Python implementations of some of the fundamental LLM based Agent method from scratch, like CoT, ReAct, etc.
The purpose of this project is to understand the core methods of agent development using Python, without relying on frameworks like Langchain.
Currently, we are only these four methods:
| Method | Link |
|---|---|
| CoT | code |
| Self-Refine | code |
| Plan and Execute | code |
| tools | code1 ,code2, code3 |
| P&E with tool | code1, code2 |
| React | code |
Currently, we're only using DeepSeek as the LLM, so you first needs your own API key. If you have an OpenAI key, you can also follow llm_based_agent_from_scratch