> ## Documentation Index > Fetch the complete documentation index at: https://ibm-llm-runtime-aaf3a78b.mintlify.site/llms.txt > Use this file to discover all available pages before exploring further. # Mellea — build predictable AI without guesswork > A Python library for writing reliable generative programs.
Mellea mascot

Mellea helps you manage the unreliable part of every AI-powered pipeline: the LLM call itself. It replaces ad-hoc prompt chains and brittle agents with structured generative programs — Python code where LLM calls are first-class operations governed by type annotations, requirement verifiers, and principled repair loops.

```bash theme={null} uv pip install mellea ``` Install Mellea and run your first generative program in minutes. Build a complete program with generation, validation, and repair. Runnable examples: RAG, agents, sampling, MObjects, and more. Full public API — backends, session, components, requirements, sampling. ## How Mellea works Mellea's design rests on three interlocking ideas. `@generative` turns a typed function signature into an LLM-backed implementation. Docstrings become prompts. Type hints become output schemas. No DSL required. Declare what good output looks like with `req()`. Mellea checks every response before it leaves the session — using LLM verifiers, programmatic checks, or domain-trained adapters. When a requirement fails, Mellea feeds the failure back and tries again. Rejection sampling, majority voting, and SOFAI are built in. ## Key patterns Add `@mify` to any class to make it LLM-queryable and tool-accessible without rewriting your data model. Explicit context threading with push/pop state keeps multi-turn workflows reproducible and debuggable. `ainstruct()`, `aact()`, and token-by-token streaming for production throughput and responsive UIs. Guardian Intrinsics detect harmful, off-topic, or hallucinated outputs before they reach downstream code. Best-of-n, SOFAI, majority voting — swap strategies in one line. `@tool`, `MelleaTool`, and the ReACT loop for goal-driven multi-step agents. ## Backends Mellea is backend-agnostic. The same program runs on any inference engine. Local inference, zero cloud costs. GPT-4o, o3-mini, any OpenAI-compatible API. AWS Bedrock via Bedrock Mantle or LiteLLM. IBM WatsonX managed AI platform. Local inference with Transformers — aLoRA and constrained decoding. Google Vertex AI, Anthropic, and 100+ providers via LiteLLM. Use LangChain tools in Mellea sessions or call Mellea from LangChain chains. See [Backends and configuration](/how-to/backends-and-configuration) for the full list of supported backends and how to configure them. ## How-to guides Pydantic models, `Literal` types, and `@generative` for guaranteed schemas. Python functions, `ValidationResult`, and multi-field validation logic. `aact()`, `ainstruct()`, and token-by-token streaming output. `ChatContext`, explicit context threading, and multi-session workflows. Temperature, seed, max tokens, system prompts — cross-backend with `ModelOption`. Pass images to `instruct()` and `chat()` with any vision-capable backend. Vector search, LLM relevance filtering, and grounded generation end-to-end. *** [GitHub](https://github.com/generative-computing/mellea) · [PyPI](https://pypi.org/project/mellea/) · [Discussions](https://github.com/generative-computing/mellea/discussions)