A meeting note-taker that talks back - now on Windows and Mac.
Disclaimer & Acknowledgement: OpenCassava is a descendant of the excellent OpenOats project created by yazinsai. A huge thank you to the original creator for laying the groundwork for this application. OpenCassava has now evolved into its own dedicated project with a focus on comprehensive cross-platform support and expanded features.
OpenCassava sits next to your call, transcribes both sides of the conversation in real time, and searches your own notes to surface talking points right when you need them.
For first-time setup with LM Studio, start here: LM Studio Setup Guide
- Invisible to the other side - the overlay window is hidden from screen sharing by default, so no one knows you're using it.
- Live transcript + instant search - see both sides of the conversation in real time, search while people are speaking, and copy/export when needed.
- Smarter call capture - built-in echo cancellation suppresses speaker bleed into your mic, and a configurable mic voice threshold reduces low-level noise.
- Auto summaries while you talk - generate periodic summaries (30 seconds to 10 minutes), review summary history, and regenerate on demand.
- Contextual talking points - when the conversation hits a key moment, OpenCassava retrieves related knowledge and suggests useful responses.
- Knowledge base search - point OpenCassava to a folder of
.md/.txtnotes and retrieve relevant passages with embeddings. - Custom prompts and templates - tailor prompts for retrieval, suggestions, question generation, and post-call note formatting.
- Structured note generation - convert transcripts into clean markdown notes using built-in or custom templates.
- Session history - every session is automatically saved and accessible from the History sidebar.
- Runs 100% locally - tested primarily with LM Studio for LLM suggestions and local embeddings; no audio has to leave your device.
- Flexible AI providers - use local providers like Ollama, or cloud endpoints through OpenRouter and OpenAI-compatible APIs.
- Multi-language transcription - supports
whisper-rs, NVIDIA Parakeet, and Omni-ASR (1,600+ languages) for local speech recognition. - Omni-ASR on Windows - multilingual Omni-ASR runs via WSL2 on Windows. The app guides you through setup automatically.
OpenCassava is designed for high-pressure conversations (interviews, sales calls, support escalations, technical meetings) where speed and clarity matter.
- Stay out of the way: an always-available overlay that doesn't steal focus from your call.
- Readability first: high-contrast dark UI, clear grouping, and low visual noise so you can skim under pressure.
- Fast path to value: get useful output in seconds (transcript, suggestions, summary) with minimal setup friction.
- Progressive power: simple defaults for first-time users with advanced controls in Settings for deeper customization.
- Privacy by architecture: local-first transcription and local model support are built into the core workflow.
- Control Bar — start/stop sessions, status, and live session actions.
- Suggestions View — real-time AI guidance sourced from your knowledge base.
- Transcript View — dual-speaker timeline for "you" and "them".
- Search + Export — quickly find moments and export transcript artifacts.
- Notes View — generated notes and template-based post-call output.
- Session Sidebar — access and review prior sessions.
- Settings — provider setup, prompts, transcription options, and knowledge base pathing.
- You start a call and click Start Session.
- OpenCassava captures your microphone and (on Windows & Mac) system audio - the other side's voice is captured as "them".
- As important moments happen, it searches your notes and surfaces relevant talking points.
- During or after the call, use summaries and note templates to turn raw dialogue into structured documentation.
- Review, search, and export session artifacts from Session History.
Grab the latest release for your platform from the Releases page.
On every tag push, GitHub Actions publishes installers for all supported desktop platforms:
- Windows: EXE and MSI
- macOS: DMG
- Linux: DEB
Requirements:
# Clone the repo
git clone https://github.com/romeroej2/OpenCassava.git
cd OpenCassava
# Build the Tauri app
cd opencassava
npm install
npm run tauri -- buildThe installers are output to opencassava/src-tauri/target/release/bundle/.
- OS: Windows 10/11 (64-bit) or macOS 15+ (Apple Silicon)
- For local mode (Tested): LM Studio (or Ollama) running locally with your preferred models (e.g.,
qwen3:8bfor suggestions,nomic-embed-textfor embeddings). - For cloud mode (Untested): OpenRouter API key + Voyage AI API key.
- For OpenAI-compatible embeddings: any server implementing
/v1/embeddings.
The following engines require Python 3 to be installed on your system:
| Engine | Platform | Python Requirement |
|---|---|---|
faster-whisper |
Windows, macOS, Linux | Python 3 on native system PATH |
parakeet |
Windows, macOS, Linux | Python 3 on native system PATH |
omni-asr |
macOS, Linux | Python 3 on native system PATH |
omni-asr |
Windows | WSL2 with Python 3 inside it |
omni-asr on Windows (WSL2 setup):
- Open PowerShell as Administrator and run:
wsl --install - After reboot, open the WSL terminal (Ubuntu) and run:
sudo apt update && sudo apt install -y python3 python3-venv python3-pip - In OpenCassava, go to Settings → Advanced, select
Omni-ASR, and click Set up Omni-ASR.
The app will detect missing prerequisites automatically and show guided setup steps in the Settings UI.
- Open the app and grant microphone permissions (and system audio recording on Windows).
- Open Settings (
Cmd+,orCtrl+,) and configure your chosen cloud or local providers. - Point it at a folder of
.mdor.txtfiles - that's your knowledge base. - Click Start Session to go live. (The first run downloads the required local Whisper speech model.)
OpenCassava is built on a cross-platform Rust core with a shared React frontend.
| Component | Technology |
|---|---|
| Framework | Tauri 2 |
| Core logic | Rust (opencassava-core) |
| Transcription | whisper-rs (default) · faster-whisper · NVIDIA Parakeet · Omni-ASR |
| Audio capture | cpal (mic), WASAPI (Windows system audio) |
| LLM inference | OpenRouter API or Ollama |
| Embeddings | Voyage AI, Ollama, or OpenAI-compatible |
| Frontend UI | React 18 + TypeScript + Vite |
| Secret storage | Windows Credential Manager / macOS Keychain |
Important: OpenCassava records and transcribes audio from your microphone and system audio. Many jurisdictions have laws requiring consent from some or all participants before a conversation may be recorded.
By using this software, you acknowledge and agree that:
- You are solely responsible for determining whether recording is lawful in your jurisdiction and for obtaining any required consent.
- The developers and contributors of OpenCassava provide no legal advice and accept no liability for any unauthorized or unlawful recording conducted using this software.
Do not use this software to record conversations without proper consent where required by law.
See CHANGELOG.md for release history and upgrade notes.
MIT