rag-paper-query

rag-paper-query

Use this skill whenever the user asks about papers that may exist in the local rag-paper knowledge base.

Priority

Prefer MCP tools when available. Use direct local commands or Python access only when MCP is unavailable, incomplete, or the user explicitly asks for direct Chroma/local access.

MCP Workflow

If MCP server rag-paper is available, use these tools:

1. list_indexed_papers

   • Use first when the user asks what papers are indexed.
   • Use to confirm whether a paper exists before detailed lookup.

2. show_indexed_paper

   • Use for metadata, DOI, authors, source path, chunk IDs, or indexed-paper details.
   • Use fuzzy selectors: title fragment, filename fragment, source path fragment, or DOI fragment.

3. search_papers

   • Use for semantic or keyword-like questions over paper content.
   • Prefer concise, content-rich search queries.
   • Use filters when the user mentions author, year, tag, or file name.

4. search_by_metadata

   • Use when the user asks for papers by author, year, tag, or filename.

5. get_chunk

   • Use when exact chunk text is needed.

6. export_context

   • Use when synthesizing an answer from multiple chunks.
   • Export only relevant chunks.

Direct Local Workflow

If MCP is unavailable, use local rag-paper commands from the project environment:

rag-paper list-indexed-papers
rag-paper show-indexed-paper "selector"
rag-paper search "query" --top-k 8
rag-paper search "query" --author "name"
rag-paper search "query" --year 2024
rag-paper search "query" --tag transformer

If commands are unavailable but Python package access works, use:

from rag_paper.config import load_config
from rag_paper.retrieval import HybridRetriever, result_to_dict

config = load_config("./config.json")
retriever = HybridRetriever(config)
results = retriever.search("your query", top_k=8)
items = [result_to_dict(result) for result in results]

For direct Chroma inspection without semantic search: