banh.

A little intelligence. A clear next step.

Small decisions.
Clear outcomes.

Give messy input a little structure. Banh makes bounded, model-backed decisions, then follows the rules you write.

Get started Read the docs

YAML workflows. TypeScript runtime. Local inference.

InputDecisionsRulesOutput
urgent-routing.yaml01 / WORKFLOW
version: 1
process: urgent_routing
input:
  type: text

decisions:
  urgent:
    decide: whether
    question: Does this need immediate attention?

flow:
  - when: urgent.probability >= 0.85
    do: return
    value:
      route: escalation
  - else:
      do: return
      value:
        route: support
EXAMPLE INPUT

“Our checkout is down.”

IF P(URGENT) ≥ 0.85

route: escalation

Illustrative decision. The model evaluates your actual input.

A SMALL, DELIBERATE RUNTIME

Let the model decide.
Let your rules take it from there.

01 / DECIDE

Keep the question bounded.

Choose an option, answer yes or no, or score a scale. Laya evaluates your decisions together in one model call.

02 / ROUTE

Make the next step explicit.

Compare decision values and probabilities in ordered rules. The first match returns a value your application can use.

03 / RUN

Start on your own machine.

Validate a YAML file, run it locally on CPU, or use the TypeScript runtime in your application. Inspect the decisions behind the result.

BANH CLOUD In development

Your workflow.
An HTTP endpoint.

We’re building a home for versioned workflows and recorded runs. The local cloud service already supports authentication, deployment, and the HTTP API.

Cloud execution currently uses a fake backend. Hosted model inference is coming next.

Explore Banh Cloud
One YAML workflowDefine the decisions and rules
An immutable versionDeploy with the Banh CLI
A recorded runInvoke and inspect over HTTP

START SMALL

One file. Your first decision.

Try Banh locally

Available from source during early development.