Dialog-LAB

Multi-Model Session Environment for A/B/C comparison, Thinking Extraction, Voice Input/Output and controlled rounds.

Model A/B/C
Thinking-Tags
Voice Input (Whisper)
TTS Output
Session Logging
Round Control
Dialog-LAB overview

What's Included?

Multi-Model Rounds

Same question to 2-3 models, then comparison and optional merge.

Thinking Extraction

Supports tags like <think>, <thinking>, <reason>.

Voice Workflow

Transcribe voice input and optionally output response as speech.

SPR Commands for Thinking

LMS.THK

Extracts thinking content from model output.

LMS.THS

Removes thinking blocks and returns cleaned text.

LMS.HTK

Checks if thinking blocks exist, including hit count.

Session Control (Recommended)

max_rounds: Set round limit to prevent infinite loops.
topic_lock: Fix topic per session to keep responses focused.
stop_repeat: Enable repetition detection.
save_logs: Store all dialogs with timestamps.

Quickstart

1. Start Dialog-LAB and assign models A/B/C.
2. Send a test question with thinking extraction enabled.
3. Compare responses, create merge and save session log.

Further Reading

For complete tool overview and deployment context see MCP Server main page.

To MCP Server Page