44 lines
1.0 KiB
Markdown
44 lines
1.0 KiB
Markdown
|
---
|
||
|
|
language:
|
||
|
|
- en
|
||
|
|
library_name: transformers
|
||
|
|
pipeline_tag: text-generation
|
||
|
|
tags:
|
||
|
|
- support
|
||
|
|
- ticket-routing
|
||
|
|
- causal-lm
|
||
|
|
---
|
||
|
|
|
||
|
|
# support_router_ai
|
||
|
|
|
||
|
|
This repository contains a fully merged causal language model for support-ticket routing.
|
||
|
|
|
||
|
|
The model is intended to classify support requests into a structured JSON format:
|
||
|
|
|
||
|
|
```json
|
||
|
|
{
|
||
|
|
"category": "...",
|
||
|
|
"priority": "...",
|
||
|
|
"team": "...",
|
||
|
|
"needs_human": true,
|
||
|
|
"short_action": "..."
|
||
|
|
}
|
||
|
|
```
|
||
|
|
|
||
|
|
## Notes
|
||
|
|
|
||
|
|
- This is a merged model artifact, not a PEFT adapter.
|
||
|
|
- Inference is expected to run with `transformers` using deterministic generation settings.
|
||
|
|
- Prompting works best when input is passed in a chat-style template with explicit JSON-only instructions.
|
||
|
|
|
||
|
|
## Intended Use
|
||
|
|
|
||
|
|
- Support intake triage
|
||
|
|
- Ticket routing assistance
|
||
|
|
- Priority suggestion for customer support workflows
|
||
|
|
|
||
|
|
## Limitations
|
||
|
|
|
||
|
|
- Output quality depends on prompt format and data used during fine-tuning.
|
||
|
|
- The model can still produce malformed JSON in some cases; a parser fallback is recommended in production.
|