deepxiv_sdk

KO | EN

gitlite — search

#python #hacktoberfest #typescript #go #android #deep-learning #macos #java #kubernetes #vue #security #pytorch #mcp #open-source

(★ 10)

Talk to research papers like talking to authors - Python package with AI agent for arXiv papers

deepxiv_sdk 최신버젼 다운로드

최종 버전 다운로드 (.zip)

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# deepxiv-sdk

A Python SDK for accessing arXiv papers with CLI and MCP server support.

**🎮 Try the live demo:** [https://1stauthor.com/](https://1stauthor.com/)

**📚 API Documentation:** [https://data.rag.ac.cn/api/docs](https://data.rag.ac.cn/api/docs)

## Features

- 🔍 **Paper Search**: Search for arXiv papers using hybrid search (BM25 + Vector)
- 📄 **Paper Access**: Retrieve paper metadata, sections, and full content
- 💻 **CLI**: Command-line interface for quick access
- 🔌 **MCP Server**: Model Context Protocol server for Claude Desktop integration
- 🤖 **Intelligent Agent**: ReAct-based agent for intelligent paper analysis
- 🔌 **Flexible LLM Support**: Compatible with OpenAI, DeepSeek, OpenRouter, and other OpenAI-compatible APIs

## Installation

```bash
# Basic install (Reader + CLI)
pip install deepxiv-sdk

# With MCP server support
pip install deepxiv-sdk[mcp]

# With Agent support
pip install deepxiv-sdk[agent]

# Full install (all features)
pip install deepxiv-sdk[all]
```

## Quick Start

### Step 1: Get Your Free API Token

Visit [https://data.rag.ac.cn/register](https://data.rag.ac.cn/register) to get your free API token (1000 requests/day).

### Step 2: Configure Your Token

```bash
# Interactive configuration (saves to ~/.env)
deepxiv config

# Or provide token directly
deepxiv config --token YOUR_TOKEN

# The CLI will automatically load token from ~/.env
```

### CLI Usage

```bash
# Show help
deepxiv help

# Get paper in different formats
deepxiv paper 2409.05591                    # Full markdown
deepxiv paper 2409.05591 --head             # Metadata (JSON)
deepxiv paper 2409.05591 --raw              # Raw markdown
deepxiv paper 2409.05591 --preview          # Preview (~10k chars)
deepxiv paper 2409.05591 --section intro    # Specific section

# Search papers
deepxiv search "agent memory" --limit 5
deepxiv search "transformer" --mode bm25 --format json
deepxiv search "LLM" --categories cs.AI,cs.CL --min-citations 100

# Start MCP server
deepxiv serve
```

### Python API

```python
from deepxiv_sdk import Reader

# Initialize the reader
reader = Reader(token="your_api_token")  # or Reader() for free papers

# Search for papers
results = reader.search("agent memory", size=10)
for paper in results['results']:
    print(f"{paper['title']} - {paper['arxiv_id']}")

# Get paper metadata
head = reader.head("2409.05591")
print(f"Title: {head['title']}")

# Read a section (case-insensitive)
intro = reader.section("2409.05591", "Introduction")
print(intro)

# Get full paper
content = reader.raw("2409.05591")
```

### Agent Usage

```python
import os
from deepxiv_sdk import Reader, Agent

reader = Reader(token="your_api_token")
agent = Agent(
    api_key=os.getenv("OPENAI_API_KEY"),
    model="gpt-4",
    reader=reader,
    print_process=True
)

answer = agent.query("What are the latest papers about agent memory?")
print(answer)
```

## MCP Server Setup

### For Claude Desktop

Add to `~/Library/Application Support/Claude/claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "deepxiv": {
      "command": "deepxiv",
      "args": ["serve"],
      "env": {
        "DEEPXIV_TOKEN": "your_token_here"
      }
    }
  }
}
```

### Available MCP Tools

| Tool | Description |
|------|-------------|
| `search_papers` | Search arXiv with hybrid search |
| `get_paper_metadata` | Get paper metadata and section TLDRs |
| `get_paper_section` | Read a specific section |
| `get_full_paper` | Get complete paper content |
| `get_paper_preview` | Get preview (~10k chars) |

## API Token

- **Get Your Free Token**: [https://data.rag.ac.cn/register](https://data.rag.ac.cn/register)
- **Daily Limit**: 1000 free requests per day
- **Test Papers**: `2409.05591` and `2504.21776` are available without authentication

### Token Configuration (3 Ways)

**1. Using `config` command (Recommended)**
```bash
deepxiv config
# Saves to ~/.env and automatically loads on every command
```

**2. Environment Variable**
```bash
export DEEPXIV_TOKEN="your_token_here"
# Add to ~/.bashrc or ~/.zshrc for persistence
```

**3. Command-line Option**
```bash
deepxiv paper 2512.02556 --token "your_token_here"
# Useful for one-time usage or multiple tokens
```

The CLI automatically loads tokens from:
1. Command-line `--token` option (highest priority)
2. `DEEPXIV_TOKEN` environment variable
3. `.env` file in current directory
4. `~/.env` file in home directory (lowest priority)

## API Reference

### Reader Methods

- `search(query, size=10, search_mode="hybrid", ...)`: Search for papers
- `head(arxiv_id)`: Get paper metadata and structure
- `section(arxiv_id, section_name)`: Get a specific section (case-insensitive)
- `raw(arxiv_id)`: Get full paper in markdown
- `preview(arxiv_id)`: Get paper preview (~10k chars)
- `json(arxiv_id)`: Get complete structured JSON

### Agent Methods

- `query(question, reset_papers=False)`: Query the agent
- `get_loaded_papers()`: Get loaded papers info
- `reset_papers()`: Reset all loaded papers
- `add_paper(arxiv_id)`: Add a paper to context

## Examples

See the [examples](examples/) directory:

- `example_reader.py`: Basic Reader usage
- `example_agent.py`: Agent usage
- `example_advanced.py`: Advanced patterns
- `quickstart.py`: Quick start guide

## License

MIT License - see [LICENSE](LICENSE) file for details.

## Support

- 🐛 **GitHub Issues**: [https://github.com/qhjqhj00/deepxiv_sdk/issues](https://github.com/qhjqhj00/deepxiv_sdk/issues)
- 📚 **API Documentation**: [https://data.rag.ac.cn/api/docs](https://data.rag.ac.cn/api/docs)
- 🎮 **Demo**: [https://1stauthor.com/](https://1stauthor.com/)

deepxiv_sdk

deepxiv_sdk 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

deepxiv_sdk

deepxiv_sdk 최신버젼 다운로드

오늘의 추천 프로젝트

Was this content helpful?

Recent Feedback

Download README