Jangada AIJangada AI

Step-by-step debug

Enable debug=True for a trace of every call: provider/model, params, retries, fallback, tokens, cost and duration — per agent.

from jangada_ai import LLM

llm = LLM("openai", "gpt-4o-mini", debug=True, name="extractor")
llm.complete("...")

The Debugger records the chain's events:

  • start — provider, model and params of the attempt
  • retry — error, attempt number and backoff delay
  • fallback — which provider/model it fell through to
  • end — resulting Completion and duration in ms
  • error — normalized error when the candidate exhausts its attempts

The name= parameter labels the agent in the trace, useful when there are several different LLMs in the same orchestration (Flow/Graph).

Related: Retry and fallback, Cost and tokens, Errors.

Example

examples/debug_params_example.py — runnable script.

Response cache

Cache LLM responses to save tokens and latency — exact (ExactCache) and semantic (SemanticCache, by embedding similarity). Plugged via LLM(..., cache=...).

Tutorial: getting started (from zero to fallback)

A ~10-minute walkthrough: install, make your first call, switch providers without changing your code, use templates, and make everything resilient with fallback. Every block is...

On this page

Example