Liatir

Appearance

Quenta ​

Quenta is a local, read-only AI inside Liatir. It helps explain app state, bioinformatics workflows, Results, Jobs, pipeline structure, and failure logs using local context.

It is designed for guidance and reports, not automation.

What it can do ​

  • Explain a Result and what its recorded output means.
  • Explain a failed or cancelled run using logs and metadata.
  • Generate a cited Markdown-style scientific report from a Result.
  • Answer questions about Liatir concepts, pipelines, AI Models, AI Tools, API Connector requests, and common bioinformatics interpretation issues.
  • Cite the local sources it used, such as a Result, Job, pipeline, or built-in bioinformatics note.

What it cannot do ​

Quenta cannot:

  • run pipelines;
  • run tools;
  • run Plugins;
  • run AI Models or AI Tools;
  • send API Connector requests;
  • edit files;
  • change workspace state;
  • execute shell commands.

The model receives text context only. Liatir does not send tool callbacks to the model.

Requirements ​

The first Quenta runtime uses Ollama running locally.

  1. Install and start Ollama.
  2. Pull a local chat model in Ollama.
  3. Open Quenta in Liatir.
  4. Set the Ollama base URL, usually http://127.0.0.1:11434.
  5. Refresh models and select the model you want to use.
  6. Save settings.

Liatir accepts only local loopback Ollama endpoints for this feature.

Explaining a Result ​

Open Results, select a run, then choose:

  • Explain with Quenta for a guided interpretation.
  • Quenta report for a structured report.
  • Explain with Quenta for a failed or cancelled run.

Quenta will open with that Result as its focus. Focused Results are forced into the retrieval context so the model sees the selected run metadata, structured output, output files, and recent logs.

Explaining a Job ​

Open Jobs and choose Explain with Quenta. Quenta uses the job status, command metadata, parent pipeline metadata when available, and buffered stdout/stderr.

How to read Quenta answers ​

Treat Quenta answers as scientific guidance. Check the cited sources and the original Result or Job before acting on conclusions.

Good Quenta answers should:

  • cite local sources;
  • distinguish observed evidence from interpretation;
  • state limitations;
  • recommend validation steps;
  • avoid inventing missing metrics;
  • avoid clinical or diagnostic conclusions.

If a source is missing, Quenta should say what is missing instead of filling the gap.