Skip to content
Junjo AI

Python API Reference

Reference for Junjo Python SDK 0.69.0.

Documentation channel: Stable release

This reference is generated from the released source revision.

The common Agent definition and binding types are also available from junjo. Provider-neutral messages, results, and typed errors live in junjo.agent, and deterministic scripted testing support is public at junjo.agent.testing.

Core API

Section titled “Core API”

Module: junjo

Agent Definitions

Section titled “Agent Definitions”

Module: junjo.agent.definition

The common definition and binding types are also available from junjo.

Agent Model Drivers

Section titled “Agent Model Drivers”

Module: junjo.agent.model_driver

Connect your provider client with an application-owned adapter. See specialist agents for model bindings and application-owned provider integration.

Agent Tools

Section titled “Agent Tools”

Module: junjo.agent.tool

Declare typed capabilities for an Agent to invoke. See compose agents and workflows for service ownership, dependencies, and nested execution.

Agent Messages

Section titled “Agent Messages”

Module: junjo.agent.messages

Agent Results

Section titled “Agent Results”

Module: junjo.agent.result

Inspect validated output, transcript, usage, and execution identity. Start with specialist agents and agent testing for successful and failed runs.

Agent State

Section titled “Agent State”

Module: junjo.agent.state

Agent JSON

Section titled “Agent JSON”

Module: junjo.agent.json

Portable JSON aliases describe the values accepted by Agent contracts. JsonValue uses mutable lists and dictionaries; FrozenJsonValue describes the recursively immutable tuples and mappings exposed in owned execution records.

Agent Errors

Section titled “Agent Errors”

Module: junjo.agent.errors

Agent Testing

Section titled “Agent Testing”

Module: junjo.agent.testing

Deterministic scripted testing support is intentionally public at junjo.agent.testing. See agent testing for scripted runtime tests and the separate live evaluation workflow.

Evaluation API

Section titled “Evaluation API”

Module: junjo.evaluation

Studio-connected evaluation orchestration for real application Nodes, Workflows, and Agents. Use datasets and local evaluation runs to register targets, run locked cases, and compare outcomes.

Studio Client API

Section titled “Studio Client API”

Module: junjo.studio

Typed, bounded evaluation-control and evidence queries for Junjo AI Studio. See datasets and local evaluation runs for client use and Junjo AI Studio for investigating the recorded evidence.

OpenAI Agents Integration

Section titled “OpenAI Agents Integration”

Module: junjo.plugins.openai_agents

Optional first-party OpenAI Agents SDK composition and OpenTelemetry integration.

OpenAI Agents Evaluation

Section titled “OpenAI Agents Evaluation”

Module: junjo.plugins.openai_agents.evaluation

Optional outer OpenAI Agent evaluation target with exact span evidence.

Hooks API

Section titled “Hooks API”

Module: junjo.hooks

Telemetry API

Section titled “Telemetry API”

Module: junjo.telemetry.junjo_otel_exporter