K8s-Chat

K8s-Chat is a project or chatbot with RAG function, which can answer questions related to kubernetes (k8s). Of course, you can also replace k8s with other fields of knowledge.

Chat Flow

The chat flow is shown in the following firgue:

First, a Query enters Chat history. Then the Chat history is processed through the LLM to get the Intent. The docs or chunks with high similarity to the intention is retrieved through the Retriever. Use the search results to update the System prompt. Finally, send the System prompt and Chat history to the LLM and get the relevant response, which will be part of the Chat history. In general, the process repeats itself.

Quick Start

Prepare

Through the operation, you can process your own documents(knowledge).
Generate summary vector index:

python summary_vector_index/doc2summary.py
python summary_vector_index/summary2vector.py
python summary_vector_index/summary2vector_gpt.py #optional gpt vector

Generate vector index:

python vector_index/vectorindex_save.py
python vector_index/vectorindex_save_gpt.py #optional gpt vector

Usage

CLI interface

Run the CLI interface demo:

python demo_cli.py

Show CLI interface:

Web UI interface

Run the Web UI interface demo:

python demo_webui.py

Access Web UI, by local ip:port : http://127.0.0.1:7860
you can also publish service by set gtadio launch(share=True) and generate share link (expires in 72 hours).
Show Web UI interface :

API interface

Use API interface demo:

python demo_api.py

then you can make request:

import requests

url = "http://127.0.0.1:7777/chat/" 
query = {"text": "你好，请做一段自我介绍。"}
response = requests.post(url, json=query)

if response.status_code == 200:
    result = response.json()
    print("BOT:", result["result"])
else:
    print("Error:", response.status_code, response.text)

Roadmap

New features are coming soon 🚀:

• Local model 🔥: Chat with local LLMs.
• Fine-tuning model 🎨: Fine-tuning model based on k8s.