Skip to content

Latest commit

 

History

History
41 lines (21 loc) · 1.15 KB

File metadata and controls

41 lines (21 loc) · 1.15 KB

Roadmap

1. Minimal Agent - Completed

Build a deterministic MockLLM, a small tool interface, a calculator tool, and a one-step agent loop.

2. Tool Agent - Completed

Add multiple tools, tool routing, input validation, and clearer action schemas.

3. ReAct Agent - Completed

Introduce thought-action-observation loops and multi-step reasoning traces.

4. Memory Agent - Completed

Add short-term memory, simple long-term memory, and memory inspection examples.

5. Planner-Executor - Completed

Separate high-level planning from execution and track task progress.

6. Mini MAS

Create a small multi-agent system with simple roles, message passing, and coordination rules.

7. Evaluation Suite

Add deterministic task cases, scoring functions, failure analysis, and result reports.

8. API-based Agent

Introduce a real LLM API behind a clean interface after the from-scratch loop is understood.

9. Lightweight RAG Agent

Add simple document retrieval with local files and lightweight indexing.

10. Agent Framework Comparison

Compare the from-scratch implementation with LangChain, LangGraph, AutoGen, and CrewAI conceptually and experimentally.