Ava, Little One's Studio chat assistant LOCAL TEST

A local server on Angelo's Mac, reaching an Ollama model on the VPS through an SSH tunnel. Not connected to app.littleones.studio.

Try it

The purple button is bottom right. Click it and ask something.

How an answer is chosen

Every question runs through the same order, fastest and most certain first.

  1. FAQ. A deterministic keyword match against the knowledge base. Instant, free, and always the team's own approved wording.
  2. Cache. If this exact question was answered by the model before, that answer is reused instantly.
  3. Retrieval and model. The question is matched against the knowledge base by meaning, and the closest few passages are handed to a small local model (through the VPS tunnel) with instructions to answer only from that material.
  4. Hand-off. If nothing above produces a safe, grounded answer, the assistant says so plainly and offers to pass the question to the team. It never guesses.

Brand

--ls-purple
--ls-purple-dark
--ls-purple-light
--ls-paper
--ls-ink
--ls-line

Guardrails

Every model answer is checked after it finishes generating, before it is marked final. If it fails a check, the visitor sees the in-progress reply disappear and a plain hand-off message take its place.