DevX: 20 GitHub repos and full‑stack templates

Two X posts from accounts systemdesignone and @_vmlops circulated a curated list of 20 GitHub repositories plus production FastAPI + LangGraph/CrewAI templates framed as a 2026 "starter kit" for deployable agents. (x.com) The curated set foregrounds major open-source stacks: LangChain for agent engineering, Dify as an end-to-end LLM app platform, CrewAI for multi‑agent orchestration, Ollama for local model hosting, and RAGFlow for retrieval‑augmented generation. (github.com) Several entries explicitly bake observability and model management into the platform: Dify documents connectors to Opik, Langfuse and Arize Phoenix, LangChain offers LangSmith tracing and evaluation for agent runs, and CrewAI’s docs describe enterprise consoles and run telemetry for flows. (github.com) Concrete FastAPI + agent templates are already proliferating in the wild: Neo4j’s LangChain starter ships a FastAPI backend for production demos, community tutorials show Ollama + LangChain + FastAPI RAG chatbots, and multiple CrewAI example repos include FastAPI-based starter projects. (github.com) The momentum is measurable: LangChain’s main repo shows roughly 130k GitHub stars, RAGFlow’s infiniflow repository reports ~75k stars and recent contributor growth, and CrewAI’s repo lists ~46.6k stars — indicators of community uptake and contributor activity. (github.com) Project messaging and Octoverse signals align on reducing shadow infra risk by shipping opinionated templates plus telemetry — RAGFlow’s own blog highlighted rapid contributor growth, Dify advertises built‑in observability and model management, and LangSmith markets tracing/evals as the operational layer for LangChain agents. (ragflow.io)