Couchbase Agent Catalog Project

Welcome to the Couchbase Agent Catalog (agentc) project! The purpose of the agentc project is to make building reliable agents more accessible.

Table of Contents

• Getting Agent Catalog
  • Building From Package
  • Building From Source (with pip)
  • Building From Source (with Poetry)
  • Verifying Your Installation
• Agent Catalog User Guide
  • Metrics Driven Development
  • Couchbase-Backed Agent Catalogs
  • Assessing Agent Quality
• Command Line Tool
  • agentc
• Agent Catalog API
  • Auditor
  • Provider
  • audit()
• Environment Variables
  • Mandatory Environment Variables
  • Optional Environment Variables
• On Catalog Entries
  • Tool Catalog Records
  • Prompt Catalog Records
• Frequently Asked Questions
  • Which Couchbase Capella or Server version should I use?
  • Do I need to use a specific agent framework (e.g., LangChain) with Agent Catalog?
  • What does Agent Catalog add to my Couchbase bucket?
  • What is Agent Catalog doing when I run agentc index?
  • How do I roll back to a previous catalog version?
  • What environment variables are required to use Agent Catalog?
  • What are the different types of tools and prompts I can create?
  • Can I write multiple tools/prompts in one file?
  • Do CLI commands need to be executed in a certain order?
  • Can I index and publish catalogs programmatically?
  • Does Agent Catalog require an OpenAI API key?
  • Does Agent Catalog work with any LLM?

Indices and Tables

• Index