Draive¶

🏎️ All-in-one, flexible Python library for building powerful LLM workflows and AI apps 🏎️

Draive gives you everything you need to turn large language models into production-ready software: agents, structured workflows, tool use, instruction refinement, guardrails, observability, and seamless multi-model integration — all in one clean, composable package.

If you've ever felt like you're stitching together glue code and hoping for the best — Draive is what you wish you'd started with!

Why Choose Draive?¶

Building LLM applications often involves complex prompt management, model switching, tool integration, and quality control. Draive simplifies these challenges through:

🧱 Composable by Design¶

Build modular flows using reusable components. Every piece is testable and fits together seamlessly.

🔄 Multi-Model Ready¶

Switch between OpenAI, Claude, Gemini, Mistral, and more in seconds. No vendor lock-in.

🛠️ Tools = Just Python¶

Define a tool by writing a function. Annotate it. Done. Draive handles serialization, context, and LLM integration.

🛡️ Production Ready¶

Built-in evaluation, guardrails, observability, and error handling for reliable applications.

🎯 Developer Experience¶

Full type safety, excellent IDE support, and clear error messages. Built on solid Haiway foundations.

Core Features¶

🤖 Multi-Provider LLM Support¶

OpenAI (GPT, DALL-E, Embeddings, Realtime API)
Anthropic (Claude, including Bedrock)
Google (Gemini via AI Studio)
Mistral (including Azure)
Cohere (including Bedrock)
Ollama (local models)
VLLM (inference servers)
AWS Bedrock (multiple providers)

🔧 Advanced Capabilities¶

Tools & Function Calling: Turn any Python function into an LLM tool
Structured Generation: Generate validated Python objects from LLMs
Multimodal Content: Handle text, images, audio, and documents
Conversation Management: Build stateful, context-aware chat applications
RAG & Vector Search: Built-in document indexing and retrieval
Evaluation Framework: Test and improve LLM outputs systematically
MCP Integration: Connect to the Model Context Protocol ecosystem

🎭 Production Features¶

Guardrails: Content moderation, privacy protection, quality control
Observability: OpenTelemetry integration and detailed metrics
Retry & Error Handling: Robust failure recovery patterns
Streaming: Real-time response generation
Context Management: Immutable state and dependency injection

Getting Started¶

📥 Installation ¶

pip install 'draive[openai]'  # or anthropic, gemini, etc.

Set up Draive with the integrations you need

🚀 Quickstart ¶

Build your first AI assistant in minutes with our hands-on tutorial

📚 First Steps ¶

Master core concepts with practical examples and patterns

What You Can Build¶

🤖 Autonomous Agents¶

Create agents that can reason, plan, and execute complex tasks using tools and structured workflows.

💬 Conversational AI¶

Build chatbots, customer support systems, and interactive assistants with memory and context.

📊 Data Processing¶

Extract structured information from documents, analyze content, and transform unstructured data.

🔍 RAG Applications¶

Create knowledge bases that can answer questions from your documents and data sources.

🛡️ Quality Control¶

Implement evaluation pipelines to ensure your LLM applications meet quality standards.

Learn Draive¶

Essential Guides¶

Basic Usage - Your first text generation and model usage
Tools & Functions - Turn Python functions into LLM tools
Conversations - Build interactive chat applications
State Management - Immutable data and context systems

Practical Cookbooks¶

RAG System - Build a document search and retrieval system
MCP Integration - Connect to external tool ecosystems
Data Extraction - Extract structured data from text
Evaluation - Test and improve your applications

When to Use Draive¶

✅ Perfect for: - LLM-powered applications and services - Multi-model AI workflows - Applications requiring tool integration - RAG and knowledge management systems - AI agents and autonomous workflows - Content generation and analysis - Conversational AI and chatbots

⚠️ Consider alternatives for: - Simple, one-off LLM API calls - Applications not using language models

Architecture¶

Draive is built on Haiway, providing:

Immutable State Management: Type-safe, validated data structures
Context-based Dependency Injection: Clean state propagation
Structured Concurrency: Automatic resource management
Functional Approach: Pure functions over stateful objects

Community & Support¶

💻 GitHub Repository ¶

Source code, issues, and contributions

🎯 Examples Repository ¶

Complete applications and use case examples

Contributing¶

Draive is open-source and growing! Whether you're fixing bugs, adding features, improving docs, or sharing examples — all contributions are welcome.

Built by Miquido Empowering developers to build amazing AI applications