DeepSeek requests fail with 400 error — app sends deprecated model name "deepseek-chat"

[dfordevrim]

When using DeepSeek as the LLM provider for moment detection, all requests fail with a 400 Bad Request error. The app appears to be hardcoded to send the model name deepseek-chat, which DeepSeek's API no longer accepts.

Error message received:

DeepSeek request failed (400 Bad Request): {"error":{"message":"The supported API model names are deepseek-v4-pro or deepseek-v4-flash, but you passed deepseek-chat.","type":"invalid_request_error","param":null,"code":"invalid_request_error"}}

Steps to reproduce:

1. Configure a valid DeepSeek API key in AutoShorts
2. Set DeepSeek as the active LLM provider
3. Import a recording and let the automated pipeline run moment detection
4. Request fails with the 400 error above