Increase mitigation eval success rate by providing localization info in the init prompts

Open Beginner friendly
#109 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
2/5
Estimated time
1-3 hours
Newbie friendliness
72/100
Issue type
Feature
Clarity
Mostly clear
Activity status
Quiet
Tech stack
python
Domain
ai

Research direction

Start at aiopslab/orchestrator/tasks/mitigation.py#L27 and inspect how mitigation task initialization builds the agent's prompts. Trace where localization information is available and determine how it should be included at the beginning of those prompts. Done means mitigation evaluations receive the localization context and failed traces no longer require the agent to locate the faulty service first.

Written by the indexing model from the issue text.

Description

https://github.com/microsoft/AIOpsLab/blob/957c550028fef82db82ac791ad345c25fe6f6596/aiopslab/orchestrator/tasks/mitigation.py#L27

Currently, the mitigation tasks have the lowest success rate(18%) when evaluated with GPT-4o-mini, ReAct.

Failed traces revealed that sometimes the agent could not even locate the faulty service correctly.

To isolate and evaluate mitigation itself, the agent should be given the localization info at the beginning.

Even if the agent manages to locate the faulty service, the localization process(reading long logs etc) takes up the context window and degrades agent performance.

Dominant language
Python
Stars
989
Forks
176
Avg merge
7d 15h
Merged PRs (30d)
3

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from microsoft/AIOpsLab

All issues in microsoft/AIOpsLab

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.