Show HN: Timelinize – Privately organize your own data from everywhere, locally

Original Article ↗ Hacker News Discussion ↗

Overall reaction & envisioned uses

  • Many commenters are enthusiastic; several say they’ve wanted this for years or built rough versions (Excel timelines, private Mastodon, homebrew dashboards).
  • Common use cases: personal history/journaling, replacing scattered photo/location tools, digital forensics, “what was I doing two weeks ago?”, and tying together life events, media, and notes.
  • People see strong synergies with finance tracking, bank feeds, car telemetry, and local LLMs for a private “personal assistant.”

Data import, Google Takeout, and real‑time updates

  • Biggest friction point: Google Takeout is cumbersome and non‑realtime; 2FA and frequent re‑auth block automation.
  • Current pattern is occasional bulk Takeouts (once or twice a year).
  • Ideas: scheduled Takeouts to Google Drive plus rclone; phone companion app that streams new photos/locations; Syncthing into a watched folder; “drop zone” directory and cron‑based imports.
  • Past attempts to use Google Photos API failed due to stripped metadata, rate limits, and “nerfed” data; Takeout is seen as the only way to get near‑originals.

Storage model, duplication, and backups

  • Timelines are just folders on disk, with SQLite for indexing; portable across OSes.
  • Author intentionally copies data into the timeline rather than only indexing external sources; duplication is framed as a feature for availability and archival.
  • Some push for decoupling index and storage (e.g., reuse Immich/Ente/Nextcloud libraries, dedup across apps); response is that this complicates reliability and is out of scope for now.
  • S3/minio for media is requested; SQLite‑on‑S3 is rejected as slow/fragile, though offsite storage of media or DB backups is considered.

Extensibility and integrations

  • Data sources implement a simple two‑method interface; third‑party sources (Immich, FindPenguins, Firefox history, HPI exporters, Signal backups, etc.) are encouraged.
  • An import HTTP API is planned so external scripts can push arbitrary data.
  • Architecture is already client–server with a JSON API, so alternative frontends are possible.

Privacy, “self‑surveillance,” and hosting model

  • Strong preference for local/home hosting, often behind WireGuard/Tailscale; remote hosting is viewed as incompatible with strong privacy.
  • Some call the idea “surveillance software”; others argue that doing this for oneself, open‑source and self‑hosted, is fundamentally different from corporate tracking.
  • Comparisons are made to Microsoft’s Recall: idea interesting, but people distrust big vendors and want a self‑governed equivalent.

Timeline focus vs other data views

  • Question raised: many data types (bookmarks, contacts, notes, ratings, ebooks, Steam library) are typically organized by context/category, not time.
  • Response: everything still has temporal aspects (when something was added, used, or changed); contacts are modeled as “entities” with attributes and time‑bounded relationships.
  • Some imagine richer, non‑temporal views in the future, but the core conceptual lens remains chronological.

Technical choices and UX feedback

  • Implemented in Go; distributed as single binaries; SQLite used internally. There’s interest in more DB‑centric storage and temporal schemas, but complexity is a concern.
  • Jquery‑like $ usage is done via a tiny shim on top of vanilla JS.
  • GPU/Apple Silicon is recommended for thumbnailing, transcoding, and semantic embedding features; M1 support is untested but likely workable.
  • One Windows path/URI bug is reported and quickly fixed; installation packaging and “setup.exe”‑style installers are requested.
  • Debate over screenshot style: some feel real data with blur looks unprofessional; others think fake data looks worse and prefer obfuscated real timelines.

Roadmap: LLMs, sharing, finance, and context enrichment

  • Local LLM integration is on the roadmap, with some suggesting a staged model (build timeline first, then selectively expose to an LLM).
  • Planned features include sharing time/geo‑based slices with others, more financial exploration pages, and richer document support.
  • Entity‑aware mapping and augmentation with public datasets (weather, news) are already envisioned to give more context to events.
  • Community suggests many feature directions; there’s clear interest in building a broader ecosystem around the core timeline engine.