llm-exe

llm-exe is a TypeScript library that lets developers build type-safe LLM agents and AI functions using modular, composable components. It abstracts away the boilerplate of LLM API calls—retries, timeouts, parsing, and validation—so you can focus on building features. The library works with any major LLM provider and supports switching providers with a single line of code. It is production-ready with 100% test coverage and built-in error handling.

• Type-Safe Everything: Full TypeScript support with inferred types throughout your LLM chains, eliminating any/unknown types and guesswork.
• Provider Agnostic: Works with OpenAI, Anthropic, Google, xAI, Ollama, AWS Bedrock, and more — switch providers by changing one line.
• Composable Executors: Build complex AI workflows by chaining simple, self-contained executors (Prompt + LLM + Parser = Executor).
• Powerful Parsers: Extract structured output in formats like JSON, lists, regex, markdown blocks, or enums — guaranteed output format or throw.
• Production Ready: Built-in retries, timeouts, schema validation, and hooks for logging and monitoring out of the box.
• Agent Building: Create autonomous agents with state management, tool calling, and dialogue tracking using createCallableExecutor and useExecutors.
• Easy Installation: Install via npm (npm install llm-exe) and import createLlmExecutor, useLlm, and createParser to get started immediately.
• Hooks & Events: Attach onSuccess, onError hooks or bind events to executors for observability and monitoring.
• Works with All Models: Agent tool-calling works even with models that don't natively support function calling, giving you full control over execution flow.