admin-valentin/wewrite
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions
104 commits 1 branch 0 tags 355 KiB
Commit graph

3 commits

Author SHA1 Message Date
wangzhuc
0871162634 feat: visual AI overhaul — style anchoring, typed illustrations, 9-provider fallback 
- Add visual anchor extraction after cover confirmation for consistent inline images
- Introduce 6 illustration types (infographic/scene/flowchart/comparison/framework/timeline) with structured prompt templates
- Expand image providers from 3 to 9: add DashScope, MiniMax, Replicate, Azure OpenAI, OpenRouter, Jimeng
- Support multi-provider auto-fallback via `image.providers` config (backward-compatible)
- Update SKILL.md Step 6 and environment check for new capabilities

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-04-01 23:51:53 +08:00
wangzhuc
5ed39b4e44 feat: improve image relevance — entity anchoring + cover-first flow 
SKILL.md Step 6 restructured:
- 6.1: extract 3-5 concrete entities from article before prompting
- 6.2: generate cover only (1 API call, test direction early)
- 6.3: validate cover (interactive: ask user; auto: self-check entities)
- 6.4: batch inline images using cover's style for consistency

visual-prompts.md:
- Add "entity anchoring" hard rule: every prompt must include ≥2 article
  entities; ban vague terms as sole subject ("科技感", "未来感")
- Add anti-pattern → good-pattern examples
- Inline images must reuse cover's style description for consistency

Addresses #9

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-04-01 14:03:33 +08:00
wangzhuc
1ab34fa450 Initial release — 公众号文章全流程 AI Skill 
热点抓取 → 选题 → 框架 → 写作 → SEO → 视觉AI → 排版 → 微信草稿箱，
一句话触发完整流程。适用于 Claude Code skill 格式。

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
 2026-03-26 22:16:18 +08:00