openJiuwen Core is a Python SDK designed for developing, running, optimizing, and evolving AI agents powered by large language models. It targets developers building LLM applications, offering a high-performance runtime and a suite of tools to streamline agent creation, workflow orchestration, and prompt engineering. The SDK aims to lower development thresholds by providing ready-to-use components and robust debugging capabilities, enabling the efficient construction of stable and flexible LLM applications.

How It Works

openJiuwen Core operates as a Python SDK with a built-in high-performance runtime. It exposes interfaces for agent creation, workflow orchestration, LLM invocation, and tool calling. The core architecture includes an Agent Engine supporting ReAct and Workflow agent paradigms, and a high-performance execution engine designed for asynchronous IO, streaming processing, and component concurrency. This approach facilitates efficient task execution, agent state saving, interruption recovery, and advanced prompt optimization techniques, balancing flexibility with stability for LLM application development.

Quick Start & Requirements

• Primary install / run command: pip install -U openjiuwen
• Non-default prerequisites and dependencies: Python 3.11 to <3.14 (3.11.4 recommended). Requires LLM API configuration (API base, key, provider, model name, SSL verification) typically set via environment variables.
• Estimated setup time or resource footprint: Moderate, involving environment variable setup and code-based workflow definition as shown in the example.
• Links: The README provides an extensive Python code example demonstrating workflow creation and agent execution.

Highlighted Details

• Agent Orchestration: Features ReActAgent for iterative reasoning and action, and WorkflowAgent for structured, multi-step task automation.
• High-Performance Execution Engine: Supports distributed deployment, asynchronous parallel graph execution, component concurrency, and streaming data transfer between components.
• Prompt Development: Offers one-click prompt generation and automatic optimization iteration using real-world datasets.
• Runtime Capabilities: Includes agent state saving, interruption recovery, conversation history context management, and basic tool sets.

Maintenance & Community

The project welcomes contributions via issues, feature suggestions, documentation improvements, and code submissions. Specific community channels (e.g., Discord, Slack), notable contributors, sponsorships, or a public roadmap are not detailed in the provided README.

Licensing & Compatibility

• License type: Apache-2.0 License.
• Compatibility notes: The Apache-2.0 license is permissive, generally allowing for commercial use and integration within closed-source projects.

Limitations & Caveats

The README does not explicitly list limitations, known bugs, or deprecation warnings. The mention of an "open-source version" may suggest that certain features or levels of polish might differ from potential commercial offerings. The provided example requires manual configuration of LLM API credentials via environment variables.