Documentation Index
Fetch the complete documentation index at: https://docs.clarifeye.ai/llms.txt
Use this file to discover all available pages before exploring further.
Overview
Tools let you define curated retrieval behaviors that your AI agent can call. A tool encapsulates three things:- Name and description visible to the agent
- A retrieval method (chunks, objects, or a Cypher template)
- A search scope (documents and tags)
1) General
On the General tab, provide:- Name: a short action‑oriented label (e.g., “Search policies by tag”)
- Description: optional guidance for the agent describing when to use the tool and what it returns
2) Retrieval Method
Choose how the tool finds information. The selector offers three types:- Search on chunks: search within chunk text
- Search on objects: search within the textual representation of extracted objects
- Templated search on objects: execute a parameterized Cypher query against the object graph
- Mode: Semantic (default), Full‑text, or Hybrid
- Query reformulation: enable HyDE‑style expansion to improve recall on vague queries
- Optimal reranking: re‑rank candidates for higher precision
- Number of objects: maximum items to return; optionally force the exact count
- Enrich with chunks (objects search): include the source chunk content alongside objects
3) Search Scope
Limit where the tool searches:- Documents: include all documents or select specific ones
- Include/Exclude tags: filter chunks and objects by tags. You can either:
- Include all tags
- Let Clarifeye decide the suitable selection
- Select explicit tags to include/exclude
Usage tips
- Create multiple focused tools instead of a single broad one (e.g., a chunk search tool for policy sections, and an object search tool for a specific schema)
- Turn on reranking for long or multi‑intent queries
- Use query reformulation when users commonly ask underspecified questions
- Prefer a Cypher template when you need precise graph constraints or field‑level matching