AI API Playbook v1.0.0
The first public research and engineering release of AI API Playbook by Faramarz Kowsari.
This repository provides a reproducible and production-oriented reference for designing, testing, securing, evaluating, and commercializing modern AI API applications.
Highlights
- Reusable Python architecture for provider-aware AI applications
- OpenAI, Google Gemini, Anthropic Claude, and Microsoft Foundry examples
- Function calling and Model Context Protocol patterns
- Retrieval-Augmented Generation workflows
- Image generation, OCR, speech, video, and realtime examples
- Hugging Face, Cohere, Pinecone, Replicate, Runway, Deepgram, and ElevenLabs integrations
- Model routing, retry handling, cost estimation, evaluation, and safety utilities
- TypeScript WebRTC realtime client
- Commercial AI product blueprints
- Google Colab-compatible notebook
- Docker and GitHub Actions support
- Tested with Python 3.11, 3.12, and 3.13
- Citation and Zenodo metadata
- 100-page infographic book production map
Reproducibility
The repository includes an offline mock mode, automated tests, strict type checking, CI validation, documented environment variables, and provider-specific implementation examples. Live provider calls require the user's own API credentials.
Author
Faramarz Kowsari
Author, Software Engineer, and AI Researcher
- ORCID: https://orcid.org/0000-0003-1692-0453
- Google Scholar: https://scholar.google.com/citations?user=G7tP5WMAAAAJ&hl=en
- GitHub: https://github.com/FaramarzKowsari
- LinkedIn: https://www.linkedin.com/in/faramarzkowsari
- Wikidata: https://www.wikidata.org/wiki/Q140389378
- Google Books: https://play.google.com/store/search?q=Faramarz_Kowsari&c=books
License
Software source code is released under the MIT License. Book text, diagrams, and the visual prompt manuscript are licensed under CC BY 4.0 unless otherwise stated.