kernal-gpt

An AI agent based on the Ollama large model, capable of executing Linux commands through natural language and invoking kernel hooks to delve into the underlying system.

Abilities & Possibilities

[✔] Support all large models locally in Ollama

[✔] langchain and langgraph integration

[✔] multi-modal LLM's support

[✔] only tool calling support

[✔] redis-vector and RAG integration

[✔] eBPF integration (Only supports Linux kernel 5.10 and above.)

[✔] dangerous command detection

[×] support memory and context （Coming soon）

[×] It will not run directly at the moment.（Currently, it is an experimental version and not suitable for direct execution.）

Mechanism

How to run

1. This project uses Redis vector as the vector database, so Redis must be started first.

docker run -p 6379:6379 docker.io/redislabs/redisearch:latest 

2. Set environment variables

export KPT_MODEL="" # "please select the llm
export KPT_OLLAMA_URL="http://127.0.0.1:11434"
export KPT_REDIS_URL="redis://127.0.0.1:6379"

3. Compile the code

make

# ./kernal_gpt
  _                                    _                           _
 | | __   ___   _ __   _ __     __ _  | |           __ _   _ __   | |_
 | |/ /  / _ \ | '__| | '_ \   / _` | | |  _____   / _` | | '_ \  | __|
 |   <  |  __/ | |    | | | | | (_| | | | |_____| | (_| | | |_) | | |_
 |_|\_\  \___| |_|    |_| |_|  \__,_| |_|          \__, | | .__/   \__|