🎤 Speaker Diarization
⚡ Real-time Streaming
🔒 100% Offline Mode
GPL-3.0 — Open Source

Your AI
Co-pilot
In Real Time

Desktop app for real-time AI assistance during interviews, meetings, and presentations. Bring your own API key or run completely offline with Ollama + Whisper.

↓ Download Free ★ Star on GitHub
Assistance — Interview Mode
Profiles
🎯 Interview
📞 Sales Call
🤝 Meeting
📊 Presentation
⚖️ Negotiation
📝 Exam

Provider
✨ Gemini 2.5 Flash
Live Session
Listening…
I
Interviewer
Can you tell me about a time you debugged a distributed system under pressure?
Y
You
Sure, at my previous role we had a Kafka consumer lag spike during peak traffic…
⚡ AI Suggestion
Use the STAR method: Situation — Kafka lag at 50k+ msgs/sec. Task — restore consumer throughput. Action — added partition rebalancing + consumer scaling. Result — lag cleared in 15 min
🏠 Local-first
✈️ Works Offline
🔌 7 AI Providers
⚡ Real-time Streaming
💻 Cross-Platform
🔓 GPL-3.0 License
🏠 Local-first
✈️ Works Offline
🔌 7 AI Providers
⚡ Real-time Streaming
💻 Cross-Platform
🔓 GPL-3.0 License
✦ Features

Built for Every Conversation

Core AI
Voice + Screen Understanding
Listens to your conversation and sees what's on your screen simultaneously. Paste screenshots, share context, or let it monitor your display in real-time.
Privacy
Fully Offline Mode
Run Ollama (any model, default llama3.1) + Whisper locally via HuggingFace Transformers. Zero network calls, zero cloud uploads. Your data never leaves your machine.
LOCAL STATUS
Ollama running (llama3.1)
Whisper transcribing
No internet required
🔒
⚡
Speed
Real-time Streaming
AI responses stream as they generate via WebSocket (Gemini Live API). Zero waiting — the AI types while it thinks.
🎙️
Transcription
Speaker Diarization
Automatically separates 2-6 speakers in real-time. Labels them as Speaker A, B, C so context is always clear.
🔑
Control
Bring Your Own Key
No subscription, no middleman. Use your own API keys from Gemini, OpenAI, Claude, or any supported provider. Pay API costs directly.
Customization
6 Built-in Profiles + Custom Instructions
Switch between Interview, Sales Call, Meeting, Presentation, Negotiation, and Exam modes. Paste your resume, job description, or any context to personalize responses.
🎯 Interview
📞 Sales Call
🤝 Meeting
📊 Presentation
⚖️ Negotiation
📝 Exam
🎨
UI/UX
9 Themes + Transparency
Dark, Light, Midnight Blue, Sepia, Catppuccin, Gruvbox, Rose Pine, Solarized, Tokyo Night. Adjustable background transparency and click-through mode.
🌍
Languages
30+ Languages
Speech transcription supports 30+ languages including English, Spanish, French, German, Chinese, Japanese, Hindi, and more.
🔍
Grounding
Google Search Integration
Toggle Google Search to ground AI responses in real-time web data. Useful for fact-checking or current events during conversations.

7 Providers + Offline Mode

Connect any provider with your own API key (BYOK). No lock-in, no subscription.

✨
Gemini Live
🌐
OpenRouter
🤖
Claude
⚡
OpenAI
🔵
DeepSeek
🚀
Groq
🧩
OpenCode
🏠
Ollama
🎙️
Whisper

Primary: Gemini 2.5 Flash (WebSocket, real-time audio) • Offline: Ollama + Whisper

→ How it Works

Three Steps to Your AI Co-pilot

01
Connect Your Provider
Add your API key from Gemini, OpenAI, Claude, OpenRouter, DeepSeek, Groq, or OpenCode. Or skip it entirely — point Assistance to a local Ollama instance and run 100% offline with Whisper transcription.
API Key Setup
ProviderGemini Live
🔑AIzaSy••••••••••••••••••
Save & Connect →
↓
02
Choose Profile & Start Listening
Select Interview, Meeting, Sales Call, Presentation, Negotiation, or Exam mode. Hit Record. Assistance begins real-time transcription with automatic speaker separation. Choose speaker-only, mic-only, or both for audio input.
Live Transcript
A
Tell me about your experience with distributed systems and event-driven architecture…
B
I've worked with Apache Kafka and microservices for three years, handling 50k+ messages per second…
↓
03
Get Live AI Coaching
As the conversation flows, Assistance streams contextual suggestions, technical answers, and real-time coaching. Responses appear instantly via WebSocket streaming — right when you need them, not after the call ends.
AI Response
⚡ Suggestion
Mention your Kafka consumer lag optimization work — you reduced lag from 2 minutes to under 15 seconds by implementing partition rebalancing and scaling consumers horizontally

Make It Yours

9 themes, adjustable transparency, keyboard shortcuts, click-through mode, and conversation history.

Dark
Light
Midnight Blue
Sepia
Catppuccin
Gruvbox
Rose Pine
Solarized
Tokyo Night
↓ Download

Get Assistance
For Free

Cross-platform desktop app. No account required. GPL-3.0 open source.

🍎 macOS (.dmg) 🪟 Windows (.exe) 🐧 Linux (.AppImage)

GPL-3.0 License • Free forever • Electron 30 + LitElement UI

Common Questions

Yes, completely free and open source under GPL-3.0. You only pay for API usage from your chosen provider (Gemini, OpenAI, Claude, etc.) which is typically pennies per session. Or use Ollama + Whisper for 100% free offline mode.
For cloud providers (Gemini, Claude, OpenAI, etc.) just download Assistance and add your API key. For offline mode, you'll need to install Ollama and Whisper locally — both are free and straightforward to set up.
Gemini offers a generous free tier. OpenAI and Anthropic provide free API credits to new accounts. Ollama is completely free for local inference with no API key needed.
Only if you use a cloud provider. In offline mode with Ollama + Whisper, all processing happens on your machine. Zero network calls, zero cloud uploads. No telemetry or analytics in any mode.
Speaker diarization works well for 2-6 speakers in a conversation. It automatically labels them (Speaker A, Speaker B, etc.) so the AI has full context of who is speaking when.
Built on Electron 30 with LitElement for the UI, marked.js for Markdown rendering, and highlight.js for code syntax. Gemini Live API via WebSocket for real-time audio streaming. Ollama for local LLM inference and Whisper via HuggingFace Transformers for offline speech-to-text.
Yes. Each of the 6 built-in profiles has tailored prompts, and you can paste custom instructions like your resume, job description, or meeting context to personalize AI responses further.