Description

This tool automatically extracts errors, warnings from a provided support bundle, compares them against existing Knowledge Base (KB) articles using semantic search, and generates a concise summary if the issue is known along with probable causes & recommended steps.

Goals

  1. Automated error extraction
  2. Semantic matching with Kbase articles (keeping it scoped to longhorn articles as of now)
  3. LLM based Summary

Resources

At present, following are the modules used:

  • LlamaIndex – to build the RAG pipeline that handles document ingestion, indexing, and querying.
  • HuggingFaceEmbedding model – to convert logs and queries into vector embeddings for semantic search.
  • Ollama (Llama 3.1 model) – to run a local LLM for reasoning over retrieved results and generating final answers.
  • ChromaDB – to store and query embeddings efficiently for fast semantic retrieval.
  • Custom query engine modules – to orchestrate retrieval logic, embed queries, and format responses.
  • Local runtime environment (Python + venv) – to integrate all components and execute the RAG workflow end-to-end.

Looking for hackers with the skills:

Nothing? Add some keywords!

This project is part of:

Hack Week 25

Activity

  • 23 days ago: ajagtap originated this project.

  • Comments

    • ajagtap
      21 days ago by ajagtap | Reply

      The code in progress exists here: https://github.com/apoorvajagtap/logbundle-analyzer

    Similar Projects

    This project is one of its kind!