Modelpublic

JisunahameD/asr-qwen-1.5b-support

sourceHugging Faceupdated 4mo agoView on Hugging Face

0likes23downloads

Model Card

🛍️ ASR World Fashion AI Support Agent (Alisha)

🤖 Model Overview

This is a highly specialized, fine-tuned language model based on Qwen2.5-1.5B-Instruct, converted to the highly optimized GGUF format (Q4_K_M). It is specifically trained to act as Alisha, the official 24/7 customer service representative for ASR World Fashion.

Engineered for seamless integration into business automation workflows (such as n8n, WhatsApp, and Facebook Messenger APIs), this model excels at handling natural Bengali and Banglish conversations with extreme accuracy.

Developed by: Jisun Ahamed | InMeTech Base Model: Qwen2.5-1.5B-Instruct Format: GGUF (qwen2.5-1.5b-instruct.Q4_K_M.gguf) Language(s): Bengali (Bangla), Banglish, English

⚡ Key Capabilities & Guardrails

This model was fine-tuned using a complex, rule-based custom dataset. It strictly adheres to enterprise guardrails:

—Strict Price Enforcement: Flawlessly handles bargaining attempts by respectfully denying discounts and presenting fixed-price policies.
—Order Collection Mode: Dynamically identifies missing order information (Size, Color, Phone, Address) and extracts them step-by-step without repeating product descriptions.
—Database Driven: Programmed to request database searches (via vector stores) before providing actual prices or stock availability.
—Zero Hallucination Guardrails: Refuses to guess prices for generic categories (e.g., "pant price") and asks for exact product identifiers.

🚀 How to Run with Ollama (Recommended)

This model is optimized for low-resource VPS deployment (8GB RAM, CPU-only or T4 GPU). The easiest way to host this as an API is via Ollama.

1. Download the Model

bash

wget [https://huggingface.co/JisunahameD/asr-qwen-1.5b-support/resolve/main/qwen2.5-1.5b-instruct.Q4_K_M.gguf](https://huggingface.co/JisunahameD/asr-qwen-1.5b-support/resolve/main/qwen2.5-1.5b-instruct.Q4_K_M.gguf)