How-To Guides

Overview

These guides are recipe-style — each one shows a complete .prompty file and the code to run it. Copy, paste, fill in your API keys, and go.

Prompty supports multiple LLM providers out of the box — OpenAI, Microsoft Foundry, and Anthropic — and its plugin architecture makes it easy to add more.

Tip

New to Prompty? Start with the Getting Started guide first, then come back here for specific use cases.

---

Guides

Guide	What you’ll build
Use with Anthropic	Chat completion using Anthropic Claude with API key auth
Use with OpenAI	Chat completion using the OpenAI API with API key auth
Use with Microsoft Foundry	Chat completion using Microsoft Foundry — API key and Azure AD options
Agent with Tool Calling	An agent that calls your Python/TypeScript functions in a loop
Structured Output	Get typed JSON responses matching a schema
Streaming Responses	Stream tokens as they arrive
Embeddings	Generate text embeddings with apiType: embedding
Image Generation	Generate images with DALL-E via apiType: image
Custom Providers	Write your own executor and processor
VS Code Extension	Author and test .prompty files in VS Code

Note

All three providers follow the same patterns — install the corresponding extra, set the provider in your frontmatter, and configure the connection. See the Python implementation docs for details on adding custom providers.

---

Common Pattern

Every guide follows the same three-step pattern:

1. Write a .prompty file     →  declarative prompt + model config
2. Call prompty.execute()     →  load → render → parse → call LLM → process
3. Use the result             →  string, JSON, embeddings, or stream

The runtime handles template rendering, role-marker parsing, provider dispatch, and response processing automatically — your code stays minimal.