When should I use BM25 search instead of vector search?

Use BM25 search for exact keyword matches, such as function names, error messages, or specific technical identifiers, where semantic similarity might return too many irrelevant results.

Can I search for exact phrases with this skill?

Yes, by enclosing your query in double quotes, such as '"connection refused"', the skill will perform an exact phrase match within the agent's memory.

Does this skill support automated index maintenance?

Yes, it includes index lifecycle configuration to handle pruning, retention, and optimization based on defined schedules and age-based levels.

How do I rebuild the search index if results are missing?

You can force a full rebuild of the BM25 index from the underlying database using the 'teleport rebuild --force' command.

BM25 Keyword Search for Agent Memory

Name: BM25 Keyword Search for Agent Memory
Author: SpillwaveSolutions

bySpillwaveSolutions

•

開発者ツール

Performs high-speed, full-text keyword searches within agent memory systems to locate exact technical terms and identifiers.

The BM25 Keyword Search skill provides a specialized retrieval mechanism for the agent-memory system, utilizing the Tantivy index for fast, precise full-text search. While vector search excels at conceptual similarity, this skill is essential for finding exact function names, specific error messages, and technical identifiers that require literal matching. It includes robust management features for index statistics, lifecycle pruning, and manual rebuilds, ensuring that an AI agent can always navigate its historical context with surgical precision through targeted keyword discovery.

主な機能

01Automated index lifecycle management and pruning configurations

024 GitHub stars

03Advanced query syntax supporting phrases, prefixes, and multiple terms

04Integrated diagnostic tools for index statistics and health monitoring

05Precise matching for technical identifiers, function names, and error codes

06Fast BM25 full-text search via Tantivy indexing

ユースケース

01Filtering agent memory when semantic/vector search returns too many conceptual matches

02Locating a specific function or variable name across large codebase histories

03Searching for exact error message strings to find previous occurrences and solutions

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add spillwavesolutions/agent-memory bm25-search

For use in Claude.ai and ChatGPT

Download Skill