v0.3.0-beta.1 — 首个公测版

Meldwise 0.3 Beta 1

Meldwise 进入 Beta 阶段。Alpha 建立了产品形态和四种核心工作流；Beta 将重点转向稳定性、兼容性、用量透明与更成熟的客户端体验。

四种模式

• 对话：单模型多轮聊天。
• 对比：两个模型独立回答同一个问题。
• 协作：初答 → 审阅 → 主模型或审阅模型综合。
• 辩论：A/B 并行初答 → 并行交叉审阅 → Judge 整理结果，完整一轮最多 5 次模型请求。

阅读体验

新增原生表格、分隔线、删除线、只读任务列表和有界嵌套列表，继续支持 Markdown 基础排版、原生数学、代码块与轻量语法高亮、JetBrains Mono、回答折叠和 Adaptive Composer。

内容渲染保持原生、有界，不使用 WebView。链接保持非联网文本；原始 HTML 和脚本样式内容不会执行。复制和历史保留原始来源文本。

Usage Foundation 与隐私

内部新增可信的运行时请求/用量基础：只记录真实发起的请求和服务商上报的 token 数值。未上报的数据保持未知，不按字符数估算，也不当作 0；部分用量保留覆盖信息。目前没有用量面板、费用显示或用量历史持久化。

凭据与历史继续本机加密保存。需要跨服务商共享可见内容时，会先明确确认，许可限于当前对话与确切的服务商集合。思考内容、凭据、诊断和 Usage 元数据不作为普通跨模型上下文传递。不自动切换模型或服务商，不自动重放失败请求。

边界与后续

这是 Beta 预发行软件，不代表稳定生产版本或 V1 功能完整。Compare 仍单轮，Collaborate 固定三阶段，Debate 固定五阶段；协作与辩论仅整轮手动重试。Markdown/TeX 为实用子集，代码高亮不是编译器分析，中文代码注释使用系统字体回退。Meldwise 无法读取或估算 ChatGPT 套餐剩余额度、百分比、重置时间或费用。SIWC compatibility remains CONDITIONAL，更多设备与长期验证仍在继续。

后续 Beta 工作可能包含 Usage/Cost 展示、历史搜索、重生成与分支、导入/导出和可靠性完善；这些不是本版新增能力。

Meldwise 自 v0.1.0-alpha.2 起采用 GPL-3.0-only。Orcex、JetBrains Mono、STIX Two Math 及其他第三方组件保留各自许可证；服务商名称与商标归各自权利人所有。

---

Meldwise 0.3 Beta 1 — First Beta

Alpha established the product shape and four core workflows. Beta now focuses on stability, compatibility, usage transparency and a more mature client experience.

Four modes

• Chat: multi-turn conversation with one model.
• Compare: two independent answers to one question.
• Collaborate: initial answer → review → synthesis by Primary or Reviewer.
• Debate: parallel A/B answers → parallel cross-reviews → Judge synthesis, with up to 5 model requests per complete round.

Reading improvements

Adds native tables, thematic breaks, strikethrough, read-only tasks and bounded nested lists. Existing Markdown basics, native math, code blocks with restrained lexical highlighting, JetBrains Mono, answer disclosures and Adaptive Composer remain available.

Rendering is bounded and native, without WebView. Links remain non-network text; raw HTML and script-like source remain inert. History and copying preserve original source.

Usage Foundation and privacy

The internal runtime foundation records actual dispatched requests and provider-reported token values only. Missing values remain unknown, never estimated from characters or replaced with zero; partial coverage stays explicit. There is no usage/cost dashboard or persisted usage history yet.

Credentials and history remain locally encrypted. Required cross-provider visible-context sharing asks for explicit confirmation scoped to the conversation and exact provider set. Reasoning, credentials, diagnostics and usage metadata are not ordinary cross-model context. No automatic provider/model fallback or failed-request replay.

Beta boundaries

This is prerelease software, not a stable production release or a complete V1. Compare remains one-shot; Collaborate has three stages and Debate five. Collaborate and Debate retry whole rounds manually only. Markdown/TeX are useful subsets, and highlighting is not compiler analysis; CJK code comments use platform font fallback. Meldwise cannot read or infer remaining ChatGPT plan quota, percentages, reset times or costs. SIWC compatibility remains CONDITIONAL; broader device and long-duration validation is ongoing.

Future Beta work may include usage/cost presentation, History Search, Regenerate/Branch, Import/Export and reliability hardening. These are not available as new features in this release.

Meldwise is GPL-3.0-only starting with v0.1.0-alpha.2. Orcex, JetBrains Mono, STIX Two Math and other third-party components retain their licenses. Provider names and marks belong to their respective owners.

v0.2.0-alpha.4 — 辩论与阅读设置

Meldwise 0.2 Alpha 4

这一版开放了辩论模式，也让多模型回答更容易阅读。欢迎试试新的聊法，分享你遇到的问题和使用感受。

新的聊法：辩论

• A 和 B 先各自回答，再互相审阅，最后由 Judge 整理成一份最终答案。两份初答、两份审阅分别并行进行，Judge 等两份审阅完成后再开始。
• A、B 需要是不同的模型；Judge 可以使用 A、B，或另一个已配置的模型。每个角色可分别设置模型支持的思考偏好。
• 完整一轮最多发起 5 次模型请求，费用和额度按各自服务计算。辩论不保证答案一定更好。
• 辩论历史可以重新打开并继续追问。取消、失败或中断时保留已收到的内容，重试按整轮、原模型快照执行。

更方便地阅读

对比的两份回答、协作的初答和审阅、辩论的两份初答都可以独立收起、展开，并带有轻量过渡动效。收起后仍显示模型与阶段状态，不影响生成或历史内容。协作综合和 Judge 最终答案保持直接展示。

隐私和已有能力

跨服务商共享可见上下文与回答前，需要明确确认；许可只用于当前对话与确切的一组服务商。思考内容、凭据与诊断信息不会作为普通跨模型上下文传递。辩论后续轮次使用用户消息与成功的 Judge 最终回答，中间阶段留在本地历史。

原生公式、代码高亮与 JetBrains Mono、流式输出、取消、本地加密保存和主题设置继续保留。不自动切换模型或服务商，不自动重放失败请求。

仍是 Alpha

• 对比仍为单轮；协作和辩论只支持整轮手动重试，不支持单阶段续跑。
• 辩论保持固定五阶段，不提供胜负评分、任意代理循环或工具调用。
• 代码高亮与数学渲染覆盖常用子集；中文代码注释由系统字体回退。
• ChatGPT 可能受套餐或模型额度限制；SIWC 兼容性仍为 CONDITIONAL。
• 更多设备、长时间使用和完整真机 AndroidTest 覆盖仍在继续验证。

本次本地完整回归：Debug 1453/1453、Release 1453/1453；构建、lint、AndroidTest 编译打包与依赖锁检查通过。验证包已由开发者手动验收。

Meldwise 自 v0.1.0-alpha.2 起采用 GPL-3.0-only。第三方组件遵循各自许可证，服务商名称和商标归各自权利人所有。

---

English

Meldwise 0.2 Alpha 4 opens Debate mode and makes multi-model answers easier to read.

Debate

• A and B answer independently, cross-review, and a Judge puts together the final answer. Each pair of initial answers and reviews runs in parallel; Judge waits for both reviews to finish.
• A and B must be different models. Judge may match either participant or use another configured model. Supported reasoning preferences are selected per role.
• A complete round can make up to 5 model requests, with usage and billing following each provider. Debate does not guarantee a better answer.
• Reopen and continue Debate conversations. Cancellation, failures and interruption preserve received content; explicit retries start a whole new round with the original model snapshots.

Reading controls

Compare answers, Collaborate initial/review stages and Debate initial answers can each collapse or expand with a lightweight transition. Model and stage status stay visible. Collapsing never stops generation or changes history. Collaborate synthesis and Judge final answers remain directly visible.

Privacy and existing capabilities

Sharing visible context and outputs across providers requires explicit confirmation for the current conversation and exact provider set. Reasoning, credentials and diagnostics are never ordinary downstream model context. Later Debate rounds use user messages and successful Judge final answers; intermediate stages stay local.

Native math, code highlighting and JetBrains Mono, streaming, cancellation, encrypted local state and themes remain available. No automatic provider/model fallback or failed-request replay is introduced.

This remains Alpha software: Compare is one-shot; Collaborate/Debate retry whole rounds only. Debate is a fixed five-stage workflow, without competitive scores, arbitrary agent loops or tools. Math/highlighting cover useful subsets; CJK code comments use platform font fallback. ChatGPT may remain subject to plan/model limits, SIWC compatibility remains CONDITIONAL, and broader device/long-running validation is ongoing.

Local full regression: 1453/1453 Debug and 1453/1453 Release. Builds, lint, AndroidTest compile/package and dependency lock checks pass. The developer manually accepted the validation build.

Meldwise is GPL-3.0-only starting with v0.1.0-alpha.2. Third-party components retain their own licenses; provider names and marks remain the property of their owners.

v0.2.0-alpha.3 — 界面与内容优化

Meldwise 0.2 Alpha 3

这次更新主要围绕两件日常小事：聊天时读得更舒服，输入时用得更顺手。模型回答里的代码也有了更清楚的展示。

聊天界面更清爽了

• 对话、对比和协作采用更一致的内容宽度与间距，回答正文更突出。
• 模型信息、思考折叠和消息操作更轻量，历史与设置页面也重新整理了疏密。
• 输入框空闲时保持紧凑，点进去输入或编辑多行文字时会自然展开。

代码更好读，也更方便复制

• 代码块有独立的原生展示，带语言标签和复制按钮。
• 长行可以横向查看，较长内容可以手动展开，复制保留原始代码。
• 加入轻量语法高亮，帮助区分关键字、字符串、数字、注释及部分明确的类型或函数位置。
• 代码块与行内代码使用 JetBrains Mono，中文注释由系统字体补充显示。

公式继续通过 Orcex 原生渲染，聊天、对比和协作的原有使用方式也保持一致。历史和凭据仍加密保存在本机；跨服务商共享的范围仅限可见消息和回答，并会先征求你的确认。发送、重试和更换模型由你决定，重新打开历史会恢复已有内容和状态。

试用前了解一下

Meldwise 仍在 Alpha 阶段，欢迎把使用中遇到的问题和感受告诉我们。

代码高亮适合常见语法的阅读辅助，复杂或难以确定的结构会保持普通代码样式。公式支持常用 TeX 子集。对比目前围绕单个问题展开，协作保持三阶段、重试按整轮进行。Debate 和 Judge 留待后续；ChatGPT 的可用额度取决于套餐与模型。SIWC 兼容性目前为 CONDITIONAL（有条件通过）。

Meldwise 自 v0.1.0-alpha.2 起采用 GPL-3.0-only。Orcex、JetBrains Mono、STIX Two Math 及其他第三方组件继续遵循各自许可证。

---

Meldwise 0.2 Alpha 3

This update focuses on two everyday things: making conversations easier to read and typing more comfortable. Code in model responses also gets a clearer presentation.

A calmer chat interface

• Chat, Compare and Collaborate share more consistent content widths and spacing, keeping the answer front and center.
• Model details, reasoning disclosure and message actions are lighter, with cleaner History and Settings layouts.
• The Composer stays compact at rest and expands naturally when you start typing or edit multiline text.

Code that's easier to read and copy

• Dedicated native code blocks come with language labels and copy controls.
• Long lines scroll horizontally, longer blocks can be expanded, and copying keeps the original code.
• Lightweight highlighting helps distinguish keywords, strings, numbers, comments and some clearly identifiable type or function positions.
• Code blocks and inline code use JetBrains Mono. Chinese comments use platform font fallback.

Math continues through native Orcex rendering, and the existing Chat, Compare and Collaborate workflows stay familiar. History and credentials remain encrypted on your device. Cross-provider sharing is limited to visible messages and answers and asks for your confirmation first. Sending, retrying and changing models remain your choices; reopening history restores existing content and status.

A few things to know

Meldwise is still in Alpha. We'd love to hear what works well and what could feel better.

Highlighting is a reading aid for common lexical patterns; uncertain or complex syntax stays plain. Math covers a useful TeX subset. Compare handles one question at a time, while Collaborate uses three stages and retries whole rounds. Debate and Judge are future directions. ChatGPT access is subject to plan and model limits. SIWC compatibility remains CONDITIONAL.

Meldwise is GPL-3.0-only starting with v0.1.0-alpha.2. Orcex, JetBrains Mono, STIX Two Math and other third-party components retain their own licenses.

v0.2.0-alpha.2 — 社区反馈更新

Meldwise 0.2 Alpha 2

这一版根据 0.2 Alpha 发布后的首轮社区反馈，集中改善内容显示、交互反馈和协作配置。

新增与改进

• 原生渲染常见 $...$、$$...$$、\(...\)、\[...\] 公式，支持分数、上下标、根式等常用表达。未闭合、解析失败或不支持时保留原文，不改写历史。
• 点击反馈改为轻柔的整面色层，减少深色模式下默认大面积水波纹的颗粒感。
• 模型选择器一次只展开一个服务商。对比/协作的 A/B 浏览不再意外改变单模型对话的服务商。
• 协作新增“简洁 / 标准 / 严格”审阅和最终综合角色，可选 A → B → A 或 A → B → B。
• 顶栏下拉箭头紧贴标题，长模型名称单行省略。

审阅强度是 Meldwise 的协作指令策略，不等同于模型思考强度。旧记录默认保持标准审阅、主模型综合（A → B → A）。

技术与行为

使用 Kotlin 2.3.21、AGP 8.13.2、compileSdk 36、Orcex 0.5.0 和 STIX Two Math；targetSdk 仍为 35。

思考内容不作为普通上下文跨模型传递。不会自动切换服务商或模型，也不会自动重放失败请求。本地凭据与历史继续加密保存；对比仍是单轮，协作仍是三阶段。

已知限制

这是 Alpha 软件。数学渲染支持常用 TeX 子集，不是完整 TeX 引擎；协作仍只支持整轮手动重试。Debate / Judge 未实现，对比暂不支持多轮。ChatGPT 可能受套餐或模型额度限制，SIWC 兼容性仍为 CONDITIONAL。更多设备、长期运行和完整真机 AndroidTest 覆盖仍待验证。

Meldwise 自 v0.1.0-alpha.2 起采用 GPL-3.0-only。Orcex（Apache-2.0）、STIX Two Math（OFL-1.1）与其他第三方组件继续遵循各自许可证。服务商名称和商标归各自权利人所有；Meldwise 是独立社区项目，与相关服务商不存在隶属、赞助或背书关系。

---

English — Meldwise 0.2 Alpha 2

This update addresses the first round of community feedback after the 0.2 Alpha release, focusing on content display, interaction feedback and collaboration settings.

New and improved

• Native math rendering for common $...$, $$...$$, \(...\) and \[...\] notation, including supported fractions, scripts and radicals. Incomplete, malformed or unsupported expressions retain their source without rewriting history.
• Subtle full-surface press feedback replaces noisy large radial ripples, especially in dark mode.
• Only one provider group opens at a time. Browsing A/B models in Compare or Collaborate no longer changes the Single Chat provider.
• Concise / Standard / Strict review and selectable final synthesis support A → B → A or A → B → B.
• The dropdown chevron stays beside the title; long model labels are ellipsized on one line.

Review Intensity is a Meldwise orchestration policy, separate from provider thinking effort. Existing records retain Standard review and Primary synthesis (A → B → A).

Technical update and behavior

Kotlin 2.3.21, AGP 8.13.2, compileSdk 36, Orcex 0.5.0 and STIX Two Math. targetSdk remains 35.

Reasoning is not normal cross-model context. There is no automatic provider/model fallback or failed-request replay. Credentials and history remain locally encrypted. Compare stays one-shot; Collaborate remains exactly three stages.

Known limitations

This is Alpha software. Math covers a useful TeX subset, not full TeX. Collaborate retries whole rounds only. Debate / Judge and multi-turn Compare are not implemented. ChatGPT may be subject to plan/model usage limits. SIWC compatibility remains CONDITIONAL. Broader device coverage, long-duration validation and full real-device AndroidTest coverage remain incomplete.

Meldwise is GPL-3.0-only starting with v0.1.0-alpha.2. Orcex (Apache-2.0), STIX Two Math (OFL-1.1) and other third-party components retain their own licenses. Provider names and trademarks belong to their respective owners. Meldwise is an independent community project, not affiliated with, sponsored by or endorsed by the referenced providers.

v0.2.0-alpha.1 — 0.2 预览版

Meldwise 0.2 Alpha

现在可以在同一段对话里一直聊下去，也可以请两个模型一起审阅和整理答案。

这一版有什么

• 单模型对话支持多轮，打开同一条历史就能继续，重启后也会保留。
• 跨服务商继续对话前，会先确认可见上下文共享；选择只用于当前对话。
• 新增协作：主模型初答 → 审阅模型检查和补充 → 主模型综合。
• 协作支持后续轮次、取消、失败保留和安全恢复。
• 历史按对话 / 对比 / 协作分类，支持删除和调整顺序。
• Meldwise 原创模式图标、应用图标和主题单色图标。
• 面板切换更连贯，模型检查提示在 1.5 秒后淡出。

ChatGPT、DeepSeek、双模型 Compare、DeepSeek Thinking、流式输出、本地加密凭据与历史、浅色 / 深色 / 自定义重点色继续保留。

隐私与使用

思考内容独立展示，不进入后续模型请求。共享可见上下文和协作回答前需要明确确认。服务商和模型由你选择；失败请求不会自动重放，恢复历史不会自动续发。

凭据和历史加密保存在本机。DeepSeek 按 API 计费，ChatGPT 使用账号可用的套餐权限。正式签名沿用此前 Alpha，可保留数据覆盖升级。

使用前了解一下

这是 Alpha 软件。

• Compare 仍为单轮。
• 协作目前只支持整轮手动重试，尚不支持单阶段续跑。
• Debate / Judge、工具、图片输入和 OpenAI API Key 接入未实现。
• ChatGPT 可能受套餐或所选模型额度限制；隐藏内部思维链不可用。
• SIWC 兼容性仍为 CONDITIONAL（有条件通过）。
• 更多设备、长期使用和完整真机 AndroidTest 覆盖仍在验证。不支持降级到旧版 journal 读取格式。

Meldwise 自 v0.1.0-alpha.2 起采用 GPL-3.0-only；Alpha 1 及其源码仍按原始 MPL-2.0 提供。第三方依赖保留各自许可，服务商名称和商标归各自权利人所有。官方服务商 Logo 未打包。Meldwise 是独立社区项目，与相关服务商没有隶属、赞助或背书关系。

---

Meldwise 0.2 Alpha

Continue the same conversation across restarts, or ask two models to review and refine an answer together.

New

• True multi-turn Single Chat, with reopen-and-continue history.
• Explicit, conversation-local confirmation before visible context crosses providers.
• Collaborate: Primary initial answer → Reviewer review → Primary synthesis.
• Persistent follow-up rounds, cancellation, failure retention and safe recovery.
• Chat / Compare / Collaborate history categories, with deletion and reordering.
• Original Meldwise mode, adaptive launcher and themed monochrome icons.
• Consistent lightweight panel motion; catalog notices start fading after 1.5 seconds.

ChatGPT, DeepSeek, independent Compare, DeepSeek Thinking, streaming, encrypted local credentials/history and Light / Dark / custom accent themes remain available.

Privacy and use

Reasoning stays separate from downstream request context. Cross-provider sharing of visible context and collaboration outputs requires explicit confirmation. There is no automatic provider/model fallback, failed-request replay or inference resume after recovery.

Credentials and history remain encrypted locally. DeepSeek incurs API charges; ChatGPT uses available account-plan access. The production signing identity is unchanged, supporting in-place upgrades with data retained.

Limitations

This remains Alpha software.

• Compare is one-shot.
• Collaborate supports whole-round manual retry, without single-stage resume.
• Debate/Judge, tools, image input and an OpenAI API-key provider are not implemented.
• ChatGPT plan/model usage limits may apply; hidden internal chain-of-thought is unavailable.
• SIWC compatibility remains CONDITIONAL.
• Broader device, long-duration and full real-device AndroidTest coverage remain incomplete. Downgrade to older journal readers is unsupported.

Meldwise is GPL-3.0-only starting with v0.1.0-alpha.2. Alpha 1 and its source retain their original MPL-2.0 terms. Third-party dependencies retain their own licenses; provider names and trademarks belong to their owners. Official provider logos are not bundled. Meldwise is independent and is not affiliated with, sponsored by, or endorsed by the referenced providers.

v0.1.0-alpha.2 — 第二版预览

Meldwise 0.1.0 Alpha 2

Alpha 2 主要改善日常交互和模型目录恢复。

改进

• 历史记录支持删除、上移和下移，调整后的顺序会保存。
• 右上角“+”打开模式选择，可新建对话或对比。
• 模型和思考强度选中后立即应用，面板保持打开。
• Android 返回键按界面层级逐步返回。
• 启动优先读取本地加密模型缓存；空目录或超过 24 小时后，在已配置的提供方下后台更新。
• 更新失败保留旧模型列表，模型和提供方保持原样。
• 模型更新使用轻量成功、警告和错误提示，点开可看脱敏详情。
• 更新提示显示在选择面板上方，面板打开时也能看到。

ChatGPT、DeepSeek、单模型聊天、双模型 Compare、DeepSeek Thinking、本地加密历史、主题与重点色、流式输出和取消功能继续保留。

使用前了解一下

这是 Alpha 软件，SIWC 兼容性仍为 CONDITIONAL。Collaborate、Debate、工具调用、图片输入和 OpenAI API Key 接入留待后续。

ChatGPT 的隐藏内部思维链不可用。更多设备、长期使用和完整真机 AndroidTest 覆盖仍在验证。

凭据加密保存在本机。发送消息由你操作；DeepSeek 按 API 计费，ChatGPT 使用账号可用的套餐权限。

正式签名沿用 Alpha 1，支持保留数据的覆盖升级。

许可证变更

从 v0.1.0-alpha.2 起，Meldwise 自有代码采用 GPL-3.0-only。Alpha 1 及其对应源码继续按原始 MPL-2.0 提供。

第三方依赖保留各自许可。OpenAI、ChatGPT、DeepSeek 等名称和商标归各自权利人所有，仅用于识别服务集成；除非另获许可，它们不属于 Meldwise 的 GPL 授权范围。本版已移除官方 Logo，Meldwise 与这些服务商没有隶属、赞助或背书关系。

---

Meldwise 0.1.0 Alpha 2

Alpha 2 focuses on day-to-day UX and model catalog recovery.

Improvements

• Delete and reorder history with persisted ordering.
• Top-right mode chooser for new Chat and Compare sessions.
• Model and reasoning selections apply immediately while the sheet stays open.
• Hierarchical Android Back navigation.
• Encrypted, cache-first model discovery; configured providers refresh empty or 24-hour-old catalogs in the background.
• Failed refreshes preserve existing catalogs and provider/model identity.
• Lightweight success, warning and error notices with sanitized details.
• Refresh notices stay visible above open selection sheets.

ChatGPT, DeepSeek, Single Chat, Compare, DeepSeek Thinking, encrypted local history, themes, streaming and cancellation remain available.

Limitations and privacy

This remains Alpha software. SIWC compatibility is CONDITIONAL. Collaborate, Debate, tools, image input and an OpenAI API-key provider are not implemented. ChatGPT hidden internal chain-of-thought is unavailable. Broader device coverage, long-duration validation and full real-device AndroidTest coverage remain incomplete.

Credentials are encrypted locally. Sending is explicit. DeepSeek incurs API charges; ChatGPT uses available account-plan access.

The production signing identity is unchanged from Alpha 1, supporting in-place upgrades with local data retained.

License change

Starting with v0.1.0-alpha.2, Meldwise-owned code is distributed under GPL-3.0-only. Alpha 1 and its corresponding source retain their original MPL-2.0 terms.

Third-party dependencies retain their respective licenses. Provider names and trademarks belong to their owners and identify integrations; they are excluded from Meldwise's GPL grant unless separately licensed. Official provider logos have been removed. Meldwise is independent and is not affiliated with, sponsored by, or endorsed by these providers.

v0.1.0-alpha.1 — 0.1.0 预览版

Meldwise 0.1.0 Alpha

第一版公开 Alpha 来了。你可以连接 ChatGPT，也可以用自己的 DeepSeek API Key，单独聊或同时问两个模型。

这一版有什么

• 流式单模型聊天与双模型对比，各自显示回答和状态。
• DeepSeek 思考强度设置，可见思考内容与正文分开展示。
• 中途取消，保留已经收到的回答。
• 凭据、聊天和对比记录在本机加密保存，按服务商和模型区分。
• 浅色、深色、跟随系统和自定义重点色。
• 轻量输入框、原生 Markdown、代码和纯文本复制。

使用前了解一下

这是测试版。SIWC 兼容性仍为 CONDITIONAL（有条件通过）。ChatGPT 隐藏的内部思维链无法查看，手动思考设置取决于当前登录路径和模型支持。

Collaborate、Debate 和 OpenAI API Key 接入留待后续。更多设备、长时间使用和完整真机 AndroidTest 套件尚未全部验证。旧 UI 存储格式的降级恢复暂未支持。

凭据加密保存在设备上，请求由你操作后发出。消息会传给所选服务商；ChatGPT 使用可用的账号套餐权限，DeepSeek 按 API 计费。

下载 APK 后，可以用同名 .sha256 文件核对校验值。后续升级请使用相同签名的正式包。

---

English

First public Alpha. Connect a ChatGPT account or configure your own DeepSeek API key, then chat with one model or compare two answers.

Highlights:

• Streaming Single Chat and independent two-model Compare.
• DeepSeek thinking controls and provider-visible reasoning, separate from the final answer.
• Cancellation with partial-response preservation.
• Encrypted, provider/model-bound local credentials and histories.
• Light, Dark, System and custom accent themes.
• Compact composer, native Markdown, code and plain-text copying.

This is Alpha software. SIWC compatibility remains CONDITIONAL. Hidden ChatGPT chain-of-thought is unavailable; manual reasoning controls depend on route/model support. Collaborate, Debate and an OpenAI API-key provider are not implemented. Broader device and long-duration validation remain incomplete; full device AndroidTest coverage is not claimed. Downgrading to older UI journal formats is unsupported.

Credentials remain encrypted on-device. Requests require explicit user actions and send content to the selected provider. DeepSeek may incur API charges; ChatGPT uses available account-plan access.

Use the accompanying .sha256 file to verify the APK. Future upgrades must retain the production signing identity.

Source license: MPL-2.0. Meldwise is an unofficial community client; provider names and icons belong to their respective owners.