From zero to a searchable knowledge graph of your entire digital life. Three steps, about 30 minutes of hands-on time. The GDPR exports take longer (1-3 days), but you only request them once.
Under GDPR (and similar laws worldwide), every platform is legally required to give you your data in a machine-readable format. Each platform's flow is slightly different. Below are step-by-step instructions, accurate as of April 2026, with the direct links you need.
Do all of these first. Most exports take a few hours to a few days to arrive. Request them all now in one sitting; by the time you run the installer in Step 2, they will be ready in your inbox.
LinkedIn does not let you request a data archive from the mobile app. You must use a desktop browser.
~/Downloads.Facebook moved data exports into Meta's Accounts Center in 2024. The flow below is the current one.
~/Downloads.Instagram exports go through the same Meta Accounts Center as Facebook. If you already started a Facebook export above, you can request Instagram from the same screen at the same time.
~/Downloads when it arrives.If you prefer the Instagram mobile app: Profile > menu > Accounts Center > Your information and permissions > Download your information. Same final flow.
WhatsApp exports can only be requested from the WhatsApp app on your phone. There is no web flow.
~/Downloads. On Android, AirDrop, USB cable, or Google Drive all work.~/Downloads.We use Google Takeout for Gmail content. Calendar, contacts and photos come from Apple Calendar, Contacts and Photos directly during install – no Google OAuth needed.
~/Downloads.You don't need every export to start. The installer is fine with whichever exports have arrived; it just skips the ones that aren't there yet. You can re-run the import later when more arrive (Step 3 is repeatable).
Open Terminal on your Mac and paste:
The installer checks your hardware, installs Docker and Ollama if needed, starts the database services, downloads the embedding model, and sets up the import pipeline. It also walks you through security setup: creating a strong passphrase (this encrypts your databases) and saving a recovery key.
What you need:
Which Mac should be your Hub?
Mac Mini or Mac Studio (always-on desktop) is the tidiest experience. Lives on a shelf, plugged in, always ready. Your iPhone always reaches it. Morning briefings fire at 9am whether you remembered or not. Marvin replies in real time, not after you open your laptop.
Your MacBook is the same product, portable. Take it to a coffee shop and the Hub goes with you. When you close the lid, Ostler goes to sleep with it. Captures queue locally and catch up the moment the lid opens. Docker and your local model pause on battery so your laptop is not cooked by the time you reach the office. Your iOS companion shows a live Online / Catching up / Offline status, so you always know the state.
Both Macs? One license, one Hub. Pick the Mac that wears the Hub hat. The other can reach the Hub’s web UI (wiki, Doctor dashboard) from Safari over your LAN or Tailscale, just like your iPhone can. A dedicated macOS companion app is on the roadmap.
macOS permissions the installer will request:
You can decline any permission – that feature is skipped and everything else still works. These are standard Apple security prompts that every Mac app must use.
Docker Desktop takes a minute to start. If the installer says Docker is not running, the installer will open it automatically and wait.
Once your GDPR exports have arrived, download them all and put them in one folder:
The importer scans for available exports and runs the appropriate parser for each platform. Missing exports are skipped gracefully. You will see a summary at the end showing how many contacts were imported from each platform.
What happens during import:
Message content is not stored. LinkedIn message metadata (who, when, how many) is imported, but the actual message text is not. This is a deliberate privacy choice.
Your knowledge graph is live. You can explore it via:
http://localhost:6333/dashboard – see your vector databasehttp://localhost:7878 – query your knowledge graph with SPARQLThe personal wiki and AI assistant (Marvin) are coming in the next phase. For now, your data is imported, connected, and searchable.
Once Ostler Doctor ships, it will check all services, identify problems, and give you copy-paste fix commands. Until then, the common fixes below cover 90% of cases.
Common issues:
ollama serve in a separate Terminal windowdocker system prune -f) and Ollama modelsIf none of the above solves it, email support at support@ostler.ai.