Custom OpenAI base URL — show models from the configured endpoint, not filter by name prefix #1320
Description
Feature Request
Summary
When a custom OpenAI base URL is configured, Pulse should display all models
returned by that endpoint's /v1/models in the model picker, rather than
filtering by known provider name prefixes (gpt-, claude-, deepseek-, etc.).
Current Behaviour
Pulse's model picker uses getProviderFromModelId() to assign each model to a
provider based on its name prefix. When using a custom base URL pointing to a
local AI server (e.g. LocalAI, LM Studio, Ollama OpenAI-compat, or a custom
proxy), models with names like llama3-8b, qwen3.5-27b, or gemma-3-4b are
invisible in the OpenAI dropdown because they don't match any known prefix.
Expected Behaviour
When a custom OpenAI base URL is set, all models returned by that endpoint's
GET /v1/models should appear in the model picker under the OpenAI provider,
regardless of name prefix.