BANDURAPractical guides
Local AI workflow guides
Set up Ollama or LM Studio, run a downloadable workflow, and inspect the output and checks in Bandura.
Updated 2026-10-01
Choose the runtime you already use, then start with a single synthetic input.
- Build a local AI workflow with Ollama: connect your model, classify a support message, and inspect a failing assertion.
- Build and debug a workflow with LM Studio: extract JSON, validate fields, and distinguish parsing failures from incorrect values.
- Evaluate a classifier with CSV test cases: compare predictions with three expected labels.
- Bandura vs n8n for local workflows: compare desktop use with scheduled or hosted automation.
Start without a model
The API request-chain template needs internet access but no AI key or account. For a completely local script and assertion, use the order-process starter guide.
Browse complete examples
The workflow library includes ZIP downloads with source, instructions, and any needed dataset. Share the complete folder with teammates; each person supplies their own model settings and credentials.