Contact person: Business Client

Phone:Show

Email:Show

Location: Campohermoso, Spain

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"I need a Forge cloud app that embeds WebLLM so every language-model call executes inside Atlassian’s runtime, with no external API dependency. The app must automate three key actions:

• Create new issues when triggered by a slash command or webhook.
• Post contextual comments (e.g., summaries, next-step checklists) on demand or at a chosen workflow transition.
• Review an issue’s title, description, and fields, then inject a comment that suggests concrete improvements in line with industry best practices for agile ticket writing.

Because the feedback mode is “Suggestions for improvement,” I only expect concise, actionable tips rather than full audits or numeric scoring. The LLM should reference patterns such as INVEST, clear acceptance criteria, reproducible steps, and appropriate labeling—no company-specific policies are required.

Core expectations
– Written in Forge (custom-UI or UI kit; I’m open, but keep the bundle lean).
– WebLLM model assets loaded locally and called through a small wrapper so the Forge function stays within the memory/time limits.
– Triggers: manual command, workflow transition, and schedule (for nightly backlog grooming).
– Permissions limited to read/write on issues and comments only.
– Config page where I can toggle each feature and set default prompts.

Deliverables
1. Complete, commented source code in a Git repo.
2. Forge manifest and deployment script.
3. README covering setup, model download, and permission scopes.
4. Short loom/demo video proving: issue creation, comment generation, and quality-feedback running entirely inside Forge.

Acceptance criteria
A. Forge deploy succeeds in my test site with no external keys.
B. Model returns suggestions under 2 s for a standard user story.
C. Creating, commenting, and feedback actions each operate from their respective triggers without errors.

If you already work with Forge, TypeScript, and WebLLM, this should be straightforward. I’m happy to clarify any edge cases once you review the scope." (client-provided description)