wiki-retrieve
wiki-retrieve upgrades vault search from page-level to chunk-level granularity, combining contextual prefixes, BM25 sparse retrieval, and dense reranking to reduce retrieval failures. Setup is opt-in with built-in privacy gates: run fully on-machine with synthetic prefixes, or opt into Anthropic API or claude CLI tiers for better context generation.
wiki-retrieve enables hybrid chunk-level search across Obsidian vaults with contextual prefixes and reranking.
AI-generated summary based on this skill's SKILL.md
Install
AgriciDaniel/claude-obsidian/wiki-retrieve · repository language: Python
git clone https://github.com/AgriciDaniel/claude-obsidian
cp -r claude-obsidian/skills/wiki-retrieve ~/.claude/skills/wiki-retrieveFrequently asked questions
AI-generated answers based on this skill's SKILL.md and metadata
How do I search chunks in my vault with wiki-retrieve?
wiki-retrieve enables chunk-level search of your vault by breaking pages into smaller passages and scoring them with both BM25 (sparse) and cosine similarity (dense) retrieval. You configure it to search these chunks using hybrid scoring, then optionally apply reranking to surface the most relevant passages. All processing can run on-machine if you use synthetic contextual prefixes, keeping your vault data private.
What is wiki-retrieve and how does it improve search?
wiki-retrieve upgrades vault search from page-level to chunk-level granularity by combining contextual prefixes, BM25 sparse retrieval, and dense reranking to reduce retrieval failures. It works with Obsidian vaults and other wiki systems. Setup is opt-in with built-in privacy gates: run fully on-machine with synthetic prefixes, or opt into Anthropic API or claude CLI tiers for better context generation.
How do I set up hybrid retrieval for my wiki?
wiki-retrieve's hybrid setup combines BM25 and cosine similarity scoring. Configure tier-1/tier-2/tier-3 contextual prefix generation to add semantic context to chunks before retrieval. You can run this entirely locally with synthetic prefixes, or enable Anthropic API integration for improved prefix quality. The setup process is opt-in, letting you choose your privacy and performance tradeoff.
Can wiki-retrieve run entirely on my machine?
Yes. wiki-retrieve is designed with privacy gates that let you run fully on-machine using synthetic contextual prefixes and local embedding models like nomic-embed-text. You retain complete control over your vault data. Alternatively, you can opt into Anthropic API or claude CLI tiers for better context generation if you prefer enhanced retrieval quality over full local operation.
What license does wiki-retrieve use?
wiki-retrieve is released under the MIT license, making it free to use, modify, and distribute for both personal and commercial projects.
How does contextual prefix generation improve retrieval?
wiki-retrieve generates tier-1/tier-2/tier-3 contextual prefixes that add semantic context to each chunk before retrieval. These prefixes help the BM25 and cosine similarity scorers better understand passage meaning, reducing retrieval failures. You can configure prefix generation to run synthetically on-machine, or use Anthropic API or claude CLI for richer, model-generated context.
SKILL.md
rendered from the published skill — quoted content, verbatim
wiki-retrieve: Hybrid Retrieval over the Vault
The v1.6 query path was Read(hot.md) → Read(index.md) → Read(3-5 pages) → synthesize. It worked, but page-level granularity loses to chunk-level granularity any time the answer lives in a specific passage rather than a whole page. The v1.7 wiki-retrieve skill is the chunk-level upgrade — opt-in, feature-gated, and replaces nothing if you don't run the setup.
Origin: This skill is original to claude-obsidian. There is no upstream kepano equivalent. The technique is from [Anthropic's Sept 2024 Contextual Retrieval
(truncated - see the full file via the links below)
Read as markdown · JSON record · Browse the source repository
File tree — 1 file
skills/wiki-retrieve/SKILL.md