Upload New Dataset (Manual)

Overview

Flow ID: corpus-upload
Category: Dataset Management
Estimated Duration: 1 minute
User Role: Admin
Complexity: Low

Purpose: Create an empty container for knowledge. This flow creates the logical “Dataset” object (name + embedding configuration) but usually directs the user immediately to a Job to fill it.

Trigger

What initiates this flow:

• User manually initiates

Specific trigger: Datasets > Create New.

Step-by-Step Flow

Main Path (Happy Path)

Step 1: Create

• User Action: Click Create New Dataset.
• System Response: Modal asks for Name and Embedding Model.

Step 2: Configure

• User Action:
  • Name: “Legal Contracts 2024”
  • Model: “nomic-embed-text-v1” (Dropdown)

Step 3: Save

• User Action: Click Create.
• System Response: Dataset created (Empty).
• Auto-Redirect: System prompt: “Dataset created. Do you want to upload files now?” -> Redirects to Create Blockify Job.

Design Considerations

• Empty State: An empty dataset is useless. The UX should heavily bias towards “Create & Fill” rather than just “Create”.

Related Flows

• Create Blockify Job - The primary way to actually fill this.

Technical References

• src/localdb/corpus.js (createCorpus)