| --- |
| license: apache-2.0 |
| language: |
| - en |
| base_model: Qwen/Qwen2.5-1.5B-Instruct |
| tags: |
| - text-generation |
| - instruction-following |
| - crm |
| - tether business |
| pipeline_tag: text-generation |
| inference: true |
| --- |
| |
| # Derry 1.5B Instruct (Tether Business Backend Brain) |
|
|
| Derry is a highly optimized, lightweight 1.5 Billion parameter instruction-following model built to handle real-time text summarization and CRM structuring for the **Tether** social platform. |
|
|
| It is based on the robust `Qwen2.5-1.5B-Instruct` architecture, delivering lightning-fast text processing speeds making it ideal for live mobile applications. |
|
|
| ## Intended Use Case (Tether Business App) |
| Derry reads unstructured live chat interactions and crunches them into clean, structured data blocks for business management tracking. |
|
|
| ### Expected Input Format |
| The model expects data formatted via the standard Qwen chat template system: |
| - **System Prompt**: Defining Derry's persona and summary guidelines. |
| - **User Prompt**: The raw customer interaction transcript. |
|
|
| ### Expected Output Format |
| - **Core Problem**: (1 sentence summary) |
| - **Customer Sentiment**: (Frustrated / Satisfied / Neutral) |
| - **Actions Taken**: (Bullet points of actions) |
| - **Follow-up Required**: (Pending tasks) |
|
|
| ## How to Test via API |
| You can send raw POST requests containing a json payload with `inputs` directly to the Hugging Face Inference API endpoint linked to this repository. |
|
|
