docs: 补充 AI 打字机输出、小程序灵感卡片、脚本主页布局等设计文档和计划

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
2026-05-24 18:39:26 +08:00
parent 886f04046b
commit 6e5a379bef
12 changed files with 722 additions and 0 deletions
@@ -0,0 +1,45 @@
+# Mini Program Script Generation Feedback Design
+
+## Goal
+
+Improve the mini-program `心愿实现中` screen so users can clearly tell that generation is running, understand long waits, and recover gently if the AI stream fails or produces no output.
+
+## Approved Approach
+
+Use option A: add a scene-friendly loading state with phased guidance and a graceful failure panel.
+
+## Behavior
+
+When script generation starts:
+
+1. The page stays on the current `心愿实现中` screen.
+2. The existing chat bubble and orbit motif remain.
+3. The orbit gets breathing/rotating animation and small star-dot motion.
+4. A rotating hint line appears while no stream text has arrived.
+5. If no stream text arrives after about 8 seconds, show a gentle slow-wait message.
+6. If no stream text arrives after about 20 seconds, show a warmer network-slow message, while still waiting for the stream.
+7. Once stream text arrives, switch to the existing typewriter text output.
+8. If the call fails or returns empty output, keep the user on the generation screen and show a soft failure state with `再试一次` and `返回修改`.
+
+## Implementation Notes
+
+Only `mini-program/src/pages/main/ScriptView.vue` needs to change.
+
+Add a small local generation feedback state:
+
+- `generationStatus`: `idle | waiting | slow | verySlow | streaming | failed`
+- `generationError`: user-facing failure copy
+- timers for rotating hints and long-wait thresholds
+
+Do not change backend API calls or stream protocol. `streamAiScene` remains the source of stream output. Existing `useTypewriterStream` continues to render text progressively after the first output arrives.
+
+## Validation
+
+Run:
+
+```bash
+cd mini-program
+npm run build:mp-weixin
+```
+
+Expected result: build succeeds with no new errors.