Discover scientific papers. Inspect source-matched excerpts.
Keep the research auditable.

AI-generated literature summaries are useful, but researchers need to inspect the sources behind them.

TraceReview combines literature discovery, AI-assisted reading, source-matched excerpts, and workflow auditing in one application, so every claim can be followed back to the paper it came from.

Prerequisites: Docker with Docker Compose, and a Groq API key.

1. Clone the repository

git clone https://github.com/samavinasir/tracereview.git
cd tracereview

2. Configure the environment

Copy .env.example to .env, then set GROQ_API_KEY and your database password.

cp .env.example .env

3. Start everything

docker compose up --build -d

4. Open the app

• 🌐  Web app: http://localhost:5173
• 📖  API docs: http://localhost:8000/docs
• 💚  Health check: http://localhost:8000/health

Settings live in .env (start from .env.example).

Variable	Required	Purpose
GROQ_API_KEY	Yes	API key for the Groq-hosted models used for summaries and excerpt selection.
Database password	Yes	Password for the PostgreSQL container.

# Lint and format check
ruff check .

# Type checking
mypy src

# Tests
pytest

For production, use docker-compose.prod.yml (with the optional Caddy reverse proxy).

Warning

TraceReview is under active development. Please keep these limitations in mind:

• Ranking is currently deterministic and lexical; embedding-based ranking is not implemented.
• Excerpt matching checks wording, not scientific truth.
• Research runs cannot resume after an API restart.

• Embedding-based ranking (pgvector is already part of the stack)
• Resumable research runs that survive an API restart
• Exportable reports (Markdown / PDF)
• Additional literature providers

See LICENSE for details.

---

Discover carefully. Read critically. Trace every excerpt.

_{Built for researchers who want more transparent literature workflows.}