---

✨ Features at a Glance

---

📸 Application Showcase

Meeting Archive & Detail View

Review audio recordings with interactive seekbars, voice enhancement indicators, speaker identification, and multi-version AI summaries with raw markdown copy & preview.

---

LLM Provider & Neural ASR Settings

Switch between cloud and local AI providers on the fly with granular control over reasoning levels and diarization clustering sensitivities.

---

Ingestion & Hardware Integration

Easily upload local audio files or connect Plaud hardware devices via automated OAuth token sync.

---

Customizable Summary Templates

Define structured summary formats tailored to specific meeting workflows (1-on-1s, Standups, Executive Briefs, Technical Architecture Reviews).

---

Real-Time Observability & System Logs

Monitor ingestion queues, diarization steps, model inference latency, and Plaud polling in real time with live streaming logs.

---

Full Backup & Disaster Recovery

Export and restore database snapshots, settings, templates, transcripts, notes, and raw audio files in a single self-contained archive.

---

🏗️ Architecture

SOURCES                           PIPELINE & ARTIFACT STORE                       INTELLIGENCE & EXPORT
───────                           ─────────────────────────                       ─────────────────────

 [ Plaud Note AI ] ──(OAuth/PCS)─┐
                                 │      ┌─────────────────────────────┐
 [ iOS Shortcut ]  ──(Ingest)───┼────► │   Meeting Aggregate Root    │
                                 │      │  ─────────────────────────  │
 [ Web UI Upload ] ──(Upload)───┘      │  • sha256 deduplicated      │
                                        │  • audio/<id>.<ext>         │
                                        │  • SQLite metadata + WAL    │
                                        └──────────────┬──────────────┘
                                                       │
                                 ┌─────────────────────┴─────────────────────┐
                                 ▼                                           ▼
                    ┌─────────────────────────┐                 ┌─────────────────────────┐
                    │     Transcription &     │                 │   LLM Summarization &   │
                    │   Speaker Diarization   │                 │    Action Item Engine   │
                    │ ─────────────────────── │                 │ ─────────────────────── │
                    │ • Local: Qwen3-ASR 1.7B │                 │ • pi-ai Unified Catalog │
                    │ • Sherpa-ONNX 3D-Spkr   │                 │ • Gemini / Claude / GPT │
                    │ • Cloud: Speechmatics   │                 │ • Local Ollama / vLLM   │
                    │ • Voiceprint Enrollment │                 │ • Structured Templates  │
                    └─────────────────────────┘                 └─────────────────────────┘
                                 │                                           │
                                 └─────────────────────┬─────────────────────┘
                                                       │
                                                       ▼
                                        ┌─────────────────────────────┐
                                        │     Interactive Web App     │
                                        │  Markdown • Audio • Backup  │
                                        └─────────────────────────────┘

---

🚀 Quickstart

Prerequisites

• Bun (v1.4+)
• ffmpeg (for local audio decoding)
• (Optional) Vulkan drivers / GPU for local ASR acceleration

Local Installation

# 1. Clone the repository
git clone https://github.com/mcowger/olive.git
cd olive

# 2. Install dependencies
bun install

# 3. Download local speech & diarization models (Qwen3-ASR & Sherpa-ONNX)
bun run scripts/download-models.ts

# 4. Build web frontend
bun run build

# 5. Start the server
bun run start

Visit http://localhost:4471 in your browser.

---

🐳 Docker Deployment

Olive includes a production-ready multi-stage Dockerfile with bundled llama.cpp Vulkan acceleration, sherpa-onnx, and ffmpeg.

# Build Docker image
docker build -t olive:latest .

# Run container with persistent data volume
docker run -d \
  --name olive \
  -p 4471:4470 \
  -v olive_data:/app/data \
  --restart unless-stopped \
  olive:latest

---

⚙️ Configuration & Environment Variables

Olive stores dynamic settings in <data>/config/settings.json, and accepts environment variables for bootstrap and secrets:

The default transcription engine is configured in LLM Provider & Model Settings → Automatic Transcription Engine. Local transcription is the default; choose Speechmatics there only when SPEECHMATICS_API_KEY is configured.

MCP integration

Olive exposes a read-only Model Context Protocol endpoint at /mcp on the main server port. It provides tools for listing meetings, retrieving meeting details, searching transcripts, extracting summary checklist items, and looking up speakers.

Local loopback clients can connect without a token. Set OLIVE_MCP_TOKEN to protect the endpoint; it falls back to OLIVE_INGEST_TOKEN. A token is required when OLIVE_BIND_HOST is not a loopback address.

Example client configuration:

{
  "mcpServers": {
    "olive": {
      "url": "http://127.0.0.1:4471/mcp",
      "headers": {
        "Authorization": "Bearer ${OLIVE_MCP_TOKEN}"
      }
    }
  }
}

---

📱 Integrations

1. Apple Voice Memos (iOS Share-Sheet Shortcut)

Upload voice memos directly from an iPhone or iPad to Olive in one tap:

1. Open Shortcuts on iOS and create a new shortcut.
2. Set Receive: Audio & Files from the Share Sheet.
3. Add action: Get Contents of URL
   • URL: https://<your-olive-instance>/api/ingest
   • Method: POST
   • Headers: Authorization: Bearer <YOUR_OLIVE_INGEST_TOKEN>
   • Request Body: Form
     • file $\rightarrow$ Shortcut Input
     • autoTranscribe $\rightarrow$ true
     • source $\rightarrow$ ios-shortcut
4. Tap Share on any Voice Memo $\rightarrow$ select your shortcut.

2. Plaud Note & Plaud AI Cloud

1. Click the Plaud button in the header.
2. Complete OAuth authentication.
3. Olive polls in the background, automatically pulling recordings, PCS transcripts, and summaries when ready.

---

💾 Backup & Disaster Recovery

Olive provides database-consistent snapshots combining SQLite data, configuration, templates, and raw audio files into a single .tar.gz archive.

Create a Backup

• Web UI: Navigate to the Backup tab $\rightarrow$ click Download Full Backup (.tar.gz) or Save Server Snapshot.
• CLI:

  bun run scripts/backup.ts ./my-backup.tar.gz
  

Restore from Backup

• Web UI: Upload any Olive .tar.gz archive in the Backup tab.
• CLI:

  bun run scripts/restore.ts ./my-backup.tar.gz
  

---

🛠️ Development

# Start backend and web client concurrently with live watch and HMR
bun run dev

# Run only the backend API server with watch mode
bun run dev:server

# Run only the web frontend with Bun HMR
bun run dev:web

# Run full test suite
bun test

# Run TypeScript typechecks across all monorepo workspaces
bun run typecheck

---

📄 License

MIT © Matt Cowger