Overview
Flow ID: embedding-model-upload
Category: Model Management
Estimated Duration: 5-15 minutes
User Role: All Users
Complexity: Simple
Purpose: Upload an embedding model that converts text into numerical vectors for semantic search. Required for dataset query functionality and creating searchable datasets from documents.
Related Flows
- Upload Large Language Model - Similar process
- Select Active Embedding Model - Use uploaded model
- Upload New Dataset - Requires embedding model
- Create New Blockify Job - Requires embedding model
Prerequisites
Before starting, users must have:
- Application running
- Embedding model file (compatible format)
- Sufficient disk space (1-5GB typically)
Step-by-Step Flow
Main Path (Happy Path)
Step 1: Navigate to Settings
- User Action: Click "Settings" in navigation
- System Response: Settings page loads
- UI Elements Visible: Settings tabs including "Embedding Models" or similar
Step 2: Access Embedding Models Section
- User Action: Click "Embedding Models" tab
- System Response: Embedding models list displays
- UI Elements Visible:
- List of existing embedding models (if any)
- "Add Model" or "Upload Model" button
- Model details: name, type, path
Step 3: Click Add Model
- User Action: Click "Add Model" button
- System Response: Upload modal appears
- UI Elements Visible:
- Upload modal with file selector
- Model name field
- Model type (auto: "Embeddings")
- Save/Cancel buttons
Step 4: Select Model File
- User Action: Click "Choose File", navigate to model file, select, click "Open"
- System Response:
- Filename appears
- Model name auto-populates
- Model type auto-set to "Embeddings"
Step 5: Confirm and Upload
- User Action: Review name and type, click "Save"
- System Response:
- Upload begins
- Progress bar shows advancement
- May take 5-15 minutes for large models
Step 6: Upload Completes
- User Action: Wait for upload to finish
- System Response:
- Success message
- Modal closes
- Model appears in list
- UI Elements Visible: New embedding model in models list
Final Step: Embedding Model Available
- Success Indicator:
- Model in list
- Can be selected for datasets
- Can be used for blockify jobs
- System State Change:
- Embedding model stored
- Available for selection
- Can create embeddings
- Next Possible Actions:
- Select as active embedding model
- Create dataset or blockify job
- Upload additional models
Error States & Recovery
Error 1: Incompatible Format
Cause: File not compatible embedding model format
User Experience:
- Error: "Incompatible format" or "Not an embedding model"
- Upload fails
Recovery Steps:
- Verify file is embedding model (not LLM)
- Check compatible formats
- Obtain correct model file
- Retry upload
Error 2: Model Loading Fails
Cause: Corrupted file or insufficient resources
User Experience:
- Upload succeeds but model won't load for use
- Error when trying to use model
Recovery Steps:
- Delete model
- Re-download model file
- Verify file integrity
- Re-upload
Version History
| Date | Version | Author | Changes |
|---|---|---|---|
| 2025-10-04 | 1.1 | Iternal Technologies | Initial comprehensive documentation |
Notes
What Are Embedding Models: Embedding models convert text into numerical vectors (arrays of numbers) that capture semantic meaning. Required for dataset search functionality.
Common Embedding Models:
- Jina Embeddings (general purpose)
- BGE models (high quality)
- MiniLM models (fast, smaller)
Best Practices:
- Upload embedding model before creating datasets
- Use same embedding model for all datasets if possible
- Smaller embedding models process faster
- Larger models may provide better search quality
Common User Questions:
- "What's the difference from chat models?" - Embeddings for search, LLMs for conversation
- "Do I need this if I don't use datasets?" - No, only needed for dataset query features
- "Can I use multiple embedding models?" - Yes, different datasets can use different models
- "Which embedding model is best?" - Depends on language and accuracy needs; start with Jina
Trigger
What initiates this flow:
- User manually initiates
Specific trigger: User needs embedding model for dataset features, typically because:
- Want to use dataset query in conversations
- Creating blockify job that requires embeddings
- System prompts for missing embedding model
User Intent Analysis
Primary Intent
Add embedding model to enable dataset search and document processing capabilities.
Secondary Intents
- Enable RAG (dataset query) features
- Prepare for blockify jobs
- Test different embedding models