Tepa Documentation

Tepa is a TypeScript framework for building autonomous AI agent pipelines that plan, execute, and self-correct. These docs cover everything from first install to full API reference.

Learn

#	Section	What you'll find
1	Introduction	What Tepa is, the problems it solves, key differentiators, and who it's for
2	Getting Started	Prerequisites, installation, a minimal working example, and understanding the result
3	How Tepa Works	The Plan-Execute-Evaluate cycle, self-correction, scratchpad, event system overview, and package architecture

Build

#	Section	What you'll find
4	The Pipeline in Detail	Deep dive into prompt structure, Planner, Executor, Evaluator, lifecycle events, cycle termination, and tool resolution
5	Configuration	TepaConfig structure, defaults, defineConfig(), model assignment, limits, and logging levels
6	Tool System	Defining tools, parameter schemas, the tool registry, all built-in tools, and creating third-party tool packages
7	Event System Patterns	Callback contract, execution order, error handling, and patterns for human-in-the-loop, safety filters, progress tracking, and more
8	LLM Providers	Provider interface, built-in providers (Anthropic, OpenAI, Gemini), native tool use, logging system, and building a custom provider

Explore

#	Section	What you'll find
9	Examples and Demos	Three runnable demos — autonomous code generation, data analysis pipeline, and human-in-the-loop study planner

Reference

#	Section	What you'll find
10	Contributing	Development setup, code conventions, how to add tools and providers, PR guidelines
11	API Reference	Complete reference for every exported class, interface, type, function, and error across all packages
12	Future Iterations	Vision and ideas for possible future directions — CLI, parallel execution, streaming, multi-agent, and more