feat(vision): #2200 V2图片分析快速路径 OCR+lite JSON VLM并行 目标单图<3s/8图<15s #2200
Merged
xiaoxia
merged 1 commits from 2026-10-05 17:08:53 +08:00
feat/vision-v2-fast-path into develop
1 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
 xiaoxia
|
ed20fbad49 |
feat(vision): V2 快速图片分析路径 — OCR+lite JSON VLM并行，单图<3s目标
CI/CD Pipeline / Check push changed paths (pull_request) Has been skipped
CI/CD Pipeline / Dedup Check - skip PR tests when covered by push pipeline (pull_request) Successful in 1s
CI/CD Pipeline / Check if frontend-only change (pull_request) Successful in 1s
CI/CD Pipeline / Frontend Lint (pull_request) Has been skipped
CI/CD Pipeline / Frontend Unit Tests (pull_request) Has been skipped
CI/CD Pipeline / PR Build Web Image (pull_request) Has been skipped
CI/CD Pipeline / Build Staging API Image (pull_request) Has been skipped
CI/CD Pipeline / Build Staging Web Image (pull_request) Has been skipped
CI/CD Pipeline / Build Staging Worker Image (pull_request) Has been skipped
CI/CD Pipeline / Retag skipped Staging API Image (pull_request) Has been skipped
CI/CD Pipeline / Retag skipped Staging Web Image (pull_request) Has been skipped
CI/CD Pipeline / Retag skipped Staging Worker Image (pull_request) Has been skipped
CI/CD Pipeline / Deploy Staging (Watchtower auto-deploy) (pull_request) Has been skipped
CI/CD Pipeline / Staging E2E Tests (pull_request) Has been skipped
CI/CD Pipeline / Staging API Integration Tests (pull_request) Has been skipped
CI/CD Pipeline / ACR Image Cleanup (pull_request) Has been skipped
CI/CD Pipeline / PR Build API Image (pull_request) Successful in 58s
CI/CD Pipeline / PR Build Worker Image (pull_request) Successful in 1m19s
Preview Deploy / Deploy Preview Environment (pull_request) Successful in 1m41s
ACR Cleanup / ACR Image Cleanup (pull_request_target) Waiting to run
Preview Cleanup / Cleanup Preview Environment (pull_request) Successful in 43s
PR Automation / Auto Approve on CI Green (pull_request) Successful in 2m59s
CI/CD Pipeline / Unit Tests (pull_request) Failing after 4m43s
CI/CD Pipeline / Validate - Python (mypy + alembic) (pull_request) Successful in 5m49s
CI/CD Pipeline / Integration Tests (pull_request) Successful in 5m50s
CI/CD Pipeline / Validate - Style (pull_request) Failing after 6m5s
AI Code Review / AI Code Review (pull_request) Successful in 6m59s
PR Automation / Auto Merge on CI Green + Approved (pull_request) Successful in 1m3s
CI/CD Pipeline / Validate - Security (pull_request) Successful in 12m21s
CI/CD Pipeline / Build Production API Image (pull_request) Has been skipped
CI/CD Pipeline / Build Production Web Image (pull_request) Has been skipped
CI/CD Pipeline / Build Production Worker Image (pull_request) Has been skipped
CI/CD Pipeline / CI Gate (pull_request) Failing after 6s
CI/CD Pipeline / Deploy Production (pull_request) Has been skipped
CI/CD Pipeline / Canary Release to Production (pull_request) Has been skipped
CI/CD Pipeline / Production Browser E2E (pull_request) Has been skipped
灵应直接指令(10-05)：图片分析专用API组合方案直接干，不要等方案确认。
API现实说明：
- 火山引擎云端真实可用的视觉专用HTTP API：OCR（MediaKit tools-sync/ocr，Bearer鉴权）
- 人体属性/商品检测/图像标签：火山云端无公开HTTP API，仅有移动端SDK（智能美化特效，年费6-60万）
- 务实方案：OCR专用API + doubao-seed-2.1-lite强约束JSON-only prompt（替代3类缺失的专用API），
pro VLM保留为终极兜底
架构：
- 新模块 apps/worker/worker_app/tasks/vision/：
- vlm_fast_json.py：lite VLM极简JSON schema prompt，max_tokens=350，temp=0.1，timeout=8s
- ocr_volc.py：MediaKit同步OCR封装，返回文本列表
- assembler.py：字段映射+portrait_prompt模板拼接，输出格式与旧_normalize完全一致
- fast_path.py：analyze_image_v2/analyze_images_v2，单图2路并行(OCR+lite JSON)，
外层8图全并发，置信度低/失败自动降级旧lite/pro竞速VLM
- _step_image_analysis增加VISION_V2_ENABLED环境变量开关：
- true→走V2快速路径
- false(默认)→走V1 #2198/#2199 lite/pro竞速路径（过渡期兜底）
- 下游信任链/t2i零改动：输出dict字段(name/brand/category/appearance/key_features/
scene/mood/portrait_prompt/summary/_source)与旧格式完全兼容
性能目标：
- 单图fast路径目标<3s（OCR+lite JSON并行取最慢）
- 8图全并发<15s（较当前V1的~115s/3图提升10倍+）
- 兜底路径仍复用现有V1竞速，最坏情况不劣化
|