Skip to main content
All functions are deployed to supabase/functions/ and run on the Deno runtime. The base endpoint for calls is: 
https://<PROJECT_REF>.supabase.co/functions/v1/<function-name>

​
inventory-oracle

File: supabase/functions/inventory-oracle/index.ts
Model: gemini-2.5-flash-lite (v1beta API) Analyses 30 days of order history for a product and predicts when stock will run out. Returns urgency level and messages for both customer-facing display and admin restock planning.

​
Request body

{
    productId: string;     // Product UUID to analyse
    currentStock: number;  // Current stock level
}

​
Response

{
    daysUntilOut: number;           // Estimated days until stock = 0
    depletionDate: string;          // ISO date string
    customerMessage: string;        // Short persuasive message for storefront urgency
    adminRecommendation: string;    // Technical restock recommendation
    urgencyLevel: 'low' | 'medium' | 'high' | 'critical';
}

​
Error response

{
    error: string;
    context: 'inventory-oracle';
    gemini_key_present: boolean;
    full_error: string;
}

// Usage from admin
const { data } = await supabase.functions.invoke('inventory-oracle', {
    body: { productId: 'uuid', currentStock: 3 }
});
console.log(data.urgencyLevel); // 'critical'

​
dashboard-intelligence

File: supabase/functions/dashboard-intelligence/index.ts
Model: gemini-2.5-flash-lite (v1beta API) Generates a C-Level executive narrative and health score from dashboard KPIs. Used by getDashboardPulse() in admin-dashboard.service.ts.

​
Request body

{
    action: 'get_pulse';  // Must be exactly 'get_pulse'
    stats: {
        todaySales: number;
        pendingOrders: number;
        lowStockProducts: number;
        totalCustomers: number;
        totalProducts: number;
        totalOrders: number;
        topProducts: { name: string; revenue: number }[];
    };
}

​
Response

{
    narrative: string;      // Max 40 words, cyberpunk corporate tone
    anomalies: string[];    // 2 key alerts
    health_score: number;   // 0–100
}

​
customer-intelligence

File: supabase/functions/customer-intelligence/index.ts
Models: gemini-2.5-flash (Analyst + Sommelier engines) The most complex edge function. Handles multiple actions dispatched via an action field in the request body.

​
Actions

The main AI concierge pipeline. Runs a two-engine Analyst→Sommelier architecture:
  1. Analyst classifies intent and dispatches tool calls
  2. Sommelier generates the empathetic response
Returns capability capsule routing payloads for client-side handling of product search, policy lookup, and cart operations. 
// Request
{
    action: 'concierge_chat',
    query: string;                // User's message
    customerId?: string;
    history?: { role: string; content: string }[];   // Last N messages
    customerContext?: { id: string; [key: string]: unknown };
}

// Response (example product search routing)
{
    requires_client_capsule: true,
    capsule_name: 'product_search_integrity',
    tool_args: { query: string; is_ambiguous: boolean; requires_semantic_expansion: boolean },
    debug: { intent: string; raw_analyst: string; runtime_truth: object }
}

​
parse_admin_intent

Converts an admin natural language command into a structured action. 
// Request
{ action: 'parse_admin_intent', query: string }

// Response
{
    action: 'search' | 'navigate' | 'filter' | 'unknown';
    target: string;
    params: Record<string, unknown>;
    message: string;
}

​
generate_supplier_copy

Generates a professional WhatsApp restock message for a supplier. 
// Request
{ action: 'generate_supplier_copy', productName: string, currentStock: number, sku: string }

// Response
{ message: string }

​
generate_whatsapp_copy

Generates a personalised WhatsApp marketing message for a customer segment. 
// Request
{ action: 'generate_whatsapp_copy', customerId: string, context?: string }

// Response
{ message: string }

​
generate_proactive_insights

Analyses low-stock products and at-risk customers to produce 3 strategic insights. 
// Request
{ action: 'generate_proactive_insights' }

// Response
{
    insights: [
        { type: 'warning' | 'info' | 'success', title: string, description: string }
    ]
}

​
voice-intelligence

File: supabase/functions/voice-intelligence/index.ts
Model: gemini-2.5-flash-lite (v1beta API) Converts a natural language query (text or base64 audio) into a clean search keyword string.

​
Request body

{
    transcript?: string;   // Text input
    audio?: string;        // Base64-encoded audio data
    mimeType?: string;     // e.g. 'audio/webm' (default)
}

​
Response

{
    original: string;       // Original transcript or 'Audio Input'
    searchQuery: string;    // Extracted keywords, e.g. 'mango ice pod'
    isComplex: boolean;     // True if the query needs advanced filtering
}
On error, the function returns HTTP 200 with searchQuery set to the original transcript, so the frontend can fall back to a plain text search without breaking.

​
product-intelligence

File: supabase/functions/product-intelligence/index.ts
Model: gemini-2.5-flash-lite (v1beta API) Generates SEO-friendly product descriptions and marketing copy.

​
Request body

{
    action: 'generate_copy';   // Currently the only supported action
    name: string;              // Product name (required)
    description?: string;      // Existing description to improve (optional)
}

​
Response

{
    description: string;         // Full description, minimum 3 paragraphs
    short_description: string;   // Max 20 words
    tags: string[];              // 5 suggested tags
}

const { data } = await supabase.functions.invoke('product-intelligence', {
    body: { name: 'Aegis Legend 2', action: 'generate_copy' }
});

​
loyalty-intelligence

File: supabase/functions/loyalty-intelligence/index.ts
Model: gemini-2.5-flash-lite (v1beta API) Evaluates a customer’s RFM segment from the customer_intelligence_360 view and generates a personalised discount offer. Creates a real coupon in the coupons table and records it in smart_loyalty_propositions.

​
Request body

{
    customerId: string;   // UUID of the customer
}

​
Response

Returns the full smart_loyalty_propositions row: 
{
    id: string;
    customer_id: string;
    coupon_id: string;
    generated_code: string;         // e.g. 'IA-RECOVER-abc123'
    personalized_message: string;   // Max 30 words, campaign tone
    discount_value: number;         // 10–25 depending on segment
    discount_type: 'percentage' | 'fixed';
    segment_at_generation: string;
    // ... other DB fields
}
Discount guide by segment:
  • Campeón → 10%
  • Leal → 15%
  • En Riesgo / Casi Perdido → 20–25%
  • Nuevo → 15%
Coupons expire in 48 hours and have a minimum purchase of $200 MXN.

​
customer-narrative

File: supabase/functions/customer-narrative/index.ts
Model: gemini-2.5-flash-lite (v1beta API) Generates a 2-sentence executive CRM narrative for a customer, combining customer_intelligence_360 data with the last 5 order amounts and statuses.

​
Request body

{
    customerId: string;
}

​
Response

{
    narrative: string;   // Max 40 words, professional tone
}

const { data } = await supabase.functions.invoke('customer-narrative', {
    body: { customerId: 'user-uuid' }
});
console.log(data.narrative);
// "Este cliente es un motor de ingresos Leal; su ticket promedio
//  es alto pero no ha comprado en 30 días..."

​
bundle-intelligence

File: supabase/functions/bundle-intelligence/index.ts
Model: gemini-2.5-flash-lite (v1beta API) Suggests a complementary product from a compatible category, generates a creative bundle name, and creates a 15% discount coupon valid for 2 hours.

​
Request body

{
    product: {
        id: string;
        name: string;
        price: number;
        category_id: string;
    };
    cartTotal?: number;
}

​
Compatibility rules

// Built-in category slug compatibility map
{
    'mods': ['liquidos', 'coils'],
    'atomizadores': ['liquidos', 'coils'],
    'liquidos': ['coils', 'mods'],
    'vaporizers': ['accesorios-420'],
    'fumables': ['accesorios-420'],
}

​
Response

{
    bundleName: string;           // Creative name, max 4 words
    suggestedProduct: {
        id: string;
        name: string;
        price: number;
        cover_image: string | null;
        category_id: string;
    };
    couponCode: string;           // e.g. 'BUNDLE-abc123'
    discountPercentage: 15;       // Always 15%
}

​
embeddings-processor

File: supabase/functions/embeddings-processor/index.ts
Model: gemini-embedding-001 (v1beta API — intentionally not migrated) Converts a text string into a vector embedding array for semantic search.

​
Request body

{
    text: string;   // Text to embed
}

​
Response

{
    embedding: number[];   // Dense vector from gemini-embedding-001
}
Note: This function deliberately uses the v1beta endpoint (https://generativelanguage.googleapis.com/v1beta/models/gemini-embedding-001:embedContent) because gemini-embedding-001 is only available on the beta API. 
const { data } = await supabase.functions.invoke('embeddings-processor', {
    body: { text: 'vape de sabor mango hielo' }
});
const vector = data.embedding; // number[]

​
Deployment

# Deploy a single function
npx supabase functions deploy inventory-oracle --project-ref <PROJECT_REF>

# Deploy all functions at once
npx supabase functions deploy --project-ref <PROJECT_REF>
Set secrets before deploying: 
npx supabase secrets set GEMINI_API_KEY=your-key-here --project-ref <PROJECT_REF>
npx supabase secrets set SUPABASE_SERVICE_ROLE_KEY=your-key --project-ref <PROJECT_REF>

Build docs developers (and LLMs) love

Get started for freeTalk to us