Workspace Technical Specifications

┌──────────────────────────────┐ │ Workspace Editor (UI) │ └──────────────┬───────────────┘ │ ┌───────────────────────┼───────────────────────┐ ▼ ▼ ▼ ┌─────────┐ ┌──────────┐ ┌───────────┐ │ WS │ │ Record │ │ Canvas │ │ relay + │◄──────────►│ + Field │ │ (2D/3D) │ │ snapshot│ │ types │ │ │ └────┬────┘ └────┬─────┘ └─────┬─────┘ │ │ │ ▼ ▼ ▼ ┌────────────────────────────────────────────────────────┐ │ Search / Knowledge layer │ │ │ │ pgvector cosine similarity • 5 embedding providers │ │ Hybrid keyword + semantic • Wiki links + backlinks │ │ Knowledge graph + network view │ └──────────────────────────┬─────────────────────────────┘ │ ┌─────────────────────┼─────────────────────┐ ▼ ▼ ▼ ┌─────────┐ ┌──────────────┐ ┌──────────┐ │ Import │ │ AI Assistant │ │ Export │ │ Notion │ │ + Field │ │ MD / PDF │ │ CSV/XLS │ │ Agents + │ │ HTML/CSV │ │ MD/JSON │ │ Canvas Lens │ │ JSON/ZIP │ └─────────┘ └──────────────┘ └──────────┘

Every layer is native to Aeion OS — the same relational database, event bus, and identity model your CRM and Commerce data already use. That's what lets a Rollup column aggregate live deal values or order totals without an export, a webhook, or a Zapier subscription in between.

Supported sources: Notion, CSV, Excel, Markdown, JSON. Every import runs through the same pipeline regardless of source: parse, map fields, preview, commit.

• Notion: Connect via OAuth, then pick specific pages/databases or your whole workspace. Aeion walks the full page and database tree — including nested sub-pages, synced blocks, and child databases — and preserves the original hierarchy. Supported block types: paragraph, headings, bulleted/numbered lists, to-dos, toggles, code, quotes, callouts, images, video, files, dividers, tables, columns, embeds, bookmarks, link previews, synced blocks, and templates.
• CSV / Excel: Header detection with type inference, plus per-column transforms — date (with a custom format, e.g. MMM DD, YYYY), number, boolean, array (custom delimiter), JSON, or a custom mapping.
• Duplicate handling: choose skip, update, or create for re-imports.
• Dry-run mode: preview exactly what will be created or updated before committing anything.

Live progress tracking:

Job ID: import_1704067200000_abc123 Status: processing Progress: ████████████████░░░░░░░ 67% Total: 1,247 items Processed: 836 Succeeded: 829 Failed: 7 Current: Importing database "Product Roadmap" rows...

Assign an agent — summarize, classify, extract, sentiment, or translate — to any database column. The agent fires automatically whenever a record is saved: it reads a source field, runs it through AI with your custom prompt, and writes the result to a target field. Every run — and its timestamp — is tracked per column.

Example — auto-summarize meeting transcripts:

Source field: rawTranscript Target field: summary Prompt: "Summarize this meeting in 3 bullet points, highlighting key decisions" Result: When a meeting record is saved, the summary column auto-populates.

Agents run in the background with automatic retries, so a slow AI response never blocks saving a record.

Two Notion/Tana-parity capabilities power connected knowledge work:

• Knowledge Graph — pages and records form a backlink/node graph you can visualize and navigate. Surface what links to a doc, traverse relationships, and see the shape of a knowledge base rather than a flat folder tree.
• Synced Blocks — a block of content can be mirrored across many pages; edit it once and every instance updates. Single source of truth for boilerplate, definitions, or status snippets that live in multiple docs.