NirDiamant/agents-towards-production: End-to-end, code-first tutorials for building production-grade GenAI agents. From prototype to enterprise deployment.

End-to-end, code-first tutorials for building production-grade GenAI agents. From prototype to enterprise deployment. - NirDiamant/agents-towards-production

Summary

Agents Towards Production is a collection of 28 runnable Jupyter Notebook tutorials that walk through building, deploying, and operating LLM-powered agents. Each tutorial lives in its own folder under tutorials/ and targets a specific production concern, from LangGraph orchestration and RAG to memory, security, GPU deployment, and observability. The repository metadata lists the license as "NOASSERTION," so the actual license terms are not clearly declared in the provided material.

What it is useful for

The repo serves as a reference cookbook for engineers moving a GenAI agent from notebook to deployable service. Tutorial categories documented in the README include:

• Orchestration: a LangGraph-based agent framework under tutorials/LangGraph-agent.
• Memory and retrieval: Redis-based agent memory, Contextual AI RAG, Mem0 self-improving memory, and an Inngest durable RAG ingestion pipeline under tutorials/durable-rag-ingestion-inngest.
• Tools and data: an Arcade tutorial for OAuth2-secured tool calling against services like Gmail and Slack, Tavily for real-time web search, Bright Data for web scraping, and a hushvert tutorial for on-device file conversion.
• Deployment: a RunPod tutorial for GPU-based agent hosting under tutorials/runpod-gpu-deploy, with Docker and FastAPI patterns also listed as tutorial topics.
• Cross-language: a JetBrains Koog tutorial for building agents in Kotlin on the JVM.

How engineers can use it

Start by opening the relevant subfolder under tutorials/, for example tutorials/agent-memory-with-redis or tutorials/agent-with-tavily-web-access, and read its own README. The top-level README does not provide a single canonical install command; per-tutorial setup steps, API keys, and external service accounts are documented inside each subfolder, so check there before running anything. Plan to budget time for the dependencies a given tutorial pulls in: Redis, an MCP runtime, an OAuth2-enabled tool service, or a GPU instance each add their own setup overhead. Because the license is marked "NOASSERTION" in the repository metadata and the README does not specify permitted use, confirm licensing terms with the maintainer before shipping any of the code in a commercial product.