v0.10.1 Release Notes

Released 2026-03-09

v0.10.0 -> v0.10.1

Summary

Refactors CLI provider setup to use an embedded model registry for interactive selection, and standardizes temperature to 0.2 across all providers.

翻譯

重構 CLI Provider 設定流程，使用內嵌模型登錄檔實現互動式選擇，並將所有 Provider 的 temperature 統一設定為 0.2。

Changes

REFACTOR

• Refactor cmd/cli/addProvider.go to delegate model selection to selectModelFromList, using the provider registry to display an interactive promptui.Select list with descriptions before falling back to manual input
• Propagate model description through addCopilot, addCompat, and getModelName return values so upsertModel receives accurate per-model descriptions
• Add confirmation prompt when an existing API key is detected, preventing accidental overwrite

翻譯

• 重構 cmd/cli/addProvider.go，將模型選擇委派給 selectModelFromList，使用 Provider 登錄檔顯示帶有描述的互動式 promptui.Select 列表，再退回手動輸入
• 透過 addCopilot、addCompat、getModelName 的回傳值傳遞模型描述，使 upsertModel 能取得精確的每模型描述
• 新增偵測到既有 API Key 時的確認提示，防止意外覆蓋

ADD

• Add internal/agents/provider/provider.go — provider model registry exposing Get, Models, Default, InputBytes, OutputTokens APIs backed by embedded JSON configs
• Add embedded model definitions for Claude, Copilot, Gemini, NVIDIA, and OpenAI with input/output context limits and descriptions (internal/agents/provider/embed/*.json)

翻譯

• 新增 internal/agents/provider/provider.go — Provider 模型登錄檔，透過內嵌 JSON 設定提供 Get、Models、Default、InputBytes、OutputTokens API
• 新增 Claude、Copilot、Gemini、NVIDIA、OpenAI 的內嵌模型定義，包含 Input/Output Context 限制與描述（internal/agents/provider/embed/*.json）

UPDATE

• Set temperature=0.2 uniformly across all provider send implementations (Claude, Copilot, Gemini, NVIDIA, OpenAI, Compat)

翻譯

• 在所有 Provider 的 send 實作中（Claude、Copilot、Gemini、NVIDIA、OpenAI、Compat）統一將 temperature 設為 0.2

---

Files Changed

---

Generated by SKILL