Antriksh Jain (antriksh30) · Oct 12, 2026 · Oct 12, 2026 · Oct 12, 2026 · Oct 12, 2026
@@ -21,17 +21,19 @@ control events.
 - `src/voice-live-bridge-basic-python/model_contract.py` — model backend protocol shared without
   circular imports
 - `src/voice-live-bridge-basic-python/model_client.py` — Foundry Responses model adapter
-- `src/voice-live-bridge-basic-python/requirements.txt` — fully pinned runtime dependency graph
+- `src/voice-live-bridge-basic-python/pyproject.toml` — runtime and development dependencies
+- `src/voice-live-bridge-basic-python/uv.lock` — fully pinned dependency graph
 - `sample.yaml` — repository build-readiness validation
 
 Do not commit `.vscode` launch or task files generated by the Foundry Toolkit.
 
 ## Dependency workflow
 
-Follow `samples/python/hosted-agents/DEPENDENCY_POLICY.md`. The runtime
-`requirements.txt` is the portable consumer artifact and must pin the complete direct and
-transitive graph. Regenerate it whenever a dependency input changes, and run the static and
-resolver-backed policy checks before completing dependency changes.
+Follow `samples/python/hosted-agents/DEPENDENCY_POLICY.md`. Runtime and development
+dependencies are declared in `pyproject.toml`; `uv.lock` is the portable, fully
+pinned consumer artifact. Regenerate the lock whenever a dependency input changes,
+and run the static and resolver-backed policy checks before completing dependency
+changes.
 
 ## Runtime logging
 

@@ -44,10 +44,12 @@ history mutation, or hosted-agent handoff.
 
 ## Prerequisites
 
-1. Python 3.13.
-2. Azure Developer CLI (`azd`) 1.27.1 or later with the Foundry extension.
-3. Azure CLI authentication for local model access.
-4. A Foundry project and model deployment, or permission for `azd provision` to create them.
+1. Python 3.13 or later.
+2. [uv](https://docs.astral.sh/uv/getting-started/installation/) 0.11.7 or later,
+   installed and available on `PATH`.
+3. Azure Developer CLI (`azd`) 1.27.1 or later with the Foundry extension.
+4. Azure CLI authentication for local model access.
+5. A Foundry project and model deployment, or permission for `azd provision` to create them.
 
 Install and authenticate the tools:
 
@@ -113,9 +115,8 @@ sample therefore uses the protocol-aware smoke client instead of the standard In
 From the sample source directory:
 
 ```bash
-python3.13 -m venv .venv
-source .venv/bin/activate
-python -m pip install -r requirements.txt -r requirements-dev.txt
+python -m pip install uv
+uv sync --frozen
 cp .env.example .env
 ```
 
@@ -129,7 +130,7 @@ Load the environment and start the server:
 set -a
 source .env
 set +a
-python main.py
+uv run --no-sync python main.py
 ```
 
 The server listens on `PORT` or port `8088` and exposes `GET /readiness` and the
@@ -140,7 +141,7 @@ The server listens on `PORT` or port `8088` and exposes `GET /readiness` and the
 The smoke test sends `/help`, so it validates the Bridge wire flow without calling a model:
 
 ```bash
-python scripts/smoke_test.py
+uv run --no-sync python scripts/smoke_test.py
 ```
 
 Expected message sequence:
@@ -182,8 +183,8 @@ the Voice wrapper and Voice Live.
 Tests use deterministic fake model streams and make no cloud calls:
 
 ```bash
-python -m pytest -q
-ruff check .
+uv run --no-sync python -m pytest -q
+uv run --no-sync ruff check .
 ```
 
 From the repository root, run the standard per-sample validator:
@@ -214,7 +215,7 @@ bash .github/scripts/validate-sample.sh \
   Protocol `1.0`. This is separate from the `invocations_ws` transport version in `azure.yaml`.
 - If model calls fail locally, verify `FOUNDRY_PROJECT_ENDPOINT`,
   `AZURE_AI_MODEL_DEPLOYMENT_NAME`, and the active Azure CLI identity.
-- If the smoke test cannot connect, start `python main.py` first and verify that port `8088` is
+- If the smoke test cannot connect, start `uv run --no-sync python main.py` first and verify that port `8088` is
   available, or set the smoke client's URL to match `PORT`.
 
 ## Next steps

@@ -2,14 +2,13 @@ name: Voice Live Bridge Basic Python
 description: A hosted text agent implementing Voice Live Bridge Protocol 1.0.
 
 build: >-
-  python -m pip install --disable-pip-version-check
-  -r src/voice-live-bridge-basic-python/requirements.txt
-  -r src/voice-live-bridge-basic-python/requirements-dev.txt
+  cd src/voice-live-bridge-basic-python &&
+  uv sync --frozen
 validate: >-
   cd src/voice-live-bridge-basic-python &&
-  python -m compileall -q . &&
-  python -c "import azure.ai.agentserver.invocations.voice" &&
-  python -m ruff check .
+  uv run --no-sync python -m compileall -q -x '[.]venv' . &&
+  uv run --no-sync python -c "import azure.ai.agentserver.invocations.voice" &&
+  uv run --no-sync ruff check .
 test: >-
   cd src/voice-live-bridge-basic-python &&
-  python -m pytest -q
+  uv run --no-sync python -m pytest -q
@@ -1,3 +1,28 @@
+[project]
+name = "voice-live-bridge-basic-python"
+version = "0.1.0"
+description = "A Microsoft Foundry hosted voice bridge agent."
+requires-python = ">=3.13"
+dependencies = [
+    "azure-ai-agentserver-core==2.1.0",
+    "azure-ai-agentserver-invocations==1.1.0",
+    "azure-identity==1.25.1",
+    "openai==3.1.0",
+    "opentelemetry-api==1.43.0",
+]
+
+[dependency-groups]
+dev = [
+    "iniconfig==2.3.0",
+    "pluggy==1.6.0",
+    "Pygments==2.21.0",
+    "pytest==9.1.1",
+    "pytest-asyncio==1.4.0",
+    "ruff==0.16.5",
+    "starlette==1.6.0",
+    "websockets==17.1",
+]
+
 [tool.pytest.ini_options]
 testpaths = ["tests"]
 asyncio_mode = "auto"
@@ -7,4 +32,4 @@ line-length = 100
 target-version = "py313"
 
 [tool.ruff.lint]
-select = ["E", "F", "I", "B", "ASYNC"]
+select = ["E", "F", "I", "B", "ASYNC"]