Capture actionable messages from anywhere — Discord, iOS share sheet, screenshots, or any HTTP client — into a LAN-hosted, AI-summarised todo list. No cloud. No API key. Runs on a Raspberry Pi.

• Receives text or images via HTTP POST from any source on your LAN
• OCRs screenshots with Tesseract (no external service)
• Summarises with a local Ollama model if configured (no API key, no cloud)
• Stores everything in SQLite
• Serves a mobile-first PWA: filter, group, drag to reorder, check off items
• Installable on iPhone/Android via "Add to Home Screen"

[iOS Shortcuts]  ──share sheet──▶ ┐
[Discord bot]    ──forward──────▶ │  POST /ingest
[Any LAN client] ──HTTP POST────▶ │  (YOUR_HOST)
                                  ▼
                           FastAPI (Docker)
                                  │
                    ┌─────────────┴──────────────┐
                    ▼                            ▼
               Ollama (local LLM)          SQLite DB
               summarise + tag             /db/ingest.db
                                            │
                                            ▼
                                      PWA (YOUR_HOST)
                                      filter / group / done

• Docker + Docker Compose

That's it. Everything else is optional.

git clone https://github.com/edleeman17/ingest
cd ingest
cp .env.example .env
docker compose up -d api

By default the API runs on port 8000. Visit http://localhost:8000 or set up a reverse proxy.

With Traefik and a *.lan cert, set INGEST_HOST=ingest.lan in .env and visit https://ingest.lan.

Since .shortcut files must be signed per-device, create them manually:

1. Open Shortcuts → tap +
2. Add action: Receive Text from Share Sheet
3. Add action: Get Contents of URL
   • URL: https://YOUR_HOST/ingest
   • Method: POST
   • Headers: Content-Type: application/json
   • Body (JSON):

     {"source": "ios-text", "raw_text": "[Shortcut Input]"}

4. Name it "Ingest Text"
5. Enable in Share Sheet (shortcut settings → Show in Share Sheet)

1. Open Shortcuts → tap +
2. Add action: Receive Images from Share Sheet
3. Add action: Convert Shortcut Input to Base64
4. Add action: Get Contents of URL
   • URL: https://YOUR_HOST/ingest
   • Method: POST
   • Headers: Content-Type: application/json
   • Body (JSON):

     {"source": "ios-screenshot", "raw_text": "", "image_data": "[Base64 Encoded]"}

5. Name it "Ingest Screenshot"

Note: If using a self-signed *.lan cert, install your CA certificate on the device first:

• Serve it at https://YOUR_HOST/ca and visit that URL in Safari on the device
• Go to Settings → Profile Downloaded → Install, then Settings → General → About → Certificate Trust Settings → enable it

1. Create a bot at discord.com/developers
2. Enable Message Content Intent under Bot → Privileged Gateway Intents
3. Invite to your private server with Send Messages + Add Reactions permissions
4. Set DISCORD_TOKEN and DISCORD_CHANNEL_ID in .env
5. docker compose up -d bot

Usage: Forward any Discord message to the watched channel. The bot adds ✅ when ingested.

Interactive docs available at https://YOUR_HOST/docs (Swagger UI) once running.

# Ingest text
curl -X POST https://YOUR_HOST/ingest \
  -H "Content-Type: application/json" \
  -d '{"source": "cli", "raw_text": "Follow up with Jake about the invoice"}'

# List todo items
curl https://YOUR_HOST/api/items?done=0

# Mark done
curl -X PATCH https://YOUR_HOST/api/items/42 \
  -H "Content-Type: application/json" \
  -d '{"done": true}'

# Create group
curl -X POST https://YOUR_HOST/api/groups \
  -H "Content-Type: application/json" \
  -d '{"name": "Work"}'

# Assign item to group
curl -X PATCH https://YOUR_HOST/api/items/42 \
  -H "Content-Type: application/json" \
  -d '{"group_id": 1}'

{
  "source": "string",        // free-form label (discord, ios-text, cli, etc.)
  "raw_text": "string",      // text to ingest (optional if image_data set)
  "image_data": "string",    // base64-encoded image — OCR'd with Tesseract (optional)
  "image_url": "string",     // URL hint (logged only, not fetched) (optional)
  "metadata": {}             // arbitrary extra data (optional)
}

Any service on your LAN can capture context with a one-liner:

# Shell
curl -s -X POST https://YOUR_HOST/ingest \
  -H "Content-Type: application/json" \
  -d '{"source": "my-service", "raw_text": "Thing to follow up on"}'

// Node.js / browser
await fetch('https://YOUR_HOST/ingest', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({ source: 'publisher', raw_text: text }),
})

# Python
import httpx
httpx.post('https://YOUR_HOST/ingest', json={'source': 'script', 'raw_text': text})

• Signal integration (signal-cli linked device)
• Voice memo → Whisper transcription
• macOS share sheet shortcut
• Email ingestion (forwarding alias)
• Browser bookmarklet

• API: FastAPI + uvicorn
• Database: SQLite via aiosqlite
• OCR: Tesseract (pytesseract)
• Summarisation: Ollama (local, configurable model)
• Frontend: Vanilla JS PWA — no framework, no build step
• Infra: Docker (Traefik optional)

MIT