Knowledge Studio (WIP)

Ingest, ask, generate — all powered by AI.

Features

• Ingest — Upload PDF, Markdown, plain text, or import from URLs. Content is automatically parsed, chunked, and indexed for search.
• Ask — Ask questions in natural language and get answers with source citations. Supports multi-turn conversation with streaming output.
• Generate — Produce structured multi-chapter reports from the content in your knowledge base.
• Observe — Built-in tracing across every LLM call and retrieval step.

Quick Start

Prerequisites: Docker & Docker Compose, an OpenAI-compatible API key.

git clone https://github.com/siuhui/knowledge-studio.git
cd knowledge-studio

cp apps/api/.env.example apps/api/.env
# Edit .env: fill in KS_LLM__API_KEY and KS_EMBEDDING__API_KEY

docker compose -f infra/docker-compose.yml up -d

Open http://localhost:3000, register an account, create a knowledge base, and upload your first document.

Service	URL
Web UI	http://localhost:3000
API docs (Swagger)	http://localhost:8000/docs
MinIO console	http://localhost:9001 (minioadmin / minioadmin)

Development

Two ways to run in development

Docker Compose (hot reload)

Full stack with live reload for both API and frontend. Database and MinIO data are bind-mounted to infra/data/ so they survive container teardown.

cp apps/api/.env.example apps/api/.env
docker compose -f infra/docker-compose.yml -f infra/docker-compose.dev.yml up -d

Local (manual)

Run only infrastructure in Docker, start backend and frontend manually for the fastest feedback loop.

Requirements: Python 3.12+, pnpm (for Node.js 20+).

docker compose -f infra/docker-compose.yml -f infra/docker-compose.dev.yml up -d db minio

# Backend
cd apps/api
cp .env.example .env
python -m venv .venv
source .venv/Scripts/activate   # Windows Git Bash (Linux/macOS: .venv/bin/activate)
pip install -r requirements-dev.txt
uvicorn app.main:app --reload --port 8000

# Frontend (separate terminal)
cd apps/web
pnpm install && pnpm dev

Tech Stack

Layer	Technology
Backend	Python 3.12, FastAPI, SQLAlchemy
Database	PostgreSQL 16 + pgvector
Storage	MinIO (S3-compatible)
Frontend	Next.js 15, React 19, Tailwind CSS v4
AI	OpenAI-compatible API, Anthropic
Telemetry	Langfuse (OpenTelemetry)

Documentation

• Product Requirements
• Architecture & Design
• Data Model
• Engineering Standards

License

MIT