Select Active Embedding Model

Overview

Flow ID: select-embedding-model
Category: Model Management
Estimated Duration: < 5 seconds
User Role: Power User
Complexity: High (Consequences)

Purpose: Choose the model used to create new vector embeddings. CRITICAL WARNING: This setting only affects newly created Datasets. Existing datasets are permanently locked to the model that created them (to ensure vector space compatibility).

Trigger

What initiates this flow:

• User manually initiates

Specific trigger: Settings > Embedding Models > Active Checkbox.

Prerequisites

Before starting, users must have:

• Detailed understanding of vector spaces
• Justification for switching (e.g., upgrading from all-minilm to nomic-embed for better accuracy)

Step-by-Step Flow

Main Path (Happy Path)

Step 1: Open Settings

• User Action: Settings > Embedding Models.

Step 2: Select New Model

• User Action: Click the radio button or “Make Active” next to a model.
• System Response: “Active Model Updated.”

Step 3: Verification

• System Awareness: Future “Blockify” or “Chunking” jobs will default to this model.

Error States & Recovery

Error 1: Deleting Active Model

Cause: User tries to delete the currently selected embedding model.
User Experience: “Cannot delete active model. Please select another first.”
Recovery: Switch active status, then delete.

Pain Points & Friction

1. “Can I switch the model for an existing dataset?”: No. This is mathematically impossible without re-processing all documents.
   • Mitigation: Lock the embedding model dropdown in Blockify Job Creation when “Add to Existing” is selected.

Technical References

• src/localdb/config.js (active_embedding_model)