Upload project files

Dockerfile

@@ -0,0 +1,35 @@
+# Use a slim Python base image
+FROM python:3.12-slim
+
+ENV PYTHONDONTWRITEBYTECODE=1 \
+    PYTHONUNBUFFERED=1
+
+# Create non-root user (optional but good practice)
+RUN useradd -m -u 1000 appuser
+WORKDIR /app
+
+# System deps (kept minimal since psycopg2-binary is used)
+RUN apt-get update && apt-get install -y --no-install-recommends \
+    curl ca-certificates && \
+    rm -rf /var/lib/apt/lists/*
+
+# Install Python deps first (better layer caching)
+COPY requirements.txt ./
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy application code
+COPY app ./app
+COPY README.md ./
+
+# Ensure the non-root user can write to /app (for ./data sqlite, logs, etc.)
+RUN chown -R appuser:appuser /app
+
+# Hugging Face Spaces exposes $PORT; default to 7860 if not set
+ENV PORT=7860
+EXPOSE 7860
+
+# Ensure we run as non-root
+USER appuser
+
+SHELL ["/bin/bash", "-lc"]
+CMD uvicorn app.main:app --host 0.0.0.0 --port ${PORT:-7860}

README.md

@@ -1,12 +1,109 @@
----
-title: TDS Project1 API
-emoji: 📉
-colorFrom: blue
-colorTo: yellow
-sdk: docker
-pinned: false
-license: mit
-short_description: FastAPI service to generate static, minimal GitHub repos
----
-
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# LLM Code Deployment Builder (FastAPI + GitHub Pages)
+
+A FastAPI service that:
+- In round 1: generates a minimal static app with an LLM, creates a GitHub repo, sets up Pages on the `gh-pages` branch, deploys, then notifies the evaluation URL.
+- In round 2: fetches current repo files, requests minimal changes from the LLM, updates the repo, waits for Pages to be live again, then notifies the evaluation URL.
+
+Built for IITM BS Tools in Data Science course. Uses an OpenAI-compatible endpoint (e.g., AIpipe) for LLM calls and PyGithub for GitHub automation.
+
+## Requirements
+
+- Python 3.10+
+- A GitHub personal access token with repo scopes
+- An OpenAI-compatible LLM endpoint and API key (AIpipe or similar)
+
+## Setup
+
+1) Create a virtual environment and install deps
+
+- Using pip and `requirements.txt`:
+  - python -m venv .venv && source .venv/bin/activate
+  - pip install -r requirements.txt
+
+2) Copy `.env.example` to `.env` and fill values
+
+3) Run the server
+
+- or: uvicorn app.main:app --host 0.0.0.0 --port 8000 --reload
+
+## Configuration
+
+Required environment variables:
+- APP_SECRET: Shared secret required in requests
+- GITHUB_TOKEN: GitHub token with repo permissions
+- GITHUB_USERNAME: Your GitHub username
+- GITHUB_EMAIL: Email used for commits
+- OPENAI_API_KEY: LLM provider API key (e.g., AIpipe token)
+
+Recommended/Optional:
+- OPENAI_BASE_URL: OpenAI-compatible base URL (default: https://aipipe.org/openai/v1)
+- AIPIPE_MODEL: Model ID (default: qwen/qwen3-coder-30b-a3b-instruct)
+- APP_DB_PATH: Override path for local SQLite file (default: ./data/state.db)
+
+See `.env.example` for a complete template.
+
+## API
+
+### POST /tasks
+
+Accepts a task request for either round 1 (initial build) or round 2 (revision).
+
+Request schema (simplified):
+- email: string
+- secret: string (must match APP_SECRET)
+- task: string (task name; used to key round-2 revisions)
+- round: 1 or 2
+- nonce: unique string per request
+- brief: short description of what to build/change
+- checks: list of acceptance checks
+- evaluation_url: URL to notify upon completion
+- attachments: optional list of { name, url } where url may be a data: URI or https URL
+
+Immediate response (always returned right away if inputs are valid):
+- { status: "accepted", task, round, message }
+
+Background processing then:
+- Round 1: Generates files, creates repo, configures GitHub Pages (gh-pages), waits for 200 OK, then POSTs to evaluation_url with {status, task, round, nonce, repo_url, commit_sha, pages_url}.
+- Round 2: Verifies a prior repo exists for the same task, fetches current files, requests minimal changes, updates repo, waits for Pages 200 OK, then POSTs to evaluation_url with the same payload.
+
+Important behaviors:
+- Evaluation notify requires a strict 200 OK; redirects are followed.
+- Pages readiness is polled until 200 OK or timeout, using internal defaults (no env required).
+- For round 2, the HTTP 200 response from this service is returned immediately after validating the secret and the existence of a prior round-1 repo for the task; all network work runs in the background.
+
+### GET /healthz
+
+Returns {"status": "ok"} for simple liveness checks.
+
+## How it works
+
+- LLM: Uses an OpenAI-compatible API. Prompts enforce an “attachment handling contract” ensuring data: URIs and base64 content aren’t truncated or modified.
+- GitHub: Creates a repo, commits workflow and files to `main`, deploys to `gh-pages` via JamesIves/github-pages-deploy-action. Pages is configured to `gh-pages` before the first content push to reduce noisy “Pages failed for main” entries.
+- Persistence: Stores task→repo mapping in local SQLite so the app can survive restarts (e.g., on Hugging Face Spaces).
+- Notifications: Exponential backoff with strict 200 OK requirement; redirects followed.
+
+## Deploying on Hugging Face Spaces (Docker)
+
+This repo includes a `Dockerfile` for Spaces (type: Docker).
+
+1) Create the Space and add Secrets (see `.env.example`).
+
+2) Push this repo to the Space; the app will bind to the provided `PORT`.
+
+3) Health check: GET /healthz
+
+Notes:
+- Ensure outbound internet is allowed to reach GitHub and the LLM provider.
+- Keep generation modest; Pages can take time to turn green.
+
+## Troubleshooting
+
+- Invalid secret: The service returns HTTP 400.
+- Round 2 without round 1: The service returns HTTP 400 indicating no prior repo for the task.
+- Pages never reaches 200 OK: Verify GitHub Pages is configured to the `gh-pages` branch and the action ran successfully. Check repo Settings → Pages.
+- Evaluation URL not accepting: The notifier retries with backoff and follows redirects, but requires a final 200 OK.
+- Data URLs truncated or modified: Prompts include an attachment handling contract; ensure your inputs use proper data: URIs or https links.
+
+## License
+
+MIT

requirements.txt

@@ -0,0 +1,14 @@
+fastapi
+uvicorn[standard]
+pydantic
+pydantic-settings
+httpx
+tenacity
+sqlalchemy
+psycopg2-binary
+PyGithub
+python-dotenv
+pytest
+pytest-asyncio
+ruff
+mypy