Skip to main content

Overview

ChatbotAI-Free offers two distinct conversation modes, each optimized for different use cases:
  • Classic Chat Mode — Turn-by-turn conversations with full markdown rendering and streaming responses
  • Live Mode — Continuous hands-free conversation with barge-in detection

Classic Chat Mode

The default mode for deliberate, thoughtful conversations.

How It Works

1

Record your message

Click the 🎤 Mic button to start recording. Speak your message, then click the mic button again to send.
You can enable auto-send in Settings to automatically send your message when you stop speaking (detected via Voice Activity Detection).
2

View transcription

Your speech is transcribed using faster-whisper (multilingual STT) and displayed in a user bubble aligned to the right.
3

See the AI response

The AI generates a streaming response displayed with:
  • Markdown formatting (headings, bold, italic, code blocks, tables)
  • Live text updates as each token arrives
  • Collapsible reasoning panel (for thinking-capable models like DeepSeek R1)
  • TTS playback starting as soon as the first sentence is ready

Key Features

Markdown Rendering All bot responses are rendered with full GitHub-flavored markdown support:
  • Headers (#, ##, ###, ####)
  • Bold and italic text
  • Inline code and multi-line code blocks with syntax highlighting
  • Tables with alternating row colors
  • Horizontal rules
Streaming Responses Tokens stream in real-time from your Ollama model. The UI updates instantly, and TTS generation runs in parallel — audio playback begins before the full response completes. Reasoning Panel Thinking-capable models (like DeepSeek R1) show their reasoning in a collapsible panel:
  • Hidden by default (toggle with ✨ Show Reasoning ▾)
  • Streams reasoning tokens in real-time
  • Italic gray text to distinguish from the final answer
Context Window Indicator The context donut (bottom-right corner) shows how much of your model’s context window is used:
  • Click it to see details: used_tokens / total_tokens (percentage)
  • Color-coded: 🟢 Green (under 50%), 🟡 Yellow (50-80%), 🔴 Red (over 80%)

Live Mode

Hands-free continuous conversation mode with natural interruption support.

How It Works

1

Enter Live Mode

Click the ✨ Live button in the main window. The status changes to “listening”.
2

Speak naturally

Just start talking. Voice Activity Detection (VAD) automatically captures your speech and stops when you pause for 1.5 seconds.
Live Mode uses a higher VAD threshold to avoid triggering on background noise.
3

AI responds immediately

Your speech is transcribed and sent to the AI. The response starts playing as soon as the first sentence is ready.
4

Interrupt anytime (barge-in)

Start speaking while the AI is talking to interrupt immediately. The AI stops mid-sentence, and the conversation resumes from your new input.
Barge-in detection runs continuously in a background thread, monitoring audio levels every 30ms for ultra-responsive interruptions.

Barge-In Detection

Live Mode continuously monitors your microphone even while the AI is speaking. When you start talking:
  1. Detection threshold: RMS audio level > 2× the normal silence threshold
  2. Consecutive frames: Requires 4 consecutive frames (120ms) to confirm speech
  3. Instant stop: Audio playback terminates immediately via player.stop()
  4. Resume listening: The mode loops back to listening for your next input
Live Mode does not display transcriptions or responses in the chat history UI. It’s designed for fluid, voice-only interaction. Use Classic Chat if you need written records.

Mute Toggle

While in Live Mode, you can mute the microphone:
  • The status changes to “muted”
  • No new user input is processed
  • The AI can finish speaking if a response is already in progress

When to Use Each Mode

Best for:
  • Reading and referencing AI responses
  • Complex queries requiring markdown formatting
  • Conversations with PDF documents
  • Debugging code or reviewing tables
  • Saving chat history for later review
UI Controls:
  • 🎤 Mic button (tap to record, tap to send)
  • ⏹ Stop button (interrupt AI response)
  • Context donut (click for usage stats)

Keyboard Shortcuts

In Classic Chat mode, you can also type messages using the text input field:
  • Enter — Send message
  • Shift+Enter — New line (multi-line input)
Text messages skip transcription and go directly to the AI, useful for pasting code or precise queries.

Build docs developers (and LLMs) love