Retrieve wiki chunks using contextual-prefix, BM25, and cosine rerank.
npx skills add https://github.com/deox420/second-brain --skill wiki-retrieve-deox420
Please help me install this Agent Skill. Skill: wiki-retrieve Source: https://github.com/deox420/second-brain/tree/main/skills/wiki-retrieve Command: npx skills add https://github.com/deox420/second-brain --skill wiki-retrieve-deox420
Hybrid retrieval over the vault replaces coarse page-level reads with chunk-level analysis to deliver more precise answers.
Run bash bin/setup-retrieve.sh to provision the retrieval setup, then test a query with wiki-query to verify retrieval.
High-intent search queries and answers about installing and using this skill.
Chunk-level vault retrieval replaces coarse page-level reads by applying contextual-prefix generation, BM25 indexing, and dense reranking to surface precise passages with citations.
Hybrid retrieval uses BM25 indexing for fast keyword-based passage retrieval, then applies a dense cosine rerank using a local or API-backed LLM to surface the most contextually relevant top passages.
No, dense reranking can use either a local LLM or an API-backed rerank service. The required setup is configured during the bin/setup-retrieve.sh provisioning step.
Integrating chunk-level hybrid retrieval into wiki-query and autoresearch workflows improves search quality by replacing page-level reads with contextual-prefix generation and dense passage reranking.
Page-level retrieval returns irrelevant context because it lacks chunking and indexing granularity; applying chunk-level BM25 indexing and dense reranking resolves this by isolating precise passages.
Yes, wiki-retrieve handles candidate curation by applying chunking, indexing, and dense reranking across wiki pages to extract and cite contextually precise passages.
byriderImplement chunk-level contextual prefix, BM25, and cosine-rerank retrieval for Obsidian Compound Vault.
AgriciDanielSelect and rerank chunk-level passages from an Obsidian Compound Vault.
danmestasManage a persistent vault-based knowledge graph with Obsidian-compatible markdown tooling.
vasilyu1983Design and implement RAG retrieval systems with hybrid search and reranking.
