BANDURAWorkflow library
Runnable AI workflow templates
Download complete examples for local classification, JSON extraction, CSV evaluation, and API request chaining. Inspect every step in Bandura.
Choose a task. Start with synthetic data.
Every ZIP includes the workflow, instructions, and any dataset it needs. Extract it into a folder, open that folder in Bandura, and run workflow.aether. No signup is required to download an example.
Ollama, LM Studio, or your own model
Classify support messages with a local model
Classify a synthetic support message, reject unexpected categories, and inspect the result in Bandura.
View workflow and setup →Ollama, LM Studio, or your own model
Extract structured data and validate AI JSON
Extract an order ID and city, parse the model response, and validate fields before using the result.
View workflow and setup →A configured model; includes cases.csv
Evaluate an AI classifier with CSV test cases
Run three labelled synthetic messages through an AI workflow and check each predicted category.
View workflow and setup →Internet access; no model or account
Chain API requests and check a captured token
Capture a demo token from an echo response, reuse it in a request, and assert the header. No account or AI key.
View workflow and setup →Before adding a model
The API example needs internet access but no model or account. AI examples need a configured provider: follow the Ollama guide or LM Studio guide. Local scripts and workflow files stay on your machine; remote models and APIs receive requests you send them.
What has been checked
The flow format and checks are tested against Bandura’s source engine with deterministic model responses and a local HTTP echo fixture. These checks verify workflow behavior, including failures; they do not measure real-model accuracy. Each example page explains its validation limits.