Document "Draft" Status & AI-Powered Vector Search for Evidence Repository

Context & Problem Statement As we roll out the new Centralized Document Repository, users need a way to stage incomplete or unapproved files. Currently, if a file is uploaded, it is immediately processed. We need a "Draft" status so users can upload work-in-progress evidence without it being prematurely embedded by OCS for policy generation or task completion.

Additionally, as the repository grows, standard keyword search won't be enough. We need to leverage our OCS vector embeddings to allow users to semantically search for documents by title (e.g., searching "employee onboarding" should return the "New Hire Access Register").

Acceptance Criteria (AC)

• Evidence State Machine: Add a Status field to document/evidence uploads with two primary states: Draft and Active.

• Draft Logic & Isolation: * Documents in the Draft state must be visible in the UI but clearly marked with a "Draft" pill/badge.

  • Draft documents MUST NOT be sent to the OCS embedding pipeline and MUST NOT be used for task auto-completion or AI questionnaire answering.

• Publishing Workflow: Add a "Publish" or "Mark as Active" button for Draft documents. Clicking this transitions the status to Active and triggers the OCS vector embedding webhook.

• Semantic Vector Search: * Add a search bar to the Document Upload repository UI.

  • Instead of a standard SQL LIKE query, wire this search bar to the OCS vector database to perform a semantic search against the embedded document titles and metadata.

  • Return search results ranked by vector similarity score.