DomainTune: Autonomous Ticket Triage LLM

DomainTune is an enterprise-grade fine-tuned model based on Qwen/Qwen2.5-1.5B. It is specifically optimized to read messy, unstructured software bug reports and support tickets and output strict, deterministic JSON triage decisions.

🎯 Model Capabilities

Given a raw support ticket or GitHub issue, DomainTune extracts:

• priority: P1 (critical outage) | P2 (high severity) | P3 (normal) | P4 (minor/cosmetic)
• category: bug | feature_request | documentation | performance | security | infra
• affected_component: The specific system or module impacted (e.g. auth_service, payment_gateway, database)
• sentiment: neutral | frustrated | urgent | satisfied
• resolution_required: true | false

🚀 Quickstart & Inference

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer

model_id = "jasirjru/DomainTune-Qwen2.5-1.5B-Triage"

tokenizer = AutoTokenizer.from_pretrained(model_id)
model = AutoModelForCausalLM.from_pretrained(
    model_id,
    torch_dtype=torch.float16,
    device_map="auto"
)

# Example ticket
ticket = """Title: Production payment gateway timeout on checkout
Description: Customers in us-east are receiving HTTP 504 errors when processing Stripe payments. Multiple checkout failures recorded in Datadog."""

prompt = f"<|im_start|>system\nYou are a support ticket triage assistant. Output only valid JSON.<|im_end|>\n<|im_start|>user\n{ticket}<|im_end|>\n<|im_start|>assistant\n"

inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
with torch.no_grad():
    outputs = model.generate(**inputs, max_new_tokens=150, do_sample=False)

print(tokenizer.decode(outputs[0][inputs.input_ids.shape[1]:], skip_special_tokens=True))