v0.preview.89

v0.preview.89 - 2026-08-26

Added

• GLM 5.3 Flash on Z.ai (glm-5.3-flash) and OpenRouter (z-ai/glm-5.3-flash), with multimodal input, a 1M context window, and graded low/high/max reasoning effort; GLM 5.3 remains the default Z.ai model

Changed

• Run arguments supplied alongside a subcommand are now rejected instead of silently ignored (ante --model X catalog errors)
• Tool filters take one value per occurrence — --tools Read,Write or a repeated flag; space-separated lists now error, and --allowed-tools is removed in favor of --tools
• --profile is process-scoped: it rides before or after any subcommand, and serve, gateway, doctor, update, rage, and external apps honor it instead of reading default settings; an unknown profile name falls back to settings.json rather than creating one
• WebFetch drops its unused prompt argument — Ante has no extraction sub-model, so the interface is URL-only

Fixed

• WebFetch identifies itself with an ante/<version> User-Agent — GitHub's API rejects UA-less requests with an opaque 403 — and a failed request now carries up to 2KB of the response body instead of a bare status line
• ante serve handles a client that closes stdout as a logged disconnect instead of hanging, stops reading stdin after an explicit Shutdown, and bounds runtime teardown