Hey everyone,
I wanted to share my first Atlassian Forge app — Veronica, an AI-powered Jira assistant that I built as part of my internship preparation. It's a Jira Issue Panel app powered by NVIDIA's Nemotron model via the NIM API.
What Veronica can do:
AI Issue Summarizer — Generates a structured summary of any Jira issue, including progress, risks/blockers, and recommended next actions by analyzing the issue description, status, and recent comments.
Smart Comment Assistant — Click "Suggest Reply" on any comment and Veronica generates 3 reply suggestions in different tones: collaborative, clarifying, and action-oriented. Saves time crafting responses during standups and async reviews.
Auto-Escalation Bot — A scheduled trigger runs every hour and scans for issues stuck "In Progress" for more than 3 days without updates. When it finds one, it automatically posts an escalation comment and records it in Forge KV Storage to prevent duplicates. JSM ticket creation is supported as an optional add-on.
Sprint Health Dashboard — Analyzes all issues in the current sprint and produces a health assessment (Healthy / At Risk / Critical), identifies at-risk issues, lists blockers, and gives actionable focus recommendations.
Tech stack:
Atlassian Forge (UI Kit with
@Forge/react)
NVIDIA NIM API (Nemotron 3.5 Lightning model)
Forge Key-Value Storage (@forge/kvs)
Forge Scheduled Triggers
Jira REST API v3 + Agile API
What I learned:
Building on Forge's runtime constraints (no arbitrary npm, no DOM access, UI Kit components only)
Working with Atlassian Document Format (ADF) for comments
Scheduled triggers and background automation without user context
Duplicate prevention patterns with KV storage
Integrating external AI APIs within Forge's egress permission model
Update: Veronica is now fully built and tested. I'll be sharing a walkthrough of how it works with real test cases in the next 3 days — stay tuned!
If anyone has feedback or suggestions for improvement, I'd love to hear it.
Cheers, Shivain
