Repository navigation
release/v1.2.4 - #692
Merged
Merged
release/v1.2.4#692
Conversation
syzsunshine219
commented
Oct 21, 2026
syzsunshine219
commented
Collaborator
- fix(providers): raise default model call timeout to 300s
- feat(models): add Custom protocols, thinking budget, and image declarations
- Let memory summary and evolution call Bedrock Converse and OpenAI Responses.
- Add official thinking levels for the September model releases.
- Remove the brain icon from the thinking mode toggle.
- Explain custom model capability options on hover.
- Save Custom as its own provider and route each protocol to the matching client.
- Let Custom connection tests follow the selected protocol.
- Send max_completion_tokens for newer OpenAI model names on Custom Chat Completions.
- Hide thinking for OpenAI Chat Completions on newer reasoning models, and warn Custom users to turn it off.
- Send reasoning_effort none for OpenAI Chat Completions when newer reasoning models are called with tools.
- Send reasoning_effort none on Chat Completions for any GPT or o-series model with tools, including Custom when thinking is off.
- Recommend the Responses API for GPT models in the Custom thinking hint, and show 自定义 for that provider in Chinese.
- Give the Custom provider its own puzzle mark so it is no longer shown with the OpenAI logo.
- feat: package agent runtime for npm
- chore: relocate agent npm package output
- fix: sync generated agent npm package version
- fix(chat): keep copy actions on completed historical replies
- fix: reuse degraded development memory service
- fix: forward native progress events over SSE
- fix: 支持 apply_patch 工作区内绝对路径
- fix: add correlated API and model latency diagnostics
- fix: wake exec waits on process completion and buffered output
- fix: locate bundled Codex CLI for hook trust installation
- fix: preserve hook removals and roll back failed Codex trust
- fix: allow models to share identical endpoints
- fix: reuse saved BYOK endpoint in setup connection tests
- fix: apply explicit API model selection to resumed sessions
- fix: 拒绝绝对路径中的父级跳转
- fix: 恢复登录账号的内置模型
- fix: 修复登录账号内置模型配置恢复
- fix: 保留账号模型恢复结果并校验账号一致性
- fix: 保留补丁规范化路径中的文件名空格
- fix: trim API keys when matching pending presets
- fix: build bundled agent and memory runtimes in Memmy
- fix: deduplicate native source captures across account changes
- test: clarify native conversation scope regression name
- fix: collapse duplicate cross-account L1 recall candidates
- style(desktop): place social sign-in below email login
- feat(desktop): refine model thinking controls and custom capabilities
- Let external plugins participate in the agent lifecycle
- feat: build and verify runtime artifacts for all platforms
- Keep installed Memmy CLI directly executable
- feat(auth): enable Google and GitHub social login
- feat: build all Memory runtime platforms with one command
- fix(runtime): skip third-party install scripts when packaging the agent runtime
- fix: 保留迁移后的模型模式并刷新记忆配置
- fix: guard embedding inputs and isolate rejected batches (MEMMY-488)
- fix: delete generated spans with task memory
- fix: emit SHA-256 checksum files for Agent runtime packages
- fix: clarify missing ASR model error
- fix: remove whole-turn deadlines from Memmy API
- fix: cancel Memmy turns only through explicit stop requests
- fix(desktop): render skill markdown and restore slash command suggestions
- fix(desktop): limit command palette height and surface stop in search
- fix(desktop): match slash suggestions by command name prefix
- feat(desktop): complete slash subcommands and model presets
- fix(desktop): expand exact commands and place closest suggestions last
- fix(desktop): match commands and subcommands as flat candidates
- test: distinguish unexpected memory endpoint fixtures
- test(windows): model non-Memory health responses
- fix: 统一 BYOK 显示、测试与保存的凭据来源
- fix: refine embedding failure isolation and empty chunk handling (MEMMY-488)
- fix: 统一设置页凭据校验并保留连续复用关系
- fix: find task spans by their own ownership links
- fix: 恢复记忆模型编辑后的连接重测并隔离卡片请求
- fix: isolate onboarding connection test responses
- test: cover BYOK requests completing after page unmount
- Register DeepSeek-V4.1-Flash-1017 so exact model lookups keep its token limits and vision support.
- fix: close PR 668 merge verification
- fix: 兼容 OpenCode V2 并更新已有记忆插件
- fix(wechat): clarify Memmy Full Disk Access permission copy
- fix: 修复 OpenCode V2 连续对话记忆漏采集
- fix: 退出时取消 OpenCode 采集重试等待
- fix(wechat): retry snapshots after transient source changes
- fix: 补齐 OpenCode 生命周期与迁移记忆采集
- fix: 保持 L3 暂不可用时的 OpenCode 记忆采集
- test: 验证 OpenCode 真实扫描先到时的采集去重
- chore(release): prepare v1.2.4 metadata
- fix(release): align provider lint and model key copy
- fix(installer): let non-C Windows installs continue after an explicit risk confirmation
The OpenAI-compatible provider defaulted to a 120s request timeout, which caused frequent timeouts during evaluation runs where model responses legitimately take longer. Raise the default to 300s. The MEMMY_AGENT_OPENAI_COMPAT_TIMEOUT_S env override is unchanged; only the fallback default changes. Update the two default-timeout assertions accordingly. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
…ations Expose OpenAI Completions/Responses and Custom providers in the model workspace, persist thinking/image capability on Custom presets, and send thinking fields through the chat turn so the backend can map them onto reasoningEffort. Co-authored-by: Cursor <cursoragent@cursor.com>
…ponses. Settings can save those protocols for both memory roles, and the memory runtime now sends the matching request instead of rejecting the assignment or falling through to chat completions. Co-authored-by: Cursor <cursoragent@cursor.com>
GPT-6, Claude Opus 5.5, Gemini 3.8 Flash, and DeepSeek V4.1 Flash were missing from the picker table, so the composer had no effort control for those model ids. Co-authored-by: Cursor <cursoragent@cursor.com>
The label and switch already identify the control, and the icon crowded the composer action bar. Co-authored-by: Cursor <cursoragent@cursor.com>
The thinking and image checkboxes did not say what they change, and the checkbox itself did not show a pointer. Co-authored-by: Cursor <cursoragent@cursor.com>
…ng client. Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
…t Completions. Co-authored-by: Cursor <cursoragent@cursor.com>
…and warn Custom users to turn it off. Co-authored-by: Cursor <cursoragent@cursor.com>
…soning models are called with tools. Co-authored-by: Cursor <cursoragent@cursor.com>
…s model with tools, including Custom when thinking is off. Co-authored-by: Cursor <cursoragent@cursor.com>
…t, and show 自定义 for that provider in Chinese. Co-authored-by: Cursor <cursoragent@cursor.com>
… with the OpenAI logo. Co-authored-by: Cursor <cursoragent@cursor.com>
merge main into zy_npm_1008
…into zy_npm_1008
Zy npm 1008
fix(providers): raise default model call timeout to 300s
…opy-v1.2.2 fix(chat): 修复多轮对话中历史助手回复缺少复制按钮 #666
…s-copy-v1.2.4 fix(wechat): 明确完全磁盘访问授权对象（1.2.4）
fix(wechat): retry snapshots after transient source changes
fix: 兼容 OpenCode V2 并修复已有记忆插件更新（Bug678）
… risk confirmation Per-user Setup cannot set an owner on the shared drive data root, so v1.2.3 disabled Install for every non-C path. Create that directory as the current user, and ask before continuing when ownership still cannot be verified. Co-authored-by: Cursor <cursoragent@cursor.com>
Collaborator
Author
|
Project Harness review handoff marker. Only reviews submitted strictly after this GitHub timestamp qualify. |
…ory-quotes fix: stop macOS native agents on quit and recover History evidence quotes
wechat-message-batch.ts imports fzstd, but App/memmy-agent/package.json did not list it. Local builds resolve it from the root workspace, while agent:runtime:package installs only the agent's own dependencies, so the bundle step failed with: Could not resolve "fzstd". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
fix: declare fzstd for the Memmy agent runtime package
syzsunshine219
marked this pull request as ready for review
October 22, 2026 13:53
syzsunshine219
requested review from
Wang-Daoji,
ZongYue99,
hijzy and
wustzdy
as code owners
October 22, 2026 13:53
syzsunshine219
had a problem deploying
to
release
October 22, 2026 14:01 — with 
GitHub Actions
Failure
GitHub Actions
Failure
syzsunshine219
had a problem deploying
to
release
October 22, 2026 15:51 — with
GitHub Actions
Failure
GitHub Actions
Failure
This branch had an error being deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.