ContextSync — Institutional Memory as a Service

Stop coding in the dark. ContextSync bridges your IDE and your team's conversations, surfacing critical historical context (Slack threads, Jira tickets, Confluence pages, Notion docs) right where you write code.

Live backend: https://contextsync-backend.onrender.com (health check at /)

The problem

You are reviewing code that looks perfect. It has retry logic. It catches exceptions. It passes the linter. It passes CI/CD. But it's wrong. Because months ago, a staff engineer mentioned in a Slack thread that "Gateway V2 requires an Idempotency Key". You didn't see that message. ContextSync did.

Features

• Explain intent — highlight code and get an explanation of why it exists based on historical context, not just syntax.
• Context cards — raw Slack threads and Jira tickets surfaced directly in a VS Code sidebar.
• RAG engine — ChromaDB vector search across your engineering history (Slack, Jira, Confluence, Notion).
• Background sync — polls connected sources every 60 seconds; manual sync via POST /context/sync.
• Chat — ask questions with optional code context attached.

Architecture

graph LR
    IDE[VS Code extension] -->|code snippet| API[FastAPI backend]
    API -->|vector search| DB[(ChromaDB)]
    SRC[Slack / Jira / Confluence / Notion] -->|60s poll| API
    DB -->|retrieved context| LLM[Gemini]
    LLM -->|insight| IDE

Loading

API

Method	Path	Purpose
GET	/	Health check
POST	/explain	Markdown explanation of a code snippet using retrieved context
POST	/context/retrieve	Structured context objects for the IDE
POST	/context/stats	Slack/Jira counts for a batch of snippets
POST	/chat	Chat with Gemini, optional context
POST	/context/sync	Manually trigger ingestion from connected sources
POST	/context/ingest	Webhook receiver for external events

Quick start

1. Backend

cd backend
python -m venv venv
source venv/bin/activate        # Windows: venv\Scripts\activate
pip install -r requirements.txt -r requirements-dev.txt
cp .env.example .env            # then fill in GOOGLE_API_KEY (required)
uvicorn app.main:app --reload

The server starts even without integration credentials — only the root health check and webhook endpoints work until GOOGLE_API_KEY is set.

2. VS Code extension

cd vscode-extension
npm install
npm run compile
# Press F5 to launch the Extension Development Host

Set contextsync.apiBaseUrl in VS Code settings to point at your backend (default http://127.0.0.1:8000).

3. Run the tests

cd backend
pytest -q
ruff check app tests ingest.py

Environment variables

Copy .env.example to backend/.env. Only GOOGLE_API_KEY is required; everything else enables an optional integration.

Variable	Required	Used for
GOOGLE_API_KEY	Yes	Gemini embeddings + chat
SLACK_BOT_TOKEN	No	Live Slack ingestion
SLACK_SIGNING_SECRET	No	Slack webhook verification
SLACK_CHANNEL_ID	No	Which Slack channel to poll
JIRA_DOMAIN	No	Jira cloud domain
JIRA_EMAIL	No	Jira auth email
JIRA_API_TOKEN	No	Jira API token
JIRA_JQL	No	Jira query (has sensible default)
CONFLUENCE_URL	No	Confluence base URL
CONFLUENCE_USERNAME	No	Confluence auth
CONFLUENCE_API_TOKEN	No	Confluence API token
NOTION_API_KEY	No	Notion integration token
NOTION_SEARCH_QUERY	No	Notion search query

Deployment

The backend ships with a Dockerfile (in backend/) and a render.yaml blueprint.

Render (used for the live demo)

1. In the Render dashboard, create a Web Service from this repo.
2. Root directory backend, runtime Python 3.11, build pip install -r requirements.txt, start uvicorn app.main:app --host 0.0.0.0 --port $PORT.
3. Add the environment variables from the table above (at minimum GOOGLE_API_KEY).
4. Free-tier instances sleep after inactivity; the first request after a cold start may take ~30s.

Docker

cd backend
docker build -t contextsync-backend .
docker run -p 8000:8000 --env-file .env contextsync-backend

The VS Code extension is not deployable to a URL — it runs inside VS Code via F5 or a packaged .vsix.

Demo scenario

The demo/ folder contains a deliberately dangerous payment processor (payment_processor.py). Use it to see ContextSync detect the missing "Idempotency Key" by cross-referencing Slack history.

Roadmap

• Per-relevance scoring surfaced in context cards (currently a placeholder 0.0)
• Slack webhook-driven ingestion instead of 60s polling
• Extension test suite (VS Code extension host tests)
• Persisted ChromaDB on a managed volume for cloud deploys

Contributing

1. Fork the repo and create a branch (git checkout -b feature/my-feature)
2. Make your change and add tests
3. Run pytest -q and ruff check in backend/
4. Open a pull request

License

MIT