LLM SDKs
Vercel AI SDK, native provider SDKs, Pydantic AI, LiteLLM, Portkey — the thin layers between your code and the model.
Prompt management
PromptLayer, Braintrust prompts, Vellum, Pezzo, Latitude, Helicone Prompts, plus git-backed self-hosted — versioning, testing, and deploying prompts as first-class artifacts.
Programmatic prompt optimization (DSPy, GEPA)
Stop hand-tuning prompt strings — program your LLM pipeline as signatures and modules, then let an optimizer (DSPy teleprompters, GEPA) search for the best prompts and few-shot examples against a metric.
RAG frameworks
LangChain, LlamaIndex, Haystack, DSPy, Llmware — and when to skip a framework entirely.
Agent frameworks
LangGraph, CrewAI, AutoGen/AG2, OpenAI Agents SDK, Pydantic AI, Vercel AI SDK — managed agent loops with state, retries, and tracing.