Guides & Recipes

Task-focused walkthroughs for refund and purchase caps, dangerous tools, local tracing, experiments, multi-agent systems, cost optimization, and troubleshooting.

Cap Refunds And Purchases

Stop an agent from refunding or buying above a dollar amount you set, before that function runs.

Stop A Dangerous Tool

Deny a named function, or a shell command that contains text you listed, before that function runs.

Multi-Agent Tracing

Trace complex multi-agent systems and understand agent collaboration.

Parallel Runs in One Process

Run multiple agents safely from a single process using run_identity() and force_flush().

Local Jaeger Setup

Set up Jaeger locally to view your traces.

Grafana Tempo Setup

Set up Grafana Tempo locally for trace visualization with powerful querying.

Cost Attribution & Optimization

Reduce LLM costs with Traccia insights and SDK features.

Run Experiments From Code

Install the SDK, point it at a dataset, call evaluate(), and open the saved run under Evaluate → Experiments.

Compare Two Experiments

Pick a baseline and a candidate on the same dataset, then read score, cost, and latency deltas before you promote.

Troubleshooting Guide

Diagnose and fix common Traccia SDK issues.

© 2026 Traccia.