Skip to content

v2.0.3 — 2026 model bank, Azure GPT-5.6 tools, Safari Connectivity Check

Choose a tag to compare

View all tags
@lqdflying lqdflying released this 03 Sep 06:12
· 167 commits to main since this release
v2.0.3
5a1a08f

Patch after v2.0.2: the built-in model pickers for the major providers are now 2026-only, plus GLM-5.3, Azure GPT-5.6 tool-chat none, and Connectivity Check fixes for Safari and MiniMax. Upgrade like any patch — pull 2.0.3 or :latest, restart, let migrations run.

Highlights

  • Built-in OpenAI, Anthropic, Google, Azure, Moonshot, Zhipu, and MiniMax cards pruned to 2026 ids (GPT-5.6 Sol/Terra/Luna, Claude 5, Gemini 3.x Flash/Lite, Foundry GPT-5.x, GLM-5.3)
  • GPT-5.5 Pro uses non-streaming Responses; Claude 4.7+ is adaptive-only; Azure GPT-5.6 tool chats send reasoning_effort: none even on custom deployments
  • Connectivity Check survives Safari “Load failed”; MiniMax short checks skip the mobile tokenizer

Upgrade

  • Image: docker.io/lqdflying/chathub:2.0.3 or :latest
  • Database migrations run on container start
  • Optional sidecars unchanged (Code Interpreter, MarkItDown)
  • Docker Deployment and Upgrades

Note

Saved agents that still use removed picker ids keep working against the vendor. Re-pick a 2026 card only if you want the new defaults.

Added

Models

  • OpenAI GPT-5.6 Terra / Luna; Anthropic Claude Opus 5 / Sonnet 5 / Fable 5.1
  • Google Gemini 3.x Flash / Lite and native image ids
  • Azure Foundry GPT-5.6 / 5.5 / 5.4
  • Zhipu GLM-5.3 / GLM-5.3-Flash; DeepSeek V4 Flash Vision Exp

Changed

Models

  • Pre-2026 built-in picker cards removed; compatible clones stay on gpt-5.6-sol and gpt-5.5 only
  • GPT-5.6 Sol quality floor remains high / xhigh / max
  • Claude 4.7+ leftover budgets coerce to adaptive; Fable 5.1 is always-on; fetched GLM-4.7 keeps Preserved Thinking only
  • GLM-5.3 summaries stay at low effort; Vision Exp prices omitted

Fixed

Azure

  • GPT-5.6 Chat Completions with tools send reasoning_effort: none from the trusted catalog id (server and Browser Request), including custom deployment names
  • Azure Connectivity Check no longer forwards responseMode in the request body

Connectivity Check

  • JSON Check path instead of wrapping MiniMax JSON as SSE
  • Safari abort recovered from captured bytes; Check always settles
  • MiniMax Check disables thinking; short checks bypass the mobile tokenizer
  • MiniMax vision detail is low / default / high (not OpenAI auto)

Docs

Full diff: v2.0.2...v2.0.3

Assets 2
Loading