HALO Assistant

Trigger HALO

Interact with the HALO system through a clean prompt-based interface. Use document grounding only when you want HALO to use a Markdown document for the current request.

Prompt → Response

A simple interface for interacting with HALO.

Document Grounding

Use a sample or custom Markdown document for this request.

RAG disabled
No document grounding is being used for this request.
Loaded document: —

LLM

Configure model and advanced chat behavior.

Response will appear here...
Trace ID: —

Answer Trace

Grounding, retrieval, and runtime details for the latest RAG response.

No trace loaded yet.

Trace Summary

  • Trace ID —
  • Model —
  • Search Mode —
  • Grounding Mode —
  • Top K —
  • Context Budget —

Retrieval

  • Semantic Candidates —
  • Lexical Candidates —
  • Fused Candidates —
  • Used Chunks —

Sources and Warnings

—
—

System Telemetry

Live hardware load and latest runtime information

Collapse

System Load

Real-time hardware load information

CPU: —
—

GPU: —
—

RAM: —

Runtime Info

Information about the latest response

  • Model —
  • Created At —
  • Total Duration —
  • Load Duration —
  • Prompt Eval Duration —
  • Eval Count —
  • Eval Duration —
RAG Options
Choose a document source and adjust grounding behavior for this request.
Document for this request
Choose one sample document for this request.
The selected file will be used only for this request.
—
No file loaded
Advanced RAG Settings
Controls how many chunks HALO retrieves before final selection.
RAG v1.0.0
  • Per-request document grounding
  • Sample and custom Markdown support
  • Preset-based retrieval configuration
  • top_k and search mode controls
  • Strict, balanced, and explain grounding modes
  • Configurable context budget
  • Optional reranking and citations
  • Trace output for retrieval diagnostics
  • Inline file preview and full file modal

Image preview

Click outside or press Esc to close.