prototype: use pyproject and uv to build distribution

Goals:

* remove the need of a custom tool to install a collection of python
  packages AKA `llama stack build`
* use the power of 'uv', which was designed to manage dependencies
* `llama stack build` can "probably" go away and be replaced with uv

Howto, with the pyproject, you can install an Ollama distribution in a
virtual env like so:

```
uv venv --python 3.10 ollama-distro
source ollama-distro/bin/activate
uv sync --extra ollama
llama stack run llama_stack/templates/ollama/run.yaml
```

Caveats:

* external provider, we could still use a build file or add
the known external providers to the pyproject?
* growth of the uv.lock?

We create a requirements.txt for convenience as some users are most
familiar with this format than looking at pyproject.

Signed-off-by: Sébastien Han <seb@redhat.com>

.github/actions/setup-runner/action.yml

@@ -5,6 +5,10 @@ inputs:
     description: The Python version to use
     required: false
     default: "3.11"
+  install-ollama:
+    description: Install ollama
+    required: false
+    default: true
 runs:
   using: "composite"
   steps:
@@ -17,11 +21,13 @@ runs:
 
     - name: Install dependencies
       shell: bash
+      env:
+        INSTALL_OLLAMA: ${{ inputs.install-ollama }}
+      if: ${{ env.INSTALL_OLLAMA == 'true' }}
       run: |
-        uv sync --all-groups
-        uv pip install ollama faiss-cpu
+        uv sync --all-groups --extra ollama
+
         # always test against the latest version of the client
         # TODO: this is not necessarily a good idea. we need to test against both published and latest
         # to find out backwards compatibility issues.
         uv pip install git+https://github.com/meta-llama/llama-stack-client-python.git@main
-        uv pip install -e .

.github/workflows/integration-auth-tests.yml

@@ -33,9 +33,6 @@ jobs:
       - name: Install dependencies
         uses: ./.github/actions/setup-runner
 
-      - name: Build Llama Stack
-        run: |
-          llama stack build --template ollama --image-type venv
 
       - name: Install minikube
         if: ${{ matrix.auth-provider == 'kubernetes' }}

.github/workflows/integration-tests.yml

@@ -41,16 +41,12 @@ jobs:
       - name: Setup ollama
         uses: ./.github/actions/setup-ollama
 
-      - name: Build Llama Stack
-        run: |
-          uv run llama stack build --template ollama --image-type venv
-
       - name: Start Llama Stack server in background
         if: matrix.client-type == 'http'
         env:
           INFERENCE_MODEL: "meta-llama/Llama-3.2-3B-Instruct"
         run: |
-          LLAMA_STACK_LOG_FILE=server.log nohup uv run llama stack run ./llama_stack/templates/ollama/run.yaml --image-type venv --env OLLAMA_URL="http://0.0.0.0:11434" &
+          LLAMA_STACK_LOG_FILE=server.log nohup uv run llama stack run ./llama_stack/templates/ollama/run.yaml --env OLLAMA_URL="http://0.0.0.0:11434"  &
 
       - name: Wait for Llama Stack server to be ready
         if: matrix.client-type == 'http'

.github/workflows/providers-build.yml

@@ -36,7 +36,7 @@ jobs:
       - name: Generate Template List
         id: set-matrix
         run: |
-          templates=$(ls llama_stack/templates/*/*build.yaml | awk -F'/' '{print $(NF-1)}' | jq -R -s -c 'split("\n")[:-1]')
+          templates=$(ls llama_stack/templates/*/*build.yaml | grep -v "experimental-post-training" | awk -F'/' '{print $(NF-1)}' | jq -R -s -c 'split("\n")[:-1]')
           echo "templates=$templates" >> "$GITHUB_OUTPUT"
 
   build:
@@ -54,16 +54,42 @@ jobs:
 
       - name: Install dependencies
         uses: ./.github/actions/setup-runner
+        with:
+          install-ollama: false
 
-      - name: Print build dependencies
+      - name: Print dependencies in the image
+        if: matrix.image-type == 'venv'
         run: |
-          uv run llama stack build --template ${{ matrix.template }} --image-type ${{ matrix.image-type }} --image-name test --print-deps-only
+          uv pip list
 
-      - name: Run Llama Stack Build
+      - name: Run Llama Stack Build - VENV
+        if: matrix.image-type == 'venv'
         run: |
-          # USE_COPY_NOT_MOUNT is set to true since mounting is not supported by docker buildx, we use COPY instead
-          # LLAMA_STACK_DIR is set to the current directory so we are building from the source
-          USE_COPY_NOT_MOUNT=true LLAMA_STACK_DIR=. uv run llama stack build --template ${{ matrix.template }} --image-type ${{ matrix.image-type }} --image-name test
+          uv sync --no-default-groups --extra ${{ matrix.template }}
+
+      # TODO
+      - name: Run Llama Stack Build - CONTAINER
+        if: matrix.image-type == 'container'
+        run: |
+          # TODO: use llama_stack/templates/Containerfile when we have a new release!
+          cat << 'EOF' > Containerfile
+          FROM registry.access.redhat.com/ubi9
+          WORKDIR /app
+          ARG TEMPLATE
+
+          RUN dnf -y update \
+              && dnf install -y python3.11 python3.11-pip python3.11-wheel python3.11-setuptools python3.11-devel gcc make \
+              && ln -s /bin/pip3.11 /bin/pip \
+              && ln -s /bin/python3.11 /bin/python \
+              && dnf clean all
+
+          RUN mkdir -p /.llama/providers.d /.cache
+          COPY . /app/llama-stack
+          RUN cd llama-stack && pip install --no-cache .[${TEMPLATE}]
+          RUN chmod -R g+rw /app /.llama /.cache
+          ENTRYPOINT ["python", "-m", "llama_stack.distribution.server.server", "--config", "/app/llama-stack/templates/${TEMPLATE}/run.yaml"]
+          EOF
+          docker build --build-arg TEMPLATE=${{ matrix.template }} -f Containerfile -t ${{ matrix.template }} .
 
       - name: Print dependencies in the image
         if: matrix.image-type == 'venv'

.github/workflows/update-readthedocs.yml

@@ -43,7 +43,7 @@ jobs:
       - name: Build HTML
         run: |
           cd docs
-          uv run make html
+          uv run --group docs make html
 
       - name: Trigger ReadTheDocs build
         if: github.event_name != 'pull_request'