GenarrativeAI/Genarrative
11
0
Fork 0
Code Issues 14 Pull Requests 3 Actions Packages Projects 3 Releases Wiki Activity

已生成资源/资产的持久化与画布完成操作不具备原子性，也不具备请求幂等性。 #134

New Issue
Closed
opened 2026-08-04 20:03:10 +08:00 by k88936 · 2 comments
k88936 commented 2026-08-04 20:03:10 +08:00
Member
Copy Link
Copy Source

范围：

  • 这是编辑器-生成共享架构的问题，并非角色动作特有。
  • 图片、视频、音效、背景音乐及角色动作流程均先持久化生成的项目资源/账户资产，再于独立的 SpacetimeDB 事务中调用 complete_editor_canvas_generation_with_items。

现有防护：

  • 画布完成时读取最新画布并通过版本 CAS 保存。
  • 若同一 resourceId 已存在，完成操作会抑制重复放置。
  • 已完成的生成对话框可防止后续过期的布局保存覆盖。

缺失防护：

  • taskId 仅为元数据且属于确定性图层 ID 的一部分，并非唯一持久化键，也未用于查找或复用已生成的资源/资产。
  • 每次持久化尝试都会重新生成资源 ID 和资产 ID。
  • 若资源/资产创建成功但画布完成失败，先前记录不会回滚。
  • 重试 HTTP 生成请求会产生新的服务端任务 ID，导致现有的资源级去重检查无法使整个操作具备幂等性。

所需架构决策：

  • 优先采用单个 SpacetimeDB 事务，利用持久化生成作业提供的稳定操作键，幂等地 upsert 最终生成的资源/资产并完成画布。
  • 若外部生成工作流无法调用单一事务，则增加以所有者/项目/操作 ID 为键的持久化完成记录，确保资源创建、资产创建和画布完成在添加清理补偿逻辑前各自具备重放安全性。
  • 禁止添加仅针对角色动作的清理逻辑或 taskId 兜底方案；修复必须覆盖所有编辑器生成类型，并保留版本冲突处理机制。

所需测试：

  • 资源创建后、资产创建前发生故障。
  • 资源/资产创建后、画布完成前发生故障。
  • 发生版本冲突后，使用相同操作 ID 进行重试。
  • 成功后重复执行完成操作，应返回原始资源/资产/图层且不产生重复项。
  • 进程重启后重试，应保持相同的操作标识。
范围： - 这是编辑器-生成共享架构的问题，并非角色动作特有。 - 图片、视频、音效、背景音乐及角色动作流程均先持久化生成的项目资源/账户资产，再于独立的 SpacetimeDB 事务中调用 `complete_editor_canvas_generation_with_items`。 现有防护： - 画布完成时读取最新画布并通过版本 CAS 保存。 - 若同一 `resourceId` 已存在，完成操作会抑制重复放置。 - 已完成的生成对话框可防止后续过期的布局保存覆盖。 缺失防护： - `taskId` 仅为元数据且属于确定性图层 ID 的一部分，并非唯一持久化键，也未用于查找或复用已生成的资源/资产。 - 每次持久化尝试都会重新生成资源 ID 和资产 ID。 - 若资源/资产创建成功但画布完成失败，先前记录不会回滚。 - 重试 HTTP 生成请求会产生新的服务端任务 ID，导致现有的资源级去重检查无法使整个操作具备幂等性。 所需架构决策： - 优先采用单个 SpacetimeDB 事务，利用持久化生成作业提供的稳定操作键，幂等地 upsert 最终生成的资源/资产并完成画布。 - 若外部生成工作流无法调用单一事务，则增加以所有者/项目/操作 ID 为键的持久化完成记录，确保资源创建、资产创建和画布完成在添加清理补偿逻辑前各自具备重放安全性。 - 禁止添加仅针对角色动作的清理逻辑或 taskId 兜底方案；修复必须覆盖所有编辑器生成类型，并保留版本冲突处理机制。 所需测试： - 资源创建后、资产创建前发生故障。 - 资源/资产创建后、画布完成前发生故障。 - 发生版本冲突后，使用相同操作 ID 进行重试。 - 成功后重复执行完成操作，应返回原始资源/资产/图层且不产生重复项。 - 进程重启后重试，应保持相同的操作标识。
k88936 added this to the 美术画布 project 2026-08-04 20:03:10 +08:00
k88936 commented 2026-08-07 10:58:13 +08:00
Author
Member
Copy Link
Copy Source

plan: 默认采用“全部 provider 生成作业化”，包括开发用 inline 模式。 ...

Q: ??? inline 仅同步领取并等待同一 job , this is fucking crazy ,,,, in what cases does inline was called ???

A: You’re right. Turning inline into “enqueue, claim, and wait” was an unjustified expansion and defeats the purpose of that diagnostic path. I’m checking every actual branch now so I can state precisely where inline is used and revise the plan around the real scope.

• Confirmed: inline here is only the explicit GENARRATIVE_EXTERNAL_GENERATION_MODE=inline diagnostic mode. It directly executes eight provider-backed handlers inside the HTTP request—image generation/edit, icon/UI extraction, character animation, video, sound effect,
and background music. It creates no job and has no lease. Production/default configs are queue; background removal and External v1 already always queue.

Perfect pixel and manual atlas splitting are also synchronous, but they do not consult this mode flag; I incorrectly conflated those with configured provider inline. The revised plan will preserve direct inline execution and will not enqueue/claim/wait.

plan: 默认采用“全部 provider 生成作业化”，包括开发用 inline 模式。 ... Q: ??? inline 仅同步领取并等待同一 job , this is fucking crazy ,,,, in what cases does inline was called ??? A: You’re right. Turning inline into “enqueue, claim, and wait” was an unjustified expansion and defeats the purpose of that diagnostic path. I’m checking every actual branch now so I can state precisely where inline is used and revise the plan around the real scope. • Confirmed: inline here is only the explicit GENARRATIVE_EXTERNAL_GENERATION_MODE=inline diagnostic mode. It directly executes eight provider-backed handlers inside the HTTP request—image generation/edit, icon/UI extraction, character animation, video, sound effect, and background music. It creates no job and has no lease. Production/default configs are queue; background removal and External v1 already always queue. Perfect pixel and manual atlas splitting are also synchronous, but they do not consult this mode flag; I incorrectly conflated those with configured provider inline. The revised plan will preserve direct inline execution and will not enqueue/claim/wait.
kdletters commented 2026-08-07 19:21:35 +08:00
Member
Copy Link
Copy Source

已经由 6c53eda71c 修复

已经由 6c53eda71c487526a8b421fa881ce2faa8c572b6 修复
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
Compat/Breaking
Breaking change that won't be backward compatible
 Kind/Bug
Something is not working
 Kind/Documentation
Documentation changes
 Kind/Enhancement
Improve existing functionality
 Kind/Feature
New functionality
 Kind/Security
This is security issue
 Kind/Testing
Issue or pull request related to testing
Priority
Critical
1
The priority is critical
Priority
High
2
The priority is high
Priority
Low
4
The priority is low
Priority
Medium
3
The priority is medium
Reviewed
Confirmed
1
Issue has been confirmed
Reviewed
Duplicate
2
This issue or pull request already exists
Reviewed
Invalid
3
Invalid issue
Reviewed
Won't Fix
3
This issue won't be fixed
Status
Abandoned
3
Somebody has started to work on this but abandoned work
Status
Blocked
1
Something is blocking this issue or pull request
Status
Need More Info
2
Feedback is required to reproduce issue or to continue work
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project 美术画布
Assignees
Clear assignees
HuangJiaJia JenkenB k88936 (王德宇) kdletters (段舒康) laura2008 (王乐) lhk229 (孔令弘) Loer (农洛儿) loong suzmii (董羽秦) taobaisu (叶东旭) wuxiangwanzi (王子民) YixuanLiu ziyu
No Assignees
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: GenarrativeAI/Genarrative#134
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?