Eval Driven Development
Use pytest to test each workflow node against unlimited test inputs and outputs, with asserts and LLM powered evals.
Junjo 順序 AI
Junjo is an open source Python library for designing, executing, testing, and debugging complex, graph-based AI workflows.
Whether you’re building a simple chatbot, a complex data manipulation pipeline, or a sophisticated workflow with dynamic branching and parallel execution, Junjo provides the tools to define your logic as a clear graph of nodes and edges.
Junjo remains decoupled from any specific AI model or framework. Simply wrap your existing business logic in a Junjo node, organize them into a Graph with conditional edges, and then execute the graph. Junjo will handle the rest, including task orchestration, error handling, and logging to any OpenTelemetry destination.
Features
OpenTelemetry Native
Execution traces are grouped into workflows and nodes. Optional self-hostable companion telemetry server.
State Management
Redux-inspired state management for your workflows. Use Junjo Server to visually step through workflow state changes.
Read the docs
Check out the Python library API the docs. Find Junjo on PyPI.