[GSoC 2026 INTRODUCTION] Kammari Ashritha - AI Chatbot to Guide User Workflow

Contributing GSoC
1

Hi @Kris Stern, @Shivay Lamba, and @Chirag Gupta. I am Ashritha, a 2nd-year CSE student and Top 10 Finalist in the TechSprint AI Challenge (LectureLoop AI). I am focusing on the ‘AI Chatbot for Workflow Guidance’ project. I have already explored the resources-ai-chatbot-plugin repository and am setting up the local environment to analyze the current RAG (Retrieval-Augmented Generation) implementation. Looking forward to discussing the transition from a resource-seeker to a workflow-guider.

2

Update: I have successfully set up the development environment for resources-ai-chatbot-plugin on native Windows. I identified a platform-specific dependency issue with the Linux-only NVIDIA wheels in requirements.txt and have submitted PR #300 to resolve this. I’ve also updated the README to clarify the setup for other Windows/Git Bash contributors. Looking forward to feedback!

3

Hi Jenkins community,

I’m Kammari Ashritha, a second-year Computer Science student at KL University, Hyderabad. I’m applying for the GSoC 2026 project -AI Chatbot to Guide User Workflow.

What I’m proposing

The current plugin is reactive - it waits for questions. My proposal makes it proactive. I’m calling this the Anticipatory Context Engine: a system that wires Jenkins’ own build-lifecycle APIs (RunListener, FlowNode, BuildData) into the RAG pipeline so that when a build fails, the diagnosis panel is already populated before the user types a single word.

The architecture closes all four mentor-filed GSoC 2026 issues under one coherent system:

  • #77 - Jenkins Context Bridge (Java-side FlowNode walk → FastAPI backend)

  • #69 - Temporal Delta Analyzer + Grounded Response Generator

  • #71 - GraphRAG Plugin Ecosystem Index

  • #70 - LLM-as-Judge Evaluation Pipeline

What I’ve already done

  • Submitted PR #300 (open, under review) - fixed Windows installation failure caused by Linux-specific NVIDIA wheel environment markers, incorporated reviewer feedback within 24 hours

  • Building a ground-truth failure log dataset (28 annotated cases so far from open jenkinsci CI runs) that will underpin the evaluation pipeline

Proposal draft: https://docs.google.com/document/d/1tJRGgPZnH8JyQrwnTXPCArbKwCSunWh3/edit?usp=sharing&ouid=106129629037158857621&rtpof=true&sd=true

I’d genuinely appreciate early feedback from mentors, especially on the Java-side Context Bridge scope and the GraphRAG storage layer design. Happy to adjust based on what the mentors think is realistic for the timeline.

Thanks,
Ashritha
GitHub: kammari-ashritha (KAMMARI ASHRITHA) · GitHub