autolog

Automatically logs car trips from OwnTracks Recorder and sends notifications via Telegram (or stdout). Trips are stored in a local SQLite database with reverse-geocoded start/end locations.

How it works

Autolog is event-driven for live trips:

1. A trip start event records the explicit start time immediately.
2. Stop and start events away from home are treated as intermediate journey events.
3. The journey remains active until a stop event arrives while the device is back in a home zone.
4. The complete home-to-home journey is then classified, stored, and notified.

Requirements

• An OwnTracks Recorder instance with location history
• A Telegram bot (optional — stdout mode available for testing)

Configuration

Configuration is read from config.yaml, environment variables, or a .env file. Environment variables take precedence.

Copy config.yaml.example to config.yaml and edit:

owntracks:
  url: "https://your-otrecorder-host/otrecorder"
  user: "yourname"
  device: "yourphone"

telegram:
  bot_token: "123456:ABC-DEF..."
  chat_id: "123456789"

http:
  addr: ":8080"
  trip_event_token: "${TRIP_EVENT_TOKEN}"

filters:
  max_train_speed_kmh: 150   # speeds above this classify the trip as train
  min_distance_km: 5.0       # trips shorter than this are discarded
  explicit_min_distance_km: 3.0 # phone-triggered trips shorter than this are discarded
  max_acc_m: 100             # GPS points with accuracy worse than this are dropped
  # A live trip must start and finish inside one of these home zones.
  exclusion_zones:
    - name: "home"
      lat: 0.0000
      lon: 0.0000
      radius_m: 300

store:
  path: "/data/autolog.db"

log:
  level: info

Environment variables

Variable	Description
OWNTRACKS_URL	OwnTracks Recorder base URL
OWNTRACKS_USER	OwnTracks username
OWNTRACKS_DEVICE	OwnTracks device name
TELEGRAM_BOT_TOKEN	Telegram bot token
TELEGRAM_CHAT_ID	Telegram chat ID to send messages to
FILTERS_EXPLICIT_MIN_DISTANCE_KM	Minimum distance for phone-triggered trips (default: 3.0)
HTTP_ADDR	Address for the phone trip-event API; empty disables it
TRIP_EVENT_TOKEN	Bearer token required by the phone trip-event API
STORE_PATH	Path to the SQLite database (default: autolog.db)
NOTIFY_STDOUT	Set to true to print notifications to stdout instead of Telegram
LOG_LEVEL	Log level: debug, info, warn, error (default: info)

Running

Local

cp config.yaml.example config.yaml
# edit config.yaml

go run ./cmd/autolog

Test without Telegram:

NOTIFY_STDOUT=true go run ./cmd/autolog

Docker / Dokploy

docker compose up -d

The compose file expects a dokploy-network external network and an .env file for configuration. The database is stored in a named volume autolog-data.

The included Traefik labels expose the API at:

https://wraeclast.bishop-bass.ts.net/autolog/api/trips/start
https://wraeclast.bishop-bass.ts.net/autolog/api/trips/stop

Traefik strips /autolog before forwarding the request to the container.

Generate the event token

Generate a high-entropy token locally and add it to .env:

openssl rand -hex 32

Store the resulting value as:

TRIP_EVENT_TOKEN=the-generated-value

Keep .env out of git and restrict its permissions:

chmod 600 .env

Do not put the token in Traefik labels or the Shortcut URL. The Shortcut should send it in the Authorization: Bearer ... header.

Backfill historical data

To process historical location data from a specific date:

go run ./cmd/autolog -backfill -from 2026-01-01
# or via env:
BACKFILL=true BACKFILL_FROM=2026-01-01 go run ./cmd/autolog

Inspect or reprocess one trip

Review a past trip using the current OwnTracks data and classifier without changing the database:

go run ./cmd/autolog -inspect-date 2026-08-10 -inspect-start 17:58

Add -inspect-points to print every replayed GPS point and its OwnTracks metadata.

Replace the matching stored trip's points, classification, and stops without sending a duplicate notification:

go run ./cmd/autolog -inspect-date 2026-08-10 -inspect-start 17:58 -reprocess

The time is interpreted in Australia/Sydney. The review prints the selected trip's mode, distance, tag counts, and each detected stop with duration, confidence, and evidence.

Phone-triggered trips

With HTTP_ADDR and TRIP_EVENT_TOKEN configured, iOS Shortcuts can call:

POST /autolog/api/trips/start
POST /api/trips/stop
Authorization: Bearer <TRIP_EVENT_TOKEN>
Content-Type: application/json
{"timestamp":"2026-08-13T00:50:00Z"}

The start call records the explicit start time without waiting for a GPS fix. Repeated starts and stops away from home are continuation events; each away-from-home stop is stored as an explicit stop and the response is {"status":"ongoing"}. The next start closes that stop's departure time. A stop at home fetches the complete OwnTracks window, stores the final result as a car trip with the explicit stops, and returns {"status":"completed"}. Authenticated event requests always return HTTP 200; processing failures are written to the application log.

Query trips through MCP

Autolog includes a read-only MCP server for asking an AI client about trip history. When HTTP_ADDR is configured, it is available at /mcp on the same HTTP server as the phone event API. It reads the same SQLite database and exposes list_trips, trip_details, and trip_days tools. The existing TRIP_EVENT_TOKEN is required as Authorization: Bearer ....

Build it once:

go build -o bin/autolog-mcp ./cmd/autolog-mcp
MCP_TOKEN=the-generated-value ./bin/autolog-mcp -db /path/to/autolog.db -addr :8081

Register it with an MCP client using a configuration like:

{
  "mcpServers": {
    "autolog": {
      "url": "https://your-autolog-host/autolog/mcp",
      "headers": {
        "Authorization": "Bearer ${TRIP_EVENT_TOKEN}"
      }
    }
  }
}

The trip_days tool is intended for questions such as "where did we take trips between Monday and Friday?"; list_trips handles more detailed filtering by date, location, and transport mode.

Development

go test ./...