Web application

A browser-based reference implementation of conversational RAG, built on a Starlette backend with pydantic-ai's AGUIAdapter and a Next.js / CopilotKit frontend. It lives in the app/ directory of the haiku.rag repository.

It is a starting point for your own deployment. The terminal equivalent is Chat.

It streams answers and tool calls, shows expandable citations with visual grounding, restricts searches through a document filter, and keeps sessions in the browser with markdown export.

Quick start

cd app
cp .env.example .env
cp haiku.rag.yaml.example haiku.rag.yaml
docker compose up -d --build
# with hot reload instead: docker compose -f docker-compose.dev.yml up -d --build

The frontend is at http://localhost:3000 and the backend at http://localhost:8001. The app needs an existing database, created with the haiku-rag CLI.

Configuration

.env in the app/ directory holds the secrets and the database path:

# API keys for the providers haiku.rag.yaml uses (none for Ollama)
ANTHROPIC_API_KEY=your-anthropic-key
OPENAI_API_KEY=your-openai-key

# Host path of the LanceDB database, mounted at /data
DB_VOLUME=/path/to/your/haiku.rag.lancedb

# Optional: Logfire for observability
LOGFIRE_TOKEN=your-logfire-token

haiku.rag.yaml, copied from haiku.rag.yaml.example, places the database at /data and configures the models. The compose files point HAIKU_RAG_CONFIG_PATH at it. The example uses Ollama on the host through providers.ollama.base_url: http://host.docker.internal:11434, which takes precedence over OLLAMA_BASE_URL:

lancedb:
  databases:
    haiku.rag: /data

qa:
  model:
    provider: ollama
    name: gpt-oss

embeddings:
  model:
    provider: ollama
    name: qwen3-embedding:4b
    vector_dim: 2560

Outside compose, the backend loads its configuration like the CLI: HAIKU_RAG_CONFIG_PATH, then ./haiku.rag.yaml, then the platform directory.

API endpoints

Development

The backend reloads automatically on file changes. For frontend changes:

docker compose -f docker-compose.dev.yml up -d --build frontend

If LOGFIRE_TOKEN is set, LLM calls are traced and available in the Logfire dashboard.