NewRoboDiag Harness
Jev classifies the operator’s text into query, diagnosis, action, or chat.
Added to Jevfast
NewHow it uses Jev
Jev classifies the operator’s text into query, diagnosis, action, or chat. For diagnosis, a separate Jev router can propose the next diagnostic action using the symptom and collected tool evidence, with further parameter choices for runtime tests or topics. Deterministic registries decide which capabilities may execute.
What Jev decides
Illustrative source-derived request: route an operator requestExample answers · not a recorded Jev response · Source ↗
Question 1 · request mode
YOUR APP
INSTRUCTION
Classify the operator request as query, diagnosis, action, or chat.
STATE
The state is the operator’s input text under the field input.
JEV · CHOICE
- query
- diagnosis
- action
- chat
App workflow
- Classify the operator request
A Choice routes the user text to query, diagnosis, action, or chat.
- Check the route
Capability/action labels are checked against deterministic registries before a tool runs.
- Investigate a diagnosis
For diagnosis, Jev can select the next diagnostic action from the symptom and prior evidence; parameterized tools may require a later selection.
- Execute bounded tools
The harness runs only registered diagnostic capabilities and adds their results to the evidence state.
Why it is interesting
It separates language-based request routing and diagnostic planning from deterministic capability checks and the code that executes tools.