AI Engineering: Advent of AI with goose Day 6

Source: Dev.to

Day 6: Festival Feedback, Automated – AI, GitHub Actions, & goose

The Problem: Feedback Overload

Hours after a festival launches, feedback floods in: bug reports, urgent facility requests, feature ideas, and a swarm of questions. The result? Chaos in GitHub Issues. This year I set out to solve it once and for all with AI, automation, and a bit of Goose magic.

Introducing: My Festival Feedback System

Turn GitHub Issues into a self‑organizing, AI‑triaged feedback powerhouse.

Tech Stack

• goose CLI (by Block) – conversational AI sidekick
• GitHub Actions – seamless issue and comment workflows
• OpenAI API & Anthropic API – LLM brains (GPT‑4o‑mini, Claude)
• Python – custom triage, sentiment analysis, summaries
• YAML – glue workflows and schedule automations
• PyGithub – Python ↔ GitHub API integration

What My System Does

• Issue Categorization – AI instantly classifies issues as bug, feature, question, or urgent.
• Priority Detection – From critical to low‑impact.
• Sentiment Analysis – Tracks positive, neutral, or negative vibes.
• Automated Responses – Human‑sounding replies directly on issues.
• Comment Handling – AI follows up on new comments to keep the conversation flowing.
• Custom Issue Templates – Streamlined forms for bugs, features, and questions.
• Daily Summaries – Each morning AI posts a status update with insights, trends, and urgent flags.

How I Built It

GitHub Secrets 🔑

Store OpenAI and Anthropic API keys securely in repository secrets. The YAML workflows can then access them safely.

Workflows as Automation Engines ⚙️

Two core YAML workflows power the system:

• issue-triage.yml – handles new issues and comments.
• daily-summary.yml – generates the morning stats.

Python as the AI Brain 🧠

Python scripts, invoked by the workflows, perform analysis, categorization, summarization, and reply generation using LLM‑powered prompts.

Testing Like a Festival Pro 🎪

Created template‑based issues for urgent facility problems, feature requests, and lost‑and‑found questions to verify that the system:

• Assigns correct labels and priority.
• Replies in a friendly, actionable way.
• Summarizes daily feedback surges.

Result: It nailed all of it! 🚀

Why This Is Game‑Changing

• Automates triage so no critical issue goes unseen.
• Transforms feedback into actionable data (sentiment, trends, priorities).
• 24/7 AI responder eliminates “sorry for the late reply” delays.
• Adaptable to any project – open source, events, support, product feedback.
• Easy to customize – edit Python prompts, labels, or templates.

Real‑World Example

Issue posted:

[URGENT] Heating system not working in storytelling tent

• Labeled as 🚨 urgent, 🟠 high priority, and 😟 sentiment: negative.
• Instant response with stats and next steps for staff.
• Appeared in the next day’s daily summary under “urgent issues”.

What I Learned

• Goose CLI is fantastic for rapid prompt and script prototyping; let AI help design workflow logic.
• GitHub Actions + AI = superpowers for project management.
• Good prompt design yields helpful (not spammy) automated replies.
• Daily summaries keep the team focused on trends, not just individual tasks.
• Secret and config management in GitHub is easier than ever.

Powered By

• Goose by Block
• OpenAI & Anthropic LLM APIs
• GitHub Actions + PyGithub

Final Thoughts

Using Goose CLI with GitHub Actions created a fast, AI‑assisted workflow where prompts, scripts, and automated summaries evolve in real time. Strong prompt design, combined with streamlined secrets management, made the pipeline cleaner, more reliable, and far easier to maintain. The automation turned issue “chaos” into calm, empowering the festival team. I look forward to adapting this setup for future development and sharing it with the world! 😊

Day 6: Automated. Feedback: Tamed. Community: Happy!