mcp/alerts.py

# mcp/alerts.py
"""
Saved-query alert helper.
• Stores the last 30 PubMed/arXiv links per query.
• Returns a dict of {query: [new_links]} when fresh papers appear.
Implementation is intentionally simple: JSON on disk + orchestrate_search.
"""

import json, asyncio
from pathlib import Path
from typing import List, Dict

from mcp.orchestrator import orchestrate_search

_ALERT_DB = Path("saved_alerts.json")
_MAX_IDS  = 30   # keep last N links per query


def _read_db() -> Dict[str, List[str]]:
    if _ALERT_DB.exists():
        return json.loads(_ALERT_DB.read_text())
    return {}


def _write_db(data: Dict[str, List[str]]):
    _ALERT_DB.write_text(json.dumps(data, indent=2))


async def check_alerts(queries: List[str]) -> Dict[str, List[str]]:
    """
    For each saved query, run a quick orchestrate_search and detect new paper links.
    Returns {query: [fresh_links]}  (empty dict if nothing new).
    """
    db      = _read_db()
    new_map = {}

    async def _check(q: str):
        res   = await orchestrate_search(q)
        links = [p["link"] for p in res["papers"]]
        prev  = set(db.get(q, []))
        fresh = [l for l in links if l not in prev]
        if fresh:
            new_map[q] = fresh
            db[q] = links[:_MAX_IDS]  # save trimmed
    # run in parallel
    await asyncio.gather(*[_check(q) for q in queries])
    _write_db(db)
    return new_map