Skip to content

Functualize

AI Outbound

raicing-ai/functualize

AI Outbound — Standalone Job¶

A job that actively calls an LLM using functualize's AI capability. Demonstrates outbound AI usage — the job is the caller, not the callee.

Source¶

examples/standalone/showcase/jobs/ai_jobs.py (the ai_write job)

Running¶

pip install "functualize[cli]" functualize-ai
cd examples/standalone/showcase
func ai-write --topic "Python async patterns" --style concise
func ai-write --topic "Error handling" --max-tokens 500

Key Concepts¶

AI.complete() — Call an LLM with optional structured output (response_model)
AI.run() — Multi-turn AI execution with tool access
ToolScope.functions() — Expose plain Python callables as AI tools
AILimits — Budget and token limits (max_tokens, max_tool_calls, timeout_seconds)
MockAI — Glob-pattern matching for deterministic test responses

Testing Without API Keys¶

The example uses MockAI which matches prompts to responses using glob patterns:

ai = MockAI(responses={
    "*concise*": GeneratedContent(title="Quick Guide", ...),
    "*detailed*": GeneratedContent(title="Deep Dive", ...),
    "*": GeneratedContent(title="Default", ...),
})

This makes tests fast, deterministic, and free of external dependencies.