scriptoriumadmin/litellm
1
0
Fork 0
mirror of https://github.com/BerriAI/litellm.git synced 2026-08-28 05:25:59 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 36

Merge remote-tracking branch 'origin/litellm_internal_staging' into litellm_fix_openai_cache_token_details_loss

Browse source 
# Conflicts:
#	litellm/litellm_core_utils/llm_cost_calc/utils.py
#	litellm/litellm_core_utils/streaming_chunk_builder_utils.py
#	litellm/litellm_core_utils/streaming_handler.py
#	tests/test_litellm/litellm_core_utils/test_streaming_chunk_builder_utils.py
This commit is contained in:
mateo-berri 2026-08-05 01:23:54 -07:00
commit 7288247682
3198 changed files with 158461 additions and 88713 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

176
.circleci/config.yml
View file

@ -88,6 +88,59 @@ commands:
rm -f /tmp/uv-install.sh
echo 'export PATH="$HOME/.local/bin:$PATH"' >> "$BASH_ENV"
export PATH="$HOME/.local/bin:$PATH"
install_node:
description: "Install the Node.js version pinned in ui/litellm-dashboard/.nvmrc (24.19.0, which bundles npm 11.17.0) with checksum verification, and prepend it to PATH. Run this on any executor whose image does not already ship that version, or `npm ci` in ui/litellm-dashboard fails EBADENGINE against the engines floor. Installs into /opt/node rather than over /usr/local on purpose: cimg/python:*-browsers ships its own node there, and unpacking the tarball on top of it leaves npm 11.17 files merged with the image's npm 11.9 tree, which reports the new version and then exits 1 on `npm ci` with no error text at all. Requires checkout, which the .nvmrc drift check reads."
steps:
- run:
name: Install Node.js 24.19.0
command: |
NODE_VERSION="24.19.0"
NODE_TARBALL="node-v${NODE_VERSION}-linux-x64.tar.xz"
NODE_EXPECTED_SHA="14b342e71204f811bde6153be8e04b62aef63c236fef92b55f9c83154b409647"
NVMRC_VERSION="$(tr -d '[:space:]' < ui/litellm-dashboard/.nvmrc)"
if [ "$NVMRC_VERSION" != "$NODE_VERSION" ]; then
echo "install_node: ui/litellm-dashboard/.nvmrc pins ${NVMRC_VERSION} but this command pins ${NODE_VERSION}; update NODE_VERSION and NODE_EXPECTED_SHA together" >&2
exit 1
fi
curl -sSLf -o "/tmp/${NODE_TARBALL}" "https://nodejs.org/dist/v${NODE_VERSION}/${NODE_TARBALL}"
echo "${NODE_EXPECTED_SHA} /tmp/${NODE_TARBALL}" | sha256sum -c -
sudo mkdir -p /opt/node
sudo tar -xJf "/tmp/${NODE_TARBALL}" -C /opt/node --strip-components=1
rm -f "/tmp/${NODE_TARBALL}"
echo 'export PATH="/opt/node/bin:$PATH"' >> "$BASH_ENV"
export PATH="/opt/node/bin:$PATH"
node --version
npm --version
install_rust:
description: "Install pinned rustup (1.28.2) and Rust toolchain (1.97.1) with checksum verification. Adds ~/.cargo/bin to PATH. Run this before any `uv sync` or `uv build` of the workspace: the root package builds litellm-rust through maturin, and on an image without cargo maturin fetches an unpinned rustup and a floating toolchain by itself."
steps:
- run:
name: Install Rust (rustup 1.28.2, toolchain 1.97.1)
command: |
case "$(uname -m)" in
x86_64)
RUSTUP_TRIPLE=x86_64-unknown-linux-gnu
RUSTUP_SHA256=20a06e644b0d9bd2fbdbfd52d42540bdde820ea7df86e92e533c073da0cdd43c
;;
aarch64)
RUSTUP_TRIPLE=aarch64-unknown-linux-gnu
RUSTUP_SHA256=e3853c5a252fca15252d07cb23a1bdd9377a8c6f3efa01531109281ae47f841c
;;
*)
echo "install_rust: unsupported architecture $(uname -m)" >&2
exit 1
;;
esac
curl -sSLf -o /tmp/rustup-init \
"https://static.rust-lang.org/rustup/archive/1.28.2/${RUSTUP_TRIPLE}/rustup-init"
echo "${RUSTUP_SHA256} /tmp/rustup-init" | sha256sum -c -
chmod +x /tmp/rustup-init
/tmp/rustup-init -y --no-modify-path --profile minimal --default-toolchain 1.97.1
rm -f /tmp/rustup-init
echo 'export PATH="$HOME/.cargo/bin:$PATH"' >> "$BASH_ENV"
export PATH="$HOME/.cargo/bin:$PATH"
rustc --version
cargo --version
start_postgres:
description: "Start a postgres-db container on port 5432 and wait until it accepts connections."
parameters:
@ -163,6 +216,26 @@ commands:
done
echo "fake OpenAI endpoint did not become ready" >&2
exit 1
start_cost_center_service:
description: "Start the stand-in cost center validation service (tests/store_model_in_db_tests/cost_center_service.py) on host port 9414 and wait until healthy. The proxy's team-metadata validator (team_metadata_validator_e2e.py, impl 'http') reaches it via TEAM_METADATA_VALIDATION_SERVICE_URL=http://host.docker.internal:9414/validate. Run after uv deps are synced."
steps:
- run:
name: Start cost center validation service
background: true
command: |
uv run --no-sync python tests/store_model_in_db_tests/cost_center_service.py --host 0.0.0.0 --port 9414
- run:
name: Wait for cost center validation service
command: |
for i in $(seq 1 30); do
if curl -sf http://localhost:9414/health >/dev/null 2>&1; then
echo "cost center validation service is up"
exit 0
fi
sleep 1
done
echo "cost center validation service did not become ready" >&2
exit 1
setup_litellm_enterprise_pip:
steps:
- run:
@ -178,6 +251,7 @@ commands:
- checkout
- setup_google_dns
- install_uv
- install_rust
- restore_cache:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
@ -281,6 +355,33 @@ jobs:
uv build --wheel --out-dir dist
uv run --no-sync python tests/windows_tests/check_windows_wheel_install.py
base_sdk_install:
docker:
- image: cimg/python:3.12@sha256:9c796c23c84e84a66a964acb508d39dc5433c81a47e07efd56dccbbc2427e07c
auth:
username: ${DOCKERHUB_USERNAME}
password: ${DOCKERHUB_PASSWORD}
working_directory: ~/project
steps:
- checkout
- setup_google_dns
- install_uv
- install_rust
- run:
name: Build the wheel
environment:
UV_HTTP_TIMEOUT: "300"
command: |
uv build --wheel --out-dir dist
- run:
name: Install the wheel with no extras and smoke-check it
environment:
UV_HTTP_TIMEOUT: "300"
command: |
uv venv /tmp/base-sdk --python 3.12
VIRTUAL_ENV=/tmp/base-sdk uv pip install dist/*.whl
/tmp/base-sdk/bin/python tests/base_sdk_tests/check_base_sdk_install.py
local_testing_part1:
docker:
- &python312_image
@ -298,6 +399,7 @@ jobs:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -371,6 +473,7 @@ jobs:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -445,6 +548,7 @@ jobs:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -496,6 +600,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -562,6 +667,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -602,6 +708,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -643,6 +750,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -676,6 +784,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- restore_cache:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
@ -726,6 +835,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- restore_cache:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
@ -777,6 +887,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -810,6 +921,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- restore_cache:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
@ -856,6 +968,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -902,6 +1015,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -944,6 +1058,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -990,6 +1105,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1037,6 +1153,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- restore_cache:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
@ -1077,6 +1194,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1122,6 +1240,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1166,6 +1285,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1198,6 +1318,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1241,6 +1362,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1285,6 +1407,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1329,6 +1452,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1360,6 +1484,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1406,6 +1531,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1451,6 +1577,7 @@ jobs:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1501,6 +1628,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1525,6 +1653,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1551,6 +1680,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1652,6 +1782,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1747,6 +1878,7 @@ jobs:
at: ~/project
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1835,6 +1967,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -1918,6 +2051,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -2050,6 +2184,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -2136,6 +2271,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -2232,12 +2368,14 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
uv sync --frozen --all-groups --all-extras --python 3.12
- start_postgres
- start_fake_openai_endpoint
- start_cost_center_service
- attach_workspace:
at: ~/project
- run:
@ -2257,11 +2395,13 @@ jobs:
-e STORE_MODEL_IN_DB="True" \
-e LITELLM_MASTER_KEY="sk-1234" \
-e FAKE_OPENAI_API_BASE=http://host.docker.internal:8190 \
-e TEAM_METADATA_VALIDATION_SERVICE_URL=http://host.docker.internal:9414/validate \
-e LITELLM_LICENSE=$LITELLM_LICENSE \
-e LITELLM_LOG=ERROR \
--add-host host.docker.internal:host-gateway \
--name my-app \
-v $(pwd)/litellm/proxy/example_config_yaml/store_model_db_config.yaml:/app/config.yaml \
-v $(pwd)/litellm/proxy/example_config_yaml/team_metadata_validator_e2e.py:/app/team_metadata_validator_e2e.py \
litellm-docker-database:ci \
--config /app/config.yaml \
--port 4000
@ -2307,6 +2447,7 @@ jobs:
- setup_google_dns
# Remove Docker CLI installation since it's already available in machine executor
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -2388,6 +2529,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -2473,20 +2615,9 @@ jobs:
bundle exec rspec
no_output_timeout: 30m
# Install Node.js directly from nodejs.org with SHA256 verification,
# instead of piping NodeSource's setup_18.x apt-repo installer into
# instead of piping NodeSource's setup_24.x apt-repo installer into
# sudo bash (which runs a mutable upstream script unattended).
- run:
name: Install Node.js 18.20.8
command: |
NODE_VERSION="18.20.8"
NODE_TARBALL="node-v${NODE_VERSION}-linux-x64.tar.xz"
NODE_EXPECTED_SHA="5467ee62d6af1411d46b6a10e3fb5cacc92734dbcef465fea14e7b90993001c9"
curl -sSLf -o "/tmp/${NODE_TARBALL}" "https://nodejs.org/dist/v${NODE_VERSION}/${NODE_TARBALL}"
echo "${NODE_EXPECTED_SHA} /tmp/${NODE_TARBALL}" | sha256sum -c -
sudo tar -xJf "/tmp/${NODE_TARBALL}" -C /usr/local --strip-components=1
rm -f "/tmp/${NODE_TARBALL}"
node --version
npm --version
- install_node
- run:
name: Install Node.js test dependencies
@ -2527,6 +2658,7 @@ jobs:
- skip_if_unrelated_changes
- setup_google_dns
- install_uv
- install_rust
- run:
name: Install Dependencies
command: |
@ -2614,7 +2746,7 @@ jobs:
ui_build:
docker:
- image: cimg/node:20.19@sha256:35e64883e8d21bc345b0a7b04c35ee46442c127607ed1d8d7d37d8a1ed76db81
- image: cimg/node:24.19@sha256:8966565f07189a67d64d6808a2b127f31dafae566508e3547f55640e1070bfad
auth:
username: ${DOCKERHUB_USERNAME}
password: ${DOCKERHUB_PASSWORD}
@ -2658,7 +2790,7 @@ jobs:
ui_unit_tests:
docker:
- image: cimg/node:20.19@sha256:35e64883e8d21bc345b0a7b04c35ee46442c127607ed1d8d7d37d8a1ed76db81
- image: cimg/node:24.19@sha256:8966565f07189a67d64d6808a2b127f31dafae566508e3547f55640e1070bfad
auth:
username: ${DOCKERHUB_USERNAME}
password: ${DOCKERHUB_PASSWORD}
@ -2716,7 +2848,9 @@ jobs:
- skip_if_unrelated_changes:
category: client
- setup_google_dns
- install_node
- install_uv
- install_rust
- restore_cache:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
@ -2731,7 +2865,7 @@ jobs:
- ~/.cache/uv
- restore_cache:
keys:
- ui-e2e-node-deps-v3-{{ checksum "ui/litellm-dashboard/package-lock.json" }}-{{ checksum "tests/e2e/ui/package-lock.json" }}
- ui-e2e-node-deps-v4-{{ checksum "ui/litellm-dashboard/package-lock.json" }}-{{ checksum "tests/e2e/ui/package-lock.json" }}
- run:
name: Install Node dependencies and Playwright
# The cimg/python:3.12-browsers image already ships the Chromium system
@ -2746,7 +2880,7 @@ jobs:
npm ci
npx playwright install chromium
- save_cache:
key: ui-e2e-node-deps-v3-{{ checksum "ui/litellm-dashboard/package-lock.json" }}-{{ checksum "tests/e2e/ui/package-lock.json" }}
key: ui-e2e-node-deps-v4-{{ checksum "ui/litellm-dashboard/package-lock.json" }}-{{ checksum "tests/e2e/ui/package-lock.json" }}
paths:
- ui/litellm-dashboard/node_modules
- tests/e2e/ui/node_modules
@ -2858,7 +2992,9 @@ jobs:
- skip_if_unrelated_changes:
category: client
- setup_google_dns
- install_node
- install_uv
- install_rust
- restore_cache:
keys:
- v1-uv-cache-{{ checksum "uv.lock" }}
@ -2873,7 +3009,7 @@ jobs:
- ~/.cache/uv
- restore_cache:
keys:
- ui-e2e-node-deps-v3-{{ checksum "ui/litellm-dashboard/package-lock.json" }}-{{ checksum "tests/e2e/ui/package-lock.json" }}
- ui-e2e-node-deps-v4-{{ checksum "ui/litellm-dashboard/package-lock.json" }}-{{ checksum "tests/e2e/ui/package-lock.json" }}
- run:
name: Install Node dependencies and Playwright
command: |
@ -2883,7 +3019,7 @@ jobs:
npm ci
npx playwright install chromium
- save_cache:
key: ui-e2e-node-deps-v3-{{ checksum "ui/litellm-dashboard/package-lock.json" }}-{{ checksum "tests/e2e/ui/package-lock.json" }}
key: ui-e2e-node-deps-v4-{{ checksum "ui/litellm-dashboard/package-lock.json" }}-{{ checksum "tests/e2e/ui/package-lock.json" }}
paths:
- ui/litellm-dashboard/node_modules
- tests/e2e/ui/node_modules
@ -3031,6 +3167,8 @@ workflows:
only:
- main
- /litellm_.*/
- base_sdk_install:
filters: *main_branches
- local_testing_part1:
filters: *main_branches
- local_testing_part2:

46
.flake8
View file

@ -1,46 +0,0 @@
[flake8]
ignore =
# The following ignores can be removed when formatting using black
W191,W291,W292,W293,W391,W504
E101,E111,E114,E116,E117,E121,E122,E123,E124,E125,E126,E127,E128,E129,E131,
E201,E202,E221,E222,E225,E226,E231,E241,E251,E252,E261,E265,E271,E272,E275,
E301,E302,E303,E305,E306,
# line break before binary operator
W503,
# inline comment should start with '# '
E262,
# too many leading '#' for block comment
E266,
# multiple imports on one line
E401,
# module level import not at top of file
E402,
# Line too long (82 > 79 characters)
E501,
# comparison to None should be 'if cond is None:'
E711,
# comparison to True should be 'if cond is True:' or 'if cond:'
E712,
# do not compare types, for exact checks use `is` / `is not`, for instance checks use `isinstance()`
E721,
# do not use bare 'except'
E722,
# x is imported but unused
F401,
# 'from . import *' used; unable to detect undefined names
F403,
# x may be undefined, or defined from star imports:
F405,
# f-string is missing placeholders
F541,
# dictionary key '' repeated with different values
F601,
# redefinition of unused x from line 123
F811,
# undefined name x
F821,
# local variable x is assigned to but never used
F841,
# https://black.readthedocs.io/en/stable/guides/using_black_with_other_tools.html#flake8
extend-ignore = E203

1
.github/pull_request_template.md vendored
View file

@ -40,6 +40,7 @@ If you're seeing a delay in your PR being merged, ping the LiteLLM Team on [Slac
The proof must be completely e2e with no mocks, using, for example, actual LLM calls costing real $. `pytest` commands are not enough
For bug fixes: show reproduction before the fix and passing behavior after
Include the commit hash each proof was captured at, for both the before and the after runs
If the change applies to all three LLM endpoints (/v1/responses, /v1/chat/completions, /v1/messages), include proof for every single one of them, not just one
For new features: show the feature working end-to-end
For UI changes: include before/after screenshots -->

5
.github/workflows/auto_update_price_and_context_window.yml vendored
View file

@ -24,9 +24,12 @@ jobs:
- name: Update JSON Data
run: |
uv run --frozen --with 'aiohttp==3.13.3' python ".github/workflows/auto_update_price_and_context_window_file.py"
- name: Regenerate JSON Schema
run: |
uv run --frozen python ci_cd/generate_model_prices_schema.py
- name: Create Pull Request
run: |
git add model_prices_and_context_window.json
git add model_prices_and_context_window.json model_prices_and_context_window.schema.json
git commit -m "Update model_prices_and_context_window.json file: $(date +'%Y-%m-%d')"
gh pr create --title "Update model_prices_and_context_window.json file" \
--body "Automated update for model_prices_and_context_window.json" \

2
.github/workflows/check-ui-api-types.yml vendored
View file

@ -56,7 +56,7 @@ jobs:
- name: Set up Node.js
uses: actions/setup-node@a0853c24544627f65ddf259abe73b1d18a591444 # v5.0.0
with:
node-version: "20"
node-version-file: ui/litellm-dashboard/.nvmrc
cache: "npm"
cache-dependency-path: ui/litellm-dashboard/package-lock.json

33
.github/workflows/image-scan.yml vendored
View file

@ -9,6 +9,8 @@ on:
- "litellm_**"
paths:
- docker/Dockerfile.non_root
- migrations/Dockerfile
- migrations/run.py
- tests/proxy_migration_tests/test_offline_image_migration.py
- uv.lock
- ui/litellm-dashboard/package-lock.json
@ -83,3 +85,34 @@ jobs:
--only-fixed \
--fail-on high \
--output table
migrations-image:
name: migrations-image
runs-on: ubuntu-latest
if: >-
github.event_name != 'pull_request' ||
github.event.pull_request.head.repo.full_name == github.repository
timeout-minutes: 30
permissions:
contents: read
steps:
- uses: actions/checkout@08eba0b27e820071cde6df949e0beb9ba4906955 # v4.3.0
with:
persist-credentials: false
- name: Build migrations image
run: docker build -f migrations/Dockerfile -t litellm-migrations-scan:${{ github.sha }} .
- name: Set up Python
uses: actions/setup-python@a26af69be951a213d495a4c3e4e4022e16d87065 # v5.6.0
with:
python-version: "3.12"
- name: Verify offline migration as a non-root uid
env:
LITELLM_IMAGE: litellm-migrations-scan:${{ github.sha }}
LITELLM_MIGRATION_INTERPRETER: python3
LITELLM_MIGRATION_SCRIPT: /app/run.py
run: |
python -m pip install "pytest==9.0.3"
python -m pytest tests/proxy_migration_tests/test_offline_image_migration.py -v

8
.github/workflows/test-code-quality.yml vendored
View file

@ -7,13 +7,17 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
code-quality:

2
.github/workflows/test-linting.yml vendored
View file

@ -105,7 +105,7 @@ jobs:
env:
BASE_SHA: ${{ github.event.pull_request.base.sha }}
run: |
(uv run --no-sync basedpyright --outputjson || true) | uv run --no-sync python scripts/type_check_gate.py --base "$BASE_SHA"
uv run --no-sync python scripts/type_check_gate.py --base "$BASE_SHA"
- name: Check tests/e2e basedpyright (zero errors)
env:

2
.github/workflows/test-litellm-ui-build.yml vendored
View file

@ -27,7 +27,7 @@ jobs:
- name: Setup Node.js
uses: actions/setup-node@a0853c24544627f65ddf259abe73b1d18a591444 # v5.0.0
with:
node-version: "20"
node-version-file: ui/litellm-dashboard/.nvmrc
cache: "npm"
cache-dependency-path: ui/litellm-dashboard/package-lock.json

2
.github/workflows/test-litellm-ui-lint.yml vendored
View file

@ -61,7 +61,7 @@ jobs:
if: steps.changed.outputs.has_files == 'true'
uses: actions/setup-node@a0853c24544627f65ddf259abe73b1d18a591444 # v5.0.0
with:
node-version: "20"
node-version-file: ui/litellm-dashboard/.nvmrc
cache: "npm"
cache-dependency-path: ui/litellm-dashboard/package-lock.json

2
.github/workflows/test-litellm-ui-unit.yml vendored
View file

@ -35,7 +35,7 @@ jobs:
- name: Setup Node.js
uses: actions/setup-node@a0853c24544627f65ddf259abe73b1d18a591444 # v5.0.0
with:
node-version: "20"
node-version-file: ui/litellm-dashboard/.nvmrc
cache: "npm"
cache-dependency-path: ui/litellm-dashboard/package-lock.json

9
.github/workflows/test-model-map.yaml vendored
View file

@ -22,3 +22,12 @@ jobs:
- name: Validate model_prices_and_context_window.json
run: |
jq empty model_prices_and_context_window.json
- name: Set up uv
uses: ./.github/actions/setup-uv-with-retries
with:
version: "0.10.9"
- name: Check model_prices_and_context_window.schema.json is in sync
run: |
uv run --frozen python ci_cd/generate_model_prices_schema.py --check

8
.github/workflows/test-unit-core-utils.yml vendored
View file

@ -7,6 +7,10 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
@ -14,8 +18,8 @@ permissions:
pull-requests: write
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
core-utils:

8
.github/workflows/test-unit-documentation.yml vendored
View file

@ -7,13 +7,17 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
documentation:

8
.github/workflows/test-unit-enterprise-routing.yml vendored
View file

@ -7,6 +7,10 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
@ -14,8 +18,8 @@ permissions:
pull-requests: write
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
enterprise-routing:

8
.github/workflows/test-unit-integrations.yml vendored
View file

@ -7,6 +7,10 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
@ -14,8 +18,8 @@ permissions:
pull-requests: write
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
integrations:

8
.github/workflows/test-unit-llm-providers.yml vendored
View file

@ -7,13 +7,17 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
vertex-ai:

9
.github/workflows/test-unit-misc.yml vendored
View file

@ -7,6 +7,10 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
@ -14,8 +18,8 @@ permissions:
pull-requests: write
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
misc:
@ -27,6 +31,7 @@ jobs:
tests/test_litellm/a2a_protocol
tests/test_litellm/anthropic_interface
tests/test_litellm/completion_extras
tests/test_litellm/compression
tests/test_litellm/containers
tests/test_litellm/experimental_mcp_client
tests/test_litellm/models

8
.github/workflows/test-unit-proxy-auth.yml vendored
View file

@ -7,6 +7,10 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
@ -14,8 +18,8 @@ permissions:
pull-requests: write
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
proxy-auth:

8
.github/workflows/test-unit-proxy-db.yml vendored
View file

@ -7,13 +7,17 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
# Semantic matrix: each shard groups tests by concern (auth, server, logging, …)
# rather than alphabetical letter ranges. Adding a new test file means adding it

8
.github/workflows/test-unit-proxy-endpoints.yml vendored
View file

@ -7,14 +7,18 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
workflow_dispatch:
permissions:
contents: read
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
proxy-endpoints:

8
.github/workflows/test-unit-proxy-infra.yml vendored
View file

@ -7,6 +7,10 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
@ -14,8 +18,8 @@ permissions:
pull-requests: write
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
proxy-infra:

8
.github/workflows/test-unit-proxy-legacy.yml vendored
View file

@ -7,13 +7,17 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
test:

8
.github/workflows/test-unit-responses-caching-types.yml vendored
View file

@ -7,6 +7,10 @@ on:
- litellm_internal_staging
- litellm_oss_staging
- "litellm_**"
push:
branches:
- main
- litellm_internal_staging
permissions:
contents: read
@ -14,8 +18,8 @@ permissions:
pull-requests: write
concurrency:
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.ref }}
cancel-in-progress: true
group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.sha }}
cancel-in-progress: ${{ github.event_name == 'pull_request' }}
jobs:
responses-caching-types:

1
.gitignore vendored
View file

@ -141,3 +141,4 @@ crash.*.log
.coverage
ui/litellm-dashboard/out/
litellm.log

11
CLAUDE.md
View file

@ -29,7 +29,7 @@ Never use `pytest` commands or the like as "Screenshots / Proof of Fix". We pref
If you ever make public-facing PR descriptions, comments, issues, commit messages, etc., always follow these guidelines to sound less AI-y:
- don't use emojis
- don't use "—". Instead, reach for ";", ".", etc.
- don't use "—". Instead, reach for ",", ".", conjunction words, ":", ";", etc. in descending order of preference: vary among them, weighted toward the front of the list, and skip "," where it would cause a comma splice or the sentence is getting long. Overusing any one of them, ";" especially, also feels AI-y
- don't use the pattern "It's not X, it's Y", "You're not X, you're Y", etc.
- don't use bulleted or numbered lists unless it would be nonsensical not to. Instead, prefer prose
- don't add a trailing "." at the end of paragraphs (just like this file). That means every paragraph, not just the last one (of the markdown file, PR description, GitHub comment, etc.). Rule of thumb: if you're adding new line(s) before the next sentence, don't add a "."
@ -39,15 +39,11 @@ Don't hesitate to use values in .env to get needed API keys and other secrets, a
Python max line length is 120, not 88
On a fresh worktree or clone, run `make bootstrap` before anything else. It provisions everything tests, `make pre-commit`, and a local proxy need
Run tests before you commit. Also, run `make pre-commit` right before each commit, which generates types (as needed) and formats/lints your code. Any errors found must be fixed. It only runs when there are staged frontend and/or backend changes and calculates violations, generates types, etc. based on the worktree, so stage what you need or stash/delete unwanted files in litellm/ or ui/ (where backend and frontend lint run, respectively) before running it. If it fails because dashboard api types are stale, it already regenerated them for you. You just need to stage the schema.d.ts, re-run `make pre-commit` to confirm it passes, and commit
When you fix violations gated by `ruff-strict-budget.json`, `type-discipline-budget.json`, or `basedpyright-code-budget.json`, run `make lint-budget-update` and commit the lowered limits so the ceilings ratchet down instead of leaving stale headroom. It measures the working tree, so it must contain exactly the fixes you're committing
If you're trying to create a new function that relies on untyped stuff, instead of adding more Any's and pushing `reportAny` / `reportExplicitAny` closer to their basedpyright ceilings, just validate it in the caller with Pydantic (a model or `TypeAdapter` that returns the typed thing or raises will do) and then pass the now typed variable in
If you get an LIT001 or LIT002 fail, refactor the code to follow functional programming best practices rather than introducing mutable data structures. For example, build values in one shot with comprehensions or generators wrapped in `tuple()` / `frozenset()` instead of seeding an empty `list`/`dict`/`set` and mutating it over time. Ideally `# mutable-ok` is never used; reach for it only as a genuine last resort when an immutable rewrite is truly impossible, and always pair it with a real reason
If you get an LIT001 or LIT002 fail, refactor the code to follow functional programming best practices rather than introducing mutable data structures. For example, build values in one shot with comprehensions or generators wrapped in `tuple()` / `frozenset()` instead of seeding an empty `list`/`dict`/`set` and mutating it over time. Ideally, `# mutable-ok` is never used; reach for it only as a genuine last resort when an immutable rewrite is truly impossible, and always pair it with a real reason
Every lint or type suppression must name the exact rule inside brackets and carry a reason comment, e.g. `# pyright: ignore[reportArgumentType] # stubs lack async overload` or `# noqa: TID251 # <reason>`. `# type: ignore` is banned (LIT009): pyrightconfig.json sets `enableTypeIgnoreComments` to false, so it silently does nothing
@ -61,6 +57,8 @@ Do not add `Co-Authored-By: Claude` or any Claude attribution to commit messages
When working on a PR, keep the PR description in sync with new commits being made
Replies/rebuttals to AI PR review bots must be 15-25 word human-readable replies
Monkeypatching attributes of a class to do testing is an anti-pattern. Prefer dependency-injecting things into classes. That way, at unit test time, you can pass a mocked dependency in
Do not put names of customers or customer company names in code, PR descriptions, issue bodies, etc. This means never mention literally any company name. Especially if you're about to say a sentence mentioning that the reason the PR exists was a feature/model/bug fix/etc. requested by a company. That's the indication that you should replace that company name with "the customer". e.g. not "Model request from Acme (Pylon #1234)" but "Model request from a customer (Pylon #1234)". This is because the codebase is public. The only exception is for publicly known providers or vendors such as OpenAI, Anthropic, AWS Bedrock, etc. only IF we're adding support for that provider/vendor in general and NOT if that PR or whatnot was a request by one of them, and they're actually one of our customers.
@ -73,6 +71,7 @@ Follow these coding conventions for new/updated code (a three-line fix in a lega
- Never-nester: early returns over deep nesting
- Don't throw; model failures as values (One function (e.g., raise_public) maps error union to existing public exception contracts via exhaustive match + assert_never)
- No mutation; don't reassign variables, global or local. Instead of mutable lists and dicts, prefer tuples, frozen dataclasses (with slots=True), etc.
- Annotate every variable with `: Final` (LIT010). Unpacking and walrus targets cannot carry the annotation, so they are implicitly final. Don't rebind them. Never rebind or mutate function parameters (LIT011); `self`/`cls` attribute stores are the exception. If rebinding or in-place mutation is truly unavoidable, suppress with `# rebind-ok: <reason>` explaining why
- Use dependency injection
- Fully typed; no `Any` or coarse types like `dict[str, Any]` or just `dict`. Every function parameter must be strongly typed
- Use tagged unions + match

2
Dockerfile
View file

@ -7,7 +7,7 @@ ARG LITELLM_BUILD_IMAGE=cgr.dev/chainguard/wolfi-base@sha256:42df77a9974d6ec8b17
ARG LITELLM_RUNTIME_IMAGE=cgr.dev/chainguard/wolfi-base@sha256:42df77a9974d6ec8b17a5ee8bc23b532600a44d705acef2409e0933c1251b45f
ARG UV_IMAGE=ghcr.io/astral-sh/uv:0.11.7@sha256:240fb85ab0f263ef12f492d8476aa3a2e4e1e333f7d67fbdd923d00a506a516a
# Pinned by digest like the other base images; bump explicitly on Node upgrades.
ARG UI_BUILD_IMAGE=node:20.18-alpine3.20@sha256:3488b10bf958af7125a176419d2d8a9937d895bf124012aae811651988d2ffe6
ARG UI_BUILD_IMAGE=node:24.19-alpine3.24@sha256:d32cdf619f63fe0471182d08996dd516c6275bb5fd31ae06e55a570bd9e1ad43
FROM $UV_IMAGE AS uvbin

8
Makefile
View file

@ -75,7 +75,7 @@ install-dev:
bootstrap:
$(UV) sync --inexact --frozen --extra proxy --group proxy-dev --group e2e-dev
$(UV_RUN) python scripts/prisma_generate_if_needed.py
cd ui/litellm-dashboard && npm ci --no-audit --no-fund
cd ui/litellm-dashboard && ../../scripts/with_dashboard_node.sh npm install --no-audit --no-fund
@main_root=$$(git worktree list --porcelain | head -1 | sed 's/^worktree //'); \
if [ "$$main_root" != "$$(git rev-parse --show-toplevel)" ] && [ -f "$$main_root/.env" ] && [ ! -f .env ]; then \
cp "$$main_root/.env" .env && echo "bootstrap: copied .env from $$main_root"; \
@ -177,7 +177,7 @@ lint-ruff-FULL-dev: install-dev
else echo "No changed .py files to check."; fi
lint-basedpyright: $(LINT_DEP_INSTALL) $(LINT_DEP_BASE)
($(UV_RUN) basedpyright --outputjson || true) | $(UV_RUN) python scripts/type_check_gate.py --base origin/litellm_internal_staging
$(UV_RUN) python scripts/type_check_gate.py --base origin/litellm_internal_staging
lint-e2e-basedpyright: $(LINT_E2E_DEP_INSTALL)
$(UV_RUN) basedpyright tests/e2e
@ -190,7 +190,7 @@ lint-type-discipline: $(LINT_DEP_INSTALL) $(LINT_DEP_BASE)
# --update lowers each limit by what this branch fixed since its branch point, so
# it needs the base ref fetched to resolve the merge-base.
lint-basedpyright-budget-update: install-dev lint-fetch-base
($(UV_RUN) basedpyright --outputjson || true) | $(UV_RUN) python scripts/type_check_gate.py --update
$(UV_RUN) python scripts/type_check_gate.py --update
lint-format: format-check
@ -237,7 +237,7 @@ lint-dev: lint-format-changed check-circular-imports check-import-safety
# test-linting.yml (Python), test-litellm-ui-build.yml's frontend-lint (dashboard), and
# check-ui-api-types.yml (API-type drift), skipping any whose files you didn't stage.
# Not auto-installed as a git hook so it never slows an unrelated human commit.
pre-commit:
pre-commit: bootstrap
./scripts/pre_commit_lint.sh
# Testing targets

4
backend/Dockerfile
View file

@ -63,6 +63,8 @@ RUN mkdir -p /home/nonroot && \
HOME=/home/nonroot prisma generate --schema=./schema.prisma && \
chown -R nonroot:nonroot /home/nonroot/.cache
RUN sed -i 's/\r$//' docker/component_entrypoint.sh && chmod +x docker/component_entrypoint.sh
# ---------- Runtime ----------
FROM $LITELLM_RUNTIME_IMAGE AS runtime
@ -93,5 +95,5 @@ USER nonroot
EXPOSE 4001/tcp
ENTRYPOINT ["uvicorn", "backend.main:app"]
ENTRYPOINT ["/app/docker/component_entrypoint.sh", "uvicorn", "backend.main:app"]
CMD ["--host", "0.0.0.0", "--port", "4001"]

5
backend/routes/allowlist.py
View file

@ -44,6 +44,7 @@ BACKEND_PATH_PREFIXES: tuple[str, ...] = (
"/router/",
"/router_settings",
"/adaptive_router/",
"/auto_router/",
"/fallback",
"/fallbacks",
"/cache_settings",
@ -70,6 +71,10 @@ BACKEND_PATH_PREFIXES: tuple[str, ...] = (
"/project/",
"/memory/",
"/mcp/",
# Control plane (see the List Endpoints + Tables standard). Every resource
# eventually moves under this prefix, so allowlist it once rather than
# per-resource.
"/management/v1/",
# Spend / analytics
"/spend/",
"/analytics/",

62
basedpyright-code-budget.json
View file

@ -1,39 +1,39 @@
{
"reportAny": {
"limit": 37484
"limit": 29809
},
"reportArgumentType": {
"limit": 2704
"limit": 2645
},
"reportAssignmentType": {
"limit": 330
"limit": 329
},
"reportAttributeAccessIssue": {
"limit": 516
},
"reportCallIssue": {
"limit": 124
"limit": 123
},
"reportConstantRedefinition": {
"limit": 59
"limit": 40
},
"reportDeprecated": {
"limit": 326
"limit": 325
},
"reportDuplicateImport": {
"limit": 42
"limit": 24
},
"reportExplicitAny": {
"limit": 10389
"limit": 9473
},
"reportFunctionMemberAccess": {
"limit": 11
},
"reportGeneralTypeIssues": {
"limit": 227
"limit": 157
},
"reportIncompatibleMethodOverride": {
"limit": 78
"limit": 77
},
"reportIncompatibleVariableOverride": {
"limit": 12
@ -42,7 +42,7 @@
"limit": 18
},
"reportIndexIssue": {
"limit": 37
"limit": 35
},
"reportInvalidTypeForm": {
"limit": 35
@ -54,13 +54,13 @@
"limit": 0
},
"reportMissingParameterType": {
"limit": 5900
"limit": 5855
},
"reportMissingTypeArgument": {
"limit": 15903
"limit": 15849
},
"reportMissingTypeStubs": {
"limit": 41
"limit": 40
},
"reportOperatorIssue": {
"limit": 0
@ -72,7 +72,7 @@
"limit": 0
},
"reportOptionalMemberAccess": {
"limit": 1085
"limit": 1079
},
"reportOptionalOperand": {
"limit": 0
@ -81,16 +81,16 @@
"limit": 0
},
"reportPossiblyUnboundVariable": {
"limit": 77
"limit": 56
},
"reportPrivateUsage": {
"limit": 2438
"limit": 2436
},
"reportRedeclaration": {
"limit": 12
"limit": 8
},
"reportReturnType": {
"limit": 225
"limit": 219
},
"reportTypedDictNotRequiredAccess": {
"limit": 27
@ -99,31 +99,31 @@
"limit": 0
},
"reportUnknownArgumentType": {
"limit": 45894
"limit": 45262
},
"reportUnknownLambdaType": {
"limit": 113
},
"reportUnknownMemberType": {
"limit": 40539
"limit": 40452
},
"reportUnknownParameterType": {
"limit": 20403
"limit": 20309
},
"reportUnknownVariableType": {
"limit": 32141
"limit": 31978
},
"reportUnnecessaryCast": {
"limit": 177
"limit": 124
},
"reportUnnecessaryComparison": {
"limit": 1025
"limit": 703
},
"reportUnnecessaryContains": {
"limit": 7
"limit": 5
},
"reportUnnecessaryIsInstance": {
"limit": 1209
"limit": 866
},
"reportUntypedBaseClass": {
"limit": 165
@ -132,15 +132,15 @@
"limit": 33
},
"reportUnusedClass": {
"limit": 33
"limit": 23
},
"reportUnusedFunction": {
"limit": 206
"limit": 139
},
"reportUnusedImport": {
"limit": 1005
"limit": 588
},
"reportUnusedVariable": {
"limit": 1297
"limit": 147
}
}

325
ci_cd/generate_model_prices_schema.py Normal file
View file

@ -0,0 +1,325 @@
from __future__ import annotations
import json
import sys
from pathlib import Path
from typing import Optional
import jsonschema
REPO_ROOT = Path(__file__).parent.parent
PRICES_PATH = REPO_ROOT / "model_prices_and_context_window.json"
SCHEMA_PATH = REPO_ROOT / "model_prices_and_context_window.schema.json"
SPECIAL_ROOT_KEYS = frozenset({"sample_spec", "fallback_generalizations"})
JsonSchema = dict
NONNEG_NUMBER: JsonSchema = {"type": "number", "minimum": 0}
NONNEG_INTEGER: JsonSchema = {"type": "integer", "minimum": 0}
BOOLEAN: JsonSchema = {"type": "boolean"}
STRING: JsonSchema = {"type": "string"}
EXTRA_BOOLEAN_KEYS = frozenset(
{
"gemini_native_audio",
"gemini_audio_only_live",
"uses_embed_content",
"use_openai_responses_path",
"bedrock_converse_supports_strict_tools",
}
)
OBJECT_KEYS: dict[str, JsonSchema] = {
"search_context_cost_per_query": {
"type": "object",
"description": "USD cost per web search query, keyed by search context size.",
"properties": {
"search_context_size_low": NONNEG_NUMBER,
"search_context_size_medium": NONNEG_NUMBER,
"search_context_size_high": NONNEG_NUMBER,
},
"additionalProperties": False,
},
"metadata": {
"type": "object",
"description": "Free-form notes about the entry (e.g. pricing derivation).",
},
"provider_specific_entry": {
"type": "object",
"description": "Provider-internal routing hints (e.g. bedrock_invocation_schema).",
},
}
ARRAY_KEYS: dict[str, JsonSchema] = {
"supported_endpoints": {
"type": "array",
"description": "OpenAI-style API routes this model can be called through, e.g. /v1/chat/completions.",
"items": STRING,
},
"supported_modalities": {
"type": "array",
"description": "Input modalities the model accepts.",
"items": {"type": "string", "enum": ["text", "image", "audio", "video"]},
},
"supported_output_modalities": {
"type": "array",
"description": "Output modalities the model can produce.",
"items": {"type": "string", "enum": ["text", "image", "audio", "video", "code"]},
},
"supported_regions": {
"type": "array",
"description": "Cloud regions the model is available in ('global' or region ids).",
"items": STRING,
},
"tiered_pricing": {
"type": "array",
"description": "Context-length or result-count tiered rates; each tier's costs apply within its range.",
"items": {
"type": "object",
"properties": {
"range": {
"type": "array",
"description": "[min, max] prompt-token span this tier applies to.",
"items": NONNEG_NUMBER,
"minItems": 2,
"maxItems": 2,
},
"max_results_range": {
"type": "array",
"description": "[min, max] result-count span this tier applies to (search models).",
"items": NONNEG_NUMBER,
"minItems": 2,
"maxItems": 2,
},
"input_cost_per_token": NONNEG_NUMBER,
"output_cost_per_token": NONNEG_NUMBER,
"output_cost_per_reasoning_token": NONNEG_NUMBER,
"cache_read_input_token_cost": NONNEG_NUMBER,
"input_cost_per_query": NONNEG_NUMBER,
},
"additionalProperties": False,
},
},
}
INTEGER_KEYS: dict[str, JsonSchema] = {
"max_tokens": {
**NONNEG_INTEGER,
"description": "Legacy field: max output tokens if the provider specifies it, else max input tokens.",
},
"max_input_tokens": {
**NONNEG_INTEGER,
"description": "Maximum prompt/context tokens the model accepts.",
},
"max_output_tokens": {
**NONNEG_INTEGER,
"description": "Maximum tokens the model can generate in one response.",
},
"output_vector_size": {
**NONNEG_INTEGER,
"description": "Embedding dimension for embedding models.",
},
"prompt_cache_min_tokens": {
**NONNEG_INTEGER,
"description": "Smallest prefix the provider will actually cache; absent means the provider default applies.",
},
"tpm": {**NONNEG_INTEGER, "description": "Provider default tokens-per-minute limit."},
"rpm": {**NONNEG_INTEGER, "description": "Provider default requests-per-minute limit."},
}
NUMBER_KEYS: dict[str, JsonSchema] = {
"regional_processing_uplift_multiplier_eu": {
"type": "number",
"minimum": 1,
"description": "Multiplier applied to all token costs for EU data residency (e.g. 1.10 = +10%).",
},
"regional_processing_uplift_multiplier_us": {
"type": "number",
"minimum": 1,
"description": "Multiplier applied to all token costs for US data residency (e.g. 1.10 = +10%).",
},
}
COST_DESCRIPTIONS: dict[str, str] = {
"input_cost_per_token": "USD per prompt token.",
"output_cost_per_token": "USD per generated token.",
"output_cost_per_reasoning_token": "USD per reasoning/thinking token, when billed separately.",
"cache_creation_input_token_cost": "USD per token written to the provider's prompt cache.",
"cache_read_input_token_cost": "USD per prompt token served from the provider's prompt cache.",
"input_cost_per_token_batches": "USD per prompt token via the provider's batch API.",
"output_cost_per_token_batches": "USD per generated token via the provider's batch API.",
}
def cost_description(key: str) -> Optional[str]:
if key in COST_DESCRIPTIONS:
return COST_DESCRIPTIONS[key]
if key.endswith("_flex"):
return "Flex service-tier rate for the same-named base field."
if key.endswith("_priority"):
return "Priority service-tier rate for the same-named base field."
if "_above_" in key:
return "Rate applied once the prompt exceeds the token threshold in the field name."
return None
def cost_schema(key: str) -> JsonSchema:
description = cost_description(key)
return {**NONNEG_NUMBER, "description": description} if description else dict(NONNEG_NUMBER)
def string_key_schemas(modes: tuple) -> dict[str, JsonSchema]:
return {
"litellm_provider": {
"type": "string",
"description": "LiteLLM provider slug; one of https://docs.litellm.ai/docs/providers.",
},
"mode": {
"type": "string",
"description": "Primary API surface / task type of the model.",
"enum": list(modes),
},
"source": {
"type": "string",
"description": "URL of the provider pricing/model page this entry was taken from.",
},
"deprecation_date": {
"type": "string",
"description": "Date the provider deprecates the model, YYYY-MM-DD.",
"format": "date",
"pattern": "^\\d{4}-(0[1-9]|1[0-2])-(0[1-9]|[12]\\d|3[01])$",
},
"web_search_billing_unit": {
"type": "string",
"description": "Whether web search is billed per query or per prompt.",
"enum": ["per_query", "per_prompt"],
},
"bedrock_output_config_effort_ceiling": {
"type": "string",
"description": "Highest reasoning effort the Bedrock output_config accepts for this model.",
"enum": ["low", "medium", "high", "max", "xhigh"],
},
"comment": STRING,
"audio_transcription_config": STRING,
}
def classify(key: str, modes: tuple) -> Optional[JsonSchema]:
curated = {**OBJECT_KEYS, **ARRAY_KEYS, **string_key_schemas(modes), **INTEGER_KEYS, **NUMBER_KEYS}
if key in curated:
return curated[key]
if key.startswith("supports_") or key in EXTRA_BOOLEAN_KEYS:
return BOOLEAN
if "cost" in key:
return cost_schema(key)
return None
def build_schema(prices: dict) -> JsonSchema:
entries = {name: entry for name, entry in prices.items() if name not in SPECIAL_ROOT_KEYS}
all_keys = tuple(sorted({key for entry in entries.values() for key in entry}))
modes = tuple(sorted({entry["mode"] for entry in entries.values() if "mode" in entry}))
unclassified = tuple(key for key in all_keys if classify(key, modes) is None)
if unclassified:
raise SystemExit(
f"Unclassified keys in {PRICES_PATH.name}: {', '.join(unclassified)}. "
f"Add them to the key tables in {Path(__file__).name} and rerun it."
)
entry_properties = {key: classify(key, modes) for key in all_keys}
return {
"$schema": "https://json-schema.org/draft/2020-12/schema",
"title": "LiteLLM model_prices_and_context_window.json",
"description": (
"Schema for LiteLLM's model price and context window registry "
"(https://github.com/BerriAI/litellm/blob/main/model_prices_and_context_window.json). "
"Every top-level key except 'sample_spec' and 'fallback_generalizations' is a model id, "
"optionally prefixed with its provider (e.g. 'azure/gpt-5.4'), mapping to a model entry. "
"All costs are USD per unit. New optional fields are added regularly, so consumers should "
"ignore unknown fields rather than reject them."
),
"type": "object",
"properties": {
"sample_spec": {
"type": "object",
"description": (
"Documentation placeholder illustrating the entry shape; not a real model and not "
"schema-conformant (several values are prose)."
),
},
"fallback_generalizations": {
"type": "object",
"description": "Regex rules that generalize unknown model ids to known families; not a model entry.",
"properties": {
"rules": {
"type": "array",
"items": {
"type": "object",
"properties": {
"name": STRING,
"pattern": STRING,
"description": STRING,
},
"required": ["name", "pattern"],
"additionalProperties": True,
},
}
},
"additionalProperties": False,
},
},
"additionalProperties": {"$ref": "#/$defs/modelEntry"},
"$defs": {
"modelEntry": {
"type": "object",
"description": (
"Pricing, limits, and capability flags for one model. Fields other than litellm_provider "
"are optional; boolean capability flags are simply omitted when unknown or false."
),
"required": ["litellm_provider"],
"properties": entry_properties,
"additionalProperties": True,
}
},
}
def render(schema: JsonSchema) -> str:
return json.dumps(schema, indent=2) + "\n"
def validation_errors(prices: dict, schema: JsonSchema) -> tuple:
validator = jsonschema.Draft202012Validator(
schema, format_checker=jsonschema.Draft202012Validator.FORMAT_CHECKER
)
return tuple(
f"{'.'.join(str(part) for part in error.absolute_path)}: {error.message}"
for error in validator.iter_errors(prices)
)
def main() -> int:
check = "--check" in sys.argv[1:]
prices = json.loads(PRICES_PATH.read_text())
rendered = render(build_schema(prices))
errors = validation_errors(prices, json.loads(rendered))
if errors:
print(f"{PRICES_PATH.name} does not validate against the generated schema:")
print("\n".join(errors[:20]))
return 1
if not check:
SCHEMA_PATH.write_text(rendered)
print(f"wrote {SCHEMA_PATH}")
return 0
if not SCHEMA_PATH.exists() or SCHEMA_PATH.read_text() != rendered:
print(
f"{SCHEMA_PATH.name} is out of sync with {PRICES_PATH.name}. "
f"Run `python {Path(__file__).relative_to(REPO_ROOT)}` and commit the result."
)
return 1
print(f"{SCHEMA_PATH.name} is in sync and {PRICES_PATH.name} validates against it")
return 0
if __name__ == "__main__":
sys.exit(main())

523
cookbook/litellm_proxy_server/grafana_dashboard/dashboard_genai_otel/grafana_dashboard.json Normal file
View file

@ -0,0 +1,523 @@
{
"annotations": {
"list": []
},
"editable": true,
"fiscalYearStartMonth": 0,
"graphTooltip": 0,
"links": [],
"panels": [
{
"type": "stat",
"title": "Requests",
"gridPos": {
"h": 4,
"w": 6,
"x": 0,
"y": 0
},
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"fieldConfig": {
"defaults": {
"unit": "short",
"decimals": 0,
"color": {
"mode": "fixed",
"fixedColor": "blue"
}
},
"overrides": []
},
"options": {
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"colorMode": "background",
"graphMode": "none"
},
"targets": [
{
"refId": "A",
"editorMode": "code",
"instant": true,
"expr": "sum(increase(gen_ai_client_operation_duration_seconds_count{service_name=~\"$service\", gen_ai_request_model=~\"$model\"}[$__range]))"
}
],
"id": 1
},
{
"type": "stat",
"title": "Spend",
"description": "LiteLLM's computed cost for the selected window, from gen_ai.usage.cost",
"gridPos": {
"h": 4,
"w": 6,
"x": 6,
"y": 0
},
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"fieldConfig": {
"defaults": {
"unit": "currencyUSD",
"decimals": 4,
"color": {
"mode": "fixed",
"fixedColor": "green"
}
},
"overrides": []
},
"options": {
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"colorMode": "background",
"graphMode": "none"
},
"targets": [
{
"refId": "A",
"editorMode": "code",
"instant": true,
"expr": "sum(increase(gen_ai_usage_cost_USD_sum{service_name=~\"$service\", gen_ai_request_model=~\"$model\"}[$__range]))"
}
],
"id": 2
},
{
"type": "stat",
"title": "Tokens",
"gridPos": {
"h": 4,
"w": 6,
"x": 12,
"y": 0
},
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"fieldConfig": {
"defaults": {
"unit": "short",
"decimals": 0,
"color": {
"mode": "fixed",
"fixedColor": "purple"
}
},
"overrides": []
},
"options": {
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"colorMode": "background",
"graphMode": "none"
},
"targets": [
{
"refId": "A",
"editorMode": "code",
"instant": true,
"expr": "sum(increase(gen_ai_client_token_usage_sum{service_name=~\"$service\", gen_ai_request_model=~\"$model\"}[$__range]))"
}
],
"id": 3
},
{
"type": "stat",
"title": "p95 request duration",
"gridPos": {
"h": 4,
"w": 6,
"x": 18,
"y": 0
},
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"fieldConfig": {
"defaults": {
"unit": "s",
"decimals": 2,
"color": {
"mode": "fixed",
"fixedColor": "orange"
}
},
"overrides": []
},
"options": {
"reduceOptions": {
"calcs": [
"lastNotNull"
],
"fields": "",
"values": false
},
"colorMode": "background",
"graphMode": "none"
},
"targets": [
{
"refId": "A",
"editorMode": "code",
"instant": true,
"expr": "histogram_quantile(0.95, sum by (le) (rate(gen_ai_client_operation_duration_seconds_bucket{service_name=~\"$service\", gen_ai_request_model=~\"$model\"}[$__range])))"
}
],
"id": 4
},
{
"type": "timeseries",
"title": "Request rate by model",
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 4
},
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"fieldConfig": {
"defaults": {
"unit": "reqpm",
"custom": {
"lineWidth": 2,
"fillOpacity": 8,
"showPoints": "never"
}
},
"overrides": []
},
"options": {
"legend": {
"displayMode": "list",
"placement": "bottom"
},
"tooltip": {
"mode": "multi",
"sort": "desc"
}
},
"targets": [
{
"refId": "A",
"editorMode": "code",
"legendFormat": "{{gen_ai_request_model}}",
"expr": "sum by (gen_ai_request_model) (rate(gen_ai_client_operation_duration_seconds_count{service_name=~\"$service\", gen_ai_request_model=~\"$model\"}[$__rate_interval])) * 60"
}
],
"id": 5
},
{
"type": "timeseries",
"title": "Spend rate by model",
"description": "USD per hour, derived from the gen_ai.usage.cost histogram",
"gridPos": {
"h": 8,
"w": 12,
"x": 12,
"y": 4
},
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"fieldConfig": {
"defaults": {
"unit": "currencyUSD",
"custom": {
"lineWidth": 2,
"fillOpacity": 8,
"showPoints": "never"
}
},
"overrides": []
},
"options": {
"legend": {
"displayMode": "list",
"placement": "bottom"
},
"tooltip": {
"mode": "multi",
"sort": "desc"
}
},
"targets": [
{
"refId": "A",
"editorMode": "code",
"legendFormat": "{{gen_ai_request_model}}",
"expr": "sum by (gen_ai_request_model) (rate(gen_ai_usage_cost_USD_sum{service_name=~\"$service\", gen_ai_request_model=~\"$model\"}[$__rate_interval])) * 3600"
}
],
"id": 6
},
{
"type": "timeseries",
"title": "Tokens per minute by model and type",
"description": "gen_ai.client.token.usage split by the gen_ai.token.type attribute",
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 12
},
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"fieldConfig": {
"defaults": {
"unit": "short",
"custom": {
"lineWidth": 2,
"fillOpacity": 8,
"showPoints": "never"
}
},
"overrides": []
},
"options": {
"legend": {
"displayMode": "list",
"placement": "bottom"
},
"tooltip": {
"mode": "multi",
"sort": "desc"
}
},
"targets": [
{
"refId": "A",
"editorMode": "code",
"legendFormat": "{{gen_ai_request_model}} {{gen_ai_token_type}}",
"expr": "sum by (gen_ai_request_model, gen_ai_token_type) (rate(gen_ai_client_token_usage_sum{service_name=~\"$service\", gen_ai_request_model=~\"$model\"}[$__rate_interval])) * 60"
}
],
"id": 7
},
{
"type": "timeseries",
"title": "p95 request duration by model",
"gridPos": {
"h": 8,
"w": 12,
"x": 12,
"y": 12
},
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"fieldConfig": {
"defaults": {
"unit": "s",
"custom": {
"lineWidth": 2,
"fillOpacity": 0,
"showPoints": "never"
}
},
"overrides": []
},
"options": {
"legend": {
"displayMode": "list",
"placement": "bottom"
},
"tooltip": {
"mode": "multi",
"sort": "desc"
}
},
"targets": [
{
"refId": "A",
"editorMode": "code",
"legendFormat": "{{gen_ai_request_model}}",
"expr": "histogram_quantile(0.95, sum by (le, gen_ai_request_model) (rate(gen_ai_client_operation_duration_seconds_bucket{service_name=~\"$service\", gen_ai_request_model=~\"$model\"}[$__rate_interval])))"
}
],
"id": 8
},
{
"type": "timeseries",
"title": "p95 time to first token (streaming)",
"description": "gen_ai.server.time_to_first_token, recorded only for streaming requests",
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 20
},
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"fieldConfig": {
"defaults": {
"unit": "s",
"custom": {
"lineWidth": 2,
"fillOpacity": 0,
"showPoints": "never"
}
},
"overrides": []
},
"options": {
"legend": {
"displayMode": "list",
"placement": "bottom"
},
"tooltip": {
"mode": "multi",
"sort": "desc"
}
},
"targets": [
{
"refId": "A",
"editorMode": "code",
"legendFormat": "{{gen_ai_request_model}}",
"expr": "histogram_quantile(0.95, sum by (le, gen_ai_request_model) (rate(gen_ai_server_time_to_first_token_seconds_bucket{service_name=~\"$service\", gen_ai_request_model=~\"$model\"}[$__rate_interval])))"
}
],
"id": 9
},
{
"type": "timeseries",
"title": "p95 provider generation time",
"description": "gen_ai.client.response.duration, upstream generation time excluding LiteLLM overhead",
"gridPos": {
"h": 8,
"w": 12,
"x": 12,
"y": 20
},
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"fieldConfig": {
"defaults": {
"unit": "s",
"custom": {
"lineWidth": 2,
"fillOpacity": 0,
"showPoints": "never"
}
},
"overrides": []
},
"options": {
"legend": {
"displayMode": "list",
"placement": "bottom"
},
"tooltip": {
"mode": "multi",
"sort": "desc"
}
},
"targets": [
{
"refId": "A",
"editorMode": "code",
"legendFormat": "{{gen_ai_request_model}}",
"expr": "histogram_quantile(0.95, sum by (le, gen_ai_request_model) (rate(gen_ai_client_response_duration_seconds_bucket{service_name=~\"$service\", gen_ai_request_model=~\"$model\"}[$__rate_interval])))"
}
],
"id": 10
}
],
"preload": false,
"refresh": "30s",
"schemaVersion": 42,
"tags": [
"litellm",
"genai",
"opentelemetry"
],
"templating": {
"list": [
{
"name": "datasource",
"label": "Prometheus",
"type": "datasource",
"query": "prometheus",
"current": {},
"hide": 0
},
{
"name": "service",
"label": "Service",
"type": "query",
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"query": "label_values(gen_ai_client_operation_duration_seconds_count, service_name)",
"refresh": 2,
"includeAll": true,
"multi": true,
"current": {
"text": "All",
"value": "$__all"
}
},
{
"name": "model",
"label": "Model",
"type": "query",
"datasource": {
"type": "prometheus",
"uid": "${datasource}"
},
"query": "label_values(gen_ai_client_operation_duration_seconds_count{service_name=~\"$service\"}, gen_ai_request_model)",
"refresh": 2,
"includeAll": true,
"multi": true,
"current": {
"text": "All",
"value": "$__all"
}
}
]
},
"time": {
"from": "now-1h",
"to": "now"
},
"timepicker": {},
"timezone": "browser",
"title": "LiteLLM GenAI (OpenTelemetry)",
"uid": "litellm-genai-otel",
"weekStart": ""
}

35
cookbook/litellm_proxy_server/grafana_dashboard/dashboard_genai_otel/readme.md Normal file
View file

@ -0,0 +1,35 @@
# LiteLLM GenAI dashboard (OpenTelemetry metrics)
Dashboard for the `gen_ai.*` metrics the OpenTelemetry v2 integration emits, as opposed to the `litellm_*` Prometheus metrics the other dashboards in this folder chart.
Import `grafana_dashboard.json` from **Dashboards > New > Import** and pick your Prometheus data source. Panels: request count, spend, token count, p95 duration, request rate by model, spend rate per hour by model, tokens per minute split by input and output, p95 duration by model, p95 time to first token, and p95 provider generation time. Template variables for data source, service, and model.
## Pre-requisites
Metrics are off by default. In the proxy environment:
```shell
LITELLM_OTEL_V2=true
LITELLM_OTEL_INTEGRATION_ENABLE_METRICS=true
OTEL_EXPORTER="otlp_http"
OTEL_ENDPOINT="<your OTLP endpoint>"
```
You also need the metric attribute filter, or the panels will plot flat lines at zero. LiteLLM's default attribute set includes per-request fields, so nearly every request lands in its own time series with a single sample, and `rate()` has nothing to compute over:
```yaml title="config.yaml"
callback_settings:
otel:
attributes:
include_list:
- gen_ai.operation.name
- gen_ai.system
- gen_ai.request.model
- gen_ai.framework
```
See [Grafana Cloud](https://docs.litellm.ai/docs/observability/grafana_cloud) for the full setup, and [OpenTelemetry v2](https://docs.litellm.ai/docs/observability/opentelemetry_v2#metrics) for the metric reference.
## Note on Grafana's AI Observability integration
Grafana Cloud ships prebuilt GenAI dashboards that query these same metric names, so they look like a drop-in alternative to this one. They are not: twenty of their twenty-two panels filter on `telemetry_sdk_name="openlit"`, a label LiteLLM does not carry and cannot be configured to add, so those panels stay empty.

4
cookbook/litellm_proxy_server/grafana_dashboard/readme.md
View file

@ -2,6 +2,10 @@
This folder contains the `json` for creating Grafana Dashboards
## [LiteLLM GenAI Dashboard (OpenTelemetry)](./dashboard_genai_otel)
Charts the `gen_ai.*` metrics from the OpenTelemetry v2 integration: spend, tokens, request rate, and latency percentiles by model. Separate from the dashboards below, which chart the `litellm_*` Prometheus metrics.
## [LiteLLM v2 Dashboard](./dashboard_v2)
<img width="1316" alt="grafana_1" src="https://github.com/user-attachments/assets/d0df802d-0cb9-4906-a679-941c547789ab">

46
db_scripts/backfill_daily_tool_spend.sql Normal file
View file

@ -0,0 +1,46 @@
-- One-shot backfill of the LiteLLM_DailyToolSpend rollup from the per-request
-- LiteLLM_SpendLogToolIndex x LiteLLM_SpendLogs tables.
--
-- This is an opt-in, manual operation. New deployments do not need it: the
-- rollup is written at request time from the moment the release is deployed.
-- Run it only if you want the Cost Optimization "Spend by tool" card to show
-- history from before the deploy, and only once.
--
-- IMPORTANT caveats before running:
--
-- 1. Pre-deploy index rows may include tools that were merely DECLARED in a
-- request body but never invoked (the release this ships with stops
-- recording those). For agentic clients that declare many tools per
-- request, backfilled history attributes each request's full spend to
-- every declared tool, overstating per-tool spend. Post-deploy rows do not
-- have this problem. If your traffic is mostly such clients, consider not
-- backfilling.
--
-- 2. Coverage is bounded by spend-log retention: rows older than
-- maximum_spend_logs_retention_period are already gone.
--
-- 3. Replace the cutover timestamp below with the time you deployed the
-- release, so backfilled per-request rows cannot double-count on top of
-- rollup rows the new writer already created. ON CONFLICT DO NOTHING is a
-- second guard for (date, tool_name) buckets the writer already touched:
-- such buckets keep the writer's numbers and skip the backfill's.
--
-- Usage:
-- psql "$DATABASE_URL" -v cutover="'2026-07-25T00:00:00Z'" -f db_scripts/backfill_daily_tool_spend.sql
SET TIME ZONE 'UTC';
INSERT INTO "LiteLLM_DailyToolSpend" (date, tool_name, spend, total_tokens, request_count, created_at, updated_at)
SELECT
to_char(ti.start_time, 'YYYY-MM-DD') AS date,
ti.tool_name,
COALESCE(SUM(sl.spend), 0) AS spend,
COALESCE(SUM(sl.total_tokens), 0) AS total_tokens,
COUNT(*) AS request_count,
now() AS created_at,
now() AS updated_at
FROM "LiteLLM_SpendLogToolIndex" ti
JOIN "LiteLLM_SpendLogs" sl ON sl.request_id = ti.request_id
WHERE ti.start_time < :cutover::timestamptz
GROUP BY 1, 2
ON CONFLICT (date, tool_name) DO NOTHING;

2
docker/Dockerfile.database
View file

@ -7,7 +7,7 @@ ARG LITELLM_BUILD_IMAGE=cgr.dev/chainguard/wolfi-base@sha256:42df77a9974d6ec8b17
ARG LITELLM_RUNTIME_IMAGE=cgr.dev/chainguard/wolfi-base@sha256:42df77a9974d6ec8b17a5ee8bc23b532600a44d705acef2409e0933c1251b45f
ARG UV_IMAGE=ghcr.io/astral-sh/uv:0.11.7@sha256:240fb85ab0f263ef12f492d8476aa3a2e4e1e333f7d67fbdd923d00a506a516a
# Pinned by digest like the other base images; bump explicitly on Node upgrades.
ARG UI_BUILD_IMAGE=node:20.18-alpine3.20@sha256:3488b10bf958af7125a176419d2d8a9937d895bf124012aae811651988d2ffe6
ARG UI_BUILD_IMAGE=node:24.19-alpine3.24@sha256:d32cdf619f63fe0471182d08996dd516c6275bb5fd31ae06e55a570bd9e1ad43
FROM $UV_IMAGE AS uvbin

2
docker/Dockerfile.non_root
View file

@ -6,7 +6,7 @@ ARG LITELLM_RUNTIME_IMAGE=cgr.dev/chainguard/wolfi-base@sha256:42df77a9974d6ec8b
ARG PROXY_EXTRAS_SOURCE=published
ARG UV_IMAGE=ghcr.io/astral-sh/uv:0.11.7@sha256:240fb85ab0f263ef12f492d8476aa3a2e4e1e333f7d67fbdd923d00a506a516a
# Pinned by digest like the other base images; bump explicitly on Node upgrades.
ARG UI_BUILD_IMAGE=node:20.18-alpine3.20@sha256:3488b10bf958af7125a176419d2d8a9937d895bf124012aae811651988d2ffe6
ARG UI_BUILD_IMAGE=node:24.19-alpine3.24@sha256:d32cdf619f63fe0471182d08996dd516c6275bb5fd31ae06e55a570bd9e1ad43
FROM $UV_IMAGE AS uvbin

5
docker/build_admin_ui.sh
View file

@ -54,8 +54,9 @@ else
fi || { echo "nvm checksum verification failed"; exit 1; }
bash "$NVM_SCRIPT"
source ~/.nvm/nvm.sh
nvm install v18.17.0
nvm use v18.17.0
NODE_VERSION="$(cat ui/litellm-dashboard/.nvmrc)"
nvm install "v${NODE_VERSION}"
nvm use "v${NODE_VERSION}"
# cd in to /ui/litellm-dashboard

8
docker/component_entrypoint.sh Executable file
View file

@ -0,0 +1,8 @@
#!/bin/sh
if [ "$USE_DDTRACE" = "true" ]; then
export DD_TRACE_OPENAI_ENABLED="False"
exec ddtrace-run "$@"
fi
exec "$@"

33
enterprise/litellm_enterprise/proxy/common_utils/check_batch_cost.py
View file

@ -17,6 +17,7 @@ if TYPE_CHECKING:
from litellm.proxy._types import LiteLLM_ManagedObjectTable
from litellm.proxy.utils import PrismaClient, ProxyLogging
from litellm.router import Router
from litellm.types.router import Deployment
from litellm.types.utils import LiteLLMBatch
@ -281,6 +282,32 @@ class CheckBatchCost:
return deployment_id
return None
@classmethod
def _get_managed_file_model_name(
cls,
job: "LiteLLM_ManagedObjectTable",
deployment_info: "Deployment",
) -> Optional[str]:
"""
Public model group name to encode as ``target_model_names`` on unified output file ids.
Key model-access checks resolve a managed file id back to a model via its
``target_model_names``, so this must be the model group the caller requested, never the
underlying provider model (e.g. ``gpt-5.5``), which no key is allowed to call.
"""
from litellm.proxy.openai_files_endpoints.common_utils import (
convert_b64_uid_to_unified_uid,
get_models_from_unified_file_id,
)
input_file_id = cls._get_input_file_id(job)
target_model_names = (
get_models_from_unified_file_id(convert_b64_uid_to_unified_uid(input_file_id)) if input_file_id else []
)
if target_model_names:
return ",".join(target_model_names)
return deployment_info.model_name or None
@staticmethod
def _get_input_file_id(job: "LiteLLM_ManagedObjectTable") -> Optional[str]:
import json
@ -406,6 +433,10 @@ class CheckBatchCost:
managed_files_hook = self.proxy_logging_obj.get_proxy_hook("managed_files")
if managed_files_hook is not None:
from litellm.proxy._types import UserAPIKeyAuth
managed_file_model_name = self._get_managed_file_model_name(
job=job, deployment_info=deployment_info
)
_minimal_auth = UserAPIKeyAuth(
user_id=job.created_by or "default-user-id",
team_id=getattr(job, "team_id", None),
@ -417,7 +448,7 @@ class CheckBatchCost:
_unified_file_id = managed_files_hook.get_unified_output_file_id(
output_file_id=_raw_file_id,
model_id=model_id,
model_name=str(model_name) if model_name else deployment_info.model_name or None,
model_name=managed_file_model_name,
)
await managed_files_hook.store_unified_file_id(
file_id=_unified_file_id,

8
enterprise/litellm_enterprise/proxy/hooks/managed_files.py
View file

@ -215,7 +215,7 @@ class _PROXY_LiteLLMManagedFiles(CustomLogger, BaseFileEndpoints):
)
if result:
return LiteLLM_ManagedFileTable(**result)
return LiteLLM_ManagedFileTable.model_validate(result)
## CHECK DB
db_object = await self.prisma_client.db.litellm_managedfiletable.find_first(
@ -223,7 +223,7 @@ class _PROXY_LiteLLMManagedFiles(CustomLogger, BaseFileEndpoints):
)
if db_object:
return LiteLLM_ManagedFileTable(**db_object.model_dump())
return LiteLLM_ManagedFileTable.model_validate(db_object.model_dump())
return None
async def delete_unified_file_id(
@ -349,7 +349,7 @@ class _PROXY_LiteLLMManagedFiles(CustomLogger, BaseFileEndpoints):
if isinstance(batch.file_object, str)
else batch.file_object
)
batch_obj = LiteLLMBatch(**batch_data)
batch_obj = LiteLLMBatch.model_validate(batch_data)
batch_obj.id = batch.unified_object_id
batch_objects.append(batch_obj)
@ -382,7 +382,7 @@ class _PROXY_LiteLLMManagedFiles(CustomLogger, BaseFileEndpoints):
"flat_model_file_ids": {"hasSome": model_object_ids},
}
)
return [OpenAIFileObject(**file_object.file_object) for file_object in file_ids]
return [OpenAIFileObject.model_validate(file_object.file_object) for file_object in file_ids]
async def check_managed_file_id_access(
self, data: Dict, user_api_key_dict: UserAPIKeyAuth

191
enterprise/litellm_enterprise/proxy/management_endpoints/project_endpoints.py
View file

@ -11,7 +11,8 @@ Endpoints for /project operations
#### PROJECT MANAGEMENT ####
import json
from typing import List, Optional, Union
from collections.abc import Mapping, Sequence
from typing import TYPE_CHECKING
from fastapi import APIRouter, Depends, HTTPException, Request
@ -25,15 +26,24 @@ from litellm.proxy.management_helpers.utils import (
)
from litellm.proxy.utils import PrismaClient, handle_exception_on_proxy
if TYPE_CHECKING:
from prisma import models as prisma_models
from prisma.actions import LiteLLM_TeamTableActions
router = APIRouter()
def _team_table(prisma_client: PrismaClient) -> "LiteLLM_TeamTableActions[prisma_models.LiteLLM_TeamTable]":
team_table: LiteLLM_TeamTableActions[prisma_models.LiteLLM_TeamTable] = prisma_client.db.litellm_teamtable
return team_table
async def _check_user_permission_for_project(
user_api_key_dict: UserAPIKeyAuth,
team_id: Optional[str],
team_id: str | None,
prisma_client: PrismaClient,
require_admin: bool = False,
team_object: Optional[LiteLLM_TeamTable] = None,
team_object: LiteLLM_TeamTable | None = None,
) -> bool:
"""
Check if user has permission to manage a project.
@ -57,9 +67,7 @@ async def _check_user_permission_for_project(
team = team_object
if team is None:
team = await prisma_client.db.litellm_teamtable.find_unique(
where={"team_id": team_id}
)
team = await _team_table(prisma_client).find_unique(where={"team_id": team_id})
if team and team.admins:
return user_api_key_dict.user_id in team.admins
@ -70,9 +78,9 @@ async def _check_user_permission_for_project(
async def _validate_team_exists(
team_id: str,
prisma_client: PrismaClient,
):
) -> "prisma_models.LiteLLM_TeamTable":
"""Validate that a team exists. Returns the team row."""
team = await prisma_client.db.litellm_teamtable.find_unique(
team = await _team_table(prisma_client).find_unique(
where={"team_id": team_id},
)
@ -89,7 +97,7 @@ async def _validate_team_exists(
def _check_team_project_limits(
team_object: LiteLLM_TeamTable,
data: Union[NewProjectRequest, UpdateProjectRequest],
data: NewProjectRequest | UpdateProjectRequest,
) -> None:
"""
Check that project limits respect its parent Team's limits.
@ -108,16 +116,12 @@ def _check_team_project_limits(
if data.max_budget is not None and data.max_budget < 0:
raise HTTPException(
status_code=400,
detail={
"error": f"max_budget cannot be negative. Received: {data.max_budget}"
},
detail={"error": f"max_budget cannot be negative. Received: {data.max_budget}"},
)
if data.soft_budget is not None and data.soft_budget < 0:
raise HTTPException(
status_code=400,
detail={
"error": f"soft_budget cannot be negative. Received: {data.soft_budget}"
},
detail={"error": f"soft_budget cannot be negative. Received: {data.soft_budget}"},
)
# --- soft_budget < max_budget ---
@ -131,7 +135,7 @@ def _check_team_project_limits(
)
# --- Validate project models are a subset of team models ---
project_models = getattr(data, "models", None)
project_models = data.models
team_models = team_object.models or []
if project_models and len(team_models) > 0:
# If team has 'all-proxy-models', skip validation as it allows all models
@ -148,11 +152,7 @@ def _check_team_project_limits(
# --- Validate project max_budget <= team max_budget ---
# Team stores budget fields directly (max_budget, tpm_limit, rpm_limit)
# unlike Project which uses a separate LiteLLM_BudgetTable relation
if (
data.max_budget is not None
and team_object.max_budget is not None
and data.max_budget > team_object.max_budget
):
if data.max_budget is not None and team_object.max_budget is not None and data.max_budget > team_object.max_budget:
raise HTTPException(
status_code=400,
detail={
@ -161,11 +161,7 @@ def _check_team_project_limits(
)
# --- Validate project tpm_limit <= team tpm_limit ---
if (
data.tpm_limit is not None
and team_object.tpm_limit is not None
and data.tpm_limit > team_object.tpm_limit
):
if data.tpm_limit is not None and team_object.tpm_limit is not None and data.tpm_limit > team_object.tpm_limit:
raise HTTPException(
status_code=400,
detail={
@ -174,11 +170,7 @@ def _check_team_project_limits(
)
# --- Validate project rpm_limit <= team rpm_limit ---
if (
data.rpm_limit is not None
and team_object.rpm_limit is not None
and data.rpm_limit > team_object.rpm_limit
):
if data.rpm_limit is not None and team_object.rpm_limit is not None and data.rpm_limit > team_object.rpm_limit:
raise HTTPException(
status_code=400,
detail={
@ -189,19 +181,19 @@ def _check_team_project_limits(
async def _create_budget_for_project(
data: NewProjectRequest,
user_id: Optional[str],
user_id: str | None,
litellm_proxy_admin_name: str,
prisma_client: PrismaClient,
) -> str:
"""Create a budget for the project and return budget_id."""
budget_params = LiteLLM_BudgetTable.model_fields.keys()
_json_data = data.json(exclude_none=True)
_json_data: Mapping[str, object] = data.json(exclude_none=True)
_budget_data = {k: v for k, v in _json_data.items() if k in budget_params}
budget_row = LiteLLM_BudgetTable(**_budget_data)
budget_row = LiteLLM_BudgetTable.model_validate(_budget_data)
new_budget = prisma_client.jsonify_object(budget_row.json(exclude_none=True))
_budget = await prisma_client.db.litellm_budgettable.create(
_budget: prisma_models.LiteLLM_BudgetTable = await prisma_client.db.litellm_budgettable.create(
data={
**new_budget,
"created_by": user_id or litellm_proxy_admin_name,
@ -214,8 +206,8 @@ async def _create_budget_for_project(
async def _set_project_object_permission(
data: NewProjectRequest,
prisma_client: Optional[PrismaClient],
) -> Optional[str]:
prisma_client: PrismaClient | None,
) -> str | None:
"""
Creates the LiteLLM_ObjectPermissionTable record for the project.
Returns the object_permission_id if created, otherwise None.
@ -224,7 +216,7 @@ async def _set_project_object_permission(
return None
if data.object_permission is not None:
created_object_permission = (
created_object_permission: prisma_models.LiteLLM_ObjectPermissionTable = (
await prisma_client.db.litellm_objectpermissiontable.create(
data=data.object_permission.model_dump(exclude_none=True),
)
@ -344,8 +336,7 @@ async def new_project(
raise HTTPException(
status_code=403,
detail={
"error": "Only premium users can add tags to projects. "
+ CommonProxyErrors.not_premium_user.value
"error": "Only premium users can add tags to projects. " + CommonProxyErrors.not_premium_user.value
},
)
@ -353,8 +344,7 @@ async def new_project(
raise HTTPException(
status_code=403,
detail={
"error": "Project management is an enterprise feature. "
+ CommonProxyErrors.not_premium_user.value
"error": "Project management is an enterprise feature. " + CommonProxyErrors.not_premium_user.value
},
)
@ -375,13 +365,11 @@ async def new_project(
)
# Validate team exists and get team object with budget
team_object = await _validate_team_exists(
team_id=data.team_id, prisma_client=prisma_client
)
team_object = await _validate_team_exists(team_id=data.team_id, prisma_client=prisma_client)
# Validate project limits against team limits
_check_team_project_limits(
team_object=LiteLLM_TeamTable(**team_object.model_dump()),
team_object=LiteLLM_TeamTable.model_validate(team_object.model_dump()),
data=data,
)
@ -391,7 +379,7 @@ async def new_project(
user_api_key_dict=user_api_key_dict,
team_id=data.team_id,
prisma_client=prisma_client,
team_object=LiteLLM_TeamTable(**team_object.model_dump()),
team_object=LiteLLM_TeamTable.model_validate(team_object.model_dump()),
)
if not has_permission:
@ -449,17 +437,13 @@ async def new_project(
value=getattr(data, field),
)
new_project_row = prisma_client.jsonify_object(
project_row.json(exclude_none=True)
)
new_project_row = prisma_client.jsonify_object(project_row.json(exclude_none=True))
# Remove budget fields (following organization_endpoints.py pattern)
new_project_row = _remove_budget_fields_from_project_data(new_project_row)
verbose_proxy_logger.info(
f"new_project_row: {json.dumps(new_project_row, indent=2)}"
)
response = await prisma_client.db.litellm_projecttable.create(
verbose_proxy_logger.info(f"new_project_row: {json.dumps(new_project_row, indent=2)}")
response: prisma_models.LiteLLM_ProjectTable = await prisma_client.db.litellm_projecttable.create(
data={
**new_project_row, # type: ignore
},
@ -469,9 +453,7 @@ async def new_project(
return response
except Exception as e:
verbose_proxy_logger.exception(
"litellm.proxy.management_endpoints.project_endpoints.new_project(): Exception occured - {}".format(
str(e)
)
"litellm.proxy.management_endpoints.project_endpoints.new_project(): Exception occured - {}".format(str(e))
)
raise handle_exception_on_proxy(e)
@ -539,8 +521,7 @@ async def update_project(
raise HTTPException(
status_code=403,
detail={
"error": "Only premium users can add tags to projects. "
+ CommonProxyErrors.not_premium_user.value
"error": "Only premium users can add tags to projects. " + CommonProxyErrors.not_premium_user.value
},
)
@ -548,8 +529,7 @@ async def update_project(
raise HTTPException(
status_code=403,
detail={
"error": "Project management is an enterprise feature. "
+ CommonProxyErrors.not_premium_user.value
"error": "Project management is an enterprise feature. " + CommonProxyErrors.not_premium_user.value
},
)
@ -576,9 +556,9 @@ async def update_project(
)
# Fetch existing project
existing_project = await prisma_client.db.litellm_projecttable.find_unique(
where={"project_id": data.project_id}
)
existing_project: (
prisma_models.LiteLLM_ProjectTable | None
) = await prisma_client.db.litellm_projecttable.find_unique(where={"project_id": data.project_id})
if existing_project is None:
raise ProxyException(
@ -595,9 +575,7 @@ async def update_project(
target_team_id = data.team_id or existing_project.team_id
target_team_obj = None
if target_team_id is not None:
target_team_obj = await _validate_team_exists(
team_id=target_team_id, prisma_client=prisma_client
)
target_team_obj = await _validate_team_exists(team_id=target_team_id, prisma_client=prisma_client)
has_permission = await _check_user_permission_for_project(
user_api_key_dict=user_api_key_dict,
@ -620,32 +598,26 @@ async def update_project(
team_id=data.team_id,
prisma_client=prisma_client,
team_object=(
LiteLLM_TeamTable(**target_team_obj.model_dump())
if target_team_obj
else None
LiteLLM_TeamTable.model_validate(target_team_obj.model_dump()) if target_team_obj else None
),
)
if not can_assign_to_target:
raise HTTPException(
status_code=403,
detail={
"error": "Cannot reassign project to a team you are not an admin of"
},
detail={"error": "Cannot reassign project to a team you are not an admin of"},
)
# Validate project limits against team limits
if target_team_obj is not None:
_check_team_project_limits(
team_object=LiteLLM_TeamTable(**target_team_obj.model_dump()),
team_object=LiteLLM_TeamTable.model_validate(target_team_obj.model_dump()),
data=data,
)
# Prepare update data
update_data = data.json(exclude_none=True, exclude={"project_id"})
update_data = prisma_client.jsonify_object(update_data)
update_data["updated_by"] = (
user_api_key_dict.user_id or litellm_proxy_admin_name
)
update_data["updated_by"] = user_api_key_dict.user_id or litellm_proxy_admin_name
# Handle budget updates
budget_fields = LiteLLM_BudgetTable.model_fields.keys()
@ -671,21 +643,17 @@ async def update_project(
if existing_project.object_permission_id:
# Update existing permission
await prisma_client.db.litellm_objectpermissiontable.update(
where={
"object_permission_id": existing_project.object_permission_id
},
where={"object_permission_id": existing_project.object_permission_id},
data=object_permission_data,
)
else:
# Create new permission
created_permission = (
created_permission: prisma_models.LiteLLM_ObjectPermissionTable = (
await prisma_client.db.litellm_objectpermissiontable.create(
data=object_permission_data,
)
)
update_data["object_permission_id"] = (
created_permission.object_permission_id
)
update_data["object_permission_id"] = created_permission.object_permission_id
# Handle metadata fields
for field in LiteLLM_ManagementEndpoint_MetadataFields:
@ -698,7 +666,7 @@ async def update_project(
update_data = _remove_budget_fields_from_project_data(update_data)
# Update project
updated_project = await prisma_client.db.litellm_projecttable.update(
updated_project: prisma_models.LiteLLM_ProjectTable | None = await prisma_client.db.litellm_projecttable.update(
where={"project_id": data.project_id},
data=update_data,
include={"litellm_budget_table": True, "object_permission": True},
@ -718,7 +686,7 @@ async def update_project(
"/project/delete",
tags=["project management"],
dependencies=[Depends(user_api_key_auth)],
response_model=List[LiteLLM_ProjectTable],
response_model=list[LiteLLM_ProjectTable],
)
@management_endpoint_wrapper
async def delete_project(
@ -749,8 +717,7 @@ async def delete_project(
raise HTTPException(
status_code=403,
detail={
"error": "Project management is an enterprise feature. "
+ CommonProxyErrors.not_premium_user.value
"error": "Project management is an enterprise feature. " + CommonProxyErrors.not_premium_user.value
},
)
@ -778,9 +745,7 @@ async def delete_project(
for project_id in data.project_ids:
# Check if project exists
existing_project = await prisma_client.db.litellm_projecttable.find_unique(
where={"project_id": project_id}
)
existing_project = await prisma_client.db.litellm_projecttable.find_unique(where={"project_id": project_id})
if existing_project is None:
raise ProxyException(
@ -791,11 +756,9 @@ async def delete_project(
)
# Check if there are any keys associated with this project
associated_keys = (
await prisma_client.db.litellm_verificationtoken.find_many(
where={"project_id": project_id}
)
)
associated_keys: Sequence[
prisma_models.LiteLLM_VerificationToken
] = await prisma_client.db.litellm_verificationtoken.find_many(where={"project_id": project_id})
if len(associated_keys) > 0:
raise ProxyException(
@ -806,9 +769,9 @@ async def delete_project(
)
# Delete the project
deleted_project = await prisma_client.db.litellm_projecttable.delete(
where={"project_id": project_id}
)
deleted_project: (
prisma_models.LiteLLM_ProjectTable | None
) = await prisma_client.db.litellm_projecttable.delete(where={"project_id": project_id})
deleted_projects.append(deleted_project)
@ -854,7 +817,7 @@ async def project_info(
)
# Fetch project
project = await prisma_client.db.litellm_projecttable.find_unique(
project: prisma_models.LiteLLM_ProjectTable | None = await prisma_client.db.litellm_projecttable.find_unique(
where={"project_id": project_id},
include={"litellm_budget_table": True, "object_permission": True},
)
@ -872,17 +835,11 @@ async def project_info(
is_team_member = False
if project.team_id and user_api_key_dict.user_id:
team = await prisma_client.db.litellm_teamtable.find_unique(
where={"team_id": project.team_id}
)
team = await _team_table(prisma_client).find_unique(where={"team_id": project.team_id})
if team:
caller_user_id = user_api_key_dict.user_id
for m in team.members_with_roles or []:
m_user_id = (
m.get("user_id")
if isinstance(m, dict)
else getattr(m, "user_id", None)
)
m_user_id = m.get("user_id") if isinstance(m, dict) else getattr(m, "user_id", None)
if m_user_id == caller_user_id:
is_team_member = True
break
@ -896,9 +853,7 @@ async def project_info(
return project
except Exception as e:
verbose_proxy_logger.exception(
"litellm.proxy.management_endpoints.project_endpoints.project_info(): Exception occured - {}".format(
str(e)
)
"litellm.proxy.management_endpoints.project_endpoints.project_info(): Exception occured - {}".format(str(e))
)
raise handle_exception_on_proxy(e)
@ -907,7 +862,7 @@ async def project_info(
"/project/list",
tags=["project management"],
dependencies=[Depends(user_api_key_auth)],
response_model=List[LiteLLM_ProjectTable],
response_model=list[LiteLLM_ProjectTable],
)
async def list_projects(
user_api_key_dict: UserAPIKeyAuth = Depends(user_api_key_auth),
@ -932,21 +887,19 @@ async def list_projects(
# If proxy admin, get all projects
if user_api_key_dict.user_role == LitellmUserRoles.PROXY_ADMIN:
projects = await prisma_client.db.litellm_projecttable.find_many(
projects: Sequence[
prisma_models.LiteLLM_ProjectTable
] = await prisma_client.db.litellm_projecttable.find_many(
include={"litellm_budget_table": True, "object_permission": True}
)
else:
# Look up the user's team memberships via the reverse-index on
# LiteLLM_UserTable.teams (maintained by team_member_add alongside
# members_with_roles). This avoids a full scan of all team rows.
user_record = await prisma_client.db.litellm_usertable.find_unique(
user_record: prisma_models.LiteLLM_UserTable | None = await prisma_client.db.litellm_usertable.find_unique(
where={"user_id": user_api_key_dict.user_id},
)
user_team_ids = (
user_record.teams
if user_record is not None and user_record.teams
else []
)
user_team_ids: Sequence[str] = user_record.teams if user_record is not None and user_record.teams else []
projects = await prisma_client.db.litellm_projecttable.find_many(
where={"team_id": {"in": user_team_ids}},

4
enterprise/pyproject.toml
View file

@ -1,6 +1,6 @@
[project]
name = "litellm-enterprise"
version = "0.1.52"
version = "0.1.53"
description = "Package for LiteLLM Enterprise features"
readme = "README.md"
requires-python = ">=3.9"
@ -26,7 +26,7 @@ required-version = ">=0.10.9"
module-root = ""
[tool.commitizen]
version = "0.1.52"
version = "0.1.53"
version_files = [
"pyproject.toml:^version",
"../pyproject.toml:litellm-enterprise==",

4
gateway/Dockerfile
View file

@ -65,6 +65,8 @@ RUN mkdir -p /home/nonroot && \
HOME=/home/nonroot prisma generate --schema=./schema.prisma && \
chown -R nonroot:nonroot /home/nonroot/.cache
RUN sed -i 's/\r$//' docker/component_entrypoint.sh && chmod +x docker/component_entrypoint.sh
# ---------- Runtime ----------
FROM $LITELLM_RUNTIME_IMAGE AS runtime
@ -95,5 +97,5 @@ USER nonroot
EXPOSE 4000/tcp
ENTRYPOINT ["sh", "-c", "exec uvicorn gateway.main:app --workers \"${NUM_WORKERS:-1}\" \"$@\"", "--"]
ENTRYPOINT ["sh", "-c", "exec /app/docker/component_entrypoint.sh uvicorn gateway.main:app --workers \"${NUM_WORKERS:-1}\" \"$@\"", "--"]
CMD ["--host", "0.0.0.0", "--port", "4000"]

1
gateway/routes/allowlist.py
View file

@ -54,6 +54,7 @@ GATEWAY_PATH_PREFIXES: tuple[str, ...] = (
"/messages",
"/v1/skills",
"/v1/a2a/",
"/a2a/",
# LiteLLM-native LLM surface
"/v1/rerank",
"/v2/rerank",

4
helm/litellm-helm/Chart.lock
View file

@ -5,5 +5,5 @@ dependencies:
- name: redis
repository: oci://registry-1.docker.io/bitnamicharts
version: 18.19.1
digest: sha256:8660fe6287f9941d08c0902f3f13731079b8cecd2a5da2fbc54e5b7aae4a6f62
generated: "2024-03-10T02:28:52.275022+05:30"
digest: sha256:38962e231f6596b93f82a8412bbe4cf5de696caecf5775dfbbd163383eb1c009
generated: "2026-07-28T10:21:22.511401-07:00"

6
helm/litellm-helm/Chart.yaml
View file

@ -18,7 +18,7 @@ type: application
# This is the chart version. This version number should be incremented each time you make changes
# to the chart and its templates, including the app version.
# Versions are expected to follow Semantic Versioning (https://semver.org/)
version: 1.1.0
version: 1.1.1
# This is the version number of the application being deployed. This version number should be
# incremented each time you make changes to the application. Versions are not expected to
@ -32,10 +32,10 @@ annotations:
dependencies:
- name: "postgresql"
version: ">=13.3.0"
version: "14.3.1"
repository: oci://registry-1.docker.io/bitnamicharts
condition: db.deployStandalone
- name: redis
version: ">=18.0.0"
version: "18.19.1"
repository: oci://registry-1.docker.io/bitnamicharts
condition: redis.enabled

12
helm/litellm-helm/README.md
View file

@ -39,7 +39,7 @@ If `db.useStackgresOperator` is used (not yet implemented):
| `livenessProbe.*` | Liveness probe settings for the LiteLLM container (`path`, `periodSeconds`, `timeoutSeconds`, thresholds, and initial delay). | See `values.yaml` |
| `readinessProbe.*` | Readiness probe settings for the LiteLLM container (`path`, `periodSeconds`, `timeoutSeconds`, thresholds, and initial delay). | See `values.yaml` |
| `startupProbe.*` | Startup probe settings for the LiteLLM container (`path`, `periodSeconds`, `timeoutSeconds`, thresholds, and initial delay). | See `values.yaml` |
| `resources.*` | CPU/memory requests and limits for the LiteLLM container. | `{}` |
| `resources.*` | CPU/memory requests and limits for the LiteLLM container. Unset by default; production deployments should set 1 CPU and 4Gi of memory per worker. | `{}` |
| `service.loadBalancerClass` | Optional LoadBalancer implementation class (only used when `service.type` is `LoadBalancer`) | `""` |
| `ingress.labels` | Additional labels for the Ingress resource | `{}` |
| `ingress.*` | See [values.yaml](./values.yaml) for example settings | N/A |
@ -130,6 +130,16 @@ Set `billingMetrics.caSecretName` only when the collector is a private or test o
| `db.deployStandalone` | Deploy a standalone, single instance deployment of Postgres, using the Bitnami postgresql chart. This is useful for getting started but doesn't provide HA or (by default) data backups. | `true` |
| `postgresql.*` | If `db.deployStandalone` is `true`, configuration passed to the Bitnami postgresql chart. See the [Bitnami Documentation](https://github.com/bitnami/charts/tree/main/bitnami/postgresql) for full configuration details. See [values.yaml](./values.yaml) for the default configuration. | See [values.yaml](./values.yaml) |
| `postgresql.auth.*` | If `db.deployStandalone` is `true`, care should be taken to ensure the default `password` and `postgres-password` values are **NOT** used. | `NoTaGrEaTpAsSwOrD` |
| `postgresql.image.*` | If `db.deployStandalone` is `true`, the image for the bundled Postgres. Pinned to a `docker.io/bitnamilegacy` build because Bitnami retired the versioned tags under `docker.io/bitnami`. | `bitnamilegacy/postgresql:16.2.0-debian-12-r6` |
| `redis.image.*` | If `redis.enabled` is `true`, the image for the bundled Redis. Pinned to a `docker.io/bitnamilegacy` build for the same reason. | `bitnamilegacy/redis:7.2.4-debian-12-r9` |
#### Bundled Postgres image
Bitnami removed the versioned tags from `docker.io/bitnami` and republished the archived builds under `docker.io/bitnamilegacy`, so the image defaults that ship inside the `postgresql` and `redis` subcharts no longer pull. The chart pins both to the `bitnamilegacy` copies of the exact builds those subchart versions were released with, which keeps the on-disk data directory layout unchanged for existing installs.
Keep `postgresql.image.tag` pinned. `docker.io/bitnami/postgresql` still publishes a floating `latest`, and pointing the bundled Postgres at a different major version starts the server against a data directory it cannot read (`database files are incompatible with server`). There is no in-place way back, so crossing a major version means dumping the database with the old image and restoring it into the new one. The chart refuses to render when the tag is empty or `latest`.
Those images no longer receive updates. For anything beyond getting started, run Postgres outside the chart and point at it with `db.useExisting`.
#### Example Postgres `db.useExisting` Secret

15
helm/litellm-helm/templates/_helpers.tpl
View file

@ -146,3 +146,18 @@ Get redis service port
{{ .Values.redis.master.service.ports.redis }}
{{- end -}}
{{- end -}}
{{/*
Reject an unpinned image tag for the bundled PostgreSQL.
A floating tag lets a chart upgrade start a newer PostgreSQL major against the
existing PersistentVolumeClaim. The server then refuses to start on a data
directory written by another major version, and the only way back is a dump
taken before the change, which by that point no longer exists.
*/}}
{{- define "litellm.validateBundledPostgresImageTag" -}}
{{- $tag := .Values.postgresql.image.tag | default "" | toString -}}
{{- $digest := .Values.postgresql.image.digest | default "" | toString -}}
{{- if and (eq $digest "") (or (eq $tag "") (eq $tag "latest")) -}}
{{- fail (printf "postgresql.image.tag must be pinned to an explicit version when db.deployStandalone is true (got %q). An unpinned tag can start a different PostgreSQL major against the existing data directory, which makes the database unreadable and is not recoverable in place. Crossing a major version requires a dump and restore." $tag) -}}
{{- end -}}
{{- end -}}

2
helm/litellm-helm/templates/migrations-job.yaml
View file

@ -35,6 +35,8 @@ spec:
{{- toYaml . | nindent 8 }}
{{- end }}
serviceAccountName: {{ include "litellm.migrationServiceAccountName" . }}
securityContext:
{{- toYaml .Values.podSecurityContext | nindent 8 }}
{{- with .Values.migrationJob.extraInitContainers }}
initContainers:
{{- tpl (toYaml .) $ | nindent 8 }}

1
helm/litellm-helm/templates/secret-dbcredentials.yaml
View file

@ -1,4 +1,5 @@
{{- if .Values.db.deployStandalone -}}
{{- include "litellm.validateBundledPostgresImageTag" . -}}
apiVersion: v1
kind: Secret
metadata:

2
helm/litellm-helm/templates/tests/test-servicemonitor.yaml
View file

@ -10,7 +10,7 @@ metadata:
spec:
containers:
- name: test
image: bitnami/kubectl:latest
image: docker.io/bitnamilegacy/kubectl:1.29.2-debian-12-r3
command: ['sh', '-c']
args:
- |

94
helm/litellm-helm/tests/bundled_db_images_tests.yaml Normal file
View file

@ -0,0 +1,94 @@
suite: test bundled database images
templates:
- charts/postgresql/templates/primary/statefulset.yaml
- charts/redis/templates/master/application.yaml
- charts/redis/templates/configmap.yaml
- charts/redis/templates/health-configmap.yaml
- charts/redis/templates/scripts-configmap.yaml
- charts/redis/templates/secret.yaml
- secret-dbcredentials.yaml
- templates/tests/test-servicemonitor.yaml
tests:
- it: should pull the bundled postgres from a repository that still publishes the pinned tag
template: charts/postgresql/templates/primary/statefulset.yaml
set:
db.deployStandalone: true
asserts:
- equal:
path: spec.template.spec.containers[0].image
value: docker.io/bitnamilegacy/postgresql:16.2.0-debian-12-r6
- it: should pull the bundled postgres metrics exporter from the same repository
template: charts/postgresql/templates/primary/statefulset.yaml
set:
db.deployStandalone: true
postgresql.metrics.enabled: true
asserts:
- equal:
path: spec.template.spec.containers[1].image
value: docker.io/bitnamilegacy/postgres-exporter:0.15.0-debian-12-r14
- it: should run the bundled postgres init container from the same repository
template: charts/postgresql/templates/primary/statefulset.yaml
set:
db.deployStandalone: true
postgresql.volumePermissions.enabled: true
asserts:
- equal:
path: spec.template.spec.initContainers[0].image
value: docker.io/bitnamilegacy/os-shell:12-debian-12-r16
- it: should pull the bundled redis from a repository that still publishes the pinned tag
template: charts/redis/templates/master/application.yaml
set:
redis.enabled: true
asserts:
- equal:
path: spec.template.spec.containers[0].image
value: docker.io/bitnamilegacy/redis:7.2.4-debian-12-r9
- it: should reject a floating postgres tag that could cross a major version on an existing volume
template: secret-dbcredentials.yaml
set:
db.deployStandalone: true
postgresql.image.tag: latest
asserts:
- failedTemplate:
errorMessage: 'postgresql.image.tag must be pinned to an explicit version when db.deployStandalone is true (got "latest"). An unpinned tag can start a different PostgreSQL major against the existing data directory, which makes the database unreadable and is not recoverable in place. Crossing a major version requires a dump and restore.'
- it: should reject an empty postgres tag
template: secret-dbcredentials.yaml
set:
db.deployStandalone: true
postgresql.image.tag: ""
asserts:
- failedTemplate:
errorMessage: 'postgresql.image.tag must be pinned to an explicit version when db.deployStandalone is true (got ""). An unpinned tag can start a different PostgreSQL major against the existing data directory, which makes the database unreadable and is not recoverable in place. Crossing a major version requires a dump and restore.'
- it: should accept an empty postgres tag when the image is pinned by digest
template: secret-dbcredentials.yaml
set:
db.deployStandalone: true
postgresql.image.tag: ""
postgresql.image.digest: sha256:0d0e2f1a5b3c4d6e7f8091a2b3c4d5e6f708192a3b4c5d6e7f8091a2b3c4d5e6
asserts:
- hasDocuments:
count: 1
- it: should run the servicemonitor test pod from a pinned image
template: templates/tests/test-servicemonitor.yaml
set:
serviceMonitor.enabled: true
asserts:
- equal:
path: spec.containers[0].image
value: docker.io/bitnamilegacy/kubectl:1.29.2-debian-12-r3
- it: should not constrain the postgres tag when the bundled database is not deployed
template: secret-dbcredentials.yaml
set:
db.deployStandalone: false
postgresql.image.tag: latest
asserts:
- hasDocuments:
count: 0

36
helm/litellm-helm/tests/migrations-job_tests.yaml
View file

@ -254,3 +254,39 @@ tests:
content:
name: sidecar-tpl
image: "ghcr.io/berriai/litellm-database:test"
- it: should render the pod-level securityContext from podSecurityContext
template: migrations-job.yaml
set:
migrationJob:
enabled: true
podSecurityContext:
fsGroup: 10000
runAsUser: 10000
runAsNonRoot: true
asserts:
- equal:
path: spec.template.spec.securityContext
value:
fsGroup: 10000
runAsUser: 10000
runAsNonRoot: true
- it: should keep the pod-level and container-level securityContext separate
template: migrations-job.yaml
set:
migrationJob:
enabled: true
podSecurityContext:
fsGroup: 10000
securityContext:
allowPrivilegeEscalation: false
readOnlyRootFilesystem: true
asserts:
- equal:
path: spec.template.spec.securityContext
value:
fsGroup: 10000
- equal:
path: spec.template.spec.containers[0].securityContext
value:
allowPrivilegeEscalation: false
readOnlyRootFilesystem: true

78
helm/litellm-helm/values.yaml
View file

@ -181,16 +181,19 @@ proxy_config:
resources:
{}
# We usually recommend not to specify default resources and to leave this as a conscious
# choice for the user. This also increases chances charts run on environments with little
# resources, such as Minikube. If you do want to specify resources, uncomment the following
# lines, adjust them as necessary, and remove the curly braces after 'resources:'.
# limits:
# cpu: 100m
# memory: 128Mi
# Unset by default so the chart installs on small clusters such as Minikube, and so an
# upgrade never leaves a running pod Pending. Production deployments should set these.
# A proxy at DB-connected steady state needs about 1 CPU and 4Gi of memory per worker;
# sizing below that gets the pod OOMKilled once traffic and DB connections ramp up.
# Scale both figures with --num_workers, then uncomment the lines below and remove the
# curly braces after 'resources:'. See "Recommended Machine Specifications" in
# https://docs.litellm.ai/docs/proxy/prod.
# requests:
# cpu: 100m
# memory: 128Mi
# cpu: "1"
# memory: 4Gi
# limits:
# cpu: "1"
# memory: 4Gi
autoscaling:
enabled: false
@ -328,8 +331,32 @@ lifecycle: {}
# Settings for Bitnami postgresql chart (if db.deployStandalone is true, ignored
# otherwise)
#
# Bitnami retired the versioned tags under docker.io/bitnami and republished the
# archived builds under docker.io/bitnamilegacy, so the subchart's own image
# defaults no longer resolve. The repository below points at the same build the
# subchart was released with, which keeps the on-disk data directory layout
# identical for existing installs.
#
# Keep the tag pinned. docker.io/bitnami still publishes a floating `latest`,
# and starting a newer PostgreSQL major against an existing data directory
# leaves the server refusing to boot ("database files are incompatible with
# server") with no way back other than a dump taken beforehand. Crossing a major
# version is a dump-and-restore, not an image bump. The chart refuses to render
# an unpinned tag for this reason
postgresql:
architecture: standalone
image:
repository: bitnamilegacy/postgresql
tag: 16.2.0-debian-12-r6
volumePermissions:
image:
repository: bitnamilegacy/os-shell
tag: 12-debian-12-r16
metrics:
image:
repository: bitnamilegacy/postgres-exporter
tag: 0.15.0-debian-12-r14
auth:
username: litellm
database: litellm
@ -359,9 +386,36 @@ postgresql:
# When `redis.sentinel.enabled` is set, the coordination block is rendered with
# `sentinel_nodes` and `service_name` (from `redis.sentinel.masterSet`) instead
# of host/port, because a plain Redis client cannot talk to the sentinel port
#
# The image repositories carry the same bitnamilegacy repoint as postgresql
# above; the versioned tags the subchart ships with are gone from
# docker.io/bitnami
redis:
enabled: false
architecture: standalone
image:
repository: bitnamilegacy/redis
tag: 7.2.4-debian-12-r9
sentinel:
image:
repository: bitnamilegacy/redis-sentinel
tag: 7.2.4-debian-12-r7
metrics:
image:
repository: bitnamilegacy/redis-exporter
tag: 1.58.0-debian-12-r4
volumePermissions:
image:
repository: bitnamilegacy/os-shell
tag: 12-debian-12-r16
sysctl:
image:
repository: bitnamilegacy/os-shell
tag: 12-debian-12-r16
kubectl:
image:
repository: bitnamilegacy/kubectl
tag: 1.29.2-debian-12-r3
coordination:
# Set to false to keep the bundled Redis for response caching only and leave
# `general_settings.coordination_redis` out of the rendered config. A
@ -381,9 +435,9 @@ migrationJob:
annotations: {}
ttlSecondsAfterFinished: 120
resources: {}
# requests:
# cpu: 100m
# memory: 100Mi
# Unset by default. This job runs the database migration and exits, so it does not
# need the steady-state headroom the proxy does; size it from your own migration
# runs rather than from the proxy figures above.
extraContainers: []
extraInitContainers: []

53
helm/litellm/templates/_helpers.tpl
View file

@ -138,6 +138,59 @@ is false the chart uses the provided name, or the namespace `default` SA.
{{- end -}}
{{- end -}}
{{/*
ServiceAccount name for the migrations Job.
The Job is a pre-install / pre-upgrade hook, so it is created before the
chart's ordinary resources. A ServiceAccount the chart creates is one of
those ordinary resources, which makes borrowing the backend name a cycle:
the hook pod is rejected because the account does not exist yet. So when
`serviceAccounts.backend.create` is true the Job falls back to the namespace
`default` account unless the operator names one that already exists. With
`create` false the backend name is either an operator-supplied existing
account or `default`, both of which are safe for the hook, so the Job keeps
sharing it.
`migrationJob.serviceAccountName` always wins when set, which is how a Job
that needs credentials of its own (IRSA / Workload Identity for IAM database
auth) gets them.
*/}}
{{- define "litellm.migrations.serviceAccountName" -}}
{{- if .Values.migrationJob.serviceAccountName -}}
{{ .Values.migrationJob.serviceAccountName }}
{{- else if .Values.serviceAccounts.backend.create -}}
default
{{- else -}}
{{ include "litellm.backend.serviceAccountName" . }}
{{- end -}}
{{- end -}}
{{/*
Extra pod labels for a component's Deployment, validated against its selector.
Invoke with a dict:
(dict "podLabels" .Values.gateway.podLabels "componentName" "gateway")
The three selector keys are also emitted on the pod template, so a podLabels
entry reusing one renders a duplicate YAML key whose later value wins. That
leaves the pod template no longer matching the (immutable) selector and the
apiserver rejects the Deployment. Fail at template time naming the key
instead, so the operator gets the reason here rather than an opaque
`selector does not match template labels` from the apiserver.
The migrations Job takes podLabels unvalidated: a Job's selector is generated
by the controller rather than declared, so nothing there can collide.
*/}}
{{- define "litellm.podLabels" -}}
{{- $componentName := .componentName -}}
{{- range $key, $value := .podLabels }}
{{- if has $key (list "app.kubernetes.io/name" "app.kubernetes.io/instance" "app.kubernetes.io/component") }}
{{- fail (printf "%s.podLabels cannot set %s: it is part of the Deployment's immutable selector" $componentName $key) }}
{{- end }}
{{- end }}
{{- toYaml .podLabels }}
{{- end -}}
{{/*
Master-key + database + redis env block — shared by gateway, backend, and the
migrations Job.

22
helm/litellm/templates/backend/deployment.yaml
View file

@ -23,9 +23,16 @@ spec:
{{- end }}
labels:
{{- include "litellm.backend.selectorLabels" . | nindent 8 }}
{{- with .Values.backend.podLabels }}
{{- include "litellm.podLabels" (dict "podLabels" . "componentName" "backend") | nindent 8 }}
{{- end }}
spec:
serviceAccountName: {{ include "litellm.backend.serviceAccountName" . }}
automountServiceAccountToken: {{ .Values.serviceAccounts.backend.automount }}
{{- with .Values.backend.podSecurityContext }}
securityContext:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- with .Values.imagePullSecrets }}
imagePullSecrets:
{{- toYaml . | nindent 8 }}
@ -34,6 +41,10 @@ spec:
- name: backend
image: "{{ .Values.backend.image.repository }}:{{ .Values.backend.image.tag | default .Chart.AppVersion }}"
imagePullPolicy: {{ .Values.backend.image.pullPolicy }}
{{- with .Values.backend.securityContext }}
securityContext:
{{- toYaml . | nindent 12 }}
{{- end }}
ports:
- name: http
containerPort: 4001
@ -70,8 +81,15 @@ spec:
readinessProbe:
{{- toYaml . | nindent 12 }}
{{- end }}
{{- with .Values.backend.lifecycle }}
lifecycle:
{{- toYaml . | nindent 12 }}
{{- end }}
resources:
{{- toYaml .Values.backend.resources | nindent 12 }}
{{- with .Values.backend.extraContainers }}
{{- tpl (toYaml .) $ | nindent 8 }}
{{- end }}
{{- if or .Values.gateway.config.create .Values.backend.volumes .Values.billingMetrics.enabled }}
volumes:
{{- if .Values.gateway.config.create }}
@ -102,4 +120,8 @@ spec:
topologySpreadConstraints:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- $gracePeriod := .Values.backend.terminationGracePeriodSeconds }}
{{- if not (or (kindIs "invalid" $gracePeriod) (eq (printf "%v" $gracePeriod) "")) }}
terminationGracePeriodSeconds: {{ $gracePeriod }}
{{- end }}
{{- end }}

22
helm/litellm/templates/gateway/deployment.yaml
View file

@ -21,9 +21,16 @@ spec:
{{- end }}
labels:
{{- include "litellm.gateway.selectorLabels" . | nindent 8 }}
{{- with .Values.gateway.podLabels }}
{{- include "litellm.podLabels" (dict "podLabels" . "componentName" "gateway") | nindent 8 }}
{{- end }}
spec:
serviceAccountName: {{ include "litellm.gateway.serviceAccountName" . }}
automountServiceAccountToken: {{ .Values.serviceAccounts.gateway.automount }}
{{- with .Values.gateway.podSecurityContext }}
securityContext:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- with .Values.imagePullSecrets }}
imagePullSecrets:
{{- toYaml . | nindent 8 }}
@ -32,6 +39,10 @@ spec:
- name: gateway
image: "{{ .Values.gateway.image.repository }}:{{ .Values.gateway.image.tag | default .Chart.AppVersion }}"
imagePullPolicy: {{ .Values.gateway.image.pullPolicy }}
{{- with .Values.gateway.securityContext }}
securityContext:
{{- toYaml . | nindent 12 }}
{{- end }}
ports:
- name: http
containerPort: 4000
@ -72,8 +83,15 @@ spec:
readinessProbe:
{{- toYaml . | nindent 12 }}
{{- end }}
{{- with .Values.gateway.lifecycle }}
lifecycle:
{{- toYaml . | nindent 12 }}
{{- end }}
resources:
{{- toYaml .Values.gateway.resources | nindent 12 }}
{{- with .Values.gateway.extraContainers }}
{{- tpl (toYaml .) $ | nindent 8 }}
{{- end }}
{{- if or .Values.gateway.config.create .Values.gateway.volumes .Values.billingMetrics.enabled }}
volumes:
{{- if .Values.gateway.config.create }}
@ -104,4 +122,8 @@ spec:
topologySpreadConstraints:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- $gracePeriod := .Values.gateway.terminationGracePeriodSeconds }}
{{- if not (or (kindIs "invalid" $gracePeriod) (eq (printf "%v" $gracePeriod) "")) }}
terminationGracePeriodSeconds: {{ $gracePeriod }}
{{- end }}
{{- end }}

2
helm/litellm/templates/ingress.yaml
View file

@ -19,7 +19,7 @@
"/v1/fine-tuning" "/fine-tuning" "/v1/responses" "/responses" "/v1/threads" "/threads"
"/v1/assistants" "/assistants" "/v1/vector_stores" "/vector_stores" "/v1/indexes"
"/v1/models" "/models" "/openai" "/engines"
"/v1/messages" "/messages" "/v1/skills" "/v1/a2a"
"/v1/messages" "/messages" "/v1/skills" "/v1/a2a" "/a2a"
"/v1/rerank" "/v2/rerank" "/rerank" "/v1/ocr" "/ocr" "/v1/rag" "/rag"
"/v1/video" "/v1/videos" "/video" "/videos" "/v1/search" "/search"
"/v1/containers" "/containers" "/v1/evals" "/v1/memory" "/queue/chat"

27
helm/litellm/templates/migrations-job.yaml
View file

@ -23,12 +23,21 @@ spec:
ttlSecondsAfterFinished: {{ .Values.migrationJob.ttlSecondsAfterFinished }}
template:
metadata:
{{- /* The Job's selector is generated by the controller rather than
declared, so podLabels may override a chart label here. Merge
instead of appending so an override replaces the key rather than
rendering it twice. */}}
{{- $chartLabels := merge (dict "app.kubernetes.io/component" "migrations") (fromYaml (include "litellm.commonLabels" .)) }}
labels:
{{- include "litellm.commonLabels" . | nindent 8 }}
app.kubernetes.io/component: migrations
{{- toYaml (merge (deepCopy .Values.migrationJob.podLabels) $chartLabels) | nindent 8 }}
spec:
restartPolicy: Never
serviceAccountName: {{ include "litellm.backend.serviceAccountName" . }}
serviceAccountName: {{ include "litellm.migrations.serviceAccountName" . }}
automountServiceAccountToken: {{ .Values.migrationJob.automountServiceAccountToken }}
{{- with .Values.migrationJob.podSecurityContext }}
securityContext:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- with .Values.imagePullSecrets }}
imagePullSecrets:
{{- toYaml . | nindent 8 }}
@ -37,10 +46,22 @@ spec:
- name: prisma-migrations
image: "{{ .Values.migrationJob.image.repository }}:{{ .Values.migrationJob.image.tag | default .Chart.AppVersion }}"
imagePullPolicy: {{ .Values.migrationJob.image.pullPolicy }}
{{- with .Values.migrationJob.securityContext }}
securityContext:
{{- toYaml . | nindent 12 }}
{{- end }}
env:
{{- include "litellm.serverEnv" (dict "root" $ "component" .Values.migrationJob) | nindent 12 }}
{{- with .Values.migrationJob.volumeMounts }}
volumeMounts:
{{- toYaml . | nindent 12 }}
{{- end }}
{{- with .Values.migrationJob.resources }}
resources:
{{- toYaml . | nindent 12 }}
{{- end }}
{{- with .Values.migrationJob.volumes }}
volumes:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- end }}

22
helm/litellm/templates/ui/deployment.yaml
View file

@ -18,9 +18,16 @@ spec:
{{- end }}
labels:
{{- include "litellm.ui.selectorLabels" . | nindent 8 }}
{{- with .Values.ui.podLabels }}
{{- include "litellm.podLabels" (dict "podLabels" . "componentName" "ui") | nindent 8 }}
{{- end }}
spec:
serviceAccountName: {{ include "litellm.ui.serviceAccountName" . }}
automountServiceAccountToken: {{ .Values.serviceAccounts.ui.automount }}
{{- with .Values.ui.podSecurityContext }}
securityContext:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- with .Values.imagePullSecrets }}
imagePullSecrets:
{{- toYaml . | nindent 8 }}
@ -29,6 +36,10 @@ spec:
- name: ui
image: "{{ .Values.ui.image.repository }}:{{ .Values.ui.image.tag | default .Chart.AppVersion }}"
imagePullPolicy: {{ .Values.ui.image.pullPolicy }}
{{- with .Values.ui.securityContext }}
securityContext:
{{- toYaml . | nindent 12 }}
{{- end }}
ports:
- name: http
containerPort: 3000
@ -58,8 +69,15 @@ spec:
readinessProbe:
{{- toYaml . | nindent 12 }}
{{- end }}
{{- with .Values.ui.lifecycle }}
lifecycle:
{{- toYaml . | nindent 12 }}
{{- end }}
resources:
{{- toYaml .Values.ui.resources | nindent 12 }}
{{- with .Values.ui.extraContainers }}
{{- tpl (toYaml .) $ | nindent 8 }}
{{- end }}
{{- with .Values.ui.volumes }}
volumes:
{{- toYaml . | nindent 8 }}
@ -80,4 +98,8 @@ spec:
topologySpreadConstraints:
{{- toYaml . | nindent 8 }}
{{- end }}
{{- $gracePeriod := .Values.ui.terminationGracePeriodSeconds }}
{{- if not (or (kindIs "invalid" $gracePeriod) (eq (printf "%v" $gracePeriod) "")) }}
terminationGracePeriodSeconds: {{ $gracePeriod }}
{{- end }}
{{- end }}

169
helm/litellm/tests/migration_job_tests.yaml Normal file
View file

@ -0,0 +1,169 @@
suite: test migrations Job ServiceAccount resolution and pod hardening
templates:
- migrations-job.yaml
values:
- ./values/required.yaml
tests:
- it: borrows the namespace default account when no ServiceAccount is configured
asserts:
- equal:
path: spec.template.spec.serviceAccountName
value: default
- it: falls back to the namespace default account when the chart creates the backend ServiceAccount
set:
serviceAccounts.backend.create: true
asserts:
- equal:
path: spec.template.spec.serviceAccountName
value: default
- notEqual:
path: spec.template.spec.serviceAccountName
value: RELEASE-NAME-litellm-backend
- it: keeps sharing an existing backend ServiceAccount the chart does not create
set:
serviceAccounts.backend.create: false
serviceAccounts.backend.name: existing-backend-sa
asserts:
- equal:
path: spec.template.spec.serviceAccountName
value: existing-backend-sa
- it: prefers an explicit migration ServiceAccount over the created backend one
set:
serviceAccounts.backend.create: true
migrationJob.serviceAccountName: migrations-sa
asserts:
- equal:
path: spec.template.spec.serviceAccountName
value: migrations-sa
- it: prefers an explicit migration ServiceAccount over an existing backend one
set:
serviceAccounts.backend.create: false
serviceAccounts.backend.name: existing-backend-sa
migrationJob.serviceAccountName: migrations-sa
asserts:
- equal:
path: spec.template.spec.serviceAccountName
value: migrations-sa
- it: mounts no ServiceAccount token by default
asserts:
- equal:
path: spec.template.spec.automountServiceAccountToken
value: false
- it: mounts a ServiceAccount token when the operator asks for one
set:
migrationJob.automountServiceAccountToken: true
asserts:
- equal:
path: spec.template.spec.automountServiceAccountToken
value: true
- it: keeps the token off the Job when the backend disables automounting
set:
serviceAccounts.backend.create: true
serviceAccounts.backend.automount: false
asserts:
- equal:
path: spec.template.spec.serviceAccountName
value: default
- equal:
path: spec.template.spec.automountServiceAccountToken
value: false
- it: renders no hardening fields by default
asserts:
- isNull:
path: spec.template.spec.securityContext
- isNull:
path: spec.template.spec.containers[0].securityContext
- isNull:
path: spec.template.spec.volumes
- isNull:
path: spec.template.spec.containers[0].volumeMounts
- equal:
path: spec.template.metadata.labels
value:
app.kubernetes.io/name: litellm
app.kubernetes.io/instance: RELEASE-NAME
app.kubernetes.io/managed-by: Helm
helm.sh/chart: litellm-0.1.0
app.kubernetes.io/component: migrations
- it: renders pod-level and container-level securityContext in their own scopes
set:
migrationJob.podSecurityContext:
runAsNonRoot: true
runAsUser: 65532
seccompProfile:
type: RuntimeDefault
migrationJob.securityContext:
allowPrivilegeEscalation: false
readOnlyRootFilesystem: true
capabilities:
drop:
- ALL
asserts:
- equal:
path: spec.template.spec.securityContext
value:
runAsNonRoot: true
runAsUser: 65532
seccompProfile:
type: RuntimeDefault
- equal:
path: spec.template.spec.containers[0].securityContext
value:
allowPrivilegeEscalation: false
readOnlyRootFilesystem: true
capabilities:
drop:
- ALL
- it: renders volumes on the pod and volumeMounts on the migration container
set:
migrationJob.volumes:
- name: tmp
emptyDir:
sizeLimit: 64Mi
migrationJob.volumeMounts:
- name: tmp
mountPath: /tmp
asserts:
- equal:
path: spec.template.spec.volumes
value:
- name: tmp
emptyDir:
sizeLimit: 64Mi
- equal:
path: spec.template.spec.containers[0].volumeMounts
value:
- name: tmp
mountPath: /tmp
- it: merges podLabels with the chart labels on the Job pod
set:
migrationJob.podLabels:
egress-policy: restricted
asserts:
- equal:
path: spec.template.metadata.labels['egress-policy']
value: restricted
- equal:
path: spec.template.metadata.labels['app.kubernetes.io/component']
value: migrations
- it: accepts a podLabel that reuses a chart label, since the Job selector is controller-generated
set:
migrationJob.podLabels:
app.kubernetes.io/component: batch-migrations
asserts:
- notFailedTemplate: {}
- equal:
path: spec.template.metadata.labels['app.kubernetes.io/component']
value: batch-migrations

298
helm/litellm/tests/pod_hardening_tests.yaml Normal file
View file

@ -0,0 +1,298 @@
suite: test pod hardening knobs on the component deployments
templates:
- gateway/deployment.yaml
- gateway/configmap.yaml
- backend/deployment.yaml
- ui/deployment.yaml
values:
- ./values/required.yaml
tests:
- it: gateway renders no hardening fields by default
template: gateway/deployment.yaml
asserts:
- isNull:
path: spec.template.spec.securityContext
- isNull:
path: spec.template.spec.containers[0].securityContext
- isNull:
path: spec.template.spec.containers[0].lifecycle
- isNull:
path: spec.template.spec.terminationGracePeriodSeconds
- lengthEqual:
path: spec.template.spec.containers
count: 1
- equal:
path: spec.template.metadata.labels
value:
app.kubernetes.io/name: litellm
app.kubernetes.io/instance: RELEASE-NAME
app.kubernetes.io/component: gateway
- it: gateway renders pod-level and container-level securityContext in their own scopes
template: gateway/deployment.yaml
set:
gateway.podSecurityContext:
runAsNonRoot: true
runAsUser: 65532
fsGroup: 65532
seccompProfile:
type: RuntimeDefault
gateway.securityContext:
allowPrivilegeEscalation: false
readOnlyRootFilesystem: true
capabilities:
drop:
- ALL
asserts:
- equal:
path: spec.template.spec.securityContext
value:
runAsNonRoot: true
runAsUser: 65532
fsGroup: 65532
seccompProfile:
type: RuntimeDefault
- equal:
path: spec.template.spec.containers[0].securityContext
value:
allowPrivilegeEscalation: false
readOnlyRootFilesystem: true
capabilities:
drop:
- ALL
- it: gateway merges podLabels with the selector labels
template: gateway/deployment.yaml
set:
gateway.podLabels:
egress-policy: restricted
team: platform
asserts:
- equal:
path: spec.template.metadata.labels
value:
app.kubernetes.io/name: litellm
app.kubernetes.io/instance: RELEASE-NAME
app.kubernetes.io/component: gateway
egress-policy: restricted
team: platform
- equal:
path: spec.selector.matchLabels
value:
app.kubernetes.io/name: litellm
app.kubernetes.io/instance: RELEASE-NAME
app.kubernetes.io/component: gateway
- it: gateway rejects a podLabel that collides with the selector
template: gateway/deployment.yaml
set:
gateway.podLabels:
app.kubernetes.io/component: not-gateway
asserts:
- failedTemplate:
errorMessage: "gateway.podLabels cannot set app.kubernetes.io/component: it is part of the Deployment's immutable selector"
- it: backend rejects a podLabel that collides with the selector
template: backend/deployment.yaml
set:
backend.podLabels:
app.kubernetes.io/name: not-litellm
asserts:
- failedTemplate:
errorMessage: "backend.podLabels cannot set app.kubernetes.io/name: it is part of the Deployment's immutable selector"
- it: ui rejects a podLabel that collides with the selector
template: ui/deployment.yaml
set:
ui.podLabels:
app.kubernetes.io/instance: not-the-release
asserts:
- failedTemplate:
errorMessage: "ui.podLabels cannot set app.kubernetes.io/instance: it is part of the Deployment's immutable selector"
- it: gateway renders lifecycle hooks on the container
template: gateway/deployment.yaml
set:
gateway.lifecycle:
preStop:
httpGet:
path: /health/drain
port: 4000
asserts:
- equal:
path: spec.template.spec.containers[0].lifecycle
value:
preStop:
httpGet:
path: /health/drain
port: 4000
- it: gateway renders terminationGracePeriodSeconds on the pod spec
template: gateway/deployment.yaml
set:
gateway.terminationGracePeriodSeconds: 90
asserts:
- equal:
path: spec.template.spec.terminationGracePeriodSeconds
value: 90
- it: gateway honors an explicit terminationGracePeriodSeconds of zero
template: gateway/deployment.yaml
set:
gateway.terminationGracePeriodSeconds: 0
asserts:
- equal:
path: spec.template.spec.terminationGracePeriodSeconds
value: 0
- it: gateway appends extraContainers after the gateway container
template: gateway/deployment.yaml
set:
gateway.extraContainers:
- name: auth-sidecar
image: registry.example.com/auth-proxy:1.2.3
args:
- --upstream
- http://127.0.0.1:4000
asserts:
- lengthEqual:
path: spec.template.spec.containers
count: 2
- equal:
path: spec.template.spec.containers[0].name
value: gateway
- equal:
path: spec.template.spec.containers[1]
value:
name: auth-sidecar
image: registry.example.com/auth-proxy:1.2.3
args:
- --upstream
- http://127.0.0.1:4000
- it: gateway templates chart context inside extraContainers
template: gateway/deployment.yaml
set:
gateway.extraContainers:
- name: auth-sidecar
image: registry.example.com/auth-proxy:1.2.3
env:
- name: RELEASE
value: "{{ .Release.Name }}"
asserts:
- equal:
path: spec.template.spec.containers[1].env[0].value
value: RELEASE-NAME
- it: backend renders every hardening knob in the right scope
template: backend/deployment.yaml
set:
backend.podLabels:
egress-policy: restricted
backend.podSecurityContext:
runAsNonRoot: true
backend.securityContext:
readOnlyRootFilesystem: true
backend.lifecycle:
preStop:
exec:
command:
- sleep
- "5"
backend.terminationGracePeriodSeconds: 60
backend.extraContainers:
- name: auth-sidecar
image: registry.example.com/auth-proxy:1.2.3
asserts:
- equal:
path: spec.template.metadata.labels['egress-policy']
value: restricted
- equal:
path: spec.template.spec.securityContext
value:
runAsNonRoot: true
- equal:
path: spec.template.spec.containers[0].securityContext
value:
readOnlyRootFilesystem: true
- equal:
path: spec.template.spec.containers[0].lifecycle
value:
preStop:
exec:
command:
- sleep
- "5"
- equal:
path: spec.template.spec.terminationGracePeriodSeconds
value: 60
- equal:
path: spec.template.spec.containers[1].name
value: auth-sidecar
- it: ui renders every hardening knob in the right scope
template: ui/deployment.yaml
set:
ui.podLabels:
egress-policy: restricted
ui.podSecurityContext:
runAsNonRoot: true
fsGroup: 101
ui.securityContext:
readOnlyRootFilesystem: true
ui.lifecycle:
preStop:
exec:
command:
- /bin/sh
- -c
- nginx -s quit
ui.terminationGracePeriodSeconds: 30
ui.extraContainers:
- name: auth-sidecar
image: registry.example.com/auth-proxy:1.2.3
asserts:
- equal:
path: spec.template.metadata.labels['egress-policy']
value: restricted
- equal:
path: spec.template.spec.securityContext
value:
runAsNonRoot: true
fsGroup: 101
- equal:
path: spec.template.spec.containers[0].securityContext
value:
readOnlyRootFilesystem: true
- equal:
path: spec.template.spec.containers[0].lifecycle
value:
preStop:
exec:
command:
- /bin/sh
- -c
- nginx -s quit
- equal:
path: spec.template.spec.terminationGracePeriodSeconds
value: 30
- equal:
path: spec.template.spec.containers[1].name
value: auth-sidecar
- it: backend and ui render no hardening fields by default
templates:
- backend/deployment.yaml
- ui/deployment.yaml
asserts:
- isNull:
path: spec.template.spec.securityContext
- isNull:
path: spec.template.spec.containers[0].securityContext
- isNull:
path: spec.template.spec.containers[0].lifecycle
- isNull:
path: spec.template.spec.terminationGracePeriodSeconds
- lengthEqual:
path: spec.template.spec.containers
count: 1

106
helm/litellm/tests/probe_tests.yaml Normal file
View file

@ -0,0 +1,106 @@
suite: test liveness and readiness probe timeouts
templates:
- gateway/deployment.yaml
- gateway/configmap.yaml
- backend/deployment.yaml
values:
- ./values/required.yaml
tests:
- it: gateway probes set an explicit timeout that outlasts a saturated event loop
template: gateway/deployment.yaml
asserts:
- equal:
path: spec.template.spec.containers[0].livenessProbe
value:
httpGet:
path: /health/liveliness
port: http
initialDelaySeconds: 10
periodSeconds: 15
timeoutSeconds: 10
failureThreshold: 6
- equal:
path: spec.template.spec.containers[0].readinessProbe
value:
httpGet:
path: /health/readiness
port: http
initialDelaySeconds: 5
periodSeconds: 10
timeoutSeconds: 10
- it: backend probes set an explicit timeout that outlasts a saturated event loop
template: backend/deployment.yaml
asserts:
- equal:
path: spec.template.spec.containers[0].livenessProbe
value:
httpGet:
path: /health/liveliness
port: http
initialDelaySeconds: 10
periodSeconds: 15
timeoutSeconds: 10
failureThreshold: 6
- equal:
path: spec.template.spec.containers[0].readinessProbe
value:
httpGet:
path: /health/readiness
port: http
initialDelaySeconds: 5
periodSeconds: 10
timeoutSeconds: 10
- it: no single-event-loop component is left on the kubernetes default 1s probe timeout
templates:
- gateway/deployment.yaml
- backend/deployment.yaml
asserts:
- isNotNullOrEmpty:
path: spec.template.spec.containers[0].livenessProbe.timeoutSeconds
- isNotNullOrEmpty:
path: spec.template.spec.containers[0].readinessProbe.timeoutSeconds
- equal:
path: spec.template.spec.containers[0].livenessProbe.timeoutSeconds
value: 10
- equal:
path: spec.template.spec.containers[0].readinessProbe.timeoutSeconds
value: 10
- it: gateway liveness tolerates a longer outage than readiness before acting
template: gateway/deployment.yaml
asserts:
- equal:
path: spec.template.spec.containers[0].livenessProbe.failureThreshold
value: 6
- notExists:
path: spec.template.spec.containers[0].readinessProbe.failureThreshold
- it: probe timeouts and thresholds stay overridable per component
template: gateway/deployment.yaml
set:
gateway.readinessProbe.timeoutSeconds: 3
gateway.readinessProbe.periodSeconds: 20
gateway.livenessProbe.timeoutSeconds: 4
gateway.livenessProbe.failureThreshold: 3
asserts:
- equal:
path: spec.template.spec.containers[0].readinessProbe
value:
httpGet:
path: /health/readiness
port: http
initialDelaySeconds: 5
periodSeconds: 20
timeoutSeconds: 3
- equal:
path: spec.template.spec.containers[0].livenessProbe
value:
httpGet:
path: /health/liveliness
port: http
initialDelaySeconds: 10
periodSeconds: 15
timeoutSeconds: 4
failureThreshold: 3

90
helm/litellm/values.yaml
View file

@ -57,6 +57,42 @@ migrationJob:
backoffLimit: 4
ttlSecondsAfterFinished: 120
resources: {}
# ServiceAccount for the Job pod only.
#
# The Job is a pre-install / pre-upgrade hook, so it runs before the chart's
# ordinary resources exist. With `serviceAccounts.backend.create: true` the
# backend ServiceAccount is one of those ordinary resources, so a Job that
# borrowed its name would reference an account that does not exist yet and
# the first install would fail with a forbidden pod creation. The name set
# here always wins; when it is empty the Job falls back to `default` if the
# chart creates the backend ServiceAccount, and to the backend
# ServiceAccount name otherwise (that name is either an existing account you
# supplied or `default`).
#
# Point this at a pre-existing ServiceAccount when the Job needs credentials
# of its own, e.g. the IRSA / Workload Identity annotations that
# `database.writer.useIAMAuth` relies on. That is also the upgrade path to
# watch: a release already running with `serviceAccounts.backend.create:
# true` used to hand the Job the created backend account on every upgrade,
# and now hands it `default` unless you name an account here.
serviceAccountName: ""
# The Job runs `prisma migrate deploy` against Postgres and never calls the
# K8s API, so it defaults to no projected ServiceAccount token, the same
# reasoning the ui SA above uses. Flip to true if your Job genuinely needs
# one; IAM database auth does not, since EKS Pod Identity injects its own
# projected token volume and GKE Workload Identity goes through the
# metadata server, neither of which is the default token mount.
automountServiceAccountToken: false
# Standard k8s pod-level and container-level securityContext for the Job
# pod. Same shape as gateway.podSecurityContext / gateway.securityContext.
podSecurityContext: {}
securityContext: {}
# Extra pod labels on the Job pod, merged into the chart's common labels.
podLabels: {}
# Additional volumes on the Job pod and volumeMounts on its container, e.g.
# the writable scratch space a read-only root filesystem needs.
volumes: []
volumeMounts: []
image:
repository: ghcr.io/berriai/litellm-migrations
tag: "" # defaults to .Chart.AppVersion
@ -180,10 +216,13 @@ gateway:
httpGet: { path: /health/liveliness, port: http }
initialDelaySeconds: 10
periodSeconds: 15
timeoutSeconds: 10
failureThreshold: 6
readinessProbe:
httpGet: { path: /health/readiness, port: http }
initialDelaySeconds: 5
periodSeconds: 10
timeoutSeconds: 10
hpa:
enabled: true
minReplicas: 1
@ -200,6 +239,37 @@ gateway:
minAvailable: ""
maxUnavailable: ""
podAnnotations: {}
# Extra pod labels, merged into the chart's selector labels. Do not
# re-declare `app.kubernetes.io/name` / `instance` / `component` here: they
# form the Deployment's immutable selector.
podLabels: {}
# Pod-level securityContext, applied to every container in the pod
# (runAsNonRoot, runAsUser, fsGroup, seccompProfile, ...). Empty by default
# so the cluster's own defaults keep applying to existing installs; clusters
# enforcing a restricted Pod Security Standard usually want at least
# `runAsNonRoot: true` and `seccompProfile.type: RuntimeDefault`.
podSecurityContext: {}
# Container-level securityContext for the gateway container. Empty by
# default for the same reason. Example:
# allowPrivilegeEscalation: false
# readOnlyRootFilesystem: true
# capabilities:
# drop:
# - ALL
# `readOnlyRootFilesystem: true` needs writable scratch space; supply it
# through `volumes` / `volumeMounts` above rather than expecting the chart
# to guess the paths your workload writes to.
securityContext: {}
# Extra sidecar containers appended to the gateway pod, e.g. an auth or
# egress proxy. Rendered through `tpl`, so entries may reference chart
# values and release metadata.
extraContainers: []
# Container lifecycle hooks (postStart / preStop) for the gateway container.
lifecycle: {}
# Grace period the kubelet allows between SIGTERM and SIGKILL. Leave empty
# to inherit the Kubernetes default of 30s. Set it a few seconds above the
# proxy's GRACEFUL_SHUTDOWN_TIMEOUT when you use a draining preStop hook.
terminationGracePeriodSeconds: ""
nodeSelector: {}
tolerations: []
affinity: {}
@ -242,10 +312,13 @@ backend:
httpGet: { path: /health/liveliness, port: http }
initialDelaySeconds: 10
periodSeconds: 15
timeoutSeconds: 10
failureThreshold: 6
readinessProbe:
httpGet: { path: /health/readiness, port: http }
initialDelaySeconds: 5
periodSeconds: 10
timeoutSeconds: 10
hpa:
enabled: true
minReplicas: 1
@ -257,6 +330,13 @@ backend:
minAvailable: ""
maxUnavailable: ""
podAnnotations: {}
# Same shape as the gateway blocks of the same name.
podLabels: {}
podSecurityContext: {}
securityContext: {}
extraContainers: []
lifecycle: {}
terminationGracePeriodSeconds: ""
nodeSelector: {}
tolerations: []
affinity: {}
@ -310,6 +390,16 @@ ui:
minAvailable: ""
maxUnavailable: ""
podAnnotations: {}
# Same shape as the gateway blocks of the same name. The nginx runtime
# writes its pid, cache, and proxy temp files under the image's root
# filesystem, so `securityContext.readOnlyRootFilesystem: true` here needs
# emptyDir volumes mounted over those paths.
podLabels: {}
podSecurityContext: {}
securityContext: {}
extraContainers: []
lifecycle: {}
terminationGracePeriodSeconds: ""
nodeSelector: {}
tolerations: []
affinity: {}

12
litellm-proxy-extras/litellm_proxy_extras/migrations/20260725000000_add_daily_tool_spend/migration.sql Normal file
View file

@ -0,0 +1,12 @@
-- CreateTable
CREATE TABLE IF NOT EXISTS "LiteLLM_DailyToolSpend" (
"date" TEXT NOT NULL,
"tool_name" TEXT NOT NULL,
"spend" DOUBLE PRECISION NOT NULL DEFAULT 0.0,
"total_tokens" BIGINT NOT NULL DEFAULT 0,
"request_count" BIGINT NOT NULL DEFAULT 0,
"created_at" TIMESTAMP(3) NOT NULL DEFAULT CURRENT_TIMESTAMP,
"updated_at" TIMESTAMP(3) NOT NULL,
CONSTRAINT "LiteLLM_DailyToolSpend_pkey" PRIMARY KEY ("date","tool_name")
);

3
litellm-proxy-extras/litellm_proxy_extras/migrations/20260729000000_add_reload_tracking_to_litellm_config/migration.sql Normal file
View file

@ -0,0 +1,3 @@
-- AlterTable
ALTER TABLE "LiteLLM_Config" ADD COLUMN IF NOT EXISTS "last_run_at" TIMESTAMP(3),
ADD COLUMN IF NOT EXISTS "reload_revision" BIGINT NOT NULL DEFAULT 0;

17
litellm-proxy-extras/litellm_proxy_extras/migrations/20260731000000_add_autorouter_savings_spend/migration.sql Normal file
View file

@ -0,0 +1,17 @@
-- AlterTable
ALTER TABLE "LiteLLM_DailyUserSpend" ADD COLUMN IF NOT EXISTS "autorouter_savings_spend" DOUBLE PRECISION NOT NULL DEFAULT 0.0;
-- AlterTable
ALTER TABLE "LiteLLM_DailyOrganizationSpend" ADD COLUMN IF NOT EXISTS "autorouter_savings_spend" DOUBLE PRECISION NOT NULL DEFAULT 0.0;
-- AlterTable
ALTER TABLE "LiteLLM_DailyEndUserSpend" ADD COLUMN IF NOT EXISTS "autorouter_savings_spend" DOUBLE PRECISION NOT NULL DEFAULT 0.0;
-- AlterTable
ALTER TABLE "LiteLLM_DailyAgentSpend" ADD COLUMN IF NOT EXISTS "autorouter_savings_spend" DOUBLE PRECISION NOT NULL DEFAULT 0.0;
-- AlterTable
ALTER TABLE "LiteLLM_DailyTeamSpend" ADD COLUMN IF NOT EXISTS "autorouter_savings_spend" DOUBLE PRECISION NOT NULL DEFAULT 0.0;
-- AlterTable
ALTER TABLE "LiteLLM_DailyTagSpend" ADD COLUMN IF NOT EXISTS "autorouter_savings_spend" DOUBLE PRECISION NOT NULL DEFAULT 0.0;

106
litellm-proxy-extras/litellm_proxy_extras/replica_identity.py Normal file
View file

@ -0,0 +1,106 @@
"""Optional post-migration step that raises Postgres REPLICA IDENTITY to FULL.
Logical-replication consumers (Neon / lakehouse sync and similar) need FULL
replica identity to reconstruct the old row of an UPDATE or DELETE. Prisma
leaves every table it creates at the Postgres default, so the setting has to be
re-applied by hand after each migration run. Setting
``LITELLM_SET_REPLICA_IDENTITY_FULL`` makes every migration run re-assert it.
The statement goes through the Prisma CLI rather than a Postgres driver because
``litellm-proxy-extras`` has no runtime dependencies, while the CLI is already
required for the migrations themselves.
"""
import subprocess
import tempfile
from pathlib import Path
from litellm_proxy_extras._logging import logger
REPLICA_IDENTITY_FULL_ENV_VAR = "LITELLM_SET_REPLICA_IDENTITY_FULL"
REPLICA_IDENTITY_FULL_SQL = r"""
DO $$
DECLARE
target regclass;
BEGIN
SET LOCAL lock_timeout = '5s';
FOR target IN
SELECT c.oid::regclass
FROM pg_class c
JOIN pg_namespace n ON n.oid = c.relnamespace
WHERE c.relkind = 'r'
AND c.relreplident <> 'f'
AND n.nspname = ANY (current_schemas(false))
AND c.relname LIKE 'LiteLLM\_%'
LOOP
BEGIN
EXECUTE format('ALTER TABLE %s REPLICA IDENTITY FULL', target);
EXCEPTION WHEN lock_not_available THEN
RAISE WARNING 'REPLICA IDENTITY FULL skipped for %: table busy, retrying next run', target;
END;
END LOOP;
END
$$;
"""
def apply_replica_identity_full(
schema_path: str,
prisma_command: str,
prisma_env: dict[str, str],
) -> bool:
"""Set REPLICA IDENTITY FULL on every LiteLLM table that is not already FULL.
Never raises. Replication metadata is not needed to serve requests, so
every failure mode is reported and stepped over rather than taking down a
migration run that already succeeded: a database that refuses the ALTER
(most often because the runtime user does not own the tables), a missing
or unrunnable Prisma CLI, a read-only temp directory, or a timeout.
Returns True when the statement was applied, False when it failed.
"""
logger.info("Applying REPLICA IDENTITY FULL to LiteLLM tables")
try:
with tempfile.TemporaryDirectory(prefix="litellm_replica_identity_") as tmp_dir:
sql_path = Path(tmp_dir) / "replica_identity_full.sql"
sql_path.write_text(REPLICA_IDENTITY_FULL_SQL)
subprocess.run(
[
prisma_command,
"db",
"execute",
"--file",
str(sql_path),
"--schema",
schema_path,
],
timeout=60,
check=True,
capture_output=True,
text=True,
env=prisma_env,
)
except subprocess.CalledProcessError as e:
logger.error(
"Failed to set REPLICA IDENTITY FULL. Logical replication "
"consumers may reject updates to these tables. Grant table "
"ownership to the migration user, or apply "
"`ALTER TABLE ... REPLICA IDENTITY FULL` by hand. Error: %s",
e.stderr,
)
return False
except subprocess.TimeoutExpired:
logger.error("Timed out setting REPLICA IDENTITY FULL on LiteLLM tables")
return False
except OSError as e:
logger.error(
"Could not run the REPLICA IDENTITY FULL statement. Logical "
"replication consumers may reject updates to these tables. "
"Error: %s",
e,
)
return False
logger.info("REPLICA IDENTITY FULL applied to LiteLLM tables")
return True

21
litellm-proxy-extras/litellm_proxy_extras/schema.prisma
View file

@ -601,6 +601,8 @@ model LiteLLM_TagTable {
model LiteLLM_Config {
param_name String @id
param_value Json?
last_run_at DateTime?
reload_revision BigInt @default(0)
}
// View spend, model, api_key per request
@ -748,6 +750,7 @@ model LiteLLM_DailyUserSpend {
compression_saved_tokens BigInt @default(0)
compression_savings_spend Float @default(0.0)
prompt_caching_savings_spend Float @default(0.0)
autorouter_savings_spend Float @default(0.0)
spend Float @default(0.0)
api_requests BigInt @default(0)
successful_requests BigInt @default(0)
@ -782,6 +785,7 @@ model LiteLLM_DailyOrganizationSpend {
compression_saved_tokens BigInt @default(0)
compression_savings_spend Float @default(0.0)
prompt_caching_savings_spend Float @default(0.0)
autorouter_savings_spend Float @default(0.0)
spend Float @default(0.0)
api_requests BigInt @default(0)
successful_requests BigInt @default(0)
@ -816,6 +820,7 @@ model LiteLLM_DailyEndUserSpend {
compression_saved_tokens BigInt @default(0)
compression_savings_spend Float @default(0.0)
prompt_caching_savings_spend Float @default(0.0)
autorouter_savings_spend Float @default(0.0)
spend Float @default(0.0)
api_requests BigInt @default(0)
successful_requests BigInt @default(0)
@ -849,6 +854,7 @@ model LiteLLM_DailyAgentSpend {
compression_saved_tokens BigInt @default(0)
compression_savings_spend Float @default(0.0)
prompt_caching_savings_spend Float @default(0.0)
autorouter_savings_spend Float @default(0.0)
spend Float @default(0.0)
api_requests BigInt @default(0)
successful_requests BigInt @default(0)
@ -882,6 +888,7 @@ model LiteLLM_DailyTeamSpend {
compression_saved_tokens BigInt @default(0)
compression_savings_spend Float @default(0.0)
prompt_caching_savings_spend Float @default(0.0)
autorouter_savings_spend Float @default(0.0)
spend Float @default(0.0)
api_requests BigInt @default(0)
successful_requests BigInt @default(0)
@ -917,6 +924,7 @@ model LiteLLM_DailyTagSpend {
compression_saved_tokens BigInt @default(0)
compression_savings_spend Float @default(0.0)
prompt_caching_savings_spend Float @default(0.0)
autorouter_savings_spend Float @default(0.0)
spend Float @default(0.0)
api_requests BigInt @default(0)
successful_requests BigInt @default(0)
@ -1097,6 +1105,19 @@ model LiteLLM_SpendLogToolIndex {
@@index([start_time])
}
// Daily tool spend rollup (one row per tool per day) – the Cost Optimization card reads this, never SpendLogs
model LiteLLM_DailyToolSpend {
date String
tool_name String
spend Float @default(0.0)
total_tokens BigInt @default(0)
request_count BigInt @default(0)
created_at DateTime @default(now())
updated_at DateTime @updatedAt
@@id([date, tool_name])
}
// Prompt table for storing prompt configurations
model LiteLLM_PromptTable {
id String @id @default(uuid())

46
litellm-proxy-extras/litellm_proxy_extras/utils.py
View file

@ -10,6 +10,10 @@ from pathlib import Path
from typing import Optional
from litellm_proxy_extras._logging import logger
from litellm_proxy_extras.replica_identity import (
REPLICA_IDENTITY_FULL_ENV_VAR,
apply_replica_identity_full,
)
def str_to_bool(value: Optional[str]) -> bool:
@ -676,6 +680,39 @@ class ProxyExtrasDBManager:
finally:
os.chdir(original_dir)
@staticmethod
def apply_replica_identity_full_if_requested() -> bool:
"""
Re-assert REPLICA IDENTITY FULL on LiteLLM's tables when the operator
opted in via LITELLM_SET_REPLICA_IDENTITY_FULL.
Prisma leaves new tables at the Postgres default, which logical
replication consumers reject, so the setting has to be re-applied after
every migration run rather than once by hand.
Returns:
bool: True if the setting was applied, False if it was not
requested or could not be applied.
"""
if not str_to_bool(os.getenv(REPLICA_IDENTITY_FULL_ENV_VAR)):
return False
try:
schema_path = ProxyExtrasDBManager._get_prisma_dir() + "/schema.prisma"
prisma_command = _get_prisma_command()
prisma_env = _get_prisma_env()
except OSError as e:
logger.error(
"Could not resolve the migrations directory for the REPLICA "
"IDENTITY FULL step, skipping it. Error: %s",
e,
)
return False
return apply_replica_identity_full(
schema_path=schema_path,
prisma_command=prisma_command,
prisma_env=prisma_env,
)
@staticmethod
def setup_database(
use_migrate: bool = False, use_v2_resolver: bool = False
@ -694,6 +731,15 @@ class ProxyExtrasDBManager:
Returns:
bool: True if setup was successful, False otherwise
"""
migrated = ProxyExtrasDBManager._run_migrations(
use_migrate=use_migrate, use_v2_resolver=use_v2_resolver
)
if migrated:
ProxyExtrasDBManager.apply_replica_identity_full_if_requested()
return migrated
@staticmethod
def _run_migrations(use_migrate: bool, use_v2_resolver: bool) -> bool:
if use_v2_resolver:
logger.info("Using v2 migration resolver (--use_v2_migration_resolver)")
return ProxyExtrasDBManager._setup_database_v2(use_migrate=use_migrate)

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

@ -1,6 +1,6 @@
[project]
name = "litellm-proxy-extras"
version = "0.4.81"
version = "0.4.83"
description = "Additional files for the LiteLLM Proxy. Reduces the size of the main litellm package."
readme = "README.md"
requires-python = ">=3.9"
@ -26,7 +26,7 @@ required-version = ">=0.10.9"
module-root = ""
[tool.commitizen]
version = "0.4.81"
version = "0.4.83"
version_files = [
"pyproject.toml:^version",
"../pyproject.toml:litellm-proxy-extras==",

11
litellm-rust/ADDING_A_PROVIDER.md
View file

@ -1,10 +1,11 @@
# Adding a provider / route to litellm-rust
Three layers, same for every route (see `ocr` and `realtime` as references):
Everything for a route lives in `crates/core/src/<route>/`; `crates/core/src/messages` is the reference. A host (the axum gateway, the Python bridge) only calls the route's entrypoint.
1. **Transform contract (pure)** — `crates/core/src/<route>/transformation.rs`: a `…ProviderConfig` trait (URL build + request/response transforms) + types in `types.rs`. No network, env, or auth.
2. **Provider config (pure)** — `crates/providers/src/<provider>/<route>/transformation.rs`: implement that trait as a `const <PROVIDER>_<ROUTE>_CONFIG`, mirroring the Python provider tree. Add parity unit tests.
3. **HTTP / transport (the host)** — `crates/providers/src/<route>.rs` (e.g. `ocr.rs`, `realtime.rs`): the callable fn (`run_ocr`, `realtime`). It resolves the key, builds the auth header, builds URL + transforms via the config, then does the network call. This is the only layer allowed to do I/O.
1. **Entrypoint** — `mod.rs`: `pub async fn <route>(request) -> CoreResult<Response>`, the Rust equivalent of `litellm.<route>()`, plus a `<route>_stream` variant when the route streams. It is the only thing a host touches.
2. **Transform contract** — `transformation.rs`: a `…ProviderConfig` trait (URL build + request/response transforms) with types in `types.rs`.
3. **Provider config** — `crates/core/src/providers/<provider>/<route>/transformation.rs`: implement that trait as a `const <PROVIDER>_<ROUTE>_CONFIG`, mirroring the Python provider tree. Add parity unit tests.
4. **Prepare + handler** — `prepare.rs` resolves provider/model, credentials, auth headers, and URL, then transforms the request; `handler.rs` performs the provider call through the shared client in `client.rs` and transforms the response.
## Coding standards
@ -25,4 +26,4 @@ variants of it. The test for a good abstraction is that adding the next provider
is a few declarative lines, not a new file of duplicated flow. Only diverge from
the base when behavior is genuinely different, and say so explicitly in the PR.
**Calling:** the host invokes the route fn — the Python bridge calls `run_ocr`; the `ai-gateway` server calls `realtime`. Register new modules in `lib.rs` / `mod.rs`, then run `cargo fmt && cargo clippy --workspace -- -D warnings && cargo test --workspace`.
**Calling:** hosts invoke the core entrypoint — the Python bridge and the `ai-gateway` route service both call `litellm_core::messages::messages`. Never add a provider handler to `ai-gateway`. Register new modules in `lib.rs` / `mod.rs`, then run `cargo fmt && cargo clippy --workspace -- -D warnings && cargo test --workspace`.

26
litellm-rust/AGENTS.md
View file

@ -4,14 +4,30 @@ litellm-rust has exactly THREE crates. A crate is a LAYER, not a route. Routes (
## Crates
| Crate | Role | Pure / I/O |
|-------|------|------------|
| litellm-core | Translation layer — types, route contracts (traits), provider transforms (modules under providers/), and the router. Builds requests/responses; no network. | Pure |
| litellm-ai-gateway | Routes + host — the only crate that touches the network. HTTP/WebSocket I/O (modules under io/) plus the axum server binary (behind the `server` feature). | I/O |
| litellm-python-bridge | PyO3 cdylib exposing Rust to the litellm Python SDK — a thin adapter over litellm-ai-gateway's I/O. | Binding |
| Crate | Role |
|-------|------|
| litellm-core | The LiteLLM SDK in Rust. One public entrypoint per top-level call (`messages::messages()`), owning types, transforms, provider resolution, auth, and the provider HTTP call. Call it, get a typed response. |
| litellm-ai-gateway | The axum server (behind the `server` feature) plus the WebSocket hosts. Translates HTTP/WS to core entrypoints; owns no provider logic and no handlers. |
| litellm-python-bridge | PyO3 cdylib exposing Rust to the litellm Python SDK — marshals Python objects and calls core entrypoints. |
Dependency direction (acyclic): litellm-core ← litellm-ai-gateway ← litellm-python-bridge.
## Where a route lives
A top-level LiteLLM call is a module under `crates/core/src/<route>/`, shaped like `messages`:
```
core/src/messages/
mod.rs # pub async fn messages(..) -> CoreResult<..> (+ messages_stream for SSE)
types.rs # request/response types, MessagesRequest
transformation.rs # the provider template trait
prepare.rs # provider resolution, auth headers, URL
handler.rs # the provider call
client.rs # the shared reqwest client
```
Handlers never live in `ai-gateway`. `ocr`, `audio_transcription`, and `realtime` are still hosted there from before this rule; they move to `core` as they are touched.
Adding a crate: default to a MODULE. New crate ONLY on a real trigger — separate artifact (binary/cdylib), proc-macro, shared foundation, or publishable standalone. A new provider or route is none of these.
Adding a crate fails crates/core/tests/workspace_crate_allowlist.rs until you update its allowlist and this file — intentional.

61
litellm-rust/CLAUDE.md
View file

@ -23,21 +23,34 @@ the base when behavior is genuinely different, and say so explicitly in the PR.
## Crates (exactly three — see AGENTS.md)
`litellm-core` describes work; `litellm-ai-gateway` executes it; `litellm-python-bridge`
exposes it to the Python SDK. A crate is a **layer**, not a route — add modules, not crates.
`litellm-core` **is** the LiteLLM SDK in Rust: it makes the LLM call.
`litellm-ai-gateway` is an HTTP/WebSocket server in front of it, and
`litellm-python-bridge` exposes it to the Python SDK. A crate is a **layer**, not
a route — add modules, not crates.
## Core Boundary
`litellm-core` is the pure translation layer; the `litellm-ai-gateway` host executes work.
`litellm-core` owns the whole call. The Rust equivalent of `litellm.messages()`
is `litellm_core::messages::messages(request).await`: you call it, it does the
provider call, and you get a typed non-streaming response back.
Route-level Rust structure mirrors LiteLLM's Python responsibilities:
- `core/src/<route>/` owns the route contract, shared types, and provider
template traits. For OCR, this means `core/src/ocr`.
- `core/src/<route>/` owns the route end to end: the public entrypoint fn named
after the route in `mod.rs`, the request/response types (`types.rs`), the
provider template trait (`transformation.rs`), the provider/auth/URL
resolution (`prepare.rs`), the HTTP client (`client.rs`), and the handler that
performs the call (`handler.rs`). `core/src/messages` is the reference.
- `core/src/providers/<provider>/<route>/transformation.rs` owns the
provider-specific transform. For Mistral OCR, this means
`core/src/providers/mistral/ocr/transformation.rs`.
- Network execution lives in the host crate `ai-gateway` (`ai-gateway/src/io/`),
never inside `core`.
provider-specific transform. For Anthropic Messages, this means
`core/src/providers/anthropic/messages/transformation.rs`.
- Handlers live in `core`, never in a host. `ai-gateway` must not contain a
route handler that talks to a provider; its axum route reads the HTTP request,
picks a deployment, and calls the `core` entrypoint. `python-bridge` marshals
Python objects and calls the same entrypoint.
Streaming keeps the same shape: the route entrypoint has a `<route>_stream`
variant in `core` that returns the upstream response so a host can splice it to
its own caller; the host still owns no provider logic.
Call-hook and lifecycle instrumentation, including phase timing, usage
accumulation, and callback payload construction, always lives in `core`.
@ -45,21 +58,31 @@ Hosts feed observed events into core and dispatch the completed payloads through
their I/O logger; hosts must not own callback orchestration.
Allowed in `core`:
- Pure request transforms
- Pure response transforms
- Pure stream chunk normalization
- The public entrypoint for a top-level LiteLLM call
- Request/response transforms and stream chunk normalization
- Provider resolution, auth header construction, and URL building
- The provider HTTP call itself, through a shared reused client with connect and
request timeouts
- Shared data types and validation errors
- Deterministic token/cost helper logic
Not allowed in `core`:
- Network calls
- Environment variable or secret reads
- Serving HTTP: axum routes, extractors, and transport concerns stay in the host
- Filesystem access
- Database or cache access
- Provider SDK signing or auth flows
- Database access
- Config file reading and rollout state
- Logging callbacks, spend writes, or custom callbacks
- Global mutable runtime state
Env reads in `core` are limited to credential fallback inside a route's
`prepare.rs` (the `env_lookup` closure), mirroring what the Python SDK does when
no key is passed. Everything else config-shaped is resolved by the host and
passed in.
Routes still hosted in `ai-gateway` (`ocr`, `audio_transcription`, `realtime`)
predate this rule and are being moved into `core` route modules; do not add new
ones there, and prefer moving one when you touch it.
Python owns rollout state and fallback while Rust is being introduced. Rust
paths must be off by default until parity tests prove equivalence with Python.
A new provider/route may instead be implemented rust-only with no Python
@ -93,10 +116,10 @@ the first PR:
- Preserve Python output shape intentionally. If a field is always serialized as
`null` for Python parity, leave a short comment explaining that parity choice.
## Host I/O Rules
## Network I/O Rules
These rules apply when adding future crates or modules that execute network I/O,
such as `ai-gateway`, router hosts, or standalone servers:
These rules apply to every module that executes network I/O, whether it is a
`core` route handler or a host such as `ai-gateway`:
- Set connect and full-request timeouts. No unbounded waits.
- Reuse HTTP clients; do not construct clients per request.

41
litellm-rust/README.md
View file

@ -2,18 +2,31 @@
This workspace contains the staged Rust implementation for LiteLLM.
Rust starts as a pure transform core used by the existing Python host. Python
continues to own auth, configuration, network I/O, retries, routing, logging,
`litellm-core` is the LiteLLM SDK in Rust: one entrypoint per top-level call
that makes the LLM call and hands back a typed response, the same shape as
`litellm.messages()` in Python.
```rust
let response = litellm_core::messages::messages(MessagesRequest {
model: "claude-sonnet-4-5",
body,
api_key: Some(key),
..
})
.await?;
```
Python continues to own configuration, retries, routing policy, logging,
callbacks, spend tracking, and customer plugins until each Rust path has parity
coverage and production evidence.
## Crates
| Crate | Role | Pure / I/O |
|-------|------|------------|
| litellm-core | Translation layer — types, route contracts (traits), provider transforms (modules under providers/), and the router. Builds requests/responses; no network. | Pure |
| litellm-ai-gateway | Routes + host — the only crate that touches the network. HTTP/WebSocket I/O (modules under io/) plus the axum server binary (behind the `server` feature). | I/O |
| litellm-python-bridge | PyO3 cdylib exposing Rust to the litellm Python SDK — a thin adapter over litellm-ai-gateway's I/O. | Binding |
| Crate | Role |
|-------|------|
| litellm-core | The SDK. Per-route entrypoints (`messages::messages()`), types, provider transforms (modules under `providers/`), provider resolution, auth, the provider HTTP call, and the router. |
| litellm-ai-gateway | The axum server (behind the `server` feature) and WebSocket hosts. Translates HTTP/WS to core entrypoints; no provider handlers. |
| litellm-python-bridge | PyO3 cdylib exposing Rust to the litellm Python SDK — marshals Python objects and calls core entrypoints. |
Dependency direction (acyclic): litellm-core ← litellm-ai-gateway ← litellm-python-bridge.
@ -21,16 +34,16 @@ Dependency direction (acyclic): litellm-core ← litellm-ai-gateway ← litellm-
```text
crates/
core/ Route contracts, shared pure types, errors, and templates.
src/ocr/
providers/ Provider-specific pure transforms.
src/mistral/ocr/transformation.rs
core/ The SDK: route modules + provider transforms.
src/messages/ mod.rs (entrypoint), types, transformation, prepare, handler, client
src/providers/anthropic/messages/transformation.rs
ai-gateway/ Axum server + WebSocket hosts; calls core entrypoints.
python-bridge/ PyO3 bridge for Python LiteLLM.
```
The folder shape should follow the Python provider tree:
`providers/src/<provider>/<route>/transformation.rs`. The bridge should expose
one function per top-level route, starting with `ocr(payload)`.
The folder shape follows the Python provider tree:
`core/src/providers/<provider>/<route>/transformation.rs`. The bridge exposes one
function per top-level route, mirroring the core entrypoints.
## Checks

10
litellm-rust/crates/CODING_STANDARDS/PROVIDER_CODING_STANDARDS.md
View file

@ -1,6 +1,6 @@
# Provider coding standards (litellm-rust)
Rules for adding or changing an LLM provider/route in `litellm-rust`. OCR (`MISTRAL_OCR_CONFIG`) is the reference; `messages` (`ANTHROPIC_MESSAGES_CONFIG`) is the next port.
Rules for adding or changing an LLM provider/route in `litellm-rust`. `messages` (`core/src/messages`, `ANTHROPIC_MESSAGES_CONFIG`) is the reference: a route is a `core` module with a public entrypoint that makes the call and returns a typed response.
## Provider resolution
@ -16,10 +16,10 @@ Rules for adding or changing an LLM provider/route in `litellm-rust`. OCR (`MIST
## Boundaries
7. Layers never cross: `core` = pure transforms/types (no network, env, secrets, auth, logging, global mutable state); `ai-gateway` = all I/O, auth headers, HTTP/SSE, lifecycle hooks; `python-bridge` = thin PyO3 adapter.
7. Layers never cross: `core` = the call itself (entrypoint, types, transforms, provider resolution, auth headers, provider HTTP, lifecycle hooks); `ai-gateway` = serving HTTP/WS (routing, extractors, auth of *our* callers, streaming to the client); `python-bridge` = thin PyO3 adapter. Hosts call the core entrypoint; they never build a provider request.
8. Generic/route files contain zero provider-specific branches. A provider is one module under `core/src/providers/<provider>/<route>/`; a route is a module, never a new crate.
9. Route entry point stays thin: `<route>()` -> `prepare_*` -> `CallLifecycle::run_request`, which owns the pre_call -> during_call -> provider call -> success/failure order and phase timing. Handlers validate and delegate; no business logic in them.
10. Constants (URLs, env-var names, API versions, error messages) live in a crate `constants.rs`, never inline. Env reads happen only at the host/config layer, with the `DEFAULT_*` fallback defined in `constants.rs`.
9. Route entry point stays thin: `core::<route>::<route>()` -> `prepare_*` -> handler (or `CallLifecycle::run_request`, which owns the pre_call -> during_call -> provider call -> success/failure order and phase timing). Axum handlers validate and delegate to a service that calls the entrypoint; no business logic in them.
10. Constants (URLs, env-var names, API versions, error messages) live in a crate `constants.rs`, never inline. Config-shaped env reads happen at the host/config layer with the `DEFAULT_*` fallback defined in `constants.rs`; the only env read in `core` is the credential fallback in a route's `prepare.rs`.
## Types and errors
@ -33,7 +33,7 @@ Rules for adding or changing an LLM provider/route in `litellm-rust`. OCR (`MIST
16. Never log request/response bodies, base64 payloads, document contents, or secrets. Truncate and bound any upstream body before it crosses a host boundary.
17. Treat empty/whitespace credentials, URLs, and config values as absent at the host resolution layer.
18. Host I/O sets connect + request timeouts (no unbounded waits), reuses a shared HTTP client, and prefers rustls TLS.
18. Network I/O sets connect + request timeouts (no unbounded waits), reuses a shared HTTP client, and prefers rustls TLS.
## Tests and rollout

9
litellm-rust/crates/ai-gateway/AGENTS.md
View file

@ -1,7 +1,9 @@
# ai-gateway — folder architecture
The Axum server that fronts the Rust gateway. It owns transport + config + auth
only; deployment selection lives in `core::router`, transforms in `core`/`providers`.
only; deployment selection lives in `core::router`, and the LLM call itself
(transforms, auth headers, provider HTTP) lives behind a `core` route entrypoint
such as `litellm_core::messages::messages`. No provider handler lives here.
```
src/
@ -32,6 +34,11 @@ src/
args; it runs during extraction. Never re-implement the check per route.
- **Handlers are thin.** A handler validates and delegates to its `service`. No
business logic, no provider calls, no transforms in handlers.
- **Services call `core`, they don't reimplement it.** A `service` picks the
deployment and calls the `core` route entrypoint. Provider resolution, auth
headers, URL building, and the HTTP call are `core`'s job; a service that
builds a provider request itself is a bug (`routes/messages/service.rs` is
the reference).
- **State is shared and cheap to clone.** Long-lived handles live behind `Arc` in
`state.rs`; read env/config only in `main.rs` when building state.

10
litellm-rust/crates/ai-gateway/README.md
View file

@ -8,11 +8,11 @@ dials OpenAI upstream, and splices the two sockets frame-by-frame.
`litellm-rust` is exactly three crates (a crate is a **layer**, not a route):
| Crate | Role | Pure / I/O |
|-------|------|------------|
| litellm-core | Translation layer — types, route contracts (traits), provider transforms (modules under `providers/`), and the router. Builds requests/responses; no network. | Pure |
| litellm-ai-gateway | Routes + host — the only crate that touches the network. HTTP/WebSocket I/O (modules under `io/`) plus the Axum server binary (behind the `server` feature). | I/O |
| litellm-python-bridge | PyO3 cdylib exposing Rust to the litellm Python SDK — a thin adapter over litellm-ai-gateway's I/O. | Binding |
| Crate | Role |
|-------|------|
| litellm-core | The LiteLLM SDK in Rust — per-route entrypoints (`messages::messages()`) that resolve the provider, transform, and make the call; plus types, provider transforms, and the router. |
| litellm-ai-gateway | The Axum server (behind the `server` feature) and WebSocket hosts. Translates HTTP/WS to core entrypoints; no provider handlers. |
| litellm-python-bridge | PyO3 cdylib exposing Rust to the litellm Python SDK — marshals Python objects and calls core entrypoints. |
Dependency direction (acyclic): litellm-core ← litellm-ai-gateway ← litellm-python-bridge.

16
litellm-rust/crates/ai-gateway/src/constants.rs
View file

@ -29,18 +29,6 @@ pub(crate) const DEFAULT_FLUSH_INTERVAL_MS: u64 = 500;
#[cfg(feature = "server")]
pub(crate) const DEFAULT_PROVIDER: &str = "openai";
/// Full-request timeout ceiling for Anthropic Messages provider calls, in
/// seconds. Mirrors the Python Anthropic Messages default. The per-request
/// timeout from `litellm_params` still overrides this on the request builder.
pub(crate) const MESSAGES_TIMEOUT_SECS: u64 = 600;
/// Connect timeout for Anthropic Messages provider calls, in seconds.
pub(crate) const MESSAGES_CONNECT_TIMEOUT_SECS: u64 = 10;
/// Max characters of an upstream error body echoed across the host boundary
/// before truncation, so provider bodies are bounded and data-minimized.
pub(crate) const MESSAGES_ERROR_BODY_MAX_CHARS: usize = 256;
pub(crate) const DEFAULT_RESPONSES_WS_CONNECT_TIMEOUT_SECS: u64 = 10;
pub(crate) const DEFAULT_RESPONSES_WS_IDLE_TIMEOUT_SECS: u64 = 300;
@ -48,10 +36,6 @@ pub(crate) const DEFAULT_RESPONSES_WS_IDLE_TIMEOUT_SECS: u64 = 300;
#[cfg(feature = "server")]
pub(crate) const MESSAGES_ROUTE_PATH: &str = "/v1/messages";
/// Provider name used by the Anthropic Messages route when a deployment's
/// provider model does not carry an explicit provider prefix.
pub(crate) const ANTHROPIC_MESSAGES_PROVIDER: &str = "anthropic";
/// Request headers owned by the gateway and never forwarded upstream.
#[cfg(feature = "server")]
pub(crate) const MESSAGES_HEADERS_NOT_FORWARDED: &[&str] =

1
litellm-rust/crates/ai-gateway/src/io/messages.rs
View file

@ -1 +0,0 @@
pub use crate::messages::{MessagesRequest, messages};

1
litellm-rust/crates/ai-gateway/src/io/mod.rs
View file

@ -1,5 +1,4 @@
pub mod audio_transcription;
pub mod messages;
pub mod ocr;
pub mod realtime;
pub mod realtime_pool;

5
litellm-rust/crates/ai-gateway/src/lib.rs
View file

@ -4,7 +4,9 @@
//! without pulling in the HTTP server:
//!
//! - Call-type modules such as [`ocr`]: provider transforms, lifecycle hooks,
//! and provider I/O. Always available — no feature required.
//! and provider I/O. Always available — no feature required. These predate the
//! rule that a route's entrypoint and handler live in `litellm-core` (see
//! `litellm_core::messages`) and move there as they are touched.
//! - [`io`]: compatibility exports and realtime WebSocket splice helpers.
//! - The server modules ([`auth`], [`routes`], [`state`]) and anything pulling
//! `axum` are gated behind the `server` feature, which the `litellm-ai-gateway`
@ -14,7 +16,6 @@
pub mod audio_transcription;
mod client;
pub mod io;
pub mod messages;
pub mod ocr;
/// GIL-activity tracking. Pure (atomics only); shared by the `server` routes and

49
litellm-rust/crates/ai-gateway/src/messages/mod.rs
View file

@ -1,49 +0,0 @@
use litellm_core::CoreResult;
use serde_json::Value;
mod client;
mod common_utils;
mod handler;
mod prepare;
mod types;
pub use types::MessagesRequest;
use handler::{execute_messages_provider_call, execute_messages_provider_stream};
use prepare::prepare_messages_call;
pub async fn messages(request: MessagesRequest<'_>) -> CoreResult<Value> {
match execute_messages(request, false).await? {
MessagesResponse::Json(body) => Ok(body),
MessagesResponse::Stream(response) => {
drop(response);
Err(litellm_core::CoreError::InvalidResponse(
"non-streaming messages execution returned a stream".to_string(),
))
}
}
}
pub(crate) enum MessagesResponse {
Json(Value),
Stream(reqwest::Response),
}
pub(crate) async fn execute_messages(
request: MessagesRequest<'_>,
stream: bool,
) -> CoreResult<MessagesResponse> {
let prepared = prepare_messages_call(request)?;
if stream {
execute_messages_provider_stream(prepared)
.await
.map(MessagesResponse::Stream)
} else {
execute_messages_provider_call(prepared)
.await
.map(MessagesResponse::Json)
}
}
#[cfg(test)]
mod tests;

24
litellm-rust/crates/ai-gateway/src/messages/types.rs
View file

@ -1,24 +0,0 @@
use std::time::Duration;
use litellm_core::messages::transformation::AnthropicMessagesProviderConfig;
use serde_json::{Map, Value};
pub struct MessagesRequest<'a> {
pub model: &'a str,
pub body: Value,
pub api_key: Option<&'a str>,
pub api_base: Option<&'a str>,
pub custom_llm_provider: Option<&'a str>,
pub extra_headers: Option<Map<String, Value>>,
pub timeout: Option<Duration>,
}
pub(crate) struct ProviderMessagesRequest {
pub(crate) provider: String,
pub(crate) model: String,
pub(crate) config: &'static dyn AnthropicMessagesProviderConfig,
pub(crate) url: String,
pub(crate) body: Value,
pub(crate) upstream_headers: Vec<(String, String)>,
pub(crate) timeout: Option<Duration>,
}

7
litellm-rust/crates/ai-gateway/src/routes/AGENTS.md
View file

@ -19,7 +19,10 @@ async fn handle(...) -> impl IntoResponse { ... }
When a route has business logic worth testing without axum, put it in a sibling
`service` (a file, or a folder if the route grows). The route file stays the
**axum surface** (router + handler + any socket/SSE adapter); `service` is plain
Rust with **no axum types**. `realtime/` is the example:
Rust with **no axum types**, and its job is to pick the deployment and call the
`core` route entrypoint (see `messages/service.rs` calling
`litellm_core::messages::messages`). Never build a provider request, resolve a
key, or perform the provider call here. `realtime/` is the older example:
```
realtime/
mod.rs # axum surface: router() + handler + the WS<->events adapter
@ -33,6 +36,8 @@ genuinely gets hard to read.
`crate::auth::RequireMasterKey` to its arguments; it runs during extraction.
Never re-implement the check per route.
- **Handlers contain no business logic; `service` contains no axum types.**
- **No provider handlers in this crate.** Transforms, auth headers, and the
provider HTTP call live in `core/src/<route>/`.
- A route owns its paths in its own `router()`; `mod.rs` only merges.
- Cross-cutting concerns (logging, CORS, timeouts) → Tower layers in `mod.rs`,
not duplicated in handlers.

23
litellm-rust/crates/ai-gateway/src/routes/messages/service.rs
View file

@ -1,12 +1,12 @@
use std::sync::Arc;
use litellm_core::constants::ANTHROPIC_MESSAGES_PROVIDER;
use litellm_core::messages::types::MessagesRequest;
use litellm_core::messages::{messages, messages_stream};
use litellm_core::router::Router;
use litellm_core::{CoreError, CoreResult};
use serde_json::{Map, Value};
use crate::constants::ANTHROPIC_MESSAGES_PROVIDER;
use crate::messages::{MessagesRequest, execute_messages};
pub(crate) enum MessagesResponse {
Json(Value),
Stream(reqwest::Response),
@ -52,13 +52,14 @@ pub async fn run(
extra_headers,
timeout: None,
};
let stream = request.body.get("stream").and_then(Value::as_bool) == Some(true);
execute_messages(request, stream)
.await
.map(|response| match response {
crate::messages::MessagesResponse::Json(body) => MessagesResponse::Json(body),
crate::messages::MessagesResponse::Stream(upstream) => {
MessagesResponse::Stream(upstream)
}
if request.body.get("stream").and_then(Value::as_bool) == Some(true) {
return messages_stream(request).await.map(MessagesResponse::Stream);
}
let response = messages(request).await?;
serde_json::to_value(response)
.map(MessagesResponse::Json)
.map_err(|err| {
CoreError::InvalidResponse(format!("failed to serialize messages response: {err}"))
})
}

8
litellm-rust/crates/core/AGENTS.md
View file

@ -1,3 +1,7 @@
litellm-core is the PURE translation layer — types, route contracts (traits), provider transforms (modules under `providers/`), and the router. No network, no I/O, no env reads.
litellm-core is the LiteLLM SDK in Rust — it makes the LLM call. Each top-level call is a module under `src/<route>/` exposing a public entrypoint named after the route (`messages::messages()`, the Rust equivalent of `litellm.messages()`): you call it and get a typed non-streaming response back.
Routes (ocr, realtime) and providers (mistral, openai) are modules, not crates.
A route module owns everything the call needs: types, the provider template trait, provider transforms (under `providers/`), provider/auth/URL resolution, and the handler that performs the HTTP call. Handlers belong here, never in a host crate.
Not here: serving HTTP (axum routes, extractors), config file reading, rollout state, databases, or callback dispatch. Env reads are limited to credential fallback in a route's `prepare.rs`.
Routes (messages, ocr, realtime) and providers (anthropic, mistral, openai) are modules, not crates.

31
litellm-rust/crates/core/CLAUDE.md
View file

@ -4,20 +4,28 @@ Rules for `litellm-rust/crates/core`.
## Responsibility
`core` owns shared data types, typed errors, and deterministic helper contracts.
It must stay pure and host-independent.
`core` is the LiteLLM SDK in Rust: it makes the LLM call. Every top-level
LiteLLM call has a public entrypoint here, named after the route
(`messages::messages()` is the Rust equivalent of `litellm.messages()`), and
calling it returns a typed non-streaming response.
Allowed:
- The public entrypoint for a route, plus its `<route>_stream` variant when the
route supports streaming.
- Provider resolution, auth header construction, URL building, and the provider
HTTP call (shared reused client, connect + request timeouts).
- Shared request/response structs.
- Typed errors with stable, non-sensitive messages.
- Deterministic validation helpers.
- Serialization helpers that intentionally mirror Python output shape.
- Route templates that match Python base config responsibilities, such as
`ocr::transformation::OcrProviderConfig`.
`messages::transformation::AnthropicMessagesProviderConfig`.
Not allowed:
- Network, filesystem, database, cache, or environment access.
- Secret reads or auth/header construction.
- Serving HTTP: axum routers, extractors, and other transport concerns.
- Filesystem, database, or cache access.
- Config file reading or rollout state; the host resolves those and passes them
in. Env reads are limited to credential fallback in a route's `prepare.rs`.
- Logging callbacks, tracing spans, spend writes, or customer callbacks.
- Provider-specific branching that belongs in `providers`.
- Panics for user/provider-controlled input.
@ -33,10 +41,21 @@ typed field on a struct, not a raw string threaded through the API.
## Structure
Use route names directly under `src/`: `ocr`, future `messages`,
Use route names directly under `src/`: `messages`, `ocr`, future
`chat_completions`, `embeddings`, and similar top-level LiteLLM calls. Do not
invent broad names like `engine` for route contracts.
`src/messages` is the reference shape for a route module:
```
mod.rs pub async fn messages(..) (+ messages_stream)
types.rs request/response types
transformation.rs the provider template trait
prepare.rs provider resolution, auth headers, URL
handler.rs the provider call
client.rs the shared reqwest client
```
## Parity Rules
- Every shared type used by a provider transform needs unit tests for

2
litellm-rust/crates/core/Cargo.toml
View file

@ -7,6 +7,7 @@ repository.workspace = true
[dependencies]
rand.workspace = true
reqwest.workspace = true
serde.workspace = true
serde_json.workspace = true
thiserror.workspace = true
@ -30,5 +31,4 @@ bedrock-auth = [
]
[dev-dependencies]
reqwest.workspace = true
tokio = { workspace = true, features = ["macros", "rt-multi-thread"] }

16
litellm-rust/crates/core/src/constants.rs
View file

@ -1,3 +1,19 @@
pub const OPENAI_DEFAULT_API_BASE: &str = "https://api.openai.com";
pub const OPENAI_RESPONSES_DEFAULT_API_BASE: &str = "https://api.openai.com/v1";
pub const OPENAI_RESPONSES_PATH: &str = "/responses";
/// Full-request timeout ceiling for Anthropic Messages provider calls, in
/// seconds. Mirrors the Python Anthropic Messages default. The per-request
/// timeout from the caller still overrides this on the request builder.
pub(crate) const MESSAGES_TIMEOUT_SECS: u64 = 600;
/// Connect timeout for Anthropic Messages provider calls, in seconds.
pub(crate) const MESSAGES_CONNECT_TIMEOUT_SECS: u64 = 10;
/// Max characters of an upstream error body echoed across the call boundary
/// before truncation, so provider bodies are bounded and data-minimized.
pub(crate) const MESSAGES_ERROR_BODY_MAX_CHARS: usize = 256;
/// Provider name used for Anthropic Messages when a deployment's provider model
/// does not carry an explicit provider prefix.
pub const ANTHROPIC_MESSAGES_PROVIDER: &str = "anthropic";

0
litellm-rust/crates/ai-gateway/src/messages/client.rs → litellm-rust/crates/core/src/messages/client.rs
View file

10
litellm-rust/crates/ai-gateway/src/messages/common_utils.rs → litellm-rust/crates/core/src/messages/common_utils.rs
View file

@ -1,11 +1,11 @@
use litellm_core::CoreResult;
use litellm_core::error::{CoreError, json_type_name};
use litellm_core::messages::transformation::AnthropicMessagesProviderConfig;
use litellm_core::providers::anthropic::messages::transformation::ANTHROPIC_MESSAGES_CONFIG;
use litellm_core::providers::azure_ai::messages::transformation::AZURE_ANTHROPIC_MESSAGES_CONFIG;
use serde_json::{Map, Value};
use crate::constants::MESSAGES_ERROR_BODY_MAX_CHARS;
use crate::error::{CoreError, CoreResult, json_type_name};
use crate::providers::anthropic::messages::transformation::ANTHROPIC_MESSAGES_CONFIG;
use crate::providers::azure_ai::messages::transformation::AZURE_ANTHROPIC_MESSAGES_CONFIG;
use super::transformation::AnthropicMessagesProviderConfig;
pub(super) fn truncate_error_body(body: &str) -> String {
if body.chars().count() <= MESSAGES_ERROR_BODY_MAX_CHARS {

17
litellm-rust/crates/ai-gateway/src/messages/handler.rs → litellm-rust/crates/core/src/messages/handler.rs
View file

@ -1,15 +1,13 @@
use litellm_core::CoreResult;
use litellm_core::error::CoreError;
use serde_json::Value;
use crate::constants::ANTHROPIC_MESSAGES_PROVIDER;
use crate::error::{CoreError, CoreResult};
use super::client::http_client;
use super::common_utils::truncate_error_body;
use super::types::ProviderMessagesRequest;
use crate::constants::ANTHROPIC_MESSAGES_PROVIDER;
use super::types::{AnthropicMessagesResponse, ProviderMessagesRequest};
pub(super) async fn execute_messages_provider_call(
request: ProviderMessagesRequest,
) -> CoreResult<Value> {
) -> CoreResult<AnthropicMessagesResponse> {
let mut request_builder = http_client().post(&request.url).json(&request.body);
for (key, value) in &request.upstream_headers {
request_builder = request_builder.header(key, value);
@ -39,12 +37,7 @@ pub(super) async fn execute_messages_provider_call(
let response = serde_json::from_str(&text).map_err(|err| {
CoreError::InvalidResponse(format!("invalid messages response JSON: {err}"))
})?;
let transformed = request
.config
.transform_response(&request.model, response)?;
serde_json::to_value(transformed).map_err(|err| {
CoreError::InvalidResponse(format!("failed to serialize messages response: {err}"))
})
request.config.transform_response(&request.model, response)
}
pub(super) async fn execute_messages_provider_stream(

30
litellm-rust/crates/core/src/messages/mod.rs
View file

@ -1,2 +1,32 @@
//! The Anthropic Messages call, the Rust equivalent of Python's
//! `litellm.messages()`.
//!
//! [`messages`] is the top-level entrypoint: give it a model, a body, and
//! credentials, and it resolves the provider, transforms the request, calls the
//! provider, and returns a typed non-streaming response. [`messages_stream`]
//! is the streaming variant; it hands the raw upstream response back so a host
//! can splice the event stream to its own caller.
mod client;
mod common_utils;
mod handler;
mod prepare;
pub mod transformation;
pub mod types;
use crate::error::CoreResult;
use handler::{execute_messages_provider_call, execute_messages_provider_stream};
use prepare::prepare_messages_call;
use types::{AnthropicMessagesResponse, MessagesRequest};
pub async fn messages(request: MessagesRequest<'_>) -> CoreResult<AnthropicMessagesResponse> {
execute_messages_provider_call(prepare_messages_call(request)?).await
}
pub async fn messages_stream(request: MessagesRequest<'_>) -> CoreResult<reqwest::Response> {
execute_messages_provider_stream(prepare_messages_call(request)?).await
}
#[cfg(test)]
mod tests;

7
litellm-rust/crates/ai-gateway/src/messages/prepare.rs → litellm-rust/crates/core/src/messages/prepare.rs
View file

@ -1,9 +1,8 @@
use litellm_core::CoreError;
use litellm_core::CoreResult;
use litellm_core::messages::transformation::MessagesAuthStrategy;
use litellm_core::routing_utils::provider::{CustomLlmProvider, get_custom_llm_provider};
use crate::error::{CoreError, CoreResult};
use crate::routing_utils::provider::{CustomLlmProvider, get_custom_llm_provider};
use super::common_utils::{has_bearer_auth, has_header, messages_provider_config, string_headers};
use super::transformation::MessagesAuthStrategy;
use super::types::{MessagesRequest, ProviderMessagesRequest};
pub(super) fn prepare_messages_call(

14
litellm-rust/crates/ai-gateway/src/messages/tests.rs → litellm-rust/crates/core/src/messages/tests.rs
View file

@ -1,14 +1,16 @@
use std::time::Duration;
use litellm_core::error::CoreError;
use serde_json::{Map, Value, json};
use tokio::io::{AsyncReadExt, AsyncWriteExt};
use tokio::net::{TcpListener, TcpStream};
use crate::error::CoreError;
use super::common_utils::{
has_bearer_auth, has_header, messages_provider_config, string_headers, truncate_error_body,
};
use super::{MessagesRequest, messages};
use super::messages;
use super::types::MessagesRequest;
async fn read_http_request(socket: &mut TcpStream) -> String {
let mut request = Vec::new();
@ -152,8 +154,8 @@ async fn messages_round_trip_builds_azure_request_and_passes_response_through()
.await
.expect("messages request succeeds");
assert_eq!(response["content"][0]["text"], "hi");
assert_eq!(response["stop_reason"], "end_turn");
assert_eq!(response.content[0]["text"], "hi");
assert_eq!(response.stop_reason.as_deref(), Some("end_turn"));
let request = server.await.expect("server task completes");
let (head, body) = request.split_once("\r\n\r\n").expect("has body");
@ -208,8 +210,8 @@ async fn messages_round_trip_builds_native_anthropic_request() {
.await
.expect("messages request succeeds");
assert_eq!(response["content"][0]["text"], "hi");
assert_eq!(response["stop_reason"], "end_turn");
assert_eq!(response.content[0]["text"], "hi");
assert_eq!(response.stop_reason.as_deref(), Some("end_turn"));
let request = server.await.expect("server task completes");
let (head, _) = request.split_once("\r\n\r\n").expect("has body");

24
litellm-rust/crates/core/src/messages/types.rs
View file

@ -1,6 +1,30 @@
use std::time::Duration;
use serde::{Deserialize, Serialize};
use serde_json::{Map, Value};
use super::transformation::AnthropicMessagesProviderConfig;
pub struct MessagesRequest<'a> {
pub model: &'a str,
pub body: Value,
pub api_key: Option<&'a str>,
pub api_base: Option<&'a str>,
pub custom_llm_provider: Option<&'a str>,
pub extra_headers: Option<Map<String, Value>>,
pub timeout: Option<Duration>,
}
pub(super) struct ProviderMessagesRequest {
pub(super) provider: String,
pub(super) model: String,
pub(super) config: &'static dyn AnthropicMessagesProviderConfig,
pub(super) url: String,
pub(super) body: Value,
pub(super) upstream_headers: Vec<(String, String)>,
pub(super) timeout: Option<Duration>,
}
#[derive(Clone, Debug, PartialEq, Serialize, Deserialize)]
#[serde(untagged)]
pub enum SystemPrompt {

4
litellm-rust/crates/python-bridge/AGENTS.md
View file

@ -1,3 +1,3 @@
litellm-python-bridge is the PyO3 cdylib that exposes Rust to the litellm Python SDK — a thin adapter (Python objects → Rust calls → Python results) over litellm-ai-gateway.
litellm-python-bridge is the PyO3 cdylib that exposes Rust to the litellm Python SDK — a thin adapter (Python objects → Rust calls → Python results) over the litellm-core route entrypoints (e.g. `litellm_core::messages::messages`).
Keep it thin: no business logic, no transforms, no I/O orchestration — just marshal in/out and call into litellm-ai-gateway.
Keep it thin: no business logic, no transforms, no I/O orchestration — just marshal in/out and call the core entrypoint.

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