Voyager Research Agent

A voyager research agent with:

• a FastAPI backend that exposes the agent over HTTP
• a Streamlit frontend UI that calls the API for chat

Install

Install dependencies with uv:

uv sync

Configure

Add your OpenAI API key to .env:

OPENAI_API_KEY=your_api_key_here

How the app works

1. API layer (api.py, src/api/v1_router.py)

   • Starts a FastAPI app.
   • Exposes POST /v1/chat (SSE streaming response) and GET /v1/health.
   • Uses a shared SessionManager stored in app.state.

2. Agent execution (src/agent/executor.py)

   • Builds the OpenAI agent (model, instructions, tools).
   • Streams token/events from the agent and formats them as SSE payloads.
   • Supports optional context injection before running the query.

3. UI layer (src/app/streamlit_app.py)

   • Renders chat + report tabs in Streamlit.
   • Sends chat requests to POST /v1/chat.
   • Reads SSE events and progressively renders assistant output.
   • Stores per-session chat history in Streamlit state.

Run

Start the API:

uv run uvicorn api:app --host 0.0.0.0 --port 8082 --reload

Then start the Streamlit app (in another terminal):

API_BASE_URL=http://localhost:8082 uv run streamlit run main.py