Threadit

A Chrome extension that reads a thread on X (Twitter) and returns a bullet-point summary powered by an LLM.

threadit/
├── extension-frontend/   Vite + React Chrome extension (Manifest V3)
└── backend-api/          FastAPI server — caching, database, Groq API

How it works

1. Open a thread on x.com and click the extension icon.
2. The popup sends a message to the content script, which reads the visible tweets from the DOM.
3. The popup POSTs the scraped tweets to the local FastAPI server.
4. The server checks an in-memory cache, then PostgreSQL. On a miss, it calls Groq's llama-3.1-8b-instant and caches the result in both.
5. The summary is returned and displayed in the popup.

---

Setup

Backend

cd backend-api

# Start a local Postgres instance
docker compose up -d

python -m venv venv
venv\Scripts\activate        # macOS/Linux: source venv/bin/activate
pip install -r requirements.txt

cp .env.example .env
# Add your GROQ_API_KEY to .env

uvicorn main:app --reload --port 8000

Verify: http://localhost:8000/health should return {"status": "ok"}.

Extension

cd extension-frontend
npm install
npm run build

Then in Chrome:

1. Go to chrome://extensions
2. Enable Developer mode
3. Click Load unpacked → select extension-frontend/dist

---

Database

Summaries are persisted to PostgreSQL via SQLAlchemy's async ORM (db.py).

Schema:

threads
  id, content_hash (unique), tweet_count, created_at, last_accessed_at, access_count
      │
      │ 1-to-many
      ▼
summaries
  id, thread_id (FK → threads.id), summary_text, model, created_at

request_log
  id, content_hash, cache_hit, created_at

request_log stores content_hash directly instead of a foreign key — it's a write-heavy audit table used only for aggregate analytics, so skipping the join keeps writes cheap.

Endpoints:

• GET /history?limit=10 — recent summaries, newest first
• GET /stats — cache hit rate and usage counts
• GET /cache/memory — current in-memory cache size

Inspect directly:

docker exec -it threadit-db psql -U postgres -d threadit
\dt
SELECT * FROM summaries ORDER BY created_at DESC LIMIT 5;

---

Caching

Two layers:

Layer	Implementation	Persistent
L1	In-memory TTL dict (SHA-256 keyed)	No — cleared on restart
L2	PostgreSQL	Yes

TTL defaults to 1 hour, configurable via CACHE_TTL_SECONDS in .env.

---

DOM selectors

content.js targets article[data-testid="tweet"] and div[data-testid="tweetText"]. These attributes are used by X's own test suite and are more stable than the obfuscated class names. If scraping breaks, open DevTools on a tweet page and verify the data-testid values haven't changed.