run generation
- `--temperature`: temperature in the sampling params to run generation
Example benchmark_config.json:
```json
@ -231,21 +327,55 @@ Example benchmark_config.json:
```
### `llama-stack-client eval run-scoring`
Run scoring from application datasets
```bash
llama-stack-client eval run-scoring <eval-task-id> --eval-task-config <config-file> --output-dir <output-dir> [--num-examples <num>] [--visualize]
llama-stack-client eval run-scoring <eval-task-id> --output-dir <output-dir> [--num-examples <num>] [--visualize]
```
Required arguments:
- `--eval-task-config`: Path to the eval task config file in JSON format
- `--output-dir`: Path to the directory where scoring results will be saved
Optional arguments:
- `--num-examples`: Number of examples to evaluate (useful for debugging)
- `--visualize`: If set, visualizes scoring results after completion
- `--scoring-params-config`: Path to the scoring params config file in JSON format
- `--dataset-id`: Pre-registered dataset_id to score (from llama-stack-client datasets list)
- `--dataset-path`: Path to the dataset file to score
## Eval Tasks
Manage evaluation tasks.
### `llama-stack-client eval_tasks list`
Show available eval tasks on distribution endpoint
```bash
llama-stack-client eval_tasks list
```
### `llama-stack-client eval_tasks register`
Register a new eval task
```bash
llama-stack-client eval_tasks register --eval-task-id <eval-task-id> --dataset-id <dataset-id> --scoring-functions <scoring-functions> [--provider-id <provider-id>] [--provider-eval-task-id <provider-eval-task-id>] [--metadata <metadata>]
```
Required arguments:
- `--eval-task-id`: ID of the eval task
- `--dataset-id`: ID of the dataset to evaluate
- `--scoring-functions`: Scoring functions to use for evaluation
Optional arguments:
- `--provider-id`: Provider ID for the eval task
- `--provider-eval-task-id`: Provider's eval task ID
## Tool Group Management
Manage available tool groups.
### `llama-stack-client toolgroups list`
Show available llama toolgroups at distribution endpoint
```bash
llama-stack-client toolgroups list
```
@ -260,17 +390,28 @@ llama-stack-client toolgroups list
```
### `llama-stack-client toolgroups get`
Get available llama toolgroups by id
```bash
llama-stack-client toolgroups get <toolgroup_id>
```
Shows detailed information about a specific toolgroup. If the toolgroup is not found, displays an error message.
Required arguments:
- `TOOLGROUP_ID`: ID of the tool group
### `llama-stack-client toolgroups register`
Register a new toolgroup at distribution endpoint
```bash
llama-stack-client toolgroups register <toolgroup_id> [--provider-id <provider-id>] [--provider-toolgroup-id <provider-toolgroup-id>] [--mcp-config <mcp-config>] [--args <args>]
```
Required arguments:
- `TOOLGROUP_ID`: ID of the tool group
Optional arguments:
- `--provider-id`: Provider ID for the toolgroup
- `--provider-toolgroup-id`: Provider's toolgroup ID
@ -278,6 +419,172 @@ Optional arguments:
- `--args`: JSON arguments for the toolgroup
### `llama-stack-client toolgroups unregister`
Unregister a toolgroup from distribution endpoint
```bash
llama-stack-client toolgroups unregister <toolgroup_id>
```
Required arguments:
- `TOOLGROUP_ID`: ID of the tool group
## Datasets Management
Manage datasets.
### `llama-stack-client datasets list`
Show available datasets on distribution endpoint
```bash
llama-stack-client datasets list
```
### `llama-stack-client datasets register`
```bash
llama-stack-client datasets register --dataset_id <dataset_id> --purpose <purpose> [--url <url] [--dataset-path <dataset-path>] [--dataset-id <dataset-id>] [--metadata <metadata>]
```
Required arguments:
- `--dataset_id`: Id of the dataset
- `--purpose`: Purpose of the dataset
Optional arguments:
- `--metadata`: Metadata of the dataset
- `--url`: URL of the dataset
- `--dataset-path`: Local file path to the dataset. If specified, upload dataset via URL
### `llama-stack-client datasets unregister`
Remove a dataset
```bash
llama-stack-client datasets unregister <dataset-id>
```
Required arguments:
- `DATASET_ID`: Id of the dataset
## Scoring Functions Management
Manage scoring functions.
### `llama-stack-client scoring_functions list`
Show available scoring functions on distribution endpoint
```bash
llama-stack-client scoring_functions list
```
```
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┓
┃ identifier ┃ provider_id ┃ description ┃ type ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━┩
│ basic::bfcl │ basic │ BFCL complex scoring │ scoring_function │
│ basic::docvqa │ basic │ DocVQA Visual Question & Answer scoring function │ scoring_function │
│ basic::equality │ basic │ Returns 1.0 if the input is equal to the target, 0.0 │ scoring_function │
│ │ │ otherwise. │ │
└────────────────────────────────────────────┴──────────────┴───────────────────────────────────────────────────────────────┴──────────────────┘
```
### `llama-stack-client scoring_functions register`
Register a new scoring function
```bash
llama-stack-client scoring_functions register --scoring-fn-id <scoring-fn-id> --description <description> --return-type <return-type> [--provider-id <provider-id>] [--provider-scoring-fn-id <provider-scoring-fn-id>] [--params <params>]
```
Required arguments:
- `--scoring-fn-id`: Id of the scoring function
- `--description`: Description of the scoring function
- `--return-type`: Return type of the scoring function
Optional arguments:
- `--provider-id`: Provider ID for the scoring function
- `--provider-scoring-fn-id`: Provider's scoring function ID
- `--params`: Parameters for the scoring function in JSON format
## Post Training Management
Post-training.
### `llama-stack-client post_training list`
Show the list of available post training jobs
```bash
llama-stack-client post_training list
```
```bash
["job-1", "job-2", "job-3"]
```
### `llama-stack-client post_training artifacts`
Get the training artifacts of a specific post training job
```bash
llama-stack-client post_training artifacts --job-uuid <job-uuid>
```
```bash
JobArtifactsResponse(checkpoints=[], job_uuid='job-1')
```
Required arguments:
- `--job-uuid`: Job UUID
### `llama-stack-client post_training supervised_fine_tune`
Kick off a supervised fine tune job
```bash
llama-stack-client post_training supervised_fine_tune --job-uuid <job-uuid> --model <model> --algorithm-config <algorithm-config> --training-config <training-config> [--checkpoint-dir <checkpoint-dir>]
```
Required arguments:
- `--job-uuid`: Job UUID
- `--model`: Model ID
- `--algorithm-config`: Algorithm Config
- `--training-config`: Training Config
Optional arguments:
- `--checkpoint-dir`: Checkpoint Config
### `llama-stack-client post_training status`
Show the status of a specific post training job
```bash
llama-stack-client post_training status --job-uuid <job-uuid>
```
```bash
JobStatusResponse(
checkpoints=[],
job_uuid='job-1',
status='completed',
completed_at="",
resources_allocated="",
scheduled_at="",
started_at=""
)
```
Required arguments:
- `--job-uuid`: Job UUID
### `llama-stack-client post_training cancel`
Cancel the training job
```bash
llama-stack-client post_training cancel --job-uuid <job-uuid>
```
```bash
# This functionality is not yet implemented for llama-stack-client
╭────────────────────────────────────────────────────────────╮
│ Failed to post_training cancel_training_job │
│ │
│ Error Type: InternalServerError │
│ Details: Error code: 501 - {'detail': 'Not implemented: '} │
╰────────────────────────────────────────────────────────────╯
```
Required arguments:
- `--job-uuid`: Job UUID

16
llama_stack/apis/tools/rag_tool.py
View file

@ -87,6 +87,20 @@ class RAGQueryGenerator(Enum):
custom = "custom"
@json_schema_type
class RAGSearchMode(Enum):
"""
Search modes for RAG query retrieval:
- VECTOR: Uses vector similarity search for semantic matching
- KEYWORD: Uses keyword-based search for exact matching
- HYBRID: Combines both vector and keyword search for better results
"""
VECTOR = "vector"
KEYWORD = "keyword"
HYBRID = "hybrid"
@json_schema_type
class DefaultRAGQueryGeneratorConfig(BaseModel):
type: Literal["default"] = "default"
@ -128,7 +142,7 @@ class RAGQueryConfig(BaseModel):
max_tokens_in_context: int = 4096
max_chunks: int = 5
chunk_template: str = "Result {index}\nContent: {chunk.content}\nMetadata: {metadata}\n"
mode: str | None = None
mode: RAGSearchMode | None = RAGSearchMode.VECTOR
ranker: Ranker | None = Field(default=None) # Only used for hybrid mode
@field_validator("chunk_template")

3
llama_stack/apis/vector_dbs/vector_dbs.py
View file

@ -19,6 +19,7 @@ class VectorDB(Resource):
embedding_model: str
embedding_dimension: int
vector_db_name: str | None = None
@property
def vector_db_id(self) -> str:
@ -70,6 +71,7 @@ class VectorDBs(Protocol):
embedding_model: str,
embedding_dimension: int | None = 384,
provider_id: str | None = None,
vector_db_name: str | None = None,
provider_vector_db_id: str | None = None,
) -> VectorDB:
"""Register a vector database.
@ -78,6 +80,7 @@ class VectorDBs(Protocol):
:param embedding_model: The embedding model to use.
:param embedding_dimension: The dimension of the embedding model.
:param provider_id: The identifier of the provider.
:param vector_db_name: The name of the vector database.
:param provider_vector_db_id: The identifier of the vector database in the provider.
:returns: A VectorDB.
"""

2
llama_stack/apis/vector_io/vector_io.py
View file

@ -346,7 +346,6 @@ class VectorIO(Protocol):
embedding_model: str | None = None,
embedding_dimension: int | None = 384,
provider_id: str | None = None,
provider_vector_db_id: str | None = None,
) -> VectorStoreObject:
"""Creates a vector store.
@ -358,7 +357,6 @@ class VectorIO(Protocol):
:param embedding_model: The embedding model to use for this vector store.
:param embedding_dimension: The dimension of the embedding vectors (default: 384).
:param provider_id: The ID of the provider to use for this vector store.
:param provider_vector_db_id: The provider-specific vector database ID.
:returns: A VectorStoreObject representing the created vector store.
"""
...

24
llama_stack/cli/stack/_build.py
View file

@ -93,7 +93,7 @@ def run_stack_build_command(args: argparse.Namespace) -> None:
)
sys.exit(1)
elif args.providers:
providers = dict()
providers_list: dict[str, str | list[str]] = dict()
for api_provider in args.providers.split(","):
if "=" not in api_provider:
cprint(
@ -112,7 +112,15 @@ def run_stack_build_command(args: argparse.Namespace) -> None:
)
sys.exit(1)
if provider in providers_for_api:
providers.setdefault(api, []).append(provider)
if api not in providers_list:
providers_list[api] = []
# Use type guarding to ensure we have a list
provider_value = providers_list[api]
if isinstance(provider_value, list):
provider_value.append(provider)
else:
# Convert string to list and append
providers_list[api] = [provider_value, provider]
else:
cprint(
f"{provider} is not a valid provider for the {api} API.",
@ -121,7 +129,7 @@ def run_stack_build_command(args: argparse.Namespace) -> None:
)
sys.exit(1)
distribution_spec = DistributionSpec(
providers=providers,
providers=providers_list,
description=",".join(args.providers),
)
if not args.image_type:
@ -182,7 +190,7 @@ def run_stack_build_command(args: argparse.Namespace) -> None:
cprint("Tip: use <TAB> to see options for the providers.\n", color="green", file=sys.stderr)
providers = dict()
providers: dict[str, str | list[str]] = dict()
for api, providers_for_api in get_provider_registry().items():
available_providers = [x for x in providers_for_api.keys() if x not in ("remote", "remote::sample")]
if not available_providers:
@ -371,10 +379,16 @@ def _run_stack_build_command_from_build_config(
if not image_name:
raise ValueError("Please specify an image name when building a venv image")
# At this point, image_name should be guaranteed to be a string
if image_name is None:
raise ValueError("image_name should not be None after validation")
if template_name:
build_dir = DISTRIBS_BASE_DIR / template_name
build_file_path = build_dir / f"{template_name}-build.yaml"
else:
if image_name is None:
raise ValueError("image_name cannot be None")
build_dir = DISTRIBS_BASE_DIR / image_name
build_file_path = build_dir / f"{image_name}-build.yaml"
@ -395,7 +409,7 @@ def _run_stack_build_command_from_build_config(
build_file_path,
image_name,
template_or_config=template_name or config_path or str(build_file_path),
run_config=run_config_file,
run_config=run_config_file.as_posix() if run_config_file else None,
)
if return_code != 0:
raise RuntimeError(f"Failed to build image {image_name}")

42
llama_stack/cli/stack/run.py
View file

@ -83,22 +83,13 @@ class StackRun(Subcommand):
return ImageType.CONDA.value, args.image_name
return args.image_type, args.image_name
def _run_stack_run_cmd(self, args: argparse.Namespace) -> None:
import yaml
from llama_stack.distribution.configure import parse_and_maybe_upgrade_config
def _resolve_config_and_template(self, args: argparse.Namespace) -> tuple[Path | None, str | None]:
"""Resolve config file path and template name from args.config"""
from llama_stack.distribution.utils.config_dirs import DISTRIBS_BASE_DIR
from llama_stack.distribution.utils.exec import formulate_run_args, run_command
if args.enable_ui:
self._start_ui_development_server(args.port)
image_type, image_name = self._get_image_type_and_name(args)
if not args.config:
return None, None
# Check if config is required based on image type
if (image_type in [ImageType.CONDA.value, ImageType.VENV.value]) and not args.config:
self.parser.error("Config file is required for venv and conda environments")
if args.config:
config_file = Path(args.config)
has_yaml_suffix = args.config.endswith(".yaml")
template_name = None
@ -123,6 +114,26 @@ class StackRun(Subcommand):
f"Config file must be a valid file path, '{config_file}' is not a file: type={type(config_file)}"
)
return config_file, template_name
def _run_stack_run_cmd(self, args: argparse.Namespace) -> None:
import yaml
from llama_stack.distribution.configure import parse_and_maybe_upgrade_config
from llama_stack.distribution.utils.exec import formulate_run_args, run_command
if args.enable_ui:
self._start_ui_development_server(args.port)
image_type, image_name = self._get_image_type_and_name(args)
# Resolve config file and template name first
config_file, template_name = self._resolve_config_and_template(args)
# Check if config is required based on image type
if (image_type in [ImageType.CONDA.value, ImageType.VENV.value]) and not config_file:
self.parser.error("Config file is required for venv and conda environments")
if config_file:
logger.info(f"Using run configuration: {config_file}")
try:
@ -138,8 +149,6 @@ class StackRun(Subcommand):
self.parser.error(f"failed to parse config file '{config_file}':\n {e}")
else:
config = None
config_file = None
template_name = None
# If neither image type nor image name is provided, assume the server should be run directly
# using the current environment packages.
@ -172,9 +181,6 @@ class StackRun(Subcommand):
run_args.extend([str(args.port)])
if config_file:
if template_name:
run_args.extend(["--template", str(template_name)])
else:
run_args.extend(["--config", str(config_file)])
if args.env:

2
llama_stack/distribution/access_control/access_control.py
View file

@ -81,7 +81,7 @@ def is_action_allowed(
if not len(policy):
policy = default_policy()
qualified_resource_id = resource.type + "::" + resource.identifier
qualified_resource_id = f"{resource.type}::{resource.identifier}"
for rule in policy:
if rule.forbid and matches_scope(rule.forbid, action, qualified_resource_id, user.principal):
if rule.when:

4
llama_stack/distribution/build_container.sh
View file

@ -96,7 +96,7 @@ FROM $container_base
WORKDIR /app
# We install the Python 3.12 dev headers and build tools so that any
# C‑extension wheels (e.g. polyleven, faiss‑cpu) can compile successfully.
# C-extension wheels (e.g. polyleven, faiss-cpu) can compile successfully.
RUN dnf -y update && dnf install -y iputils git net-tools wget \
vim-minimal python3.12 python3.12-pip python3.12-wheel \
@ -169,7 +169,7 @@ if [ -n "$run_config" ]; then
echo "Copying external providers directory: $external_providers_dir"
cp -r "$external_providers_dir" "$BUILD_CONTEXT_DIR/providers.d"
add_to_container << EOF
COPY --chmod=g+w providers.d /.llama/providers.d
COPY providers.d /.llama/providers.d
EOF
fi

8
llama_stack/distribution/configure.py
View file

@ -17,7 +17,7 @@ from llama_stack.distribution.distribution import (
builtin_automatically_routed_apis,
get_provider_registry,
)
from llama_stack.distribution.stack import replace_env_vars
from llama_stack.distribution.stack import cast_image_name_to_string, replace_env_vars
from llama_stack.distribution.utils.config_dirs import EXTERNAL_PROVIDERS_DIR
from llama_stack.distribution.utils.dynamic import instantiate_class_type
from llama_stack.distribution.utils.prompt_for_config import prompt_for_config
@ -164,7 +164,8 @@ def upgrade_from_routing_table(
def parse_and_maybe_upgrade_config(config_dict: dict[str, Any]) -> StackRunConfig:
version = config_dict.get("version", None)
if version == LLAMA_STACK_RUN_CONFIG_VERSION:
return StackRunConfig(**replace_env_vars(config_dict))
processed_config_dict = replace_env_vars(config_dict)
return StackRunConfig(**cast_image_name_to_string(processed_config_dict))
if "routing_table" in config_dict:
logger.info("Upgrading config...")
@ -175,4 +176,5 @@ def parse_and_maybe_upgrade_config(config_dict: dict[str, Any]) -> StackRunConfi
if not config_dict.get("external_providers_dir", None):
config_dict["external_providers_dir"] = EXTERNAL_PROVIDERS_DIR
return StackRunConfig(**replace_env_vars(config_dict))
processed_config_dict = replace_env_vars(config_dict)
return StackRunConfig(**cast_image_name_to_string(processed_config_dict))

20
llama_stack/distribution/routers/vector_io.py
View file

@ -5,6 +5,7 @@
# the root directory of this source tree.
import asyncio
import uuid
from typing import Any
from llama_stack.apis.common.content_types import (
@ -81,6 +82,7 @@ class VectorIORouter(VectorIO):
embedding_model: str,
embedding_dimension: int | None = 384,
provider_id: str | None = None,
vector_db_name: str | None = None,
provider_vector_db_id: str | None = None,
) -> None:
logger.debug(f"VectorIORouter.register_vector_db: {vector_db_id}, {embedding_model}")
@ -89,6 +91,7 @@ class VectorIORouter(VectorIO):
embedding_model,
embedding_dimension,
provider_id,
vector_db_name,
provider_vector_db_id,
)
@ -123,7 +126,6 @@ class VectorIORouter(VectorIO):
embedding_model: str | None = None,
embedding_dimension: int | None = None,
provider_id: str | None = None,
provider_vector_db_id: str | None = None,
) -> VectorStoreObject:
logger.debug(f"VectorIORouter.openai_create_vector_store: name={name}, provider_id={provider_id}")
@ -135,17 +137,17 @@ class VectorIORouter(VectorIO):
embedding_model, embedding_dimension = embedding_model_info
logger.info(f"No embedding model specified, using first available: {embedding_model}")
vector_db_id = name
vector_db_id = f"vs_{uuid.uuid4()}"
registered_vector_db = await self.routing_table.register_vector_db(
vector_db_id,
embedding_model,
embedding_dimension,
provider_id,
provider_vector_db_id,
vector_db_id=vector_db_id,
embedding_model=embedding_model,
embedding_dimension=embedding_dimension,
provider_id=provider_id,
provider_vector_db_id=vector_db_id,
vector_db_name=name,
)
return await self.routing_table.get_provider_impl(registered_vector_db.identifier).openai_create_vector_store(
vector_db_id,
name=name,
file_ids=file_ids,
expires_after=expires_after,
chunking_strategy=chunking_strategy,

2
llama_stack/distribution/routing_tables/vector_dbs.py
View file

@ -36,6 +36,7 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
embedding_dimension: int | None = 384,
provider_id: str | None = None,
provider_vector_db_id: str | None = None,
vector_db_name: str | None = None,
) -> VectorDB:
if provider_vector_db_id is None:
provider_vector_db_id = vector_db_id
@ -62,6 +63,7 @@ class VectorDBsRoutingTable(CommonRoutingTableImpl, VectorDBs):
"provider_resource_id": provider_vector_db_id,
"embedding_model": embedding_model,
"embedding_dimension": model.metadata["embedding_dimension"],
"vector_db_name": vector_db_name,
}
vector_db = TypeAdapter(VectorDBWithOwner).validate_python(vector_db_data)
await self.register_object(vector_db)

5
llama_stack/distribution/server/server.py
View file

@ -47,6 +47,7 @@ from llama_stack.distribution.server.routes import (
initialize_route_impls,
)
from llama_stack.distribution.stack import (
cast_image_name_to_string,
construct_stack,
replace_env_vars,
validate_env_pair,
@ -439,13 +440,13 @@ def main(args: argparse.Namespace | None = None):
logger.error(f"Error: {str(e)}")
sys.exit(1)
config = replace_env_vars(config_contents)
config = StackRunConfig(**config)
config = StackRunConfig(**cast_image_name_to_string(config))
# now that the logger is initialized, print the line about which type of config we are using.
logger.info(log_line)
logger.info("Run configuration:")
safe_config = redact_sensitive_fields(config.model_dump())
safe_config = redact_sensitive_fields(config.model_dump(mode="json"))
logger.info(yaml.dump(safe_config, indent=2))
app = FastAPI(

13
llama_stack/distribution/stack.py
View file

@ -98,6 +98,7 @@ async def register_resources(run_config: StackRunConfig, impls: dict[Api, Any]):
method = getattr(impls[api], register_method)
for obj in objects:
logger.debug(f"registering {rsrc.capitalize()} {obj} for provider {obj.provider_id}")
# Do not register models on disabled providers
if hasattr(obj, "provider_id") and obj.provider_id is not None and obj.provider_id == "__disabled__":
logger.debug(f"Skipping {rsrc.capitalize()} registration for disabled provider.")
@ -112,6 +113,11 @@ async def register_resources(run_config: StackRunConfig, impls: dict[Api, Any]):
):
logger.debug(f"Skipping {rsrc.capitalize()} registration for disabled model.")
continue
if hasattr(obj, "shield_id") and obj.shield_id is not None and obj.shield_id == "__disabled__":
logger.debug(f"Skipping {rsrc.capitalize()} registration for disabled shield.")
continue
# we want to maintain the type information in arguments to method.
# instead of method(**obj.model_dump()), which may convert a typed attr to a dict,
# we use model_dump() to find all the attrs and then getattr to get the still typed value.
@ -261,6 +267,13 @@ def _convert_string_to_proper_type(value: str) -> Any:
return value
def cast_image_name_to_string(config_dict: dict[str, Any]) -> dict[str, Any]:
"""Ensure that any value for a key 'image_name' in a config_dict is a string"""
if "image_name" in config_dict and config_dict["image_name"] is not None:
config_dict["image_name"] = str(config_dict["image_name"])
return config_dict
def validate_env_pair(env_pair: str) -> tuple[str, str]:
"""Validate and split an environment variable key-value pair."""
try:

5
llama_stack/distribution/utils/context.py
View file

@ -6,12 +6,9 @@
from collections.abc import AsyncGenerator
from contextvars import ContextVar
from typing import TypeVar
T = TypeVar("T")
def preserve_contexts_async_generator(
def preserve_contexts_async_generator[T](
gen: AsyncGenerator[T, None], context_vars: list[ContextVar]
) -> AsyncGenerator[T, None]:
"""

3
llama_stack/providers/inline/files/localfs/files.py
View file

@ -51,6 +51,9 @@ class LocalfsFilesImpl(Files):
},
)
async def shutdown(self) -> None:
pass
def _generate_file_id(self) -> str:
"""Generate a unique file ID for OpenAI API."""
return f"file-{uuid.uuid4().hex}"

3
llama_stack/providers/inline/post_training/torchtune/post_training.py
View file

@ -123,7 +123,8 @@ class TorchtunePostTrainingImpl:
training_config: TrainingConfig,
hyperparam_search_config: dict[str, Any],
logger_config: dict[str, Any],
) -> PostTrainingJob: ...
) -> PostTrainingJob:
raise NotImplementedError()
async def get_training_jobs(self) -> ListPostTrainingJobsResponse:
return ListPostTrainingJobsResponse(

34
llama_stack/providers/inline/safety/llama_guard/llama_guard.py
View file

@ -146,10 +146,9 @@ class LlamaGuardSafetyImpl(Safety, ShieldsProtocolPrivate):
pass
async def register_shield(self, shield: Shield) -> None:
if shield.provider_resource_id not in LLAMA_GUARD_MODEL_IDS:
raise ValueError(
f"Unsupported Llama Guard type: {shield.provider_resource_id}. Allowed types: {LLAMA_GUARD_MODEL_IDS}"
)
# Allow any model to be registered as a shield
# The model will be validated during runtime when making inference calls
pass
async def run_shield(
self,
@ -167,11 +166,25 @@ class LlamaGuardSafetyImpl(Safety, ShieldsProtocolPrivate):
if len(messages) > 0 and messages[0].role != Role.user.value:
messages[0] = UserMessage(content=messages[0].content)
model = LLAMA_GUARD_MODEL_IDS[shield.provider_resource_id]
# Use the inference API's model resolution instead of hardcoded mappings
# This allows the shield to work with any registered model
model_id = shield.provider_resource_id
# Determine safety categories based on the model type
# For known Llama Guard models, use specific categories
if model_id in LLAMA_GUARD_MODEL_IDS:
# Use the mapped model for categories but the original model_id for inference
mapped_model = LLAMA_GUARD_MODEL_IDS[model_id]
safety_categories = MODEL_TO_SAFETY_CATEGORIES_MAP.get(mapped_model, DEFAULT_LG_V3_SAFETY_CATEGORIES)
else:
# For unknown models, use default Llama Guard 3 8B categories
safety_categories = DEFAULT_LG_V3_SAFETY_CATEGORIES + [CAT_CODE_INTERPRETER_ABUSE]
impl = LlamaGuardShield(
model=model,
model=model_id,
inference_api=self.inference_api,
excluded_categories=self.config.excluded_categories,
safety_categories=safety_categories,
)
return await impl.run(messages)
@ -183,20 +196,21 @@ class LlamaGuardShield:
model: str,
inference_api: Inference,
excluded_categories: list[str] | None = None,
safety_categories: list[str] | None = None,
):
if excluded_categories is None:
excluded_categories = []
if safety_categories is None:
safety_categories = []
assert len(excluded_categories) == 0 or all(
x in SAFETY_CATEGORIES_TO_CODE_MAP.values() for x in excluded_categories
), "Invalid categories in excluded categories. Expected format is ['S1', 'S2', ..]"
if model not in MODEL_TO_SAFETY_CATEGORIES_MAP:
raise ValueError(f"Unsupported model: {model}")
self.model = model
self.inference_api = inference_api
self.excluded_categories = excluded_categories
self.safety_categories = safety_categories
def check_unsafe_response(self, response: str) -> str | None:
match = re.match(r"^unsafe\n(.*)$", response)
@ -214,7 +228,7 @@ class LlamaGuardShield:
final_categories = []
all_categories = MODEL_TO_SAFETY_CATEGORIES_MAP[self.model]
all_categories = self.safety_categories
for cat in all_categories:
cat_code = SAFETY_CATEGORIES_TO_CODE_MAP[cat]
if cat_code in excluded_categories:

51
llama_stack/providers/inline/vector_io/faiss/faiss.py
View file

@ -181,8 +181,8 @@ class FaissVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtocolPr
)
self.cache[vector_db.identifier] = index
# Load existing OpenAI vector stores using the mixin method
self.openai_vector_stores = await self._load_openai_vector_stores()
# Load existing OpenAI vector stores into the in-memory cache
await self.initialize_openai_vector_stores()
async def shutdown(self) -> None:
# Cleanup if needed
@ -261,42 +261,10 @@ class FaissVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtocolPr
return await index.query_chunks(query, params)
# OpenAI Vector Store Mixin abstract method implementations
async def _save_openai_vector_store(self, store_id: str, store_info: dict[str, Any]) -> None:
"""Save vector store metadata to kvstore."""
assert self.kvstore is not None
key = f"{OPENAI_VECTOR_STORES_PREFIX}{store_id}"
await self.kvstore.set(key=key, value=json.dumps(store_info))
async def _load_openai_vector_stores(self) -> dict[str, dict[str, Any]]:
"""Load all vector store metadata from kvstore."""
assert self.kvstore is not None
start_key = OPENAI_VECTOR_STORES_PREFIX
end_key = f"{OPENAI_VECTOR_STORES_PREFIX}\xff"
stored_openai_stores = await self.kvstore.values_in_range(start_key, end_key)
stores = {}
for store_data in stored_openai_stores:
store_info = json.loads(store_data)
stores[store_info["id"]] = store_info
return stores
async def _update_openai_vector_store(self, store_id: str, store_info: dict[str, Any]) -> None:
"""Update vector store metadata in kvstore."""
assert self.kvstore is not None
key = f"{OPENAI_VECTOR_STORES_PREFIX}{store_id}"
await self.kvstore.set(key=key, value=json.dumps(store_info))
async def _delete_openai_vector_store_from_storage(self, store_id: str) -> None:
"""Delete vector store metadata from kvstore."""
assert self.kvstore is not None
key = f"{OPENAI_VECTOR_STORES_PREFIX}{store_id}"
await self.kvstore.delete(key)
async def _save_openai_vector_store_file(
self, store_id: str, file_id: str, file_info: dict[str, Any], file_contents: list[dict[str, Any]]
) -> None:
"""Save vector store file metadata to kvstore."""
"""Save vector store file data to kvstore."""
assert self.kvstore is not None
key = f"{OPENAI_VECTOR_STORES_FILES_PREFIX}{store_id}:{file_id}"
await self.kvstore.set(key=key, value=json.dumps(file_info))
@ -324,7 +292,16 @@ class FaissVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtocolPr
await self.kvstore.set(key=key, value=json.dumps(file_info))
async def _delete_openai_vector_store_file_from_storage(self, store_id: str, file_id: str) -> None:
"""Delete vector store file metadata from kvstore."""
"""Delete vector store data from kvstore."""
assert self.kvstore is not None
key = f"{OPENAI_VECTOR_STORES_FILES_PREFIX}{store_id}:{file_id}"
keys_to_delete = [
f"{OPENAI_VECTOR_STORES_FILES_PREFIX}{store_id}:{file_id}",
f"{OPENAI_VECTOR_STORES_FILES_CONTENTS_PREFIX}{store_id}:{file_id}",
]
for key in keys_to_delete:
try:
await self.kvstore.delete(key)
except Exception as e:
logger.warning(f"Failed to delete key {key}: {e}")
continue

2
llama_stack/providers/inline/vector_io/milvus/config.py
View file

@ -18,7 +18,7 @@ from llama_stack.schema_utils import json_schema_type
@json_schema_type
class MilvusVectorIOConfig(BaseModel):
db_path: str
kvstore: KVStoreConfig
kvstore: KVStoreConfig = Field(description="Config for KV store backend (SQLite only for now)")
consistency_level: str = Field(description="The consistency level of the Milvus server", default="Strong")
@classmethod

14
llama_stack/providers/inline/vector_io/sqlite_vec/config.py
View file

@ -6,14 +6,24 @@
from typing import Any
from pydantic import BaseModel
from pydantic import BaseModel, Field
from llama_stack.providers.utils.kvstore.config import (
KVStoreConfig,
SqliteKVStoreConfig,
)
class SQLiteVectorIOConfig(BaseModel):
db_path: str
db_path: str = Field(description="Path to the SQLite database file")
kvstore: KVStoreConfig = Field(description="Config for KV store backend (SQLite only for now)")
@classmethod
def sample_run_config(cls, __distro_dir__: str) -> dict[str, Any]:
return {
"db_path": "${env.SQLITE_STORE_DIR:=" + __distro_dir__ + "}/" + "sqlite_vec.db",
"kvstore": SqliteKVStoreConfig.sample_run_config(
__distro_dir__=__distro_dir__,
db_name="sqlite_vec_registry.db",
),
}

319
llama_stack/providers/inline/vector_io/sqlite_vec/sqlite_vec.py
View file

@ -7,6 +7,7 @@
import asyncio
import json
import logging
import re
import sqlite3
import struct
from typing import Any
@ -24,6 +25,8 @@ from llama_stack.apis.vector_io import (
VectorIO,
)
from llama_stack.providers.datatypes import VectorDBsProtocolPrivate
from llama_stack.providers.utils.kvstore import kvstore_impl
from llama_stack.providers.utils.kvstore.api import KVStore
from llama_stack.providers.utils.memory.openai_vector_store_mixin import OpenAIVectorStoreMixin
from llama_stack.providers.utils.memory.vector_store import (
RERANKER_TYPE_RRF,
@ -40,6 +43,13 @@ KEYWORD_SEARCH = "keyword"
HYBRID_SEARCH = "hybrid"
SEARCH_MODES = {VECTOR_SEARCH, KEYWORD_SEARCH, HYBRID_SEARCH}
VERSION = "v3"
VECTOR_DBS_PREFIX = f"vector_dbs:sqlite_vec:{VERSION}::"
VECTOR_INDEX_PREFIX = f"vector_index:sqlite_vec:{VERSION}::"
OPENAI_VECTOR_STORES_PREFIX = f"openai_vector_stores:sqlite_vec:{VERSION}::"
OPENAI_VECTOR_STORES_FILES_PREFIX = f"openai_vector_stores_files:sqlite_vec:{VERSION}::"
OPENAI_VECTOR_STORES_FILES_CONTENTS_PREFIX = f"openai_vector_stores_files_contents:sqlite_vec:{VERSION}::"
def serialize_vector(vector: list[float]) -> bytes:
"""Serialize a list of floats into a compact binary representation."""
@ -108,6 +118,10 @@ def _rrf_rerank(
return rrf_scores
def _make_sql_identifier(name: str) -> str:
return re.sub(r"[^a-zA-Z0-9_]", "_", name)
class SQLiteVecIndex(EmbeddingIndex):
"""
An index implementation that stores embeddings in a SQLite virtual table using sqlite-vec.
@ -117,13 +131,14 @@ class SQLiteVecIndex(EmbeddingIndex):
- An FTS5 table (fts_chunks_{bank_id}) for full-text keyword search.
"""
def __init__(self, dimension: int, db_path: str, bank_id: str):
def __init__(self, dimension: int, db_path: str, bank_id: str, kvstore: KVStore | None = None):
self.dimension = dimension
self.db_path = db_path
self.bank_id = bank_id
self.metadata_table = f"chunks_{bank_id}".replace("-", "_")
self.vector_table = f"vec_chunks_{bank_id}".replace("-", "_")
self.fts_table = f"fts_chunks_{bank_id}".replace("-", "_")
self.metadata_table = _make_sql_identifier(f"chunks_{bank_id}")
self.vector_table = _make_sql_identifier(f"vec_chunks_{bank_id}")
self.fts_table = _make_sql_identifier(f"fts_chunks_{bank_id}")
self.kvstore = kvstore
@classmethod
async def create(cls, dimension: int, db_path: str, bank_id: str):
@ -138,14 +153,14 @@ class SQLiteVecIndex(EmbeddingIndex):
try:
# Create the table to store chunk metadata.
cur.execute(f"""
CREATE TABLE IF NOT EXISTS {self.metadata_table} (
CREATE TABLE IF NOT EXISTS [{self.metadata_table}] (
id TEXT PRIMARY KEY,
chunk TEXT
);
""")
# Create the virtual table for embeddings.
cur.execute(f"""
CREATE VIRTUAL TABLE IF NOT EXISTS {self.vector_table}
CREATE VIRTUAL TABLE IF NOT EXISTS [{self.vector_table}]
USING vec0(embedding FLOAT[{self.dimension}], id TEXT);
""")
connection.commit()
@ -153,7 +168,7 @@ class SQLiteVecIndex(EmbeddingIndex):
# based on query. Implementation of the change on client side will allow passing the search_mode option
# during initialization to make it easier to create the table that is required.
cur.execute(f"""
CREATE VIRTUAL TABLE IF NOT EXISTS {self.fts_table}
CREATE VIRTUAL TABLE IF NOT EXISTS [{self.fts_table}]
USING fts5(id, content);
""")
connection.commit()
@ -168,9 +183,9 @@ class SQLiteVecIndex(EmbeddingIndex):
connection = _create_sqlite_connection(self.db_path)
cur = connection.cursor()
try:
cur.execute(f"DROP TABLE IF EXISTS {self.metadata_table};")
cur.execute(f"DROP TABLE IF EXISTS {self.vector_table};")
cur.execute(f"DROP TABLE IF EXISTS {self.fts_table};")
cur.execute(f"DROP TABLE IF EXISTS [{self.metadata_table}];")
cur.execute(f"DROP TABLE IF EXISTS [{self.vector_table}];")
cur.execute(f"DROP TABLE IF EXISTS [{self.fts_table}];")
connection.commit()
finally:
cur.close()
@ -202,7 +217,7 @@ class SQLiteVecIndex(EmbeddingIndex):
metadata_data = [(chunk.chunk_id, chunk.model_dump_json()) for chunk in batch_chunks]
cur.executemany(
f"""
INSERT INTO {self.metadata_table} (id, chunk)
INSERT INTO [{self.metadata_table}] (id, chunk)
VALUES (?, ?)
ON CONFLICT(id) DO UPDATE SET chunk = excluded.chunk;
""",
@ -220,7 +235,7 @@ class SQLiteVecIndex(EmbeddingIndex):
for chunk, emb in zip(batch_chunks, batch_embeddings, strict=True)
]
cur.executemany(
f"INSERT INTO {self.vector_table} (id, embedding) VALUES (?, ?);",
f"INSERT INTO [{self.vector_table}] (id, embedding) VALUES (?, ?);",
embedding_data,
)
@ -228,13 +243,13 @@ class SQLiteVecIndex(EmbeddingIndex):
fts_data = [(chunk.chunk_id, chunk.content) for chunk in batch_chunks]
# DELETE existing entries with same IDs (FTS5 doesn't support ON CONFLICT)
cur.executemany(
f"DELETE FROM {self.fts_table} WHERE id = ?;",
f"DELETE FROM [{self.fts_table}] WHERE id = ?;",
[(row[0],) for row in fts_data],
)
# INSERT new entries
cur.executemany(
f"INSERT INTO {self.fts_table} (id, content) VALUES (?, ?);",
f"INSERT INTO [{self.fts_table}] (id, content) VALUES (?, ?);",
fts_data,
)
@ -270,8 +285,8 @@ class SQLiteVecIndex(EmbeddingIndex):
emb_blob = serialize_vector(emb_list)
query_sql = f"""
SELECT m.id, m.chunk, v.distance
FROM {self.vector_table} AS v
JOIN {self.metadata_table} AS m ON m.id = v.id
FROM [{self.vector_table}] AS v
JOIN [{self.metadata_table}] AS m ON m.id = v.id
WHERE v.embedding MATCH ? AND k = ?
ORDER BY v.distance;
"""
@ -312,9 +327,9 @@ class SQLiteVecIndex(EmbeddingIndex):
cur = connection.cursor()
try:
query_sql = f"""
SELECT DISTINCT m.id, m.chunk, bm25({self.fts_table}) AS score
FROM {self.fts_table} AS f
JOIN {self.metadata_table} AS m ON m.id = f.id
SELECT DISTINCT m.id, m.chunk, bm25([{self.fts_table}]) AS score
FROM [{self.fts_table}] AS f
JOIN [{self.metadata_table}] AS m ON m.id = f.id
WHERE f.content MATCH ?
ORDER BY score ASC
LIMIT ?;
@ -425,27 +440,81 @@ class SQLiteVecVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtoc
self.files_api = files_api
self.cache: dict[str, VectorDBWithIndex] = {}
self.openai_vector_stores: dict[str, dict[str, Any]] = {}
self.kvstore: KVStore | None = None
async def initialize(self) -> None:
def _setup_connection():
# Open a connection to the SQLite database (the file is specified in the config).
self.kvstore = await kvstore_impl(self.config.kvstore)
start_key = VECTOR_DBS_PREFIX
end_key = f"{VECTOR_DBS_PREFIX}\xff"
stored_vector_dbs = await self.kvstore.values_in_range(start_key, end_key)
for db_json in stored_vector_dbs:
vector_db = VectorDB.model_validate_json(db_json)
index = await SQLiteVecIndex.create(
vector_db.embedding_dimension,
self.config.db_path,
vector_db.identifier,
)
self.cache[vector_db.identifier] = VectorDBWithIndex(vector_db, index, self.inference_api)
# Load existing OpenAI vector stores into the in-memory cache
await self.initialize_openai_vector_stores()
async def shutdown(self) -> None:
# nothing to do since we don't maintain a persistent connection
pass
async def list_vector_dbs(self) -> list[VectorDB]:
return [v.vector_db for v in self.cache.values()]
async def register_vector_db(self, vector_db: VectorDB) -> None:
index = await SQLiteVecIndex.create(
vector_db.embedding_dimension,
self.config.db_path,
vector_db.identifier,
)
self.cache[vector_db.identifier] = VectorDBWithIndex(vector_db, index, self.inference_api)
async def _get_and_cache_vector_db_index(self, vector_db_id: str) -> VectorDBWithIndex | None:
if vector_db_id in self.cache:
return self.cache[vector_db_id]
if self.vector_db_store is None:
raise ValueError(f"Vector DB {vector_db_id} not found")
vector_db = self.vector_db_store.get_vector_db(vector_db_id)
if not vector_db:
raise ValueError(f"Vector DB {vector_db_id} not found")
index = VectorDBWithIndex(
vector_db=vector_db,
index=SQLiteVecIndex(
dimension=vector_db.embedding_dimension,
db_path=self.config.db_path,
bank_id=vector_db.identifier,
kvstore=self.kvstore,
),
inference_api=self.inference_api,
)
self.cache[vector_db_id] = index
return index
async def unregister_vector_db(self, vector_db_id: str) -> None:
if vector_db_id not in self.cache:
logger.warning(f"Vector DB {vector_db_id} not found")
return
await self.cache[vector_db_id].index.delete()
del self.cache[vector_db_id]
async def _save_openai_vector_store_file(
self, store_id: str, file_id: str, file_info: dict[str, Any], file_contents: list[dict[str, Any]]
) -> None:
"""Save vector store file metadata to SQLite database."""
def _create_or_store():
connection = _create_sqlite_connection(self.config.db_path)
cur = connection.cursor()
try:
# Create a table to persist vector DB registrations.
cur.execute("""
CREATE TABLE IF NOT EXISTS vector_dbs (
id TEXT PRIMARY KEY,
metadata TEXT
);
""")
# Create a table to persist OpenAI vector stores.
cur.execute("""
CREATE TABLE IF NOT EXISTS openai_vector_stores (
id TEXT PRIMARY KEY,
metadata TEXT
);
""")
# Create a table to persist OpenAI vector store files.
cur.execute("""
CREATE TABLE IF NOT EXISTS openai_vector_store_files (
@ -464,168 +533,6 @@ class SQLiteVecVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtoc
);
""")
connection.commit()
# Load any existing vector DB registrations.
cur.execute("SELECT metadata FROM vector_dbs")
vector_db_rows = cur.fetchall()
return vector_db_rows
finally:
cur.close()
connection.close()
vector_db_rows = await asyncio.to_thread(_setup_connection)
# Load existing vector DBs
for row in vector_db_rows:
vector_db_data = row[0]
vector_db = VectorDB.model_validate_json(vector_db_data)
index = await SQLiteVecIndex.create(
vector_db.embedding_dimension,
self.config.db_path,
vector_db.identifier,
)
self.cache[vector_db.identifier] = VectorDBWithIndex(vector_db, index, self.inference_api)
# Load existing OpenAI vector stores using the mixin method
self.openai_vector_stores = await self._load_openai_vector_stores()
async def shutdown(self) -> None:
# nothing to do since we don't maintain a persistent connection
pass
async def register_vector_db(self, vector_db: VectorDB) -> None:
def _register_db():
connection = _create_sqlite_connection(self.config.db_path)
cur = connection.cursor()
try:
cur.execute(
"INSERT OR REPLACE INTO vector_dbs (id, metadata) VALUES (?, ?)",
(vector_db.identifier, vector_db.model_dump_json()),
)
connection.commit()
finally:
cur.close()
connection.close()
await asyncio.to_thread(_register_db)
index = await SQLiteVecIndex.create(
vector_db.embedding_dimension,
self.config.db_path,
vector_db.identifier,
)
self.cache[vector_db.identifier] = VectorDBWithIndex(vector_db, index, self.inference_api)
async def list_vector_dbs(self) -> list[VectorDB]:
return [v.vector_db for v in self.cache.values()]
async def unregister_vector_db(self, vector_db_id: str) -> None:
if vector_db_id not in self.cache:
logger.warning(f"Vector DB {vector_db_id} not found")
return
await self.cache[vector_db_id].index.delete()
del self.cache[vector_db_id]
def _delete_vector_db_from_registry():
connection = _create_sqlite_connection(self.config.db_path)
cur = connection.cursor()
try:
cur.execute("DELETE FROM vector_dbs WHERE id = ?", (vector_db_id,))
connection.commit()
finally:
cur.close()
connection.close()
await asyncio.to_thread(_delete_vector_db_from_registry)
# OpenAI Vector Store Mixin abstract method implementations
async def _save_openai_vector_store(self, store_id: str, store_info: dict[str, Any]) -> None:
"""Save vector store metadata to SQLite database."""
def _store():
connection = _create_sqlite_connection(self.config.db_path)
cur = connection.cursor()
try:
cur.execute(
"INSERT OR REPLACE INTO openai_vector_stores (id, metadata) VALUES (?, ?)",
(store_id, json.dumps(store_info)),
)
connection.commit()
except Exception as e:
logger.error(f"Error saving openai vector store {store_id}: {e}")
raise
finally:
cur.close()
connection.close()
try:
await asyncio.to_thread(_store)
except Exception as e:
logger.error(f"Error saving openai vector store {store_id}: {e}")
raise
async def _load_openai_vector_stores(self) -> dict[str, dict[str, Any]]:
"""Load all vector store metadata from SQLite database."""
def _load():
connection = _create_sqlite_connection(self.config.db_path)
cur = connection.cursor()
try:
cur.execute("SELECT metadata FROM openai_vector_stores")
rows = cur.fetchall()
return rows
finally:
cur.close()
connection.close()
rows = await asyncio.to_thread(_load)
stores = {}
for row in rows:
store_data = row[0]
store_info = json.loads(store_data)
stores[store_info["id"]] = store_info
return stores
async def _update_openai_vector_store(self, store_id: str, store_info: dict[str, Any]) -> None:
"""Update vector store metadata in SQLite database."""
def _update():
connection = _create_sqlite_connection(self.config.db_path)
cur = connection.cursor()
try:
cur.execute(
"UPDATE openai_vector_stores SET metadata = ? WHERE id = ?",
(json.dumps(store_info), store_id),
)
connection.commit()
finally:
cur.close()
connection.close()
await asyncio.to_thread(_update)
async def _delete_openai_vector_store_from_storage(self, store_id: str) -> None:
"""Delete vector store metadata from SQLite database."""
def _delete():
connection = _create_sqlite_connection(self.config.db_path)
cur = connection.cursor()
try:
cur.execute("DELETE FROM openai_vector_stores WHERE id = ?", (store_id,))
connection.commit()
finally:
cur.close()
connection.close()
await asyncio.to_thread(_delete)
async def _save_openai_vector_store_file(
self, store_id: str, file_id: str, file_info: dict[str, Any], file_contents: list[dict[str, Any]]
) -> None:
"""Save vector store file metadata to SQLite database."""
def _store():
connection = _create_sqlite_connection(self.config.db_path)
cur = connection.cursor()
try:
cur.execute(
"INSERT OR REPLACE INTO openai_vector_store_files (store_id, file_id, metadata) VALUES (?, ?, ?)",
(store_id, file_id, json.dumps(file_info)),
@ -643,7 +550,7 @@ class SQLiteVecVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtoc
connection.close()
try:
await asyncio.to_thread(_store)
await asyncio.to_thread(_create_or_store)
except Exception as e:
logger.error(f"Error saving openai vector store file {store_id} {file_id}: {e}")
raise
@ -722,6 +629,10 @@ class SQLiteVecVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtoc
cur.execute(
"DELETE FROM openai_vector_store_files WHERE store_id = ? AND file_id = ?", (store_id, file_id)
)
cur.execute(
"DELETE FROM openai_vector_store_files_contents WHERE store_id = ? AND file_id = ?",
(store_id, file_id),
)
connection.commit()
finally:
cur.close()
@ -730,15 +641,17 @@ class SQLiteVecVectorIOAdapter(OpenAIVectorStoreMixin, VectorIO, VectorDBsProtoc
await asyncio.to_thread(_delete)
async def insert_chunks(self, vector_db_id: str, chunks: list[Chunk], ttl_seconds: int | None = None) -> None:
if vector_db_id not in self.cache:
raise ValueError(f"Vector DB {vector_db_id} not found. Found: {list(self.cache.keys())}")
index = await self._get_and_cache_vector_db_index(vector_db_id)
if not index:
raise ValueError(f"Vector DB {vector_db_id} not found")
# The VectorDBWithIndex helper is expected to compute embeddings via the inference_api
# and then call our index's add_chunks.
await self.cache[vector_db_id].insert_chunks(chunks)
await index.insert_chunks(chunks)
async def query_chunks(
self, vector_db_id: str, query: Any, params: dict[str, Any] | None = None
) -> QueryChunksResponse:
if vector_db_id not in self.cache:
index = await self._get_and_cache_vector_db_index(vector_db_id)
if not index:
raise ValueError(f"Vector DB {vector_db_id} not found")
return await self.cache[vector_db_id].query_chunks(query, params)
return await index.query_chunks(query, params)

7
llama_stack/providers/remote/inference/anthropic/models.py
View file

@ -15,8 +15,11 @@ LLM_MODEL_IDS = [
"anthropic/claude-3-5-haiku-latest",
]
SAFETY_MODELS_ENTRIES = []
MODEL_ENTRIES = [ProviderModelEntry(provider_model_id=m) for m in LLM_MODEL_IDS] + [
MODEL_ENTRIES = (
[ProviderModelEntry(provider_model_id=m) for m in LLM_MODEL_IDS]
+ [
ProviderModelEntry(
provider_model_id="anthropic/voyage-3",
model_type=ModelType.embedding,
@ -33,3 +36,5 @@ MODEL_ENTRIES = [ProviderModelEntry(provider_model_id=m) for m in LLM_MODEL_IDS]
metadata={"embedding_dimension": 1024, "context_length": 32000},
),
]
+ SAFETY_MODELS_ENTRIES
)

6
llama_stack/providers/remote/inference/bedrock/models.py
View file

@ -9,6 +9,10 @@ from llama_stack.providers.utils.inference.model_registry import (
build_hf_repo_model_entry,
)
SAFETY_MODELS_ENTRIES = []
# https://docs.aws.amazon.com/bedrock/latest/userguide/models-supported.html
MODEL_ENTRIES = [
build_hf_repo_model_entry(
"meta.llama3-1-8b-instruct-v1:0",
@ -22,4 +26,4 @@ MODEL_ENTRIES = [
"meta.llama3-1-405b-instruct-v1:0",
CoreModelId.llama3_1_405b_instruct.value,
),
]
] + SAFETY_MODELS_ENTRIES

9
llama_stack/providers/remote/inference/cerebras/models.py
View file

@ -9,6 +9,9 @@ from llama_stack.providers.utils.inference.model_registry import (
build_hf_repo_model_entry,
)
SAFETY_MODELS_ENTRIES = []
# https://inference-docs.cerebras.ai/models
MODEL_ENTRIES = [
build_hf_repo_model_entry(
"llama3.1-8b",
@ -18,4 +21,8 @@ MODEL_ENTRIES = [
"llama-3.3-70b",
CoreModelId.llama3_3_70b_instruct.value,
),
]
build_hf_repo_model_entry(
"llama-4-scout-17b-16e-instruct",
CoreModelId.llama4_scout_17b_16e_instruct.value,
),
] + SAFETY_MODELS_ENTRIES

9
llama_stack/providers/remote/inference/databricks/databricks.py
View file

@ -47,7 +47,10 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
from .config import DatabricksImplConfig
model_entries = [
SAFETY_MODELS_ENTRIES = []
# https://docs.databricks.com/aws/en/machine-learning/model-serving/foundation-model-overview
MODEL_ENTRIES = [
build_hf_repo_model_entry(
"databricks-meta-llama-3-1-70b-instruct",
CoreModelId.llama3_1_70b_instruct.value,
@ -56,7 +59,7 @@ model_entries = [
"databricks-meta-llama-3-1-405b-instruct",
CoreModelId.llama3_1_405b_instruct.value,
),
]
] + SAFETY_MODELS_ENTRIES
class DatabricksInferenceAdapter(
@ -66,7 +69,7 @@ class DatabricksInferenceAdapter(
OpenAICompletionToLlamaStackMixin,
):
def __init__(self, config: DatabricksImplConfig) -> None:
ModelRegistryHelper.__init__(self, model_entries=model_entries)
ModelRegistryHelper.__init__(self, model_entries=MODEL_ENTRIES)
self.config = config
async def initialize(self) -> None:

21
llama_stack/providers/remote/inference/fireworks/models.py
View file

@ -11,6 +11,17 @@ from llama_stack.providers.utils.inference.model_registry import (
build_hf_repo_model_entry,
)
SAFETY_MODELS_ENTRIES = [
build_hf_repo_model_entry(
"accounts/fireworks/models/llama-guard-3-8b",
CoreModelId.llama_guard_3_8b.value,
),
build_hf_repo_model_entry(
"accounts/fireworks/models/llama-guard-3-11b-vision",
CoreModelId.llama_guard_3_11b_vision.value,
),
]
MODEL_ENTRIES = [
build_hf_repo_model_entry(
"accounts/fireworks/models/llama-v3p1-8b-instruct",
@ -40,14 +51,6 @@ MODEL_ENTRIES = [
"accounts/fireworks/models/llama-v3p3-70b-instruct",
CoreModelId.llama3_3_70b_instruct.value,
),
build_hf_repo_model_entry(
"accounts/fireworks/models/llama-guard-3-8b",
CoreModelId.llama_guard_3_8b.value,
),
build_hf_repo_model_entry(
"accounts/fireworks/models/llama-guard-3-11b-vision",
CoreModelId.llama_guard_3_11b_vision.value,
),
build_hf_repo_model_entry(
"accounts/fireworks/models/llama4-scout-instruct-basic",
CoreModelId.llama4_scout_17b_16e_instruct.value,
@ -64,4 +67,4 @@ MODEL_ENTRIES = [
"context_length": 8192,
},
),
]
] + SAFETY_MODELS_ENTRIES

7
llama_stack/providers/remote/inference/gemini/models.py
View file

@ -17,11 +17,16 @@ LLM_MODEL_IDS = [
"gemini/gemini-2.5-pro",
]
SAFETY_MODELS_ENTRIES = []
MODEL_ENTRIES = [ProviderModelEntry(provider_model_id=m) for m in LLM_MODEL_IDS] + [
MODEL_ENTRIES = (
[ProviderModelEntry(provider_model_id=m) for m in LLM_MODEL_IDS]
+ [
ProviderModelEntry(
provider_model_id="gemini/text-embedding-004",
model_type=ModelType.embedding,
metadata={"embedding_dimension": 768, "context_length": 2048},
),
]
+ SAFETY_MODELS_ENTRIES
)

10
llama_stack/providers/remote/inference/groq/groq.py
View file

@ -38,24 +38,18 @@ class GroqInferenceAdapter(LiteLLMOpenAIMixin):
provider_data_api_key_field="groq_api_key",
)
self.config = config
self._openai_client = None
async def initialize(self):
await super().initialize()
async def shutdown(self):
await super().shutdown()
if self._openai_client:
await self._openai_client.close()
self._openai_client = None
def _get_openai_client(self) -> AsyncOpenAI:
if not self._openai_client:
self._openai_client = AsyncOpenAI(
return AsyncOpenAI(
base_url=f"{self.config.url}/openai/v1",
api_key=self.config.api_key,
api_key=self.get_api_key(),
)
return self._openai_client
async def openai_chat_completion(
self,

4
llama_stack/providers/remote/inference/groq/models.py
View file

@ -10,6 +10,8 @@ from llama_stack.providers.utils.inference.model_registry import (
build_model_entry,
)
SAFETY_MODELS_ENTRIES = []
MODEL_ENTRIES = [
build_hf_repo_model_entry(
"groq/llama3-8b-8192",
@ -51,4 +53,4 @@ MODEL_ENTRIES = [
"groq/meta-llama/llama-4-maverick-17b-128e-instruct",
CoreModelId.llama4_maverick_17b_128e_instruct.value,
),
]
] + SAFETY_MODELS_ENTRIES

5
llama_stack/providers/remote/inference/nvidia/models.py
View file

@ -11,6 +11,9 @@ from llama_stack.providers.utils.inference.model_registry import (
build_hf_repo_model_entry,
)
SAFETY_MODELS_ENTRIES = []
# https://docs.nvidia.com/nim/large-language-models/latest/supported-llm-agnostic-architectures.html
MODEL_ENTRIES = [
build_hf_repo_model_entry(
"meta/llama3-8b-instruct",
@ -99,4 +102,4 @@ MODEL_ENTRIES = [
),
# TODO(mf): how do we handle Nemotron models?
# "Llama3.1-Nemotron-51B-Instruct" -> "meta/llama-3.1-nemotron-51b-instruct",
]
] + SAFETY_MODELS_ENTRIES

41
llama_stack/providers/remote/inference/nvidia/nvidia.py
View file

@ -7,7 +7,6 @@
import logging
import warnings
from collections.abc import AsyncIterator
from functools import lru_cache
from typing import Any
from openai import APIConnectionError, AsyncOpenAI, BadRequestError
@ -93,42 +92,22 @@ class NVIDIAInferenceAdapter(Inference, ModelRegistryHelper):
self._config = config
@lru_cache # noqa: B019
def _get_client(self, provider_model_id: str) -> AsyncOpenAI:
@property
def _client(self) -> AsyncOpenAI:
"""
For hosted models, https://integrate.api.nvidia.com/v1 is the primary base_url. However,
some models are hosted on different URLs. This function returns the appropriate client
for the given provider_model_id.
Returns an OpenAI client for the configured NVIDIA API endpoint.
This relies on lru_cache and self._default_client to avoid creating a new client for each request
or for each model that is hosted on https://integrate.api.nvidia.com/v1.
:param provider_model_id: The provider model ID
:return: An OpenAI client
"""
@lru_cache # noqa: B019
def _get_client_for_base_url(base_url: str) -> AsyncOpenAI:
"""
Maintain a single OpenAI client per base_url.
"""
base_url = f"{self._config.url}/v1" if self._config.append_api_version else self._config.url
return AsyncOpenAI(
base_url=base_url,
api_key=(self._config.api_key.get_secret_value() if self._config.api_key else "NO KEY"),
timeout=self._config.timeout,
)
special_model_urls = {
"meta/llama-3.2-11b-vision-instruct": "https://ai.api.nvidia.com/v1/gr/meta/llama-3.2-11b-vision-instruct",
"meta/llama-3.2-90b-vision-instruct": "https://ai.api.nvidia.com/v1/gr/meta/llama-3.2-90b-vision-instruct",
}
base_url = f"{self._config.url}/v1" if self._config.append_api_version else self._config.url
if _is_nvidia_hosted(self._config) and provider_model_id in special_model_urls:
base_url = special_model_urls[provider_model_id]
return _get_client_for_base_url(base_url)
async def _get_provider_model_id(self, model_id: str) -> str:
if not self.model_store:
raise RuntimeError("Model store is not set")
@ -169,7 +148,7 @@ class NVIDIAInferenceAdapter(Inference, ModelRegistryHelper):
)
try:
response = await self._get_client(provider_model_id).completions.create(**request)
response = await self._client.completions.create(**request)
except APIConnectionError as e:
raise ConnectionError(f"Failed to connect to NVIDIA NIM at {self._config.url}: {e}") from e
@ -222,7 +201,7 @@ class NVIDIAInferenceAdapter(Inference, ModelRegistryHelper):
extra_body["input_type"] = task_type_options[task_type]
try:
response = await self._get_client(provider_model_id).embeddings.create(
response = await self._client.embeddings.create(
model=provider_model_id,
input=input,
extra_body=extra_body,
@ -283,7 +262,7 @@ class NVIDIAInferenceAdapter(Inference, ModelRegistryHelper):
)
try:
response = await self._get_client(provider_model_id).chat.completions.create(**request)
response = await self._client.chat.completions.create(**request)
except APIConnectionError as e:
raise ConnectionError(f"Failed to connect to NVIDIA NIM at {self._config.url}: {e}") from e
@ -339,7 +318,7 @@ class NVIDIAInferenceAdapter(Inference, ModelRegistryHelper):
)
try:
return await self._get_client(provider_model_id).completions.create(**params)
return await self._client.completions.create(**params)
except APIConnectionError as e:
raise ConnectionError(f"Failed to connect to NVIDIA NIM at {self._config.url}: {e}") from e
@ -398,7 +377,7 @@ class NVIDIAInferenceAdapter(Inference, ModelRegistryHelper):
)
try:
return await self._get_client(provider_model_id).chat.completions.create(**params)
return await self._client.chat.completions.create(**params)
except APIConnectionError as e:
raise ConnectionError(f"Failed to connect to NVIDIA NIM at {self._config.url}: {e}") from e

25
llama_stack/providers/remote/inference/ollama/models.py
View file

@ -12,6 +12,19 @@ from llama_stack.providers.utils.inference.model_registry import (
build_model_entry,
)
SAFETY_MODELS_ENTRIES = [
# The Llama Guard models don't have their full fp16 versions
# so we are going to alias their default version to the canonical SKU
build_hf_repo_model_entry(
"llama-guard3:8b",
CoreModelId.llama_guard_3_8b.value,
),
build_hf_repo_model_entry(
"llama-guard3:1b",
CoreModelId.llama_guard_3_1b.value,
),
]
MODEL_ENTRIES = [
build_hf_repo_model_entry(
"llama3.1:8b-instruct-fp16",
@ -73,16 +86,6 @@ MODEL_ENTRIES = [
"llama3.3:70b",
CoreModelId.llama3_3_70b_instruct.value,
),
# The Llama Guard models don't have their full fp16 versions
# so we are going to alias their default version to the canonical SKU
build_hf_repo_model_entry(
"llama-guard3:8b",
CoreModelId.llama_guard_3_8b.value,
),
build_hf_repo_model_entry(
"llama-guard3:1b",
CoreModelId.llama_guard_3_1b.value,
),
ProviderModelEntry(
provider_model_id="all-minilm:l6-v2",
aliases=["all-minilm"],
@ -100,4 +103,4 @@ MODEL_ENTRIES = [
"context_length": 8192,
},
),
]
] + SAFETY_MODELS_ENTRIES

8
llama_stack/providers/remote/inference/openai/models.py
View file

@ -48,9 +48,11 @@ EMBEDDING_MODEL_IDS: dict[str, EmbeddingModelInfo] = {
"text-embedding-3-small": EmbeddingModelInfo(1536, 8192),
"text-embedding-3-large": EmbeddingModelInfo(3072, 8192),
}
SAFETY_MODELS_ENTRIES = []
MODEL_ENTRIES = [ProviderModelEntry(provider_model_id=m) for m in LLM_MODEL_IDS] + [
MODEL_ENTRIES = (
[ProviderModelEntry(provider_model_id=m) for m in LLM_MODEL_IDS]
+ [
ProviderModelEntry(
provider_model_id=model_id,
model_type=ModelType.embedding,
@ -61,3 +63,5 @@ MODEL_ENTRIES = [ProviderModelEntry(provider_model_id=m) for m in LLM_MODEL_IDS]
)
for model_id, model_info in EMBEDDING_MODEL_IDS.items()
]
+ SAFETY_MODELS_ENTRIES
)

14
llama_stack/providers/remote/inference/openai/openai.py
View file

@ -59,9 +59,6 @@ class OpenAIInferenceAdapter(LiteLLMOpenAIMixin):
# if we do not set this, users will be exposed to the
# litellm specific model names, an abstraction leak.
self.is_openai_compat = True
self._openai_client = AsyncOpenAI(
api_key=self.config.api_key,
)
async def initialize(self) -> None:
await super().initialize()
@ -69,6 +66,11 @@ class OpenAIInferenceAdapter(LiteLLMOpenAIMixin):
async def shutdown(self) -> None:
await super().shutdown()
def _get_openai_client(self) -> AsyncOpenAI:
return AsyncOpenAI(
api_key=self.get_api_key(),
)
async def openai_completion(
self,
model: str,
@ -120,7 +122,7 @@ class OpenAIInferenceAdapter(LiteLLMOpenAIMixin):
user=user,
suffix=suffix,
)
return await self._openai_client.completions.create(**params)
return await self._get_openai_client().completions.create(**params)
async def openai_chat_completion(
self,
@ -176,7 +178,7 @@ class OpenAIInferenceAdapter(LiteLLMOpenAIMixin):
top_p=top_p,
user=user,
)
return await self._openai_client.chat.completions.create(**params)
return await self._get_openai_client().chat.completions.create(**params)
async def openai_embeddings(
self,
@ -204,7 +206,7 @@ class OpenAIInferenceAdapter(LiteLLMOpenAIMixin):
params["user"] = user
# Call OpenAI embeddings API
response = await self._openai_client.embeddings.create(**params)
response = await self._get_openai_client().embeddings.create(**params)
data = []
for i, embedding_data in enumerate(response.data):

12
llama_stack/providers/remote/inference/runpod/runpod.py
View file

@ -11,7 +11,7 @@ from llama_stack.apis.inference import * # noqa: F403
from llama_stack.apis.inference import OpenAIEmbeddingsResponse
# from llama_stack.providers.datatypes import ModelsProtocolPrivate
from llama_stack.providers.utils.inference.model_registry import ModelRegistryHelper
from llama_stack.providers.utils.inference.model_registry import ModelRegistryHelper, build_hf_repo_model_entry
from llama_stack.providers.utils.inference.openai_compat import (
OpenAIChatCompletionToLlamaStackMixin,
OpenAICompletionToLlamaStackMixin,
@ -25,6 +25,8 @@ from llama_stack.providers.utils.inference.prompt_adapter import (
from .config import RunpodImplConfig
# https://docs.runpod.io/serverless/vllm/overview#compatible-models
# https://github.com/runpod-workers/worker-vllm/blob/main/README.md#compatible-model-architectures
RUNPOD_SUPPORTED_MODELS = {
"Llama3.1-8B": "meta-llama/Llama-3.1-8B",
"Llama3.1-70B": "meta-llama/Llama-3.1-70B",
@ -40,6 +42,14 @@ RUNPOD_SUPPORTED_MODELS = {
"Llama3.2-3B": "meta-llama/Llama-3.2-3B",
}
SAFETY_MODELS_ENTRIES = []
# Create MODEL_ENTRIES from RUNPOD_SUPPORTED_MODELS for compatibility with starter template
MODEL_ENTRIES = [
build_hf_repo_model_entry(provider_model_id, model_descriptor)
for provider_model_id, model_descriptor in RUNPOD_SUPPORTED_MODELS.items()
] + SAFETY_MODELS_ENTRIES
class RunpodInferenceAdapter(
ModelRegistryHelper,

14
llama_stack/providers/remote/inference/sambanova/models.py
View file

@ -9,6 +9,14 @@ from llama_stack.providers.utils.inference.model_registry import (
build_hf_repo_model_entry,
)
SAFETY_MODELS_ENTRIES = [
build_hf_repo_model_entry(
"sambanova/Meta-Llama-Guard-3-8B",
CoreModelId.llama_guard_3_8b.value,
),
]
MODEL_ENTRIES = [
build_hf_repo_model_entry(
"sambanova/Meta-Llama-3.1-8B-Instruct",
@ -46,8 +54,4 @@ MODEL_ENTRIES = [
"sambanova/Llama-4-Maverick-17B-128E-Instruct",
CoreModelId.llama4_maverick_17b_128e_instruct.value,
),
build_hf_repo_model_entry(
"sambanova/Meta-Llama-Guard-3-8B",
CoreModelId.llama_guard_3_8b.value,
),
]
] + SAFETY_MODELS_ENTRIES

21
llama_stack/providers/remote/inference/sambanova/sambanova.py
View file

@ -7,6 +7,7 @@
import json
from collections.abc import Iterable
import requests
from openai.types.chat import (
ChatCompletionAssistantMessageParam as OpenAIChatCompletionAssistantMessage,
)
@ -56,6 +57,7 @@ from llama_stack.apis.inference import (
ToolResponseMessage,
UserMessage,
)
from llama_stack.apis.models import Model
from llama_stack.log import get_logger
from llama_stack.models.llama.datatypes import BuiltinTool
from llama_stack.providers.utils.inference.litellm_openai_mixin import LiteLLMOpenAIMixin
@ -176,10 +178,11 @@ class SambaNovaInferenceAdapter(LiteLLMOpenAIMixin):
def __init__(self, config: SambaNovaImplConfig):
self.config = config
self.environment_available_models = []
LiteLLMOpenAIMixin.__init__(
self,
model_entries=MODEL_ENTRIES,
api_key_from_config=self.config.api_key,
api_key_from_config=self.config.api_key.get_secret_value() if self.config.api_key else None,
provider_data_api_key_field="sambanova_api_key",
)
@ -246,6 +249,22 @@ class SambaNovaInferenceAdapter(LiteLLMOpenAIMixin):
**get_sampling_options(request.sampling_params),
}
async def register_model(self, model: Model) -> Model:
model_id = self.get_provider_model_id(model.provider_resource_id)
list_models_url = self.config.url + "/models"
if len(self.environment_available_models) == 0:
try:
response = requests.get(list_models_url)
response.raise_for_status()
except requests.exceptions.RequestException as e:
raise RuntimeError(f"Request to {list_models_url} failed") from e
self.environment_available_models = [model.get("id") for model in response.json().get("data", {})]
if model_id.split("sambanova/")[-1] not in self.environment_available_models:
logger.warning(f"Model {model_id} not available in {list_models_url}")
return model
async def initialize(self):
await super().initialize()

20
llama_stack/providers/remote/inference/together/models.py
View file

@ -11,6 +11,16 @@ from llama_stack.providers.utils.inference.model_registry import (
build_hf_repo_model_entry,
)
SAFETY_MODELS_ENTRIES = [
build_hf_repo_model_entry(
"meta-llama/Llama-Guard-3-8B",
CoreModelId.llama_guard_3_8b.value,
),
build_hf_repo_model_entry(
"meta-llama/Llama-Guard-3-11B-Vision-Turbo",
CoreModelId.llama_guard_3_11b_vision.value,
),
]
MODEL_ENTRIES = [
build_hf_repo_model_entry(
"meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo",
@ -40,14 +50,6 @@ MODEL_ENTRIES = [
"meta-llama/Llama-3.3-70B-Instruct-Turbo",
CoreModelId.llama3_3_70b_instruct.value,
),
build_hf_repo_model_entry(
"meta-llama/Meta-Llama-Guard-3-8B",
CoreModelId.llama_guard_3_8b.value,
),
build_hf_repo_model_entry(
"meta-llama/Llama-Guard-3-11B-Vision-Turbo",
CoreModelId.llama_guard_3_11b_vision.value,
),
ProviderModelEntry(
provider_model_id="togethercomputer/m2-bert-80M-8k-retrieval",
model_type=ModelType.embedding,
@ -78,4 +80,4 @@ MODEL_ENTRIES = [
"together/meta-llama/Llama-4-Maverick-17B-128E-Instruct-FP8",
],
),
]
] + SAFETY_MODELS_ENTRIES

17
llama_stack/providers/remote/inference/together/together.py
View file

@ -68,19 +68,12 @@ class TogetherInferenceAdapter(ModelRegistryHelper, Inference, NeedsRequestProvi
def __init__(self, config: TogetherImplConfig) -> None:
ModelRegistryHelper.__init__(self, MODEL_ENTRIES)
self.config = config
self._client = None
self._openai_client = None
async def initialize(self) -> None:
pass
async def shutdown(self) -> None:
if self._client:
# Together client has no close method, so just set to None
self._client = None
if self._openai_client:
await self._openai_client.close()
self._openai_client = None
pass
async def completion(
self,
@ -108,7 +101,6 @@ class TogetherInferenceAdapter(ModelRegistryHelper, Inference, NeedsRequestProvi
return await self._nonstream_completion(request)
def _get_client(self) -> AsyncTogether:
if not self._client:
together_api_key = None
config_api_key = self.config.api_key.get_secret_value() if self.config.api_key else None
if config_api_key:
@ -120,17 +112,14 @@ class TogetherInferenceAdapter(ModelRegistryHelper, Inference, NeedsRequestProvi
'Pass Together API Key in the header X-LlamaStack-Provider-Data as { "together_api_key": <your api key>}'
)
together_api_key = provider_data.together_api_key
self._client = AsyncTogether(api_key=together_api_key)
return self._client
return AsyncTogether(api_key=together_api_key)
def _get_openai_client(self) -> AsyncOpenAI:
if not self._openai_client:
together_client = self._get_client().client
self._openai_client = AsyncOpenAI(
return AsyncOpenAI(
base_url=together_client.base_url,
api_key=together_client.api_key,
)
return self._openai_client
async def _nonstream_completion(self, request: CompletionRequest) -> ChatCompletionResponse:
params = await self._get_params(request)

11
llama_stack/providers/remote/safety/sambanova/sambanova.py
View file

@ -33,6 +33,7 @@ CANNED_RESPONSE_TEXT = "I can't answer that. Can I help with something else?"
class SambaNovaSafetyAdapter(Safety, ShieldsProtocolPrivate, NeedsRequestProviderData):
def __init__(self, config: SambaNovaSafetyConfig) -> None:
self.config = config
self.environment_available_models = []
async def initialize(self) -> None:
pass
@ -54,18 +55,18 @@ class SambaNovaSafetyAdapter(Safety, ShieldsProtocolPrivate, NeedsRequestProvide
async def register_shield(self, shield: Shield) -> None:
list_models_url = self.config.url + "/models"
if len(self.environment_available_models) == 0:
try:
response = requests.get(list_models_url)
response.raise_for_status()
except requests.exceptions.RequestException as e:
raise RuntimeError(f"Request to {list_models_url} failed") from e
available_models = [model.get("id") for model in response.json().get("data", {})]
self.environment_available_models = [model.get("id") for model in response.json().get("data", {})]
if (
len(available_models) == 0
or "guard" not in shield.provider_resource_id.lower()
or shield.provider_resource_id.split("sambanova/")[-1] not in available_models
"guard" not in shield.provider_resource_id.lower()
or shield.provider_resource_id.split("sambanova/")[-1] not in self.environment_available_models
):
raise ValueError(f"Shield {shield.provider_resource_id} not found in SambaNova")
logger.warning(f"Shield {shield.provider_resource_id} not available in {list_models_url}")
async def run_shield(
self, shield_id: str, messages: list[Message], params: dict[str, Any] | None = None

1
llama_stack/providers/remote/vector_io/chroma/chroma.py
View file

@ -217,7 +217,6 @@ class ChromaVectorIOAdapter(VectorIO, VectorDBsProtocolPrivate):
embedding_model: str | None = None,
embedding_dimension: int | None = 384,
provider_id: str | None = None,
provider_vector_db_id: str | None = None,
) -> VectorStoreObject:
raise NotImplementedError("OpenAI Vector Stores API is not supported in Chroma")

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