insightsGPT
Application for scanning RHEL systems (also insights-report/sosreports), diagnosing and triaging issues in simple english. With the help of Red Hat’s knowledge base(KCS) + the power of LLM(Large language model) pulls out the relevant information and provides resolution.
Service can integrate with pre existing tools like insights-client, sos(sosreport) etc.
Architecture Diagram
Installation
We use poetry for dependency management so make sure that poetry binary is installed - doc.
curl -sSL https://install.python-poetry.org | python3 -Install the needed dependencies.
poetry installTo start cromadb(where all the KCS will be embedded and stored) use the below command.
podman-compose up
Start llm service using below.
make run-llm
Start insightsGPT-backend service using below.
make run-backend
Usage
There is load_data.py script which loads the KCS article from the docs/kcs directory and store them in the cromaDB.
poetry run python load_data.py
Use below make target to trigger test API.
make api-test
llm service has a /completions endpoint where one can pass the context and ask the related question to llm model.
curl -X POST -H "Content-Type: application/json" http://127.0.0.1:8080/v1/chat/completions -d '{"context": "Chandrayaan-3 launched at 14 July 2023 and landed to south pole of moon at 23 August 2023.", "question":"When did Chandrayaan-3 landed?"}' | jq