mkdir -p ~/Tools/paper-scanner
cd ~/Tools/paper-scanner
 
# Initialize environment
uv init --python 3.12
uv add pymupdf openai
 

export DEEPSEEK_API_KEY="sk-xxxx"
export DEEPSEEK_BASE_URL="https://api.deepseek.com/v1"
export DEEPSEEK_MODEL="deepseek-chat"
# Optional: Qwen or Doubao keys...
 

import os
import sys
import fitz  # PyMuPDF
import json
import argparse
from openai import OpenAI
from concurrent.futures import ThreadPoolExecutor
 
# --- 1. Config Loading ---
MODEL_CONFIG = {
    "deepseek": {
        "api_key": os.getenv("DEEPSEEK_API_KEY"),
        "base_url": os.getenv("DEEPSEEK_BASE_URL"),
        "model": os.getenv("DEEPSEEK_MODEL", "deepseek-chat")
    },
    "qwen": {
        "api_key": os.getenv("QWEN_API_KEY"),
        "base_url": os.getenv("QWEN_BASE_URL"),
        "model": os.getenv("QWEN_MODEL", "qwen-max")
    },
    "doubao": {
        "api_key": os.getenv("DOUBAO_API_KEY"),
        "base_url": os.getenv("DOUBAO_BASE_URL"),
        "model": os.getenv("DOUBAO_MODEL")
    }
}
 
# --- 2. Dual Logger ---
class DualLogger:
    def __init__(self, filepath=None):
        self.file = None
        if filepath:
            try:
                self.file = open(filepath, 'w', encoding='utf-8')
                print(f"💾 Results will be saved to: {filepath}")
            except Exception as e:
                print(f"⚠️ Failed to create output file: {e}")
 
    def log(self, message, end="\n", to_file=True):
        print(message, end=end, flush=True)
        if self.file and to_file:
            self.file.write(message + end)
            self.file.flush()
 
    def close(self):
        if self.file:
            self.file.close()
 
# --- 3. Core Functions ---
def get_client(model_name):
    config = MODEL_CONFIG.get(model_name)
    if not config or not config["api_key"]:
        print(f"❌ Error: Missing config for {model_name}")
        sys.exit(1)
    return OpenAI(api_key=config["api_key"], base_url=config["base_url"]), config["model"]
 
def extract_abstracts(pdf_path, logger):
    if not os.path.exists(pdf_path):
        logger.log(f"❌ File not found: {pdf_path}")
        sys.exit(1)
 
    logger.log(f"📖 Reading PDF: {os.path.basename(pdf_path)} ...", to_file=False)
    doc = fitz.open(pdf_path)
    toc = doc.get_toc(simple=True)
    
    start_pages = sorted(list(set([item[2] for item in toc if item[2] > 0])))
    if 1 not in start_pages: start_pages.insert(0, 1)
    
    papers = []
    logger.log(f"📑 Identified {len(start_pages)} papers. Extracting abstracts...", to_file=False)
    
    for i, page_num in enumerate(start_pages):
        try:
            # Extract first 2000 chars (Covering Title + Abstract)
            text = doc[page_num - 1].get_text()[:2000]
            papers.append({"id": i+1, "page": page_num, "text": text})
        except Exception:
            continue
    return papers
 
def analyze_paper(client, model_id, paper, keywords):
    prompt = f"""
    You are a research assistant. Read the following paper snippet (Title/Abstract).
    User is looking for papers related to: "{keywords}".
    
    Input Text:
    ---
    {paper['text']}
    ---
    
    Return strictly a JSON object:
    {{
        "title": "Paper Title",
        "is_relevant": true/false,
        "reason": "Brief reason in Chinese (Start with '匹配' or '不匹配')"
    }}
    """
    try:
        response = client.chat.completions.create(
            model=model_id,
            messages=[
                {"role": "system", "content": "You are a JSON parser helper."},
                {"role": "user", "content": prompt}
            ],
            temperature=0.1,
            response_format={"type": "json_object"}
        )
        result = json.loads(response.choices[0].message.content)
        return {**paper, **result}
    except Exception:
        return None
 
# --- 4. Main ---
def main():
    parser = argparse.ArgumentParser(description="AI Paper Scanner")
    parser.add_argument("pos_input", nargs="?", help="PDF Path")
    parser.add_argument("pos_query", nargs="?", help="Keywords")
    parser.add_argument("--in", "--input", dest="opt_input")
    parser.add_argument("--q", "--query", dest="opt_query")
    parser.add_argument("--out", "--output", dest="output_file")
    
    group = parser.add_mutually_exclusive_group()
    group.add_argument("--deepseek", action="store_true", default=True)
    group.add_argument("--qwen", action="store_true")
    group.add_argument("--doubao", action="store_true")
 
    args = parser.parse_args()
    pdf_path = args.opt_input or args.pos_input
    keywords = args.opt_query or args.pos_query
    logger = DualLogger(args.output_file)
 
    if not pdf_path or not keywords:
        logger.log("Usage: scan-issue issue.pdf 'Keywords' [--out result.txt]")
        sys.exit(1)
 
    selected_model = "deepseek"
    if args.qwen: selected_model = "qwen"
    if args.doubao: selected_model = "doubao"
 
    client, model_id = get_client(selected_model)
    
    logger.log("="*50)
    logger.log(f"🚀 Model: {selected_model.upper()}")
    logger.log(f"📂 File: {os.path.basename(pdf_path)}")
    logger.log(f"🔍 Keywords: {keywords}")
    logger.log("="*50)
 
    papers = extract_abstracts(pdf_path, logger)
    results = []
    
    logger.log(f"\nScanning {len(papers)} papers concurrently...\n", to_file=False)
 
    with ThreadPoolExecutor(max_workers=5) as executor:
        futures = [executor.submit(analyze_paper, client, model_id, p, keywords) for p in papers]
        for future in futures:
            res = future.result()
            if res and res.get('is_relevant'):
                msg = f"✅ [P.{res['page']}] {res['title']}\n   💡 {res['reason']}\n"
                logger.log(msg)
                results.append(res)
 
    logger.log("-"*50)
    logger.log(f"🎉 Done! Found {len(results)} relevant papers.")
    if args.output_file:
        print(f"💾 Saved to: {os.path.abspath(args.output_file)}")
    logger.close()
 
if __name__ == "__main__":
    main()
 

function scan-issue() {
    # Directly call the python interpreter inside the .venv created by uv
    # This preserves the current working directory (CWD)
    ~/Tools/paper-scanner/.venv/bin/python ~/Tools/paper-scanner/scanner.py "$@"
}
 

scan-issue RA-L_Issue_1.pdf "Soft Robotics, Sensors"
 

scan-issue RA-L_Issue_1.pdf "Deep Learning" --out reading_list.txt
 

scan-issue RA-L_Issue_1.pdf "SLAM" --qwen --out slam_papers.txt