v0.1.3

v0.1.3 — channel UX hotfix

Bug fixes

• channels: clicking the Test button when no Test model is configured no longer returns a cryptic "请填写测速模型后再试" error; Prism now falls back to GET /v1/models to verify connectivity and key validity. Providers without a model-list endpoint (Bedrock, Vertex AI) surface a clear "please fill the Test model field" hint instead.
• channels: every model-list-fallback failure now surfaces one actionable hint ("please fill the Test model field") with the upstream's real error in parentheses, so a 401 invalid-key is still visible.
• capture: upstream error responses with an empty body now render as (empty response body) in the Logs panel instead of a blank string.
• tray (macOS): the status-bar template icon no longer has visible stair-stepping ("毛边") on Retina displays. The PNG is rendered at 4× supersampled resolution and downsampled with LANCZOS, and each NSImageRep is tagged as 18pt so AppKit stops double-resampling the bitmap.

New features

• channels: a "Fetch from Base URL" button next to the Models textarea probes GET /v1/models on the unsaved editor form, merges the upstream model list with whatever names you have already typed, and deduplicates. Useful for populating a one-api / OpenRouter / Ollama channel in one click.
• channels: a help line under Models explains that custom client aliases (e.g. cursor-opus-4-7) must appear in this list for the distributor to route requests, and points to Model Mapping for name translation.

Internal

• appVersion constant now tracks the git tag (was stuck at 0.1.0 since the initial release; users on 0.1.1/0.1.2 binaries saw 0.1.0 in the UI footer).