feat(docker): add Dockerfile, action

Author: dineshpinto

.github/workflows/build_and_push.yml

@@ -0,0 +1,59 @@
+name: Build and Push
+
+on:
+  push:
+    # tags: ["v*.*.*"] # Match semantic versioning tags like v1.0.0
+    branches: ["main"]
+  workflow_dispatch: # Allow manual triggering of the workflow
+
+env:
+  REGISTRY: ghcr.io
+  IMAGE_NAME: ${{ github.repository }}
+
+jobs:
+  build-and-push:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: read
+      packages: write
+      attestations: write
+      id-token: write
+    steps:
+      # Checkout repository
+      - name: Checkout repository
+        uses: actions/checkout@v4
+
+      # Log in to GHCR
+      - name: Login to GitHub Container Registry
+        uses: docker/login-action@v3
+        with:
+          registry: ${{ env.REGISTRY }}
+          username: ${{ github.actor }}
+          password: ${{ secrets.GITHUB_TOKEN }}
+
+      # Generate Docker metadata
+      - name: Metadata for Docker
+        id: meta
+        uses: docker/metadata-action@v5
+        with:
+          images: ${{ env.REGISTRY }}/${{ env.IMAGE_NAME }}
+
+      # Cache Docker layers to speed up builds
+      - name: Cache Docker layers
+        uses: actions/cache@v3
+        with:
+          path: /tmp/.buildx-cache
+          key: ${{ runner.os }}-docker-${{ github.ref }}
+          restore-keys: |
+            ${{ runner.os }}-docker-
+
+      # Build and push Docker image
+      - name: Build and push Docker image
+        uses: docker/build-push-action@v6
+        id: push
+        with:
+          context: .
+          push: true
+          load: true
+          tags: ${{ steps.meta.outputs.tags }}
+          labels: ${{ steps.meta.outputs.labels }}

Dockerfile

@@ -0,0 +1,20 @@
+# Stage 1: Build Backend
+FROM ghcr.io/astral-sh/uv:python3.12-bookworm-slim AS backend-builder
+ADD . /flare-ai-social
+WORKDIR /flare-ai-social
+RUN uv sync --frozen
+
+# Stage 2: Final Image
+FROM ghcr.io/astral-sh/uv:python3.12-bookworm-slim
+
+WORKDIR /app
+COPY --from=backend-builder /flare-ai-social/.venv ./.venv
+COPY --from=backend-builder /flare-ai-social/src ./src
+COPY --from=backend-builder /flare-ai-social/pyproject.toml .
+COPY --from=backend-builder /flare-ai-social/README.md .
+
+# Allow workload operator to override environment variables
+LABEL "tee.launch_policy.allow_env_override"="GEMINI_API_KEY,GEMINI_MODEL"
+LABEL "tee.launch_policy.log_redirect"="always"
+
+CMD ["uv", "run", "start-social"]

README.md

@@ -4,16 +4,37 @@ Flare AI Kit template for Social AI Agents.
 
 ## 🏗️ Build & Run Instructions
 
+### Fine tuning a model over a dataset
+
 1. **Prepare the Environment File:**  
    Rename `.env.example` to `.env` and update the variables accordingly.
+   Some parameters are specific to model fine-tuning:
+
+   | Parameter             | Description                                                                | Default                              |
+   | --------------------- | -------------------------------------------------------------------------- | ------------------------------------ |
+   | `tuning_source_model` | Name of the model to tune.                                                 | `models/gemini-1.5-flash-001-tuning` |
+   | `epoch_count`         | Number of tuning epochs to run. An epoch is a pass over the whole dataset. | `100`                                |
+   | `batch_size`          | Number of examples to use in each training batch.                          | `4`                                  |
+   | `learning_rate`       | Step size multiplier for the gradient updates.                             | `0.001`                              |
 
-2. **Tune a new model**
+2. **Prepare a dataset:**
+   An example dataset is provided in `src/data/training_data.json`, which consists of tweet from
+   [Hugo Philion's X](https://x.com/HugoPhilion) account. You can use any publicly available dataset
+   for model fine-tuning.
+
+3. **Tune a new model**
+   Set the name of the new tuned model in `src/flare_ai_social/tune_model.py`, then:
 
    ```bash
    uv run start-tuning
    ```
 
-3. **Test the new model**
+4. **Observe loss parameters:**
+   After tuning in complete, a training loss PNG will be saved in the root folder corresponding to the new model.
+   Ideally the loss should minimize to near 0 after several training epochs.
+
+5. **Test the new model**
+   Select the new tuned model and test it:
 
    ```bash
    uv run start-social