papi-ask

分类: 调研与分析 | 上传者: hummat | 下载: 0 | 版本: v1.0（最新）

使用 RAG（PaperQA2 或 LEANN）查询论文。当用户需要论文中的综合答案、询问“论文 X 关于 Y 说了什么”或需要引用的答案时使用。

更新日志: Source: GitHub https://github.com/hummat/paperpipe

目录结构

当前层级: tree/main/skills/papi-ask/

📄 SKILL.md 1.4 KB

SKILL.md

---
name: papi-ask
description: Query papers using RAG (PaperQA2 or LEANN). Use when user needs synthesized answers from papers, asks "what does paper X say about Y", or needs cited responses.
---

# Query Papers via RAG

Use `papi ask` for questions requiring synthesis across papers or cited answers.

## Cost-Aware Retrieval

Before using RAG, consider cheaper alternatives:

1. **Exact match**: `papi search --rg "query"` — fast, no LLM
2. **Ranked search**: `papi search "query"` — BM25 ranking
3. **Direct read**: `papi show <paper> -l eq|tex|summary` — if you know the paper

Use `papi ask` only when:
- User explicitly requests RAG/synthesis
- Question spans multiple papers
- Search/show cannot answer

## Commands

```bash
# PaperQA2 (default) — full RAG with citations
papi ask "question"

# LEANN — faster semantic search + LLM
papi ask "question" --backend leann

# Structured output for programmatic use
papi ask "question" --format evidence-blocks
```

## MCP Tools (if available)

For quick retrieval without full RAG:
- `leann_search(index_name, query, top_k)` — fast semantic search
- `retrieve_chunks(query, index_name, k)` — PaperQA2 chunks with citations

Check available indexes: `leann_list()` or `list_pqa_indexes()`

## Output

RAG answers include:
- Synthesized response
- Citations with page/section references
- Confidence indicators

For general CLI commands, see `/papi`.

登录后下载/点赞/收藏 ❤ 9 | ★ 0

papi-ask

目录结构

SKILL.md

举报内容

提示