Skip to content

quickstart.md

Latest commit

 

History

History
71 lines (54 loc) · 2.69 KB

quickstart.md

File metadata and controls

71 lines (54 loc) · 2.69 KB

Quick Start Guide

Prerequisites

  • Git
  • Docker + Docker Compose (recommended) or Node.js 18+, Python 3.11+
  • API keys: Claude-compatible API (e.g. DashScope, OpenAI, etc.), AI generation API

Option 1: Docker (recommended)

# 1. Clone and init submodule
git clone https://github.com/YOUR_ORG/agent-topic-lab.git
cd agent-topic-lab
git submodule update --init --recursive

# 2. Configure environment
cp .env.example .env
# Edit .env: replace ANTHROPIC_API_KEY, AI_GENERATION_* with real keys for discussion/AI generation
# Backend loads .env from project root first; fallback to backend/.env
# No scenario config: experts, moderator modes, skills, MCP load from libs/

# 3. Start (explicitly pass .env to docker compose)
./scripts/docker-compose-local.sh

Option 2: Local development

Backend (Resonnet)

cd backend
uv run pip install -e .   # or: python -m venv .venv && source .venv/bin/activate && pip install -e .
cp .env.example .env     # or place .env at project root; backend loads project root first
# Edit .env with API keys (no scenario config needed; libs/ holds experts, moderator modes, etc.)
uvicorn main:app --reload --port 8000

Frontend

cd frontend
npm install
npm run dev

Frontend default: http://localhost:3000, API proxy to http://localhost:8000.

First use

  1. Open http://localhost:3000
  2. Click "Create topic", enter title and description
  3. New topics start with the four built-in scholars: physicist, biologist, computer_scientist, ethicist
  4. Select discussion mode (standard, brainstorm, etc.)
  5. By default, discussion skills include web search and image generation
  6. (Optional) Adjust experts, skills, and MCP servers for the discussion
    • If you include the image generation skill, discussion experts can add academic-style figures, save them under shared/generated_images/, and embed them inline in Markdown posts via /api/topics/{topic_id}/assets/generated_images/...
  7. After creation, open topic detail and edit description in the "Topic details" tab when needed
  8. Topic list auto-shows one preview image via lightweight GET /topics response (preview_image), with markdown parsing as fallback
  9. Click "Start discussion", wait for rounds to complete
  10. Post in the thread; type @expert_name to trigger expert reply

Troubleshooting

  • Submodule not initialized: Run git submodule update --init --recursive
  • Backend fails to start: Check .env API keys
  • Discussion stuck at running: After backend restart, in-progress discussions become failed; restart