The model is designed to understand customer queries, identify issue categories, determine ticket priority, and route support requests to the appropriate departments automatically.

It uses QLoRA-based parameter-efficient fine-tuning with the Unsloth framework, enabling efficient training on limited GPU resources such as NVIDIA T4 GPUs available in Google Colab.

The system helps reduce manual ticket handling, improves response efficiency, and enhances customer support workflows.

Developed by: ABI
Model type: Causal Language Model
Language(s): English
License: Apache-2.0
Finetuned from model: unsloth/Llama-3.2-1B-Instruct-bnb-4bit

Model Sources

Base Model: https://huggingface.co/unsloth/Llama-3.2-1B-Instruct-bnb-4bit
Framework: https://docs.unsloth.ai/
Transformers Library: https://huggingface.co/docs/transformers/index

Uses

Direct Use

This model can be used for:

Customer ticket classification
Automated ticket routing
Support query understanding
Priority prediction
Customer intent detection
Helpdesk workflow automation

Example use cases:

Routing billing issues to finance teams
Detecting technical support requests
Identifying urgent customer complaints
Categorizing customer queries automatically

Downstream Use

The model can be integrated into:

FastAPI backend services
AI-powered helpdesk systems
CRM platforms
Customer support chatbots
Enterprise support automation pipelines

Out-of-Scope Use

This model is not intended for:

Legal or financial decision making
Medical advice
Sensitive customer data analysis without safeguards
Multilingual support without additional fine-tuning

Human oversight is recommended for critical customer interactions.

Bias, Risks, and Limitations

The model performance depends heavily on training data quality.
Incorrect ticket routing may occur for ambiguous customer queries.
Domain-specific terminology outside training distribution may reduce accuracy.
The model may generate inconsistent outputs for extremely short or unclear tickets.

Recommendations

Use confidence thresholds before automatic routing.
Include fallback manual review systems.
Continuously retrain using real customer ticket data.
Combine with retrieval systems for better contextual understanding.

How to Get Started with the Model

python

import unsloth
from unsloth import FastLanguageModel
from peft import PeftModel
import torch

base_model = "unsloth/Llama-3.2-1B-Instruct-bnb-4bit"

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name=base_model,
    max_seq_length=2048,
    dtype=None,
    load_in_4bit=True,
)

model = PeftModel.from_pretrained(
    model,
    "/content/support_router_model"
)

prompt = f"""<|begin_of_text|><|start_header_id|>system<|end_header_id|>

You are a customer support routing AI.
Return ONLY valid JSON.

<|eot_id|><|start_header_id|>user<|end_header_id|>

Classify this support ticket:

My payment failed twice and I need refund urgently

<|eot_id|><|start_header_id|>assistant<|end_header_id|>
"""

inputs = tokenizer(prompt, return_tensors="pt").to("cuda")

outputs = model.generate(
    **inputs,
    max_new_tokens=80,
    temperature=0.1,
    do_sample=False,
    pad_token_id=tokenizer.eos_token_id
)

response = tokenizer.decode(
    outputs[0], skip_special_tokens=True
)

print(response)

S-ABISHEAK/aegisops-ai-ticket-router-llama3.2-1b

Model Card for Customer Support Ticket Routing Model

Model Details

Model Description