Haive Core Overview

Welcome to haive-core, the foundation of the Haive AI Agent Framework! 🚀

Note

Looking for other Haive packages? The complete framework documentation will be available at the Haive root documentation.

🧠 Core Engine

The brain of Haive - manages agent execution, LLM integration, and state management.

Explore Engine →

autoapi/core/engine/index
📊 Graph System

Powerful graph-based workflow orchestration with LangGraph integration.

View Graphs →

autoapi/core/graph/index
🗃️ State & Schema

Type-safe state management with Pydantic schemas and state persistence.

Schema Docs →

autoapi/core/schema/index
🛠️ Tools & Extensions

Rich tool ecosystem for agent capabilities - calculators, web search, and more.

Browse Tools →

autoapi/core/tools/index

Quick Start 🏃‍♂️

# Install haive-core
pip install haive-core

# Or with Poetry (recommended)
poetry add haive-core

from haive.core.engine import AugLLMConfig
from haive.core.schema import StateSchema

# Configure your LLM
config = AugLLMConfig(
    model="gpt-4",
    temperature=0.7
)

# Define your state
class MyState(StateSchema):
    messages: list[str]
    context: dict

from haive.core.graph import DynamicGraph
from haive.core.tools import CalculatorTool

# Build complex workflows
graph = DynamicGraph()
graph.add_node("process", process_node)
graph.add_tool(CalculatorTool())

Core Components 🎯

Engine Module 🔧

  • AugLLMConfig

  • BaseAgent

  • ToolEngine

  • ValidationEngine

Graph Module 📈

  • DynamicGraph

  • StateGraph

  • GraphNode

  • ValidationNode

Schema Module 📋

  • StateSchema

  • MetaStateSchema

  • MessagesState

  • ToolMessage

Architecture Overview 🏗️

        graph TB
   A[Application Layer] --> B[haive-core]
   B --> C[Engine]
   B --> D[Graph]
   B --> E[Schema]
   B --> F[Tools]

   C --> G[LLM Integration]
   C --> H[Tool Management]

   D --> I[Workflow Orchestration]
   D --> J[State Management]

   E --> K[Type Safety]
   E --> L[Persistence]

   F --> M[Built-in Tools]
   F --> N[Custom Tools]

   style B fill:#2563eb,stroke:#1d4ed8,color:#fff
   style C fill:#60a5fa,stroke:#3b82f6
   style D fill:#60a5fa,stroke:#3b82f6
   style E fill:#60a5fa,stroke:#3b82f6
   style F fill:#60a5fa,stroke:#3b82f6

Key Features ✨

🔄 Auto-Persistence

Automatic state saving and recovery with multiple backend support (PostgreSQL, Redis, File).

🌐 Multi-Provider Support

Works with OpenAI, Anthropic, Google, AWS Bedrock, and more LLM providers.

📊 Graph-Based Workflows

Build complex agent workflows with visual graph representation and debugging.

🛡️ Type Safety

Full Pydantic integration for type-safe state management and validation.

🧩 Modular Design

Clean separation of concerns with pluggable components and extensions.

⚡ High Performance

Optimized for production use with async support and efficient state handling.

Integration Examples 🔗

from haive.core.engine import AugLLMConfig
from haive.agents import ReactAgent

# Core provides the foundation
config = AugLLMConfig(temperature=0.7)

# Agents build on top
agent = ReactAgent(
    name="assistant",
    engine=config
)

from haive.core.tools import BaseTool
from haive.tools import WebSearchTool

# Core defines the interface
class MyTool(BaseTool):
    def __call__(self, query: str) -> str:
        return f"Processed: {query}"

from haive.core.schema import StateSchema
from haive.dataflow import StreamProcessor

# Core schemas work with dataflow
class StreamState(StateSchema):
    buffer: list[str]
    processed: int

Resources 📚

📖 API Reference

Complete API documentation with examples

API Reference
🎓 Tutorials

Step-by-step guides and examples

tutorials/index
💬 Community

Join discussions and get help

https://github.com/haive-ai/haive/discussions

Need Help? 🤝