manifest v5.50.0

✨ Minor Changes

• abd9dea: Promote LM Studio to a first-class provider tile alongside Ollama. In the self-hosted version, clicking the LM Studio tile opens a dedicated detail view that probes http://{localLlmHost}:1234/v1, lists the loaded chat models, and connects them in one click — no URL typing, no manual pricing. A refresh icon re-probes after the user swaps models on the host. When the probe fails inside Docker, the detail view surfaces a two-path "Fix it" card: GUI toggle (LM Studio → Developer → "Serve on Local Network") and CLI one-liner (lms server start --bind 0.0.0.0 --port 1234 --cors) with a Copy button and a "one-time setup" reassurance. Embedding / reranker / moderation models are filtered out of the discovered list so they don't leak into the routing tier picker. Local-only tiles are disabled in cloud mode with an "Only available on self-hosted Manifest" hint. The customProviderLogo lookup now normalizes names against the shared registry, so an "LM Studio" custom provider shows the real logo everywhere (tier cards, Messages table, Model Prices, Fallback picker).