235 lines
9.2 KiB
Python
235 lines
9.2 KiB
Python
"""
|
||
创新启发（Pillar 4）：缺口挖掘 + 合作者匹配 + 创新点 + LLM 研究方向建议
|
||
|
||
模式:
|
||
--owner O --repo R 单仓：缺口/匹配/创新点 + LLM idea
|
||
--category <域> 领域级：聚合分类下 top 仓的缺口 + 领域主题 + LLM idea
|
||
|
||
LLM 可选：配置 DEEPSEEK_API_KEY 才生成 idea；无 key 确定性产出照常。
|
||
产物：inspire.json + report.md。
|
||
"""
|
||
from __future__ import annotations
|
||
|
||
import argparse
|
||
import json
|
||
import os
|
||
import sys
|
||
from collections import Counter
|
||
|
||
sys.path.insert(0, os.path.dirname(os.path.abspath(__file__)))
|
||
import collect as c
|
||
import topics as T
|
||
import match as M
|
||
import lineage as L
|
||
import hotspot as H
|
||
import llm
|
||
|
||
SYS_PROMPT = (
|
||
"你是资深科研协作顾问。基于给定的 GitLink 仓库/领域缺口信号、可合作学者、近期创新点，"
|
||
"提出 3-5 条具体、可执行的研究方向或合作建议。要求：紧扣缺口主题、指出可切入的角度、"
|
||
"点名潜在合作者类型。用中文，分条输出（1. 2. 3.），每条不超过 80 字。"
|
||
)
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 单仓创新启发
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def analyze_repo(owner: str, repo: str, top: int = 8, pool_cap: int = 12,
|
||
context: dict | None = None) -> dict:
|
||
"""单仓：缺口 + 匹配 + 创新点 + LLM idea。"""
|
||
m = M.match(owner, repo, top=top, pool_cap=pool_cap) # gap_topics/signals/candidates
|
||
innovations: list[dict] = []
|
||
try:
|
||
innovations = (L.lineage(owner, repo) or {}).get("innovation_points") or []
|
||
except Exception as e:
|
||
sys.stderr.write(f"[inspire] lineage 取创新点失败: {e!r}\n")
|
||
|
||
idea = _llm_idea_repo(owner, repo, m, innovations, context)
|
||
|
||
return {
|
||
"scenario": "inspire",
|
||
"mode": "repo",
|
||
"repo": f"{owner}/{repo}",
|
||
"gap_topics": m.get("gap_topics") or [],
|
||
"needed_languages": m.get("needed_languages") or [],
|
||
"gap_signals": (m.get("gap_signals") or [])[:12],
|
||
"candidates": m.get("candidates") or [],
|
||
"innovation_points": innovations[:6],
|
||
"idea": idea,
|
||
"llm_used": idea is not None,
|
||
}
|
||
|
||
|
||
def _llm_idea_repo(owner: str, repo: str, m: dict, innovations: list[dict],
|
||
context: dict | None) -> str | None:
|
||
if not llm.available():
|
||
return None
|
||
gaps = ", ".join(f"{t}" for t in (m.get("gap_topics") or [])[:6]) or "(无明显缺口主题)"
|
||
cands = "; ".join(
|
||
f"{x.get('login')}(契合{x.get('score')}, 重叠{x.get('topic_overlap')})"
|
||
for x in (m.get("candidates") or [])[:3]) or "(暂无候选)"
|
||
inno = "; ".join((x.get("description") or "")[:50] for x in innovations[:3]) or "(暂无)"
|
||
cat_line = ""
|
||
if context and context.get("category"):
|
||
cat_line = f"\n所属领域热点分类：{context['category']}；领域热门主题：{', '.join(context.get('topic_heat', [])[:5])}"
|
||
prompt = (
|
||
f"仓库：{owner}/{repo}{cat_line}\n"
|
||
f"缺口主题（按权重）：{gaps}\n"
|
||
f"可合作学者：{cands}\n"
|
||
f"近期创新点：{inno}\n\n"
|
||
f"请基于以上给出 3-5 条研究方向/合作建议。"
|
||
)
|
||
return llm.chat(prompt, system=SYS_PROMPT)
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 领域级创新启发
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def analyze_category(category: str, repos_limit: int = 12, top_repos_for_gap: int = 3) -> dict:
|
||
"""领域级：聚合分类下 top 仓的缺口 + 领域主题热度 + LLM idea。"""
|
||
raw = H.collect(category=category, limit=repos_limit)
|
||
hot = H.compute(raw)
|
||
topic_heat = [t["topic"] for t in (hot.get("topic_heat") or [])[:8]]
|
||
scholars = [s["login"] for s in (hot.get("core_scholars") or [])[:5]]
|
||
|
||
# 聚合 top-N 仓的缺口（轻量：每仓 build_gap_signals，累加 gap_topics）
|
||
agg: Counter = Counter()
|
||
gap_signals_agg: list[dict] = []
|
||
repos = hot.get("trending_repos") or []
|
||
for r in repos[:top_repos_for_gap]:
|
||
full = r.get("repo") or ""
|
||
if "/" not in full:
|
||
continue
|
||
o, rr = full.split("/", 1)
|
||
try:
|
||
info = c.repo_info(o, rr) or {}
|
||
gt, _langs, sigs = M.build_gap_signals(o, rr, info, issue_sample=40)
|
||
agg.update({k: v for k, v in gt.items()})
|
||
for s in sigs[:3]:
|
||
s2 = dict(s); s2["repo"] = full; gap_signals_agg.append(s2)
|
||
except Exception as e:
|
||
sys.stderr.write(f"[inspire] {full} 缺口采集失败: {e!r}\n")
|
||
|
||
idea = _llm_idea_category(category, topic_heat, agg, scholars)
|
||
|
||
return {
|
||
"scenario": "inspire",
|
||
"mode": "category",
|
||
"category": category,
|
||
"topic_heat": topic_heat,
|
||
"gap_topics": [t for t, _ in agg.most_common(10)],
|
||
"gap_signals": gap_signals_agg[:12],
|
||
"core_scholars": scholars,
|
||
"idea": idea,
|
||
"llm_used": idea is not None,
|
||
}
|
||
|
||
|
||
def _llm_idea_category(category: str, topic_heat: list[str], gaps: Counter,
|
||
scholars: list[str]) -> str | None:
|
||
if not llm.available():
|
||
return None
|
||
gap_str = ", ".join(f"{t}({n})" for t, n in gaps.most_common(8)) or "(无明显缺口)"
|
||
prompt = (
|
||
f"GitLink 领域：{category}\n"
|
||
f"热门主题：{', '.join(topic_heat[:6])}\n"
|
||
f"缺口主题（主题词频）：{gap_str}\n"
|
||
f"核心活跃学者：{', '.join(scholars)}\n\n"
|
||
f"请基于该领域的热点与缺口，给出 3-5 条具切入价值的研究方向建议。"
|
||
)
|
||
return llm.chat(prompt, system=SYS_PROMPT)
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 报告
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def render_report(result: dict) -> str:
|
||
lines = ["# 💡 创新启发报告", ""]
|
||
if result.get("mode") == "category":
|
||
lines.append(f"> 领域：**{result.get('category')}**（领域级缺口 + LLM 建议）")
|
||
else:
|
||
lines.append(f"> 焦点仓库：`{result.get('repo')}`")
|
||
lines.append(f"> LLM 建议：{'✅ 已生成' if result.get('llm_used') else '⏭ 未启用（无 DEEPSEEK_API_KEY）'}")
|
||
lines.append("")
|
||
|
||
gaps = result.get("gap_topics") or []
|
||
if gaps:
|
||
lines.append("## 🎯 缺口主题")
|
||
lines.append(", ".join(f"`{g}`" for g in gaps[:10]))
|
||
lines.append("")
|
||
|
||
sigs = result.get("gap_signals") or []
|
||
if sigs:
|
||
lines.append("## 🚩 缺口信号（未解决 Issue）")
|
||
for s in sigs[:6]:
|
||
lines.append(f"- [{s.get('priority') or '—'}] {s.get('topic')}：{(s.get('evidence') or '')[:70]} · `{s.get('repo','')}`")
|
||
lines.append("")
|
||
|
||
cands = result.get("candidates") or []
|
||
if cands:
|
||
lines.append("## 🤝 可合作学者 Top 5")
|
||
lines.append("| 学者 | 契合度 | 主题重叠 | 语言匹配 | 活跃 | 理由 |")
|
||
lines.append("|---|---|---|---|---|---|")
|
||
for x in cands[:5]:
|
||
lines.append(f"| `{x.get('login')}` | {x.get('score')} | {x.get('topic_overlap')} | "
|
||
f"{x.get('language_match')} | {x.get('activity_level')} | {';'.join(x.get('reasons', [])[:2])} |")
|
||
lines.append("")
|
||
|
||
inno = result.get("innovation_points") or []
|
||
if inno:
|
||
lines.append("## 🌟 近期创新点")
|
||
for i in inno[:5]:
|
||
lines.append(f"- {i.get('description')} _({i.get('category','')})_")
|
||
lines.append("")
|
||
|
||
idea = result.get("idea")
|
||
if idea:
|
||
lines.append("## 🧠 LLM 研究方向建议")
|
||
lines.append("")
|
||
lines.append(idea.strip())
|
||
lines.append("")
|
||
|
||
lines.append("---\n*由 gitlink-research-inspire 生成*")
|
||
return "\n".join(lines)
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# CLI
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def main() -> None:
|
||
ap = argparse.ArgumentParser(description="创新启发 — 缺口/匹配/创新点 + LLM idea")
|
||
ap.add_argument("--owner", default="")
|
||
ap.add_argument("--repo", default="")
|
||
ap.add_argument("--category", "-c", default="", help="领域分类（领域级模式）")
|
||
ap.add_argument("--top", type=int, default=8)
|
||
ap.add_argument("--pool", type=int, default=12)
|
||
ap.add_argument("--out", "-o", default="")
|
||
args = ap.parse_args()
|
||
|
||
if args.category:
|
||
result = analyze_category(args.category)
|
||
elif args.owner and args.repo:
|
||
result = analyze_repo(args.owner, args.repo, top=args.top, pool_cap=args.pool)
|
||
else:
|
||
print(json.dumps({"ok": False, "error": "need --owner/--repo OR --category"}, ensure_ascii=False))
|
||
sys.exit(1)
|
||
|
||
text = json.dumps(result, ensure_ascii=False, indent=2)
|
||
if args.out:
|
||
os.makedirs(args.out, exist_ok=True)
|
||
with open(os.path.join(args.out, "inspire.json"), "w", encoding="utf-8") as f:
|
||
f.write(text)
|
||
with open(os.path.join(args.out, "report.md"), "w", encoding="utf-8") as f:
|
||
f.write(render_report(result))
|
||
sys.stderr.write(f"[inspire] ✓ mode={result['mode']} llm={result['llm_used']} 产物落 {args.out}\n")
|
||
else:
|
||
print(text)
|
||
|
||
|
||
if __name__ == "__main__":
|
||
main()
|