What to feed ABRIA

• python/abria_server/abr_runtime.py
• python/abria_server/file_reasoning.py

This example matches a successful ABRIA teaching flow used with the project. Feed the lesson first, then ask it to describe Nexa.

textVelune means softly glowing.
Driftal means moving smoothly through space.
Mirael means calm and focused.
Nexa is Velune.
Nexa is Driftal.
Nexa is Mirael.
Nexa has silver wings.
Nexa stores energy.

Ask: Describe Nexa.

A successful response from that teaching session was:

textNexa has silver wings, is moving smoothly through space, and is calm and focused. Nexa is softly glowing and stores energy.

Attach several related source files, then ask for a bounded diagnosis instead of only pasting one stack trace.

textAnalyze these files as one project. Identify the first failing boundary, explain what contract is violated, propose the smallest fix, and list the regression tests that prove unrelated routes still work.

ABRIA's workspace/file intelligence is suited to prompts that specify domain, interactions, constraints and validation rather than only visual adjectives.

textCreate a responsive research operations dashboard with datasets, experiments, job queue, ownership, validation states and a working search/filter flow. Keep each requested domain object distinct and validate the generated source before saving it.

When you want ABRIA itself to learn a fact, make that intent clear. 'Teach me X' is a question to the assistant; 'I am teaching you that X' is an ingestion signal.

textI am teaching you this rule: release candidates are promoted only after checksum verification and a regression pass.