Skip to content

README.md

Latest commit

 

History

History
17 lines (15 loc) · 3.1 KB

README.md

File metadata and controls

17 lines (15 loc) · 3.1 KB
Raw

0.4.2

中文版

  • [2026-08-18] The official DeepSeek registry entries re-price to the updated off-peak list. (details, #172)
  • [2026-08-17] The Kimi K2.6 client merges into the unified kimi_k3 client, branching thinking config and prompt-cache keying by generation. (details, #171)
  • [2026-08-17] The Claude 4.6 client merges into the unified claude5 client, which rejects temperature for the whole family. (details, #171)
  • [2026-08-17] Generic openai_responses and ant_messages protocol clients cover OpenAI, OpenRouter, DeepSeek, Z.AI, and MiniMax; the generic chat client renames to openai_chat. (details, #171)
  • [2026-08-17] GPT-5.6 support: the gpt5_6 client serves GPT-5.4/5.5/5.6, with official and OpenRouter registry pricing. (details, #171)
  • [2026-08-17] UniConfig.fast_mode maps to OpenAI service_tier: "priority" and Anthropic speed: "fast". (details, #171)
  • [2026-08-16] The dev skill specifies the client implementation standard: minimal replay probed against the live API, the reference-client shape, and delta-only streaming. (details, #170)
  • [2026-08-16] Changelog entries carry a metadata block with explicit PR links, and the dev skill gains a capture-serialization rule and a shared-test rule. (details, #170)
  • [2026-08-14] GLM-5.3 support (docs-only, API pre-launch) and the GLM series clients merge into one unified client. (details, #169)
  • [2026-08-14] The Gemini 3 and 3.6/3.7 clients merge into one unified client that rejects temperature for the whole family. (details, #168)
  • [2026-08-13] Gemini 3.7 generation support: gemini-3.7-flash on the shared 3.6-protocol client, with per-model thinking-level clamping. (details, #168)
  • [2026-08-03] Official MiniMax M3 direct Responses API and Token Plan Subscription Key support. (details, #167)
  • [2026-07-24] Gemini 3 clients clamp thinking levels to what each model actually supports — fixes gemini-3.1-pro rejecting ThinkingLevel.NONE with "Thinking level MINIMAL is not supported". (details, #166)