How to use Gemini to turn stories into DevOps tasks
Power Automate orchestrates Gemini generation, codebase retrieval, and Azure DevOps API calls to create story documentation and work items using rigid JSON outputs.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Turn [user story] into requirements, a technical specification, a test strategy, and actionable tasks using [retrieved codebase context]. Return output matching [JSON schema] for creating [Azure DevOps wiki pages and work items]. Identify missing context explicitly.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source poster / publisher
- antonybrahin
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports saving hours on every user story and producing consistent, high-quality documentation and tasks through codebase-aware generation.
- Human involvement
- The author configured the workflow and output schema; the post does not describe review checkpoints for generated documents or tasks.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- An Azure OpenAI model creates vector embeddings of the codebase.
- Azure AI Search retrieves actual code context for the generation workflow.
- Gemini API calls generate requirements, technical specifications, and test strategy content.
The author built a custom workflow to turn user stories into documentation and actionable tasks. Power Automate orchestrates a sequence of API calls rather than relying on a packaged assistant. The stated goal was flexibility and control across the tools used for each part of the process.
Five calls to the Gemini API handle generation steps such as requirements, a technical specification, and a test strategy. An Azure OpenAI model creates codebase vector embeddings, and Azure AI Search supplies retrieval-augmented context. The author identifies access to actual code as the key to producing technical specifications and tasks that are specific to their system rather than generic.
The workflow then makes direct Azure DevOps REST API calls, authenticated with a PAT, to create wiki pages and work items. The author chose direct calls because standard connectors were limited. Requiring rigid JSON output instead of parsing free-form text was another reported reliability improvement. They say the resulting system saves hours per story and produces consistent documentation and tasks, without providing an independent measurement.
Who it's for: Teams seeking codebase-aware story refinement across Microsoft tooling may find this orchestration pattern useful.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · antonybrahin
Frequently asked questions
How does the workflow use the existing codebase?
It creates vector embeddings with an Azure OpenAI model and uses Azure AI Search for retrieval, giving generation steps context from the actual code.
Where does the generated material go?
Direct Azure DevOps REST API calls create wiki pages and work items. The author uses these calls because standard connectors were limited.
More details
Remaining steps
- Azure DevOps REST API calls create wiki pages and work items from structured output.
About this account
Power Automate coordinates API-based generation, code retrieval, and creation of Azure DevOps artifacts.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=45103205. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: antonybrahin on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Power Automate · Gemini API · Azure OpenAI · Azure AI Search · Azure DevOps REST API
Azure DevOps · RAG · User stories
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.