Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
5 changes: 3 additions & 2 deletions README.md
Original file line number Diff line number Diff line change
Expand Up @@ -213,8 +213,9 @@ explicitly unscored; the interface distinguishes the first cold model load
from later warm reuse, and Clear resets both media tabs, the prompt, response,
and metrics without unloading the model. Running with that cleared prompt
restores the default visual-description prompt automatically. The AlvenX
wordmark is a static identity element with no hidden click interaction. Use
`oml run` for durable, comparable evidence.
wordmark button returns to the document top while preserving the URL and current
workspace state. It stays still at the top and scrolls immediately when reduced
motion is enabled. Use `oml run` for durable, comparable evidence.
See the [local interface guide and security boundary](docs/alvenx-studio.md).

## Run a real local model
Expand Down
3 changes: 2 additions & 1 deletion README.zh-CN.md
Original file line number Diff line number Diff line change
Expand Up @@ -174,7 +174,8 @@ Reports 和项目介绍后置,同时严格区分试玩结果与正式评测证
界面只监听 `127.0.0.1`,关闭公开分享和分析,并把模型调用串行化以适配
8 GB 显卡。界面会区分首次冷加载和后续热复用;Clear 会同时清空两个媒体
标签、Prompt、回答与指标,但不卸载模型;清空后直接运行会自动恢复默认视觉
描述 Prompt。Workspace 回答明确不评分;需要可保存、可比较的证据时仍使用
描述 Prompt。顶部 AlvenX 标志按钮返回文档顶部,同时保留 URL 和当前工作区状态;
已在顶部时保持不动,启用减少动态效果时立即滚动。Workspace 回答明确不评分;需要可保存、可比较的证据时仍使用
`oml run`。完整使用方法和安全边界见
[本地界面说明](docs/alvenx-studio.md)。

Expand Down
11 changes: 5 additions & 6 deletions TASKS.md
Original file line number Diff line number Diff line change
Expand Up @@ -30,13 +30,12 @@
- [x] 将原先强度不一的语料审查证据统一为全部 102 条任务逐项、审查人、日期、
检查项和数据集 SHA 绑定记录;发布检查已要求四个数据集全部通过校验。

## Next:未来 4~6 周
## Next:工程完整性与可复现入口(2026-09-26)

- 让至少 5 位非作者用户从全新环境完成核心 quick start,记录真实失败点;
这是目标,不是已经取得的采用数据。
- 使用现有公开任务与 mock 路径核对清洁环境 quick start,记录可复现的安装、输入和报告问题;先复用已通过且仍适用的检查。
- [x] 为第三方 Adapter 提供最小离线模板和契约测试;真实后端仍只接纳
revision、许可证、硬件边界和复现命令齐全的实现。
- 增加一个由明确问题驱动的 8GB 显存实验;不为追逐模型热度而堆 Adapter。
- 若已有证据无法回答具体问题,再单独定义一个 8GB 显存实验及资源上限。
- 改进报告 schema 的兼容性说明和迁移测试,确保旧证据仍可读取或明确拒绝。
- 将首次安装到离线 mock 报告的操作时间控制在 10 分钟内(不含下载);该数字
是后续验证目标,不是当前成绩。
Expand All @@ -47,9 +46,9 @@
- 只有当真实使用者需要时才考虑网页界面、远程 API 或更大数据集。
- 新数据集必须有可重建媒体、清晰许可证、逐条人工复核和版本化冻结流程。

## 每周 2~3 次有意义更新
## 按完成的阶段交付更新

一次更新必须至少满足以下一类,并附带测试、命令或原始证据:
不设每周提交数量;一次更新至少满足以下一类,并附带测试、命令或原始证据:

1. **功能**:解决一个清楚的用户问题,并有回归测试。
2. **实验**:固定输入、模型 revision、硬件和协议,保存完整失败与限制。
Expand Down
4 changes: 4 additions & 0 deletions docs/01-goals-and-success.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,9 @@
# 目标与验收标准

<!-- 2026-09-26 execution routing -->
> 原始规划/学习参考,保留当时目标、技术选择和节奏。当前工程任务以 [TASKS](../TASKS.md) 和 [post-v1 roadmap](post-v1-roadmap.md) 为准;旧MVP清单、外部用户数、工时和周更频次不作为当前执行门槛。


> 状态:v1.0 的技术与公开交付标准已经完成。原始 KR5 中“发布 v0.1、v0.5、
> v1.0 三个正式 Release”的节奏目标没有完成:实际只发布了 `v1.0.0`,不会
> 为美化记录而回溯补发。独立用户采用仍是发布后的观测指标,不作为回溯修改
Expand Down
4 changes: 4 additions & 0 deletions docs/02-scope-and-requirements.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,9 @@
# 范围与需求

<!-- 2026-09-26 execution routing -->
> 原始规划/学习参考,保留当时目标、技术选择和节奏。当前工程任务以 [TASKS](../TASKS.md) 和 [post-v1 roadmap](post-v1-roadmap.md) 为准;旧MVP清单、外部用户数、工时和周更频次不作为当前执行门槛。


## 1. MVP 用户故事

### 模型使用者
Expand Down
4 changes: 4 additions & 0 deletions docs/03-architecture.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,9 @@
# 技术架构

<!-- 2026-09-26 execution routing -->
> 原始规划/学习参考,保留当时目标、技术选择和节奏。当前工程任务以 [TASKS](../TASKS.md) 和 [post-v1 roadmap](post-v1-roadmap.md) 为准;旧MVP清单、外部用户数、工时和周更频次不作为当前执行门槛。


## 1. 总体结构

```mermaid
Expand Down
4 changes: 4 additions & 0 deletions docs/05-weekly-workflow.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,9 @@
# 每周工作方法

<!-- 2026-09-26 execution routing -->
> 原始规划/学习参考,保留当时目标、技术选择和节奏。当前工程任务以 [TASKS](../TASKS.md) 和 [post-v1 roadmap](post-v1-roadmap.md) 为准;旧MVP清单、外部用户数、工时和周更频次不作为当前执行门槛。


## 1. 固定节奏

### 更新 A:功能
Expand Down
4 changes: 4 additions & 0 deletions docs/06-quality-and-open-source.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,9 @@
# 质量与开源标准

<!-- 2026-09-26 execution routing -->
> 原始规划/学习参考,保留当时目标、技术选择和节奏。当前工程任务以 [TASKS](../TASKS.md) 和 [post-v1 roadmap](post-v1-roadmap.md) 为准;旧MVP清单、外部用户数、工时和周更频次不作为当前执行门槛。


> `v1.0.0` 的公开证据见[发布准备矩阵](public-release-readiness.md)和
> [最终验证报告](reports/final-candidate-validation.md)。以下规则继续作为未来
> 版本的可复用门槛,复选框不是 v1.0 状态记录。
Expand Down
4 changes: 4 additions & 0 deletions docs/07-learning-path.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,9 @@
# 学习路线

<!-- 2026-09-26 execution routing -->
> 原始规划/学习参考,保留当时目标、技术选择和节奏。当前工程任务以 [TASKS](../TASKS.md) 和 [post-v1 roadmap](post-v1-roadmap.md) 为准;旧MVP清单、外部用户数、工时和周更频次不作为当前执行门槛。


学习内容必须服务于当前里程碑,不采用“全部学完再开始”的方式。

## 阶段一:Python 工程基础
Expand Down
4 changes: 4 additions & 0 deletions docs/08-risks-and-decisions.md
Original file line number Diff line number Diff line change
@@ -1,5 +1,9 @@
# 风险、假设与决策

<!-- 2026-09-26 execution routing -->
> 原始规划/学习参考,保留当时目标、技术选择和节奏。当前工程任务以 [TASKS](../TASKS.md) 和 [post-v1 roadmap](post-v1-roadmap.md) 为准;旧MVP清单、外部用户数、工时和周更频次不作为当前执行门槛。


> 状态:本文件保留关键决策及其当时依据;事实部分已在 v1.0 发布后更新。
> 当前取舍见[发布后路线图](post-v1-roadmap.md)。

Expand Down
5 changes: 3 additions & 2 deletions docs/alvenx-studio.md
Original file line number Diff line number Diff line change
Expand Up @@ -196,8 +196,9 @@ Run 页先显示紧凑的四步引导和“仅本机”边界,再显示“准
两张工作卡片。卡片说明与运行状态会明确区分冷启动与热复用:第一次请求需要加载模型,成功后模型会
留在显存中供下一次请求复用;Clear 只清空媒体、Prompt、回答和指标,不卸载
模型;清空后在 Prompt 为空时点击 Run,会自动恢复默认视觉描述 Prompt。模型
选择位于媒体之后,所选后端用于下一次运行。顶部 AlvenX 标志只作为静态品牌
元素,不可点击、没有隐藏彩蛋。玻璃标题栏、分段导航、媒体控件和工作卡片采用
选择位于媒体之后,所选后端用于下一次运行。顶部 AlvenX 标志按钮返回文档顶部,
同时保留 URL 和当前工作区状态;已在顶部时保持不动,启用减少动态效果时立即滚动。
玻璃标题栏、分段导航、媒体控件和工作卡片采用
统一的嵌套圆角层级。媒体仍按原始比例
完整显示,长回答在框内滚动。

Expand Down
19 changes: 9 additions & 10 deletions docs/post-v1-roadmap.md
Original file line number Diff line number Diff line change
Expand Up @@ -52,10 +52,10 @@ An adapter that only makes another model import successfully is not enough.

### Next: evidence-led extension

- Run observed fresh-clone sessions with non-author users and fix the highest
friction points.
- Publish an adapter template only after its contract tests and license/revision
checklist are complete.
- Verify the documented fresh-install path with existing public tasks and fix observed
friction. Reuse unchanged acceptance evidence; external-user sessions can follow when available.
- Maintain the already delivered adapter template and its contract tests; change it only
for an observed integration gap.
- Add one low-VRAM comparison at a time, each motivated by a question the
existing v1.0 evidence cannot answer.

Expand All @@ -76,14 +76,13 @@ An adapter that only makes another model import successfully is not enough.

The next validation targets are:

- five successful clean-environment runs by people other than the author;
- three substantive external Issues, Discussions, or pull requests that change
a documented decision;
- a core quick start completed within 10 minutes, excluding model downloads;
- two or three substantive updates per week when evidence is ready.
- documented input-to-report behavior and understandable malformed-input failures;
- compatibility of preserved evidence and clear boundaries for new report versions;
- a reviewed release when a coherent engineering milestone is complete.

Until measured, these remain targets. Empty commits, retrospective releases,
or invented feedback do not satisfy them.
Until measured, these remain targets. External use is recorded when it occurs; recruitment
counts and a weekly commit quota are not engineering acceptance conditions.

## Stop conditions

Expand Down
Loading