v0.52.62

Highlights

• Rebased onto upstream (v0.52.61) while keeping the fork's trimmed provider catalog
• Alibaba (Singapore) enabled: served from the token-plan MaaS gateway
  (token-plan.ap-southeast-1.maas.aliyuncs.com)
• Served providers (8): asanti, synthetic, zai, umans, alibaba-singapore,
  opencode-go, opencode-zen, openrouter
• Asanti gateway: served from https://agent.asanti.dev/golem/llm/v1 with an
  auto-discovered model catalog (defaults: asanti-coder / asanti-flash)
• Container build fixed: the image builder now uses Go 1.26, so
  ghcr.io/asantigroup/catwalk publishes again (it had been failing since the 1.26 bump)

Commits

• a237340 fix: build the container image with Go 1.26
• 8fbc0f0 feat: serve the Alibaba (Singapore) provider in the catalog
• 8b84406 fix: use the Asanti gateway LLM endpoint path
• 20f09fb feat: serve Asanti with a runtime-discovered model catalog
• b8e6eb6 chore: drop Zhipu Coding in favor of Z.AI
• d12bd59 chore: refresh OpenRouter and Umans model catalogs
• 9d6430f feat: point Asanti at the agent gateway with auto-discovered models
• a5fda03 chore: align provider catalog with served providers after upstream rebase
• bd5452a fix: mark Deepseek V4 Flash models as reasoning on Umans
• 512456d feat: add reasoning levels and current models to Umans provider
• 6ffbfdf feat: trim provider catalog and add Asanti gateway
• 5bcfa23 feat: remove OpenAI from the provider catalog
• 274d813 feat: add Umans provider updater command
• bf721a2 ci: make ghcr package public after push
• 80ec6e8 ci: grant required permissions to upstream goreleaser jobs when they run
• 43950ac ci: skip upstream-only workflows on the asantigroup fork
• 6a18309 ci: build and push container image to ghcr.io/asantigroup/catwalk
• 4fc6381 feat: add Umans provider