Scri8e reads across Slack, Google Drive and Notion and turns them into one brain your team can ask. Plain answers, every source shown, and a straight warning the moment something goes out of date.
In a 10 to 50 person company, how things work is written down in three places at once: chat threads, wiki pages and shared documents. The same question has several answers of different ages.
Search tools rank results by wording, not by how current they are, and an assistant that reads everything risks showing people files they could not open in the original tool.
Scri8e pulls content from Slack, Notion and Drive into one Postgres database with vector search. Every answer comes with citations, and every result passes a permission check inside the query, so people only see what they could open in the original tool.
Answers stream over Server-Sent Events with numbered citations. Only sources the model actually cited are returned, renumbered in order, with tool, date and link.
Slack private-channel membership, per-file Google Drive sharing and restricted uploads are enforced inside every search query. Missing permission data means hidden, never visible.
Each teammate connects their own Drive and picks folders to index. The file owner's sync decides who can see each file.
Drive and Notion index nothing until folders or pages are picked. Large syncs run in groups of 100 with live progress and a safe cancel between groups.
Cited sources are scored against half-lives by document category (process, decision, reference, ephemeral). A caveat appears when an answer rests on one source or old material.
Admins and, if allowed, members upload PDF or text files with workspace or restricted visibility. Member uploads wait for approval and never appear in answers until approved.
Owner, admin and member roles on every route, with escalation and last-owner guards. Connects, uploads, approvals, role changes and every denied request are written to an audit log.
Admins see how much of the knowledge base is current or aging, the oldest documents, connector state and duplicate or orphaned data.
An admin connects Slack and Notion, and each teammate connects their own Google Drive and picks the folders to index. Anyone can then ask a question in a chat screen. The answer streams in with numbered citations, and each source card shows the tool, the last-edited date and a link back to the original. A small stamp also notes whether the sources are recent.
Setting up a workspace · User Flow
Chose One database for both content and permissions over Separate specialised stores
Access is checked in the same step that finds results, so people never see what they should not.
Chose Meaning-based and keyword search together over A single search method
Team knowledge mixes exact terms with loose wording, and either method alone misses things.
Chose A simple job queue with safe stop points over A heavier job framework
Long imports can be cancelled cleanly without leaving half-finished data behind.
Chose Permissions taken from each tool's own records over One shared connection for the whole team
Only a file's owner sees its full sharing details, so access has to be read from the source.
Chose Model choice controlled through configuration over Models fixed in the code
Switching models becomes a configuration change instead of a rebuild.
Phase 1
Phase 2
Phase 3
Phase 4
Phase 5
3
Sources connected: Slack, Notion and Google Drive
Each via Nango, with source permissions mirrored per document (backend/app/services/nango.py, services/search.py).
302
Automated backend tests
249 unit tests plus 53 integration tests that run against a real Postgres with pgvector (backend/tests/).
49
API routes checked against a written role matrix
A test reads the role dependency off every route and compares it to docs/rbac-matrix.md.
Tell us what you're building and we'll show you how we'd approach it.
Consumer apps and creator economyA studio promised its creator client an AI social app in a month with one developer. We scoped the real need and had four roles working eight days after the first call.
Gaming and PC optimizationA full-stack, design-first e-commerce build for a PC optimization brand, with cart, checkout and key delivery, that reached a client-reported 25% CTR in a low-conversion niche.
Gaming and PC optimizationA Discord bot that pulls the best gaming deals from Amazon through its Product Advertising API and posts them to the community every four hours.