Document Analysis & Retrieval
High-precision information extraction from unstructured documents via semantic search, targeted sampling, and sequential traversal.
Prompt Example
Use the 'document-reader' skill to extract information from the following document: [DOCUMENT_PATH]. Question: [QUERY]
Concrete examples
Use the 'document-reader' skill to extract information from the following document: ariadne-engine-specs.md. Question: Which version of the Ariadne Engine is currently recommended for production environments?
Specialized in retrieving precise evidence from indexed document repositories, prioritizing evidentiary depth over heuristic approximation.
The retrieval pipeline executes a multi-stage process: initial content sampling to assess coverage, extraction of keywords and structural markers, and a tiered search strategy covering literal terms, synonyms, and structural context.
When targeted searches yield insufficient results, the agent transitions to sequential browsing via `browse_document_content`, allowing granular traversal of document chunks with fine-grained element control.
Technical safeguards include: initializing unknown coverage to zero, triggering context resets after consecutive search failures, and providing full auditability through citations containing chunk indices and element offsets.