Translate natural language queries into semantic searches for claims and sources.
npx skills add https://github.com/lhl/realitycheck --skill search-lhl
Please help me install this Agent Skill. Skill: search Source: https://github.com/lhl/realitycheck/tree/main/integrations/claude/skills/search Command: npx skills add https://github.com/lhl/realitycheck --skill search-lhl
This Skill helps users locate relevant claims and supporting sources quickly by interpreting natural language queries and retrieving related items from a connected knowledge base.
Use the search skill to find all sources related to "AI safety" within TECH domain.
High-intent search queries and answers about installing and using this skill.
Semantic search across claims uses embedding models to convert your natural language query into vectors, retrieving related items from a connected knowledge base. You can find supporting sources by querying concepts rather than exact keywords.
Yes, semantic search results can be filtered by domain using categories like TECH, LABOR, ECON, GOV, SOC, RESOURCE, TRANS, GEO, INST, RISK, and META. You can also filter results by type to isolate either claims or sources.
To retrieve claims and supporting sources, your environment requires an available embedding pipeline and a vector index. The search relies on these components to process natural language queries and match them with related items.
Natural language retrieval works by translating your query into a semantic search across the knowledge base. A researcher can look up sources related to AI safety within the TECH domain and retrieve discussions complete with evidence levels.
Limitations of embedding-based search include dependency on the availability of the embedding pipeline and vector index. It requires a pre-populated knowledge base containing claims and sources to successfully discover related items across domains.
ice-droid-99Creates and queries Databricks Vector Search endpoints and indexes for semantic retrieval.
lhlExtract discrete claims from a source into a structured YAML list with metadata fields.
codysumpter-cloudIngest videos, audio, PDFs, books, screenshots, and GitHub repos into a knowledge base.
ccfCombine vector similarity with keyword matching for hybrid retrieval.
