feat/uv package manager

.github/workflows/lint.yaml

@@ -10,10 +10,11 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: astral-sh/setup-uv@v2
         with:
-          python-version: 3.12
-      - run: pip install uv
-      - run: uv pip install --system -r requirements-dev.txt
-      - run: ruff check .
-      - run: ruff format --check --diff .
+          version: "0.4.4"
+          enable-cache: true
+      - run: uv python install 3.12
+      - run: uv sync --extra dev
+      - run: uv run ruff check .
+      - run: uv run ruff format --check --diff .

.github/workflows/test.yaml

@@ -7,13 +7,17 @@ on:
   pull_request:
 jobs:
   pytest:
-    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        os: [ubuntu-latest, macos-latest]
+    runs-on: ${{ matrix.os }}
     steps:
       - uses: actions/checkout@v4
-      - uses: actions/setup-python@v5
+      - uses: astral-sh/setup-uv@v2
         with:
-          python-version: 3.12
-      - run: pip install uv
-      - run: uv pip install --no-cache --system -r requirements-dev.txt
+          version: "0.4.4"
+          enable-cache: true
+      - run: uv python install 3.12
+      - run: uv sync --extra dev
       # TODO: figure out why `pytest` doesn't discover tests in `faster_whisper_server` directory by itself
-      - run: pytest faster_whisper_server/* tests
+      - run: uv run pytest src/faster_whisper_server/* tests

Dockerfile.cpu

@@ -7,14 +7,18 @@ RUN apt-get update && \
     DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends python3.12 python3.12-distutils && \
     apt-get clean && \
     rm -rf /var/lib/apt/lists/*
+COPY --from=ghcr.io/astral-sh/uv:0.4.4 /uv /bin/uv
 WORKDIR /root/faster-whisper-server
-COPY requirements.txt .
-RUN --mount=from=ghcr.io/astral-sh/uv:latest,source=/uv,target=/bin/uv \
-    uv pip install --system --python 3.12 --no-cache -r requirements.txt
-COPY ./faster_whisper_server ./faster_whisper_server
-ENV WHISPER__MODEL=Systran/faster-whisper-medium.en
+# https://docs.astral.sh/uv/guides/integration/docker/#intermediate-layers
+RUN --mount=type=cache,target=/root/.cache/uv \
+    --mount=type=bind,source=uv.lock,target=uv.lock \
+    --mount=type=bind,source=pyproject.toml,target=pyproject.toml \
+    uv sync --frozen --no-install-project
+COPY ./src ./pyproject.toml ./uv.lock ./
+RUN --mount=type=cache,target=/root/.cache/uv \
+    uv sync --frozen
+ENV WHISPER__MODEL=Systran/faster-whisper-small
 ENV WHISPER__INFERENCE_DEVICE=cpu
-ENV WHISPER__COMPUTE_TYPE=int8
 ENV UVICORN_HOST=0.0.0.0
 ENV UVICORN_PORT=8000
-CMD ["uvicorn", "faster_whisper_server.main:app"]
+CMD ["uv", "run", "uvicorn", "faster_whisper_server.main:app"]

Dockerfile.cuda

@@ -7,13 +7,18 @@ RUN apt-get update && \
     DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends python3.12 python3.12-distutils && \
     apt-get clean && \
     rm -rf /var/lib/apt/lists/*
+COPY --from=ghcr.io/astral-sh/uv:0.4.4 /uv /bin/uv
 WORKDIR /root/faster-whisper-server
-COPY requirements.txt .
-RUN --mount=from=ghcr.io/astral-sh/uv:latest,source=/uv,target=/bin/uv \
-    uv pip install --system --python 3.12 --no-cache -r requirements.txt
-COPY ./faster_whisper_server ./faster_whisper_server
-ENV WHISPER__MODEL=Systran/faster-distil-whisper-large-v3
+# https://docs.astral.sh/uv/guides/integration/docker/#intermediate-layers
+RUN --mount=type=cache,target=/root/.cache/uv \
+    --mount=type=bind,source=uv.lock,target=uv.lock \
+    --mount=type=bind,source=pyproject.toml,target=pyproject.toml \
+    uv sync --frozen --no-install-project
+COPY ./src ./pyproject.toml ./uv.lock ./
+RUN --mount=type=cache,target=/root/.cache/uv \
+    uv sync --frozen
+ENV WHISPER__MODEL=Systran/faster-whisper-large-v3
 ENV WHISPER__INFERENCE_DEVICE=auto
 ENV UVICORN_HOST=0.0.0.0
 ENV UVICORN_PORT=8000
-CMD ["uvicorn", "faster_whisper_server.main:app"]
+CMD ["uv", "run", "uvicorn", "faster_whisper_server.main:app"]

Taskfile.yaml

@@ -12,18 +12,6 @@ tasks:
     sources:
       - Dockerfile.*
       - faster_whisper_server/*.py
-  pip-compile:
-    cmds:
-      - uv pip compile --override overrides.txt pyproject.toml {{.CLI_ARGS}} > requirements.txt
-      - uv pip compile --override overrides.txt --extra dev pyproject.toml {{.CLI_ARGS}} > requirements-dev.txt
-      - uv pip compile --override overrides.txt --all-extras pyproject.toml {{.CLI_ARGS}} > requirements-all.txt
-    sources:
-      - pyproject.toml
-  pip-sync:
-    cmds:
-      - uv pip sync requirements-all.txt
-    sources:
-      - requirements-all.txt
   create-multi-arch-builder: docker buildx create --name main --driver=docker-container
   docker-build:
     cmds:

compose.yaml

@@ -1,3 +1,4 @@
+# TODO: https://docs.astral.sh/uv/guides/integration/docker/#configuring-watch-with-docker-compose
 # NOTE: arm images haven't been tested
 services:
   faster-whisper-server-cuda:
@@ -9,7 +10,7 @@ services:
         - linux/amd64
         - linux/arm64
     volumes:
-      - ~/.cache/huggingface:/root/.cache/huggingface
+      - hugging_face_cache:/root/.cache/huggingface
     restart: unless-stopped
     ports:
       - 8000:8000
@@ -22,6 +23,12 @@ services:
         reservations:
           devices:
             - capabilities: ["gpu"]
+              # If you have CDI feature enabled use the following instead
+              # https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/cdi-support.html
+              # https://docs.docker.com/reference/cli/dockerd/#enable-cdi-devices
+              # - driver: cdi
+              #   device_ids:
+              #   - nvidia.com/gpu=all
   faster-whisper-server-cpu:
     image: fedirz/faster-whisper-server:latest-cpu
     build:
@@ -31,11 +38,13 @@ services:
         - linux/amd64
         - linux/arm64
     volumes:
-      - ~/.cache/huggingface:/root/.cache/huggingface
+      - hugging_face_cache:/root/.cache/huggingface
     restart: unless-stopped
     ports:
       - 8000:8000
     develop:
       watch:
         - path: faster_whisper_server
           action: rebuild
+volumes:
+  hugging_face_cache:

flake.nix

@@ -1,7 +1,8 @@
 {
   inputs = {
-    nixpkgs.url = "github:NixOS/nixpkgs/master";
+    nixpkgs.url = "github:NixOS/nixpkgs/nixos-unstable";
     flake-utils.url = "github:numtide/flake-utils";
+    flake-utils.inputs.nixpkgs.follows = "nixpkgs";
   };
   outputs =
     { nixpkgs, flake-utils, ... }:
@@ -11,6 +12,8 @@
         pkgs = import nixpkgs {
           inherit system;
           config.allowUnfree = true;
+          # `cudnn_8_9` is marked as broken but still seems to work as before
+          config.allowBroken = true;
         };
       in
       {

pyproject.toml

@@ -1,36 +1,40 @@
 [project]
 name = "faster-whisper-server"
+version = "0.1.0"
 requires-python = ">=3.12,<3.13"
 dependencies = [
-    "pydantic",
-    "fastapi",
-    "uvicorn",
-    "python-multipart",
-    "soundfile",
-    "pydantic-settings",
+    "pydantic==2.9.0",
+    "fastapi==0.113.0",
+    "uvicorn==0.30.6",
+    "python-multipart==0.0.9",
+    "soundfile==0.12.1",
+    "pydantic-settings==2.4.0",
     "numpy<2.0.0",
     "faster-whisper==1.0.3",
-    "huggingface-hub",
-    "httpx-sse",
-    "httpx",
-    "gradio",
-    "openai",
+    "huggingface-hub==0.24.6",
+    "httpx-sse==0.4.0",
+    "httpx==0.27.2",
+    "gradio==4.42.0",
+    "openai==1.43.1",
 ]
 
 [project.optional-dependencies]
 client = [
-    "keyboard",
+    "keyboard==0.13.5",
 ]
 dev = [
     "ruff==0.5.3",
-    "pytest",
-    "webvtt-py",
-    "srt",
+    "pytest==8.3.2",
+    "webvtt-py==0.5.1",
+    "srt==3.5.3",
     "basedpyright==1.13.0",
-    "pytest-xdist"
+    "pytest-xdist==3.6.1",
+    "pytest-asyncio>=0.24.0",
 ]
 
-other = ["youtube-dl @ git+https://github.com/ytdl-org/youtube-dl.git@37cea84f775129ad715b9bcd617251c831fcc980", "aider-chat==0.39.0"]
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
 
 # https://docs.astral.sh/ruff/configuration/
 [tool.ruff]