Production-grade AI research assistant with autonomous multi-tool reasoning. Built with LangGraph, Google Gemini, and Chainlit.
生產級 AI 研究助理，具備自主多工具推理能力。基於 LangGraph、Google Gemini 與 Chainlit 構建。
Live demo: Coming soon at Hugging Face Spaces / 即將上線
AutoResearch Agent autonomously researches any topic by combining:
- Strategic Planning — Creates research strategy
- Web Search — Tavily AI-optimized search
- Content Reading — Full-page extraction
- Synthesis — Multi-source insights
- Citation Tracking — Verifiable sources
- PDF Reports — Professional output
AutoResearch Agent 能夠自主研究任何主題：
- 策略規劃 — 制定研究策略
- 網路搜尋 — Tavily AI 優化搜尋
- 內容讀取 — 完整頁面擷取
- 資訊整合 — 多來源洞察
- 引用追蹤 — 可驗證來源
- PDF 報告 — 專業輸出
- Multi-tool Agent — 4 specialized tools / 4 個專業工具
- ReAct Pattern — Think-Act-Observe-Repeat / ReAct 推理模式
- LangGraph Workflow — Production-grade framework / 生產級框架
- Beautiful Chainlit UI — Modern AI chat interface / 現代化聊天介面
- Streaming Responses — Real-time updates / 即時更新
- Citation Tracking — Every claim sourced / 完整來源引用
- Free APIs — Generous free tiers / 慷慨的免費 API
┌─────────────────────────────────────────────────┐
│ USER QUERY │
└────────────────────┬────────────────────────────┘
▼
┌─────────────┐
│ Planner │ Research strategy
└──────┬──────┘
▼
┌─────────────┐
│ Researcher │ ←──────────┐
│ (LLM) │ │
└──────┬──────┘ │
│ Tool Call? │
▼ │
┌──────────────┐ │
│ Tools │ │
│ │ │
│ web_search │───────────┘
│ web_fetch │
│ summarize │
│ pdf_gen │
└──────────────┘
│
┌──────▼──────┐
│ Writer │ Final report
└──────┬──────┘
▼
┌─────────────┐
│ REPORT │ With citations
└─────────────┘
| Component / 元件 | Technology / 技術 |
|---|---|
| Agent Framework | LangGraph |
| LLM | Google Gemini 2.5 Flash |
| UI Framework | Chainlit (modern AI-native) |
| Web Search | Tavily AI Search |
| Web Scraping | BeautifulSoup4 |
| Vector DB | ChromaDB |
| PDF Generation | ReportLab |
- Python 3.10+
- Google account (for Gemini API)
- Tavily account (for search API)
# Clone repo
git clone https://github.com/your-username/autoresearch-agent.git
cd autoresearch-agent
# Install dependencies
pip install -r requirements.txt
# Setup environment
cp .env.example .env
# Edit .env with your API keys- Gemini API: https://aistudio.google.com/app/apikey
- Tavily Search: https://app.tavily.com (1,000 searches/month free)
# Test setup
python -m tests.test_setup
# Run with Chainlit (recommended) / 啟動 Chainlit (推薦)
chainlit run ui/chainlit_app.py -w
# Or run with Streamlit
streamlit run ui/streamlit_app.pyBrowser opens at: http://localhost:8000 (Chainlit) or http://localhost:8501 (Streamlit)
This project includes two UI implementations:
本專案包含兩種介面實作：
- Modern AI-native chat interface / 現代化 AI 聊天介面
- Built-in streaming and tool visualization / 內建串流與工具視覺化
- Beautiful animations and theming / 精美動畫與主題
- ChatGPT-like UX / 類 ChatGPT 體驗
- Simpler dashboard layout / 較簡潔的儀表板佈局
- Familiar to ML practitioners / ML 從業者熟悉
- Quick deployment / 快速部署
chainlit run ui/chainlit_app.py -wFeatures:
- 4 starter questions
- 3 chat profiles (Quick, Deep, Academic)
- Action buttons (PDF export, follow-up)
- Live tool execution visualization
- Custom purple/violet theme
from agents import ResearchAgent
agent = ResearchAgent()
result = agent.research("Latest AI agent trends 2026")
print(result["report"])
for source in result["sources"]:
print(f"- {source['title']}: {source['url']}")python -m scripts.run_cli "Your research query here"autoresearch-agent/
├── agents/
│ └── research_agent.py # LangGraph agent
├── core/
│ ├── config.py # Configuration
│ ├── state.py # Agent state
│ └── prompts.py # System prompts
├── tools/
│ ├── web_search.py # Tavily search
│ ├── web_fetch.py # URL reader
│ ├── summarizer.py # Summarizer
│ └── pdf_generator.py # PDF builder
├── ui/
│ ├── chainlit_app.py # Chainlit UI (recommended)
│ └── streamlit_app.py # Streamlit UI (alternative)
├── public/
│ ├── style.css # Custom Chainlit CSS
│ ├── logo_dark.svg
│ ├── logo_light.svg
│ └── icons/ # Starter icons
├── .chainlit/
│ └── config.toml # Chainlit configuration
├── chainlit.md # Chainlit welcome page
├── tests/
├── scripts/
├── notebooks/
├── docs/
├── .env.example
├── requirements.txt
└── README.md
-
Why LangGraph? / 為何選擇 LangGraph？
- Modern, graph-based architecture
- Production-grade with checkpointing
- 現代化圖形架構，具備檢查點機制
-
Why Chainlit over Streamlit? / 為何選擇 Chainlit？
- Designed for AI/LLM apps specifically
- Built-in streaming and tool visualization
- More professional appearance
- 專為 AI/LLM 應用設計，外觀更專業
-
Why Tavily for search? / 為何選擇 Tavily？
- AI-optimized results
- Generous free tier (1K/month)
- Better than Google Custom Search for agents
-
Why Gemini 2.5 Flash? / 為何選擇 Gemini？
- Fast and free
- 1M token context window
- Strong reasoning capabilities
- Multi-agent orchestration (researcher + fact-checker + writer)
- Long-term memory with ChromaDB
- Email integration (Gmail API)
- Voice input/output
- Citation verification agent
- Custom domain RAG (upload PDFs)
- Hugging Face Spaces deployment
- Mobile-friendly UI
MIT License — see LICENSE for details.
[Embun Ventani]
- GitHub: @Venta02
- LinkedIn: embun ventani
- Email: embunventa02@gmail.com
If this project helped you, please consider starring the repository.
若本專案對您有所幫助，歡迎為儲存庫加上星標。