Fabra WebUI

TL;DR: A visual interface for exploring your Feature Store and Context Store. Launch with fabra ui your_features.py.

Quick Start

# Install Fabra
pip install fabra-ai

# Launch the UI
fabra ui examples/demo_context.py

The UI opens at http://localhost:3001 with the API server at http://localhost:8502.

---

Demo Mode vs Production Mode

Fabra automatically detects when you're running in demo mode (local development) vs production mode.

Demo Mode Indicators

When the API returns is_demo_mode: true, it means:

Indicator	Meaning
InMemoryOnlineStore	Feature data is stored in memory — lost on restart
DuckDBOfflineStore	Using embedded DuckDB — local development only
Mock retrievers	Retrievers without index= parameter use keyword matching, not vector search

Example API Response (Demo Mode):

{
  "file_name": "demo_context.py",
  "online_store_type": "InMemoryOnlineStore",
  "offline_store_type": "DuckDBOfflineStore",
  "is_demo_mode": true,
  "demo_warning": "Demo mode active. InMemoryOnlineStore: data lost on restart; DuckDBOfflineStore: local development only. Set FABRA_ENV=production for persistent storage.",
  "retrievers": [
    {
      "name": "search_docs",
      "backend": "mock",
      "is_mock": true,
      "index_name": null
    }
  ]
}

Production Mode

Switch to production mode by setting:

export FABRA_ENV=production

This uses:

• PostgresOfflineStore — persistent storage with pgvector
• RedisOnlineStore — fast, persistent feature cache
• Real vector search retrievers with index= parameter

Example API Response (Production Mode):

{
  "online_store_type": "RedisOnlineStore",
  "offline_store_type": "PostgresOfflineStore",
  "is_demo_mode": false,
  "demo_warning": null,
  "retrievers": [
    {
      "name": "search_docs",
      "backend": "pgvector",
      "is_mock": false,
      "index_name": "knowledge_base"
    }
  ]
}

---

API Authentication

The WebUI API supports optional API key authentication.

Enabling Authentication

Set the FABRA_UI_API_KEY environment variable:

export FABRA_UI_API_KEY=your-secret-key
fabra ui examples/demo_context.py

Using the API with Authentication

Include the X-API-Key header in requests:

curl -H "X-API-Key: your-secret-key" http://localhost:8502/api/store

Without the key (when enabled):

{
  "detail": "Invalid or missing API key"
}

When to Enable Authentication

• Development: Leave unset for easy testing
• Shared environments: Set a key to prevent unauthorized access
• Production: Always set when exposing the UI publicly

---

API Endpoints

`GET /api/store`

Returns information about the loaded Feature Store.

Response:

{
  "file_name": "features.py",
  "entities": [...],
  "features": [...],
  "contexts": [...],
  "retrievers": [...],
  "online_store_type": "InMemoryOnlineStore",
  "offline_store_type": "DuckDBOfflineStore",
  "is_demo_mode": true,
  "demo_warning": "..."
}

`GET /api/features/{entity_name}/{entity_id}`

Fetch feature values for a specific entity.

curl http://localhost:8502/api/features/User/user_123

Response:

{
  "user_tier": "premium",
  "user_engagement_score": 72.5,
  "support_priority": "high"
}

`POST /api/context/{context_name}`

Assemble a context with the given parameters.

curl -X POST http://localhost:8502/api/context/chat_context \
  -H "Content-Type: application/json" \
  -d '{"user_id": "user_123", "query": "how do features work?"}'

Response:

{
  "id": "ctx_018f3a2b-...",
  "items": [...],
  "meta": {
    "token_usage": 150,
    "freshness_status": "guaranteed"
  },
  "lineage": {
    "context_id": "ctx_018f3a2b-...",
    "features_used": [...],
    "retrievers_used": [...]
  }
}

`GET /api/graph`

Generate a Mermaid diagram of the Feature Store.

Response:

{
  "code": "graph LR\n    ..."
}

---

Understanding Mock Data

The examples/demo_context.py file uses mock data for demonstration:

Mock Features

Features use deterministic hash-based values:

@feature(entity=User)
def user_tier(user_id: str) -> str:
    return "premium" if hash(user_id) % 2 == 0 else "free"

This ensures consistent values across restarts without a database.

Mock Retrievers

The demo retriever uses keyword matching instead of vector search:

MOCK_DOCS = {
    "features": ["Features in Fabra are defined using..."],
    "context": ["Context assembly in Fabra uses..."],
}

@retriever(name="demo_docs", cache_ttl=timedelta(seconds=300))
async def search_docs(query: str, top_k: int = 3) -> List[dict]:
    # Simple keyword matching - not real vector search
    if "feature" in query.lower():
        docs = MOCK_DOCS["features"]
    # ...

Real Vector Search

For production, use the index= parameter:

@retriever(index="knowledge_base", top_k=5)
async def search_docs(query: str):
    pass  # Auto-wired to pgvector

---

Startup Warnings

When loading a module with InMemoryOnlineStore, you'll see:

UserWarning: Using InMemoryOnlineStore - data will be lost on restart.
Set FABRA_ENV=production for persistent storage.

This is expected in development. To suppress:

import warnings
warnings.filterwarnings("ignore", message="Using InMemoryOnlineStore")

---

Architecture

┌─────────────────┐     ┌──────────────────┐     ┌─────────────────┐
│  Next.js UI     │────▶│  FastAPI Server  │────▶│  Your Module    │
│  (port 3001)    │     │  (port 8502)     │     │  (features.py)  │
└─────────────────┘     └──────────────────┘     └─────────────────┘
                                │
                                ▼
                        ┌───────────────────────────────────┐
                        │         FeatureStore              │
                        ├───────────────┬───────────────────┤
                        │ Online Store  │  Offline Store    │
                        │ (InMemory/    │  (DuckDB/         │
                        │  Redis)       │   Postgres)       │
                        └───────────────┴───────────────────┘

---

Troubleshooting

"No store loaded" Error

The module must contain a FeatureStore instance:

from fabra.core import FeatureStore
store = FeatureStore()  # Required!

Data Disappears on Restart

You're using InMemoryOnlineStore. Either:

1. Accept this for development (it's fine!)
2. Switch to production mode: FABRA_ENV=production

Authentication Failures

Check that X-API-Key header matches FABRA_UI_API_KEY:

# Set the key
export FABRA_UI_API_KEY=secret123

# Use the same key in requests
curl -H "X-API-Key: secret123" http://localhost:8502/api/store

---

Next Steps

• Context Store — Deep dive into context assembly
• Quickstart — Get started with Fabra
• Context Accountability — Full lineage tracking