azure-openai
Skill46 starsChanged 8 months agoMITView on GitHub
Azure OpenAI and AI integration with Power Platform. Use when: Azure OpenAI, AOAI, AI Foundry, GPT, LLM, AI integration, BYOM, bring your own model, AI Builder prompts, structured output, AI architecture decisions, Copilot Studio custom models.
In this repository40 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- accessibility-uxSkill
- ai-builderSkill
- almSkill
- architectureSkill
- business-process-flowsSkill
- canvas-appsSkill
- code-appsSkill
- code-reviewerSkill
- copilot-studioSkill
- custom-connectorsSkill
- dashboardsSkill
- data-migrationSkill
- dataverse-mcpSkill
- dataverseSkill
- dataverse-web-apiSkill
- devops-engineerSkill
- env-strategySkill
- governanceSkill
- integration-patternsSkill
- licensingSkill
- m365-integrationSkill
- model-driven-appsSkill
- observabilitySkill
- pcfSkill
- perf-optimiseSkill
- platform-builderSkill
- pluginsSkill
- power-automateSkill
- power-biSkill
- power-pagesSkill
- power-platform-teamSkill
- project-managerSkill
- securitySkill
- solutions-architectSkill
- spec-driven-devSkill
- test-engineSkill
- testingSkill
- uat-coordinatorSkill
- web-resourcesSkill
The file
We split this skill into the rules it gives an agent
5 rules. About 0 are common, 5 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈5
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 41
Use managed identity: when possible — Azure Function with managed identity to AOAI is the most secure pattern
- Distinctive: line 52
No retry logic on AI calls (AOAI returns 429 under load)
- Distinctive: line 53
No token limit management (prompt + response exceeding model context window)
- Distinctive: line 56
No cost monitoring on AOAI deployment (token costs can spike unexpectedly)
- Distinctive: line 58
Not testing with representative data (AI behavior varies with real-world input)
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Skill: Azure OpenAI Integration
When to Use
Trigger when integrating Azure OpenAI (or AI Foundry models) with Power Platform — from Power Automate flows, Code Apps, Copilot Studio agents, or when making architecture decisions about which AI approach to use.
Integration Patterns
| Pattern | Description | Best For |
| AI Builder prompts | Low-code GPT prompts in flows/apps | Simple text generation, classification, extraction |
| Azure AI Foundry connector | Power Automate connector to AOAI | Flow-based AI with premium connector |
| HTTP connector to AOAI | Direct REST calls to AOAI endpoint | Full control, streaming, advanced parameters |
| Custom API + Azure Function | Server-side AI calls exposed via Custom API | Code Apps, secure key management, streaming |
| Copilot Studio BYOM | Custom model for agent responses/prompts | Domain-specific agent behavior |
| Copilot Studio generative actions | AI selects actions dynamically | Context-dependent automation |
⚠️ REQUIRED: Load Sub-Files Before Implementation
SKILL.md is a summary only — it is NOT sufficient for implementation.
The detailed content (complete payloads, XML templates, working examples, edge-case handling) lives in sub-files in the same directory as this SKILL.md. Before writing any code, you MUST use read_file on the sub-files relevant to your task:
- Power Automate + AI — AI Foundry connector in flows, HTTP connector to AOAI endpoints, prompt patterns, retry/fallback, structured output parsing, token limit management
- Code Apps + AI — AOAI from Code Apps via Custom API or Azure Function (never client-side keys), streaming responses, chat UI patterns, React hooks for AI calls
- Copilot Studio BYOM — Bring Your Own Model for prompts and response generation, AI Foundry model configuration, testing, monitoring token usage
- AI Architecture Decisions — decision tree: Copilot Studio agent vs Azure AI Foundry agent vs Power Automate Agent Flow vs AI Builder vs direct AOAI, licensing implications, data residency
Key Principles
- Never expose API keys client-side — Code Apps and Canvas Apps must call AOAI via a server-side intermediary (Custom API, Azure Function, or Power Automate flow)
- Use managed identity when possible — Azure Function with managed identity to AOAI is the most secure pattern
- AI Builder for simple, Foundry for complex — AI Builder prompts are faster to build; direct AOAI is more flexible
- Monitor token consumption — set up alerts on Azure OpenAI usage to prevent unexpected costs
- Structured output — use JSON mode or function calling for parseable responses, not free-form text
Anti-Patterns
- Exposing AOAI API keys in client-side code (Canvas/Code Apps)
- Using GPT for deterministic tasks that don't need AI (regex, simple lookups)
- No retry logic on AI calls (AOAI returns 429 under load)
- No token limit management (prompt + response exceeding model context window)
- Free-form text output when structured data is needed (use JSON mode)
- Using AI Builder for everything (limited control, limited models)
- No cost monitoring on AOAI deployment (token costs can spike unexpectedly)
- Calling AOAI from every flow run without caching (expensive for repetitive queries)
- Not testing with representative data (AI behavior varies with real-world input)
Related Skills
copilot-studio— Agent building, generative actions, knowledge sourcespower-automate— Flow integration with AI connectorscode-apps— React UI patterns for AI featureslicensing— AI Builder credits, Copilot Credits, AOAI consumption costsintegration-patterns— Azure Functions for server-side AI
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

