docs: add model card with intent schema and training plan

README.md

@@ -0,0 +1,55 @@
+---
+tags:
+- text-classification
+- intent-detection
+- gcc
+- e-commerce
+- agentic-commerce
+- ocg-dubai
+- gulf-retail
+language:
+- en
+- ar
+pipeline_tag: text-classification
+license: mit
+---
+
+# GCC Intent Classifier v2
+
+> Built by [OCG Dubai](https://ocg-dubai.ae) — Agentic Commerce APIs for the GCC
+
+A text classification model for detecting customer intents in GCC e-commerce conversations. Supports English and Arabic queries across common retail interaction patterns.
+
+## Intents
+
+| Intent | Example |
+|--------|---------|
+| `product_search` | "Show me gold jewelry under 500 AED" |
+| `order_status` | "Where is my order?" |
+| `return_request` | "I want to return this item" |
+| `price_inquiry` | "How much is the Samsung S24?" |
+| `complaint` | "The delivery was late" |
+| `recommendation` | "What do you suggest for Eid gifts?" |
+| `store_info` | "What are your opening hours?" |
+| `payment_help` | "Can I pay with Tamara?" |
+
+## Usage
+
+```python
+from transformers import pipeline
+
+classifier = pipeline("text-classification", model="GencoDiv/intent-classifier-gcc-v2")
+result = classifier("I want to return the shoes I bought yesterday")
+# [{'label': 'return_request', 'score': 0.95}]
+```
+
+## Status
+
+⚠️ **Model weights pending upload.** This card documents the intended architecture and training plan. Model files will be uploaded after fine-tuning on GCC e-commerce conversation data.
+
+## Training Plan
+
+- **Base model:** `distilbert-base-multilingual-cased`
+- **Fine-tuning data:** GCC e-commerce customer service logs (anonymized)
+- **Languages:** English + Gulf Arabic
+- **Target accuracy:** >90% on held-out test set