SSean Nguyen
← All demos
Workflow automationDocument processingSystem integrations

Studio Knowledge & Asset Desk

Studio docs in → keyword chunk search → grounded answer with citations + permission gate → OneDrive/DAM next actions.

Grounded answers with citations and permission-aware next actions for review.

Best for: Creative studios, design ops, marketing teams using Adobe CC + OneDrive

Tom tat: Tha tai lieu studio, hoi cau hoi, nhan cau tra loi kem trich dan. Phan quyen viewer/editor.

30-second recap

  • Drop brand docs, ask a question, get a cited answer in seconds.
  • Permission gate (viewer vs editor) shows data privacy by design.
  • Mock retrieval now, vector search later, same UI, no infra lock-in.
  1. Paste or edit the 3 mock docs (Brand Guidelines, Summer Campaign, Asset Index).
  2. Ask a question and pick viewer or editor role.
  3. Run studio search: keyword chunks are scored, LLM drafts a cited answer.
  4. Copy the answer or check citations and next actions.
Try the demoLive AI only · production models

Studio Knowledge & Asset Desk

The interactive console loads on demand. Upload your own file or edit the example input. Every result comes back live from the API.

The problem

Designers hunt for brand guides, past proofs, and asset locations across OneDrive, email, and local drives. Knowledge lives in heads, not in a searchable system.

How it works

Paste docs (or OneDrive mock) → chunk 400 chars → keyword score top 4 → LLM drafts grounded answer with citations → viewer/editor permission gate → next actions (archive to DAM, share, tag in OneDrive).

System design

Mock retrieval (no DB) + LLM structured JSON answer with citations. Same surface as future pgvector RAG. Queue with 2 concurrent + per-IP serial + 3 retries, Telegram alert on exhaustion.

Human review & control

Viewer needs editor sign-off to publish externally. Editors can approve directly. Every answer shows chunks and confidence.

Integration scope

OneDrive / SharePoint shaped input, Adobe CC task and DAM archive as next actions, Slack/Teams fan-out (mock).

Stack

  • LLM APIs
  • Mock RAG
  • OneDrive/M365
  • Adobe CC
  • Next.js

What the live demo shows

Live grounded answer with citations over keyword-scored chunks. Permission gate enforced in UI.

Current limitations

Retrieval is keyword scoring over mock docs, not vector search. No OneDrive, SharePoint, or DAM connection. Vector RAG is planned, not live.