Glacivis is in closed alpha. Learn what we're building →
Glacivis

Feature · lab records

Lab records that don't go missing.

A single SQLite database on your laptop holds every entity Glacivis knows about. Open it offline, search it offline, edit it offline. The records are yours.

What Glacivis treats as a record

Glacivis ships with a small, opinionated entity model that maps onto how wet labs actually organise work — not a generic document store. Out of the box, you have:

  • Projects — long-running scopes (a grant, a thesis chapter, a pipeline programme).
  • Experiments and sub-experiments — the unit of "what I'm trying to find out". Sub-experiments nest inside experiments and inherit the parent's protocol references.
  • Runs and measurements — what actually happened on the bench, with timestamps, instrument outputs, and conditions.
  • Notes — rich-text Tiptap documents with embedded entities, tables, equations, and inline attachments.
  • Protocols — versioned step lists you can clone into a run.
  • Inventory — reagents, consumables, plasmids, animals, tissue samples, with lot, location, and depletion.
  • Attachments — content-addressed by SHA-256 so the same raw .csv attached to two notes lives once on disk.

How it works

Every entity is a row in a local SQLite file. The Tauri 2 desktop app talks to that database directly through tauri-plugin-sql; there is no intermediate sync server in the private case, and there is no proprietary file format wrapping your data. If Glacivis disappeared tomorrow, the database is a standard SQLite file you can open with sqlite3, DB Browser for SQLite, or a Python script.

Search runs through SQLite FTS5 against the same database. You don't ship your data to a vendor index; the index lives next to the rows it indexes. That keeps search latency under a hundred milliseconds for the typical academic-lab corpus, and it works fully offline.

Mutations go through a universal op-log. Every create, edit, rename, move, delete, and relation-change writes a row to the op_log table with a hybrid logical clock (HLC) timestamp, an actor identity, and a JSON payload of what changed. The op-log is the replay tape: workspace history, per-entity activity feeds, restore-from-history, and the AI-action filter all read from the same source.

Deletion is a tombstone, not a row drop. The entity disappears from the UI, but the op-log retains the deletion event. Restore-from-history brings any entity back for as long as the op-log retention policy keeps the events — universal-version retention default is 10 (set via globalVersionKeepN) and is configurable per workspace.

Hold-to-confirm, never click-and-regret

Every destructive action — delete entry, delete experiment, bulk delete, lock entry — uses a hold-to-confirm button with a visible progress fill. The default hold is 1.2 seconds; bulk operations are a flat 2 seconds; entities with no edits get a 0.3-second short-press because a misclick on an empty draft isn't worth 1.2 seconds of friction. The pattern is consistent across every page so muscle memory works in your favour, not against you.

Why this matters. Lab notebooks fail on lost data more often than on feature parity. A vendor outage, a soft-deleted record nobody can restore, a "your workspace has been archived" email — these are the failure modes wet labs cite when they leave one ELN for another. Glacivis's approach is to make the local SQLite the source of truth, and the cloud the optional caching layer.

What you can do today

  • Create projects, experiments (and sub-experiments), runs, measurements, notes, protocols, and inventory entries.
  • Embed entities inside other entities through Tiptap mentions — every embedded reference is a typed link, not a copy.
  • Attach files of any size; large attachments are content-addressed and de-duplicated across entities.
  • Full-text search across every entity in the active workspace, including attachment metadata.
  • Browse per-entity history; restore any prior version.
  • Filter the activity feed by actor (you, a teammate, AI assistant, or system).
  • Export any subtree to PDF, DOCX, Markdown, or HTML.

What's coming

  • Bulk export to .eln RO-Crate (Q3 2026). Today's per-entity export is a ZIP; the .eln format is the emerging ELN-Consortium standard for portable archives, and we're tracking that spec.
  • FTS-inside-attachments. PDFs and CSVs get text-extracted on upload; the index spans body and attachments together. (Phase 2.)
  • Sign-and-lock entries. Cryptographic record-locking for patent-defense and review workflows. (Phase 2.)
  • Inline scientific content rendering. Sequence and chemistry blocks rendered in Tiptap, not pasted as flat images.

Download Glacivis →