Small decisions. Interesting possibilities.Submit contentSubmit

Jev Workflow Builder

Build collaborative workflows that connect typed Jev decisions with language-model steps.

Added to Jevfast

How it uses Jev

A workflow stores a text input, typed Jev questions, and downstream LLM nodes. Jev receives the input plus prior answers as state; its typed answers activate connected nodes and can be referenced in prompts. The included support-ticket demo classifies intent, urgency, and frustration before routing to specialist reply nodes.

What Jev decides

Support-ticket triage example (illustrative)Example answers · not a recorded Jev response · Source ↗
Question 1 · intent
YOUR APP
INSTRUCTION

Identify the main request in the input.

STATE

The demo input describes a duplicate charge, repeat issue, customer frustration, and a request for a refund today.

JEV · CHOICE
  1. billing
  2. technical
  3. feature_request
  4. other
Question 2 · urgent
YOUR APP
INSTRUCTION

Decide whether the input needs a response today, considering repeated issues, money at stake, or an explicit deadline.

STATE

The demo input describes a duplicate charge, repeat issue, customer frustration, and a request for a refund today.

JEV · NOUL
YesNo
Question 3 · frustration
YOUR APP
INSTRUCTION

Rate how frustrated the customer appears in the input.

STATE

The demo input describes a duplicate charge, repeat issue, customer frustration, and a request for a refund today.

JEV · SCORE
02

App workflow

  1. Enter ticket

    The input node provides the support-ticket text.

  2. Classify with Jev

    A Jev node sends its configured typed questions with the input and receives answers keyed by question ID.

  3. Route and draft

    Answer handles activate connected LLM nodes; prompts can include the input and named Jev answers. The demo routes billing and technical tickets to specialist drafts and gates escalation on intent plus urgency.

  4. Collect output

    Connected outputs are gathered into named output properties.

Why it is interesting

It makes Jev decisions editable graph nodes. Users can connect individual answer handles, combine conditions with any/all activation, and pass named answers into later model prompts.