AI Frontier Explorer 🧠🔍

Track cutting-edge AI research, deep-read papers, and generate plain-language reports — powered by DeepSeek + LangChain.

✨ Features

• 📡 Track — Search arXiv for the latest papers on any AI topic
• 📖 Deep Read — Fetch full paper content via web scraping
• 🤖 Parallel Sub-agents — Delegate paper analysis to sub-agents for concurrent reading
• 📝 Structured Reports — Auto-generate reports with: summary, key points, plain-language explanation, glossary, impact score
• 💾 Persistent Memory — Remembers your preferences across conversations
• 🎯 Skills System — Built-in AI research methodology

🚀 Quick Start

Prerequisites

• Python 3.11+
• API Keys: DeepSeek + LangSmith

1. Install

git clone https://github.com/your-username/ai-frontier-explorer
cd ai-frontier-explorer
python -m venv .venv

# Windows
.venv\Scripts\activate

# macOS / Linux
source .venv/bin/activate

pip install -r requirements.txt

2. Configure

Create .env file. On Windows PowerShell, use Set-Content so the file encoding is explicit:

"DEEPSEEK_API_KEY=sk-your_deepseek_key" | Set-Content -Encoding utf8 .env

The CLI also accepts UTF-16 .env files produced by Windows PowerShell 5. Full example:

DEEPSEEK_API_KEY=sk-your_deepseek_key
DEEPSEEK_BASE_URL=https://api.deepseek.com
DEEPSEEK_MODEL=deepseek-v4-flash

LANGCHAIN_TRACING_V2=true
LANGCHAIN_API_KEY=lsv2_pt_your_langsmith_key
LANGCHAIN_PROJECT=ai-frontier-explorer

# 报告保存位置（可选，默认在用户数据目录的 reports/ 下）
# REPORTS_DIR=D:\MyReports\papers

3. Launch

# One-click (Windows)
start.bat

# Or manually
python ai_frontier.py

4. 设置 API Key（持久化，升级不丢失）

# 方式一：CLI 命令（推荐，保存到用户数据目录，升级/换目录不丢）
/key sk-你的DeepSeek密钥

# 方式二：.env 文件（首次运行会自动迁移到持久化配置）
DEEPSEEK_API_KEY=sk-你的DeepSeek密钥

🎮 Usage

╔══════════════════════════════════════════╗
║    AI Frontier Explorer  v1.0            ║
║   追踪前沿 · 深度研读 · 通俗报告          ║
╚══════════════════════════════════════════╝

You > /track AI Agent 最新进展
You > /track 多模态大模型 2026
You > /search transformer architecture
You > /reports
You > /read 2026-07-30-AI-Agent-前沿进展追踪.md
You > /remember 我关注AI安全方向

Commands

🧠 Architecture

ai_frontier.py          # CLI entry point
deep_agent/
  ├── agent.py          # create_deep_agent() with DeepSeek model
  ├── config.py         # Settings management
  └── llm.py            # LLM initialization
tools/
  ├── arxiv_tools.py    # arXiv paper search
  ├── web_tools.py      # Web page content extraction
  ├── report_tools.py   # Save reports to disk
  ├── search.py         # DuckDuckGo web search
  └── calculator.py     # Safe calculator
skills/
  └── ai-research/      # Research methodology skill
memories/               # Persistent memory files
reports/                # Generated reports output

🔧 Tech Stack

• Agent Framework: Deep Agents SDK (official)
• Orchestration: LangGraph
• LLM: DeepSeek via OpenAI-compatible API
• Memory: LangGraph StoreBackend
• CLI: Rich terminal UI

📄 Report Format

Each generated report includes:

• Name — Paper title
• Domain — AI subfield (NLP/CV/RL/Agent/LLM...)
• Summary — What the paper does
• Key Points — 3-5 core insights
• Plain Explanation — Easy-to-understand analogy-based explanation
• Glossary — Technical term definitions
• Impact Score — 1-10 rating with breakdown

📦 Install as a Package

pip install -e .
ai-frontier     # launch from anywhere

📜 License

MIT