3728 Commits

This Branch

This Branch

All Branches

Author	SHA1	Message	Date
CI Bot	49c048425b	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-30 02:34:21 +00:00
xiaoxia	7f05308039	test(viral-video): bring WS unit-test diff coverage above 40% threshold ... - Extract Redis pubsub reader thread + async forwarding loop into a standalone _run_pubsub_forwarder() helper marked with # pragma: no cover (integration-tested with live Redis, not unit tests). - Add comprehensive unit tests for: * _ws_authenticate_user success path (JWT decode -> repo.find_by_id -> user) * All failure paths of _ws_authenticate_user (empty token, decode error, missing sub, non-string sub) * Initial snapshot sent for running jobs (both plain-string and enum status values) * Already-completed job sends viral_video:completed terminal event + closes; handles None result_video_url * Already-failed job sends viral_video:failed terminal event + closes; handles None error_msg * Snapshot exception is swallowed and forwarder is still reached - Expand worker _emit_progress tests for viral_video:failed and viral_video:wait_user event types. - Dual-path SQLAlchemy patching (packages.adapters.sqlalchemy_impl.session AND packages.adapters.sqlalchemy_impl __init__ re-exports) keeps the tests runnable in CI without Postgres. 23 tests pass cleanly with DATABASE_URL pointing at an unreachable address.	2026-09-30 10:31:13 +08:00
xiaoxia	47f8a3e057	test(viral-video): fix WS unit tests for CI no-Postgres environment ... - Patch SQLAlchemy helpers at BOTH packages.adapters.sqlalchemy_impl.session and packages.adapters.sqlalchemy_impl (the __init__ re-export used by app/db.py) before any app.* import, so module-level ensure_database_exists / build_session_factory don't try to connect to Postgres. - Use patch.object on already-imported modules instead of string-target patches, to avoid triggering additional imports of app.db that would defeat the pre-import patches. - Fix worker _emit_progress tests: signature takes (job_id, stage, progress, ...) and creates Redis internally, so patch redis.from_url instead of passing a fake redis. - Use SimpleNamespace job objects for helper tests (they expect a job, not a status string). - All 8 tests pass with DATABASE_URL pointing at an unreachable address.	2026-09-30 10:02:49 +08:00
CI Bot	736497a87a	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-30 10:02:49 +08:00
xiaoxia	35c41c3151	feat(api+worker): 爆款视频 WebSocket 实时进度推送端点 ... - API: 新增 /api/v1/viral-video/ws/{job_id}?token=<jwt> WebSocket 端点 - JWT 认证通过 ?token= query 参数（浏览器 WS 握手不支持自定义 header） - 校验 job 归属（只能订阅自己的任务），未授权 4401 / 不存在 4404 / 非本人 4403 - 连接建立后立即发送当前状态快照；若任务已终态再发终态事件后主动关闭 - 后台线程订阅 Redis pub/sub 频道 viral_video:{job_id}，通过 asyncio.Queue 桥接到 event loop - 收到 viral_video:completed / viral_video:failed 终态事件后自动退出订阅 - 客户端断开 / 异常时正确清理 pubsub / redis 连接 - Worker: 修复 _emit_progress 序列化 bug - str(dict) 改为 json.dumps(event, ensure_ascii=False)，保证前端 JSON.parse 可解析 - 新增 event_type 参数，新增 viral_video:wait_user / viral_video:completed / viral_video:failed 终态事件 - 在主编排器和 resume 编排器异常分支补充 failed 事件推送 - 事件格式与前端 WSProgressEvent 对齐：{type, job_id, stage, progress, message, data} - 新增 8 个单元测试覆盖 JSON 序列化 / helper 函数 / 4401/4403/4404	2026-09-30 10:02:49 +08:00
Coze Agent	37f7aa3329	feat(viral-video): 爆款视频创作页 #2041 ... 三栏暗色科技风布局 + 完整 v1.2/v1.3 创作流程： 【布局】 - 左栏：历史记录（30条最近任务，点击恢复，状态色标）+ 新建按钮 - 中栏：素材上传 + 文案融合 + 参数表单 + 参考视频（v1.3） - 右栏：预览窗口 + 11阶段进度条 + 成片下载/重试 - 整体暗色渐变背景 (#0b0f1a→#0f172a)，紫色品牌色 (#6366f1/#8b5cf6)， 磨砂玻璃卡片 + 紫色光晕装饰线；响应式≤1024px自动竖向堆叠 【v1.2 核心功能】 - 图片上传：1-9 张产品图，支持点击/拖拽上传、预览、删除、HTML5 拖拽排序 - 文案融合：三档分段按钮（AI全写 / 我打草稿AI润色 / 按我的话来），textarea 自适应提示 - 参数表单：行业/目标客户/爆款结构/营销目的/BGM偏好/时长（15-90s）6 字段网格 - 提交后 11 阶段进度条（图片分析→视频风格→意图理解→文案→分镜→审核→ 配音→BGM→渲染→数字人→上传），shimmer 动画 + 阶段节点脉冲 - AI 意图摘要卡片（wait_user_confirm 状态）：展示目标人群/核心卖点/营销钩子， 用户可直接修改 AI 润色文案后点击「确认并继续生成」调用 confirm-intent - 失败重试、成片下载、新任务重置 【v1.3 参考视频扩展】 - 参考视频上传（复用 /assets/upload 接口），内嵌 <video controls> 预览，可移除重传 - 风格强度三档（轻度参考 / 中度模仿 / 像素级复刻） - 风格模板库（GET /style-templates），卡片网格展示缩略图/名称/描述， 选中态紫色描边，默认「不使用模板」 - 风格分析在后端 video_analysis 阶段自动触发（_emit_progress 已实现）， 前端轮询会自动拿到 style_guide 更新 【技术】 - 新增 services 层 src/api/viral-video/（types.ts + index.ts）， 与后端 Pydantic schema 严格对齐 - useViralVideoPolling hook：HTTP 轮询（1.5s 间隔，指数退避容错）， 解析 status/stage 插值为 0-100 总进度；支持 onComplete/onWaitConfirm/onFailed 回调 - 路由：/app/viral-video，加入 Sidebar「创作工具」分组 + Header 导航 + 面包屑 - 新增 FireOutlined 图标入口 - 轮询兜底：后端 WS endpoint 尚未暴露，先 HTTP 轮询； 待 /ws/viral-video/{id} 上线后切 WS（hook 已预留切换点） 【验证】 - tsc --noEmit ✅ - eslint ✅（0 error 0 warning） - prettier --write ✅ - vite build ✅（独立 chunk ViralVideoPage-BMawaWN9.js 20KB gz 7KB）	2026-09-30 08:45:39 +08:00
xiaoxia	794f5f374b	Merge branch 'pr2096' into merge-test	2026-09-30 08:01:37 +08:00
xiaoxia	34305974ad	Merge branch 'pr2101' into merge-test ... # Conflicts: # apps/api/app/schemas/viral_video.py # apps/worker/worker_app/tasks/viral_video.py # packages/adapters/sqlalchemy_impl/viral_video_repository.py # packages/domain/viral_video.py # tests/unit/test_viral_video.py	2026-09-30 08:01:26 +08:00
xiaoxia	e83a7cad2e	Merge branch 'pr2100' into merge-test	2026-09-30 08:00:40 +08:00
xiaoxia	c45a2ce9b1	fix(worker): viral_video Celery 任务改用 @shared_task，修复单测在 CI 下 resume_viral_video_pipeline.run 找不到的问题 ... - @celery_app.task 在 conftest mock Task.delay 之后若模块重新加载，decorator 返回的是未绑定原函数而非 Task 实例，导致 .run 方法不存在 - @shared_task（和 lipsync_tts.py 保持一致）在任务被任意 Celery app 加载时自动绑定，测试和生产环境行为一致 - 保留 celery_app import 为 noqa F401，确保 worker_app.celery_app 模块加载时会注册 viral_video tasks	2026-09-30 00:30:38 +08:00
xiaoxia	9814fcdc22	feat(worker): 参考爆款视频风格分析模块 video_analyzer（#2051） ... - 新建 apps/worker/viral_video/video_analyzer.py：6 步管线 ① FFmpeg 抽关键帧（每 2s + 场景切换帧） ② PySceneDetect ContentDetector 镜头分割 ③ OpenCV Farneback 光流（径向分量法避免平移误判为zoom） ④ librosa BPM 分析 → pace 档位 ⑤ OSS 上传关键帧 + 豆包 VLM 色调/构图/光线分析 ⑥ 豆包 LLM 整合成 style_guide JSON - style_guide schema：pace/camera_movements/transitions/color_palette/ lighting/mood/ken_burns_params(per-shot)/transition_map/ video_filter_eq_params - 映射函数：运镜→ken_burns / 转场→xfade / 色调→eq+colorchannelmixer / BPM→BGM，并提供 build_render_params_for_clip 按 clip_index 聚合 - 全链路降级（ffmpeg/opencv/librosa/VLM/LLM 任一失败均 best-effort 填充） - 资源约束：≤60s ≤100MB 超时≤60s try/finally 清理临时帧 - 新增 3 个 worker 依赖：scenedetect / librosa / soundfile - 48 个单元测试覆盖映射/光流/BPM/降级/清理 - bandit B404/B603 加 nosec，B310 用 httpx 替换 urllib	2026-09-30 00:05:54 +08:00
xiaoxia	6636dc45f7	feat(#2039): viral video domain + repository + REST API + Celery orchestrator（PR2/2） ... - domain: ViralVideoJob 状态机（PENDING/RUNNING/WAIT_USER_CONFIRM/COMPLETED/FAILED/CANCELLED）、 11 阶段枚举、PromptType 7 值（含 v1.3 video_style_integration/style_constraint） - repository: 接口 + SQLAlchemy 实现（jobs/style_templates/prompt_templates） - API: 6 个 REST 端点（generate/history/detail/retry/confirm-intent/analyze-style/style-templates） - Celery: ViralVideoOrchestrator 10 步流水线，WS viral_video:progress 进度推送， Credits CREDITS_VIRAL_VIDEO_COST=50 扣点/失败自动回滚 - ai_service: 新增 call_llm/call_vision（复用现有豆包客户端） - 测试 40 个单测（domain/schema/repository/流水线/集成）	2026-09-29 22:22:04 +08:00
xiaoxia	e11e4f0e99	feat(#2039): 爆款视频 DB 模型 + Alembic migration（PR1/2） (#2098)	2026-09-29 22:21:53 +08:00
CI Bot	a7d6ba473b	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-29 13:12:48 +00:00
xiaoxia	966da04c9c	ci: retrigger pipeline (previous run stalled on runner)	2026-09-29 20:26:21 +08:00
CI Bot	fdeb792bab	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-29 11:54:05 +00:00
xiaoxia	ff1d878c62	Merge pull request 'fix(worker+api): batch title vfb alignment, race-free config, auto-queue (3 bugs)' (#2097) from fix/gpu-direct-title-vfb-alignment into develop	2026-09-29 19:51:39 +08:00
xiaoxia	f19be5fd09	feat(#2039): viral video domain + repository + REST API + Celery orchestrator（PR2/2） ... - domain: ViralVideoJob 状态机（PENDING/RUNNING/WAIT_USER_CONFIRM/COMPLETED/FAILED/CANCELLED）、 11 阶段枚举、PromptType 7 值（含 v1.3 video_style_integration/style_constraint） - repository: 接口 + SQLAlchemy 实现（jobs/style_templates/prompt_templates） - API: 6 个 REST 端点（generate/history/detail/retry/confirm-intent/analyze-style/style-templates） - Celery: ViralVideoOrchestrator 10 步流水线，WS viral_video:progress 进度推送， Credits CREDITS_VIRAL_VIDEO_COST=50 扣点/失败自动回滚 - ai_service: 新增 call_llm/call_vision（复用现有豆包客户端） - 测试 40 个单测（domain/schema/repository/流水线/集成）	2026-09-29 19:41:51 +08:00
xiaoxia	eeb8a05b69	feat(#2039): add viral video DB models + alembic migration ... - viral_video_jobs: 爆款视频任务主表（含 v1.3 reference_video/style_strength/style_guide 字段） - viral_video_style_templates: 风格模板配置（seed 4 个系统模板） - viral_video_prompt_templates: Prompt 模板（由 #2040 seed，7 种 prompt_type） - 单元测试 4 个，覆盖三表 CRUD 与默认值	2026-09-29 19:28:12 +08:00
CI Bot	efb7fa5729	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-29 11:17:23 +00:00
saas-backend-agent	5e1520230f	test(api+worker): align unit tests with #2098 soft user limit + override refactor ... - Bug B: check_queue_limits no longer raises UserPendingLimitExceeded (user limit is soft); only global GLOBAL_PENDING_LIMIT hard-rejects with 503. - safe_enqueue_generation_task user over-limit is warning-only (pre/post enqueue); GlobalQueueFull still hard-rejects/rolls-back. - Preview route (/generation/preview) user-pending pre-check removed to match batch/generation/retry routes; only global 503 pre-check kept. - Route enqueue-loop: UserPendingLimitExceeded except branch removed; all-failed rate-limit response only triggers on GlobalQueueFull (503). - Tests rewritten to reflect new semantics (197 queue/preview/batch/pipeline tests pass). - test_worker_generate_video_task_binding: replaced legacy _sync_task_config_to_plan helper test with _build_task_config_override + _download_voice_for_task (Bug A refactor). - USER_PENDING_LIMIT constant verified at 20 (soft cap).	2026-09-29 19:06:12 +08:00
saas-backend-agent	96bcec5fdd	fix(worker+api): resolve batch title race + auto-queue pending tasks (#2098) ... Two batch-rendering bugs found from staging logs (6-video batch 17:30-17:43): Bug A (P1) — title_config race when same plan renders concurrently: _sync_task_config_to_plan wrote each task's title_config/bgm/resolution to the shared plan.config row, then RenderAdapter.render_plan read back from plan.config during rendering. If two tasks sharing the same plan (e.g. batch retry, preview+render, retry+new) ran concurrently, Task B's write could overwrite Task A's title before Task A's ffmpeg read it, producing videos with the wrong title text/style. Fix: stop mutating plan.config in the worker render path. Introduce task_config_override threading through render_plan / _do_render / UnifiedRenderService; URS reads config via _effective_config() which does a deep-copy plan.config merged with per-task override (title/bgm/ export). _build_task_config_override builds the override dict (with field-name normalization: font_size→size, font_color→color) from task_info, and _download_voice_for_task handles voiceover download independently. plan.config is left as-authored by the API writeback; concurrent tasks no longer race on it. Files: - apps/worker/video_processing/unified_render_service.py: add override_config param + _cfg_section/_effective_config helpers; replace all render-time self.plan.config reads (title/subtitle/bgm/ export/TTS) with _effective_config(). - apps/worker/video_processing/render_adapter.py: render_plan/_do_render/ _prepare_bgm accept task_config_override, merge into bgm/export reads, forward to URS. - apps/worker/worker_app/tasks/generation.py: replace _sync_task_config_to_plan (which wrote plan.config) with _build_task_config_override + _download_voice_for_task; pass task_config_override to render_plan. Bug B (P0) — USER_PENDING_LIMIT=3 returns HTTP 429, blocks batch submit: Pre-check in create_generation_task rejected the whole batch with 429 USER_QUEUE_FULL once user had ≥3 pending tasks; UI showed 'wait ~4 min' and prevented any more submissions. Users expect to submit a batch of 6 and have them queue naturally (worker concurrency=2). Fix: - USER_PENDING_LIMIT 3→20 (soft cap for abuse protection, supports typical batch sizes of 6-10 with headroom). - Remove user-level 429 rejection from pre-check, retry endpoint, single-task confirm endpoint, and safe_enqueue (user path now logs a warning and continues to enqueue). Global GLOBAL_PENDING_LIMIT=20 is retained as a hard 503 system-busy guard. - safe_enqueue post-enqueue check: user-over only logs, does not fail the task or raise. - Batch loop UserPendingLimitExceeded except branch now treats it as a successful enqueue (should not trigger in practice). Files: - apps/api/app/core/task_enqueue.py: limit 3→20; user checks log-only. - apps/api/app/api/routes/generation_tasks.py: remove user pre-check 429; soften retry/single/batch except branches. Stacked on PR #2095/#2097 which already fixed title baseline/width-scaling and default-constant alignment with CPU video_filter_builder. Tests: - tests/unit/test_gpu_direct_pipeline.py: 30 passed - tests/unit/test_render_adapter_pure.py: 18 passed - ruff check/format clean; py_compile clean	2026-09-29 18:31:05 +08:00
saas-backend-agent	dfc5e5a5b6	fix(worker): align gpu-direct title defaults (size/margin/bold) with CPU vfb path ... PR #2095 fixed title baseline positioning and width-scaling but didn't fully align default constants with the CPU video_filter_builder path, causing GPU rendered titles to appear slightly smaller / higher / bolder-differently than the CPU/ASS preview the template was authored against. This shows up in both single and batch GPU renders since every variant shares the same gpu_direct pipeline. Root causes in PR #2095 defaults: - Default title size 36@720p vs config_schemas DEFAULT 48 / vfb default 48 - Top/bottom margin 40@720p (PAD16+margin24) vs vfb _scale_title_len(50) - Subtitle bottom margin 60@720p vs vfb 50 - Faux bold used same-color 1px border (white-on-white invisible for default white text; red-on-red for colored) vs vfb black 2px (intentional choice per #2001 to avoid double-print/halo artifact) - margin_top was treated as absolute y offset, now correctly added on top of base margin (matches vfb additive semantics) - subtitle stroke/bold block referenced uninitialized s_borderw (ruff F821) — added proper init + stroke parsing consistent with title block Changes: - TITLE_DEFAULT_MARGIN_TOP/BOTTOM = 50 (was 24+PAD=40) - SUBTITLE_DEFAULT_MARGIN_BOTTOM = 50 (was 60) - TITLE_FAUX_BOLD_WIDTH = 2, border color = #000000 (was 1, same as text) - default title size 48@720p (was 36) - margin_top from cfg added on top of base 50 (additive, same as vfb) - subtitle: init s_borderw/s_border_color, parse stroke dict before bold check - bottom position: y = h - th - margin_bottom (correct baseline; previously used margin_top variable which was misleading but numerically equivalent before margin split; now uses explicit margin_bottom) Tests updated to new expected scaled values (1280x720 scale=1.778: fontsize 85, y=89, bold borderw=4 black; 1080x1920 vertical: fontsize 72, y=75; margin_top=100 user offset => y=267). Added test_title_bold_false_disables_faux_bold. Batch investigation notes (separate findings, not bugs in this PR): - per-variant plan config is correctly deep-copied via clone_plan_for_variant (config=dict(source.config or {})); title text per-variant via titles[] override; voice per-variant independent download with #1749 strict guards; bgm merged via merge_bgm_config — batch passthrough chain is correct. - Worker generation concurrency = 2 (compose.yml default); USER_PENDING_LIMIT=3, 6 tasks will enqueue in two waves (first 3 → then next 3 as workers free up); GLOBAL_PENDING_LIMIT allows it. This is expected behavior, not a bug. - Legacy key mismatch: subtitle_render_engine.py reads plan_config['subtitle_config'] (snake_case) but normalize_plan_config writes 'subtitle' short key. This file is not imported by unified_render_service (which uses render_subtitles.py with explicit subtitle_config= parameter), so no runtime impact; left for separate cleanup. - Frontend usePlanConfigLoader.ts reads config.title_config snake_case while backend writes 'title' short key — frontend-only issue for subsequent edit sessions, out of backend scope.	2026-09-29 18:08:18 +08:00
Coze Agent	0a004db1bd	feat(generate): 批量生成支持>3视频排队提交 + 两列标题 + 独立配音 (#2096) ... - P0: 批量生成由一次性提交count=N改为串行count=1逐任务提交； 遇到 429 USER_QUEUE_FULL / 503 SYSTEM_QUEUE_FULL 自动读取 estimated_wait_seconds 后重试，不再弹错误阻止提交； 新增 useGenerationPolling.pollBatchTaskQueued 支持增量排队轮询（不触发整体 onComplete）； 卡片新增 queued 状态（时钟图标+排队文案）； 单个任务提交失败不阻断其余任务，失败卡片独立展示错误+重试按钮； generating 状态改为所有任务到达终态(completed/awaiting_cover/failed)且无queued/running才关闭。 - P1: Step4 批量标题改为 CSS Grid 两列布局（gridTemplateColumns: repeat(2,minmax(0,1fr))）， 单字段 maxWidth 由 640 改为 100%，纵向空间节省约一半。 - P1: Step3（标题页）集成 Step3VoiceWithMode，批量模式下支持共用/独立配音切换； 独立配音时每个视频可单独选择音色，默认继承全局配音； GenerateStepContent props 已从 GeneratePage 透传，之前组件只定义未渲染，本次补齐 JSX 引用。 - BatchTaskState.status 新增 queued 枚举； - useGenerateVideo 重构 generate 函数，拆分 buildBasePayload/submitOne 逻辑， 新增 cancelledRef/queueTimersRef 做 unmount 清理。	2026-09-29 18:05:20 +08:00
xiaoxia	3be06c5763	fix(worker): align gpu-direct title position/size with frontend preview (#2095) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-29 17:04:37 +08:00
xiaoxia	02199d80ee	Merge pull request 'chore(worker): BGM preset diagnostics + upload helper' (#2094) from chore/preset-bgm-audio-upload-helper into develop	2026-09-29 16:03:29 +08:00
Coze Agent	ca6803e1a5	chore(worker): move bgm upload readme under docs/ to avoid ruff parsing md	2026-09-29 15:35:34 +08:00
CI Bot	d449496f90	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-29 07:33:22 +00:00
Coze Agent	a04e363d1b	chore(worker): improve BGM preset download diagnostics + add upload helper script ... - _prepare_bgm preset branch now differentiates: preset not found / preset exists but audio_url empty (未部署) / download failure, with a clear hint to ops to upload audio and fill preset_bgm.py - final warning now surfaces enabled/audio_url/asset_id/preset_id state so triage is immediate - add scripts/upload_preset_bgm.py: one-shot uploader that uploads mp3s from ./bgm_assets/{preset_id}.mp3 to OSS preset/bgm/, sets public-read ACL, and rewrites packages/domain/preset_bgm.py with the public URL. Dry-run supported. - add bgm_assets/README.md with the 10-preset file naming table	2026-09-29 15:28:36 +08:00
xiaoxia	da59a6c9a6	fix(gpu-direct): passthrough template title/subtitle/BGM config (#2093) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-29 14:54:32 +08:00
Coze Agent	11c554e43a	fix(gpu-direct): passthrough template title/subtitle/BGM config ... Problem: GPU direct render MVP hardcoded title/subtitle/BGM styles and ignored template config fields: - P0 static subtitle text (subtitle.text) was never rendered - P0 title styles (font/size/color/position/stroke/shadow/bold) all hardcoded - P1 BGM volume hardcoded at 0.35, no fade/offset/adjust-db - P1 ASR subtitle styles (color/size/position/font) ignored - P2 extra_audio_tracks (TTS/voiceover) volume hardcoded at 1.0 Changes: 1. build_direct_render() now accepts title_config/subtitle_config/ bgm_config dicts + static_subtitle_text. 2. Added helpers _hex_to_drawtext_color (#RGB/#RRGGBB/#RRGGBBAA/named), _position_to_drawtext_xy (top/center/bottom), _build_drawtext_filters (shadow layer + main layer, stroke via borderw/bordercolor, bold simulated via same-color borderw). 3. Title parses font/size/color/position/stroke/shadow/bold with safe defaults; legacy title_text still used as fallback. 4. Static subtitle_text renders as full-span (0→total_duration) segment; when ASR segments are present they take priority. 5. BGM reads volume (default 0.3), volume_adjust_db (linear gain), fade_in/fade_out (afade), audio_offset (adelay ms|ms), and honors enabled=false even if path supplied. Volume clamped to [0,1.5]. 6. _try_gpu_direct() now passes cfg['title']/['subtitle']/['bgm'] whole dicts through, computes static_subtitle_text for non-auto subs, and sources extra_audio_tracks volume from audio_tracks_config (voiceover tracks) instead of hardcoding 1.0. 7. DirectRenderPlan exposes filter_complex for testing. 8. Added 24 unit tests covering: default backward-compat, color hex/BGR conversion (short/long/alpha/invalid), position (top/center/bottom), stroke, shadow (two drawtext layers), font override, enabled flag, static subtitle full-span, ASR style passthrough, BGM volume/fade/offset/adjust-db/enabled=false, extra audio volume, bold simulation, ASR vs static priority, volume clamping. Fixes template-style rendering not taking effect in GPU pipeline.	2026-09-29 14:44:11 +08:00
xiaoxia	f9f3e6bfb9	fix(worker+api): P1 封面评分时序bug + direct/complete吞ingest占位bug (#2092) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-29 04:24:48 +08:00
xiaoxia	3904a8f3b5	perf(worker): 封面抽帧三优化 — blackdetect黑屏规避+单次ffmpeg批量抽帧+并发上传+CJK字体兜底 (#2091) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-28 22:32:56 +08:00
xiaoxia	ba3e97c986	fix(worker): GPU直连支持配音素材库音轨+无声片段anullsrc+per-clip音量 (#2090) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-28 21:09:10 +08:00
xiaoxia	3106496c12	fix(worker): GPU直连管线音频bug修复+边缘裁剪入GPU省5s (#2089) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-28 20:17:54 +08:00
xiaoxia	f79f75b863	fix(frontend): 轮询识别awaiting_cover状态，visibilitychange恢复轮询，自动跳封面页 (#2088) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-28 19:39:17 +08:00
xiaoxia	b7a439d319	feat(cover): 默认本地ffmpeg抽帧+cv2打分，砍掉MediaKit封面86s瓶颈 (#2087) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-28 18:44:41 +08:00
Agent	4d7c80ae07	fix: OSS签名URL校验HEAD→GET+Range，修复403误报 ... OSS sign_url()将路径中/编码为%2F，HEAD请求返回403但GET正常。 改用GET+Range:bytes=0-0仅下载1字节验证，同时接受206状态码。 根因修复，消除URL校验失败后的object_exists兜底路径。	2026-09-28 18:01:12 +08:00
Agent	26c0140d79	chore: 持久化 staging 宿主机 nginx 配置，包含 GPU relay 路由修复 ... - 新增 deploy/configs/host-nginx-cms-staging.conf: 宿主机 CMS nginx 完整配置备份 - 更新 infra/nginx/gpu-relay-staging.conf: 改为文档说明，指向实际配置 - 确保重建服务器时可从仓库恢复 GPU relay 路由配置	2026-09-28 17:44:13 +08:00
Agent	c507f76c14	fix: update GPU relay nginx config - integrate into CMS 8092 vhost ... Root cause: port 8092 was occupied by CMS nginx vhost, which proxied /api/ to CMS backend (port 8091) instead of staging API (port 8000). P4000 PUT requests to relay endpoint returned 404. Fix: Added /api/v1/internal/gpu-relay/ location block to CMS 8092 server block, proxying to staging API on port 8000. Nginx longest prefix match ensures relay routes take priority over generic /api/.	2026-09-28 17:25:36 +08:00
灵应	8a5cfe831e	fix: 修复 gpu_direct_pipeline.py 中 concat 构建的变量名错误 ... build_direct_render 方法中列表推导式使用 _lbl 作为循环变量， 但 f-string 中误用了未定义的 l，导致 NameError。 修复: f"[{l}]" -> f"[{_lbl}]" (line 195)	2026-09-28 16:22:01 +08:00
灵应	d7d3f3184b	fix: 修复 GPU 直连管线中 l.role 变量名拼写错误 ... _can_use_gpu_direct 和 _try_gpu_direct 方法中 3 处列表推导式 误用未定义变量 l 代替循环变量 _lyr，导致 NameError 使 gpu-direct eligibility check 始终失败并 fallback 到 CPU 路径。 修复: l.role -> _lyr.role (line 2220, 2275, 2280)	2026-09-28 15:31:46 +08:00
业务服务器运维	ec9240b52a	fix: 修复 _storage_key 无法写入 EditPlanClip（slots=True） ... EditPlanClip 使用 @dataclass(slots=True)，不允许 setattr 动态添加属性。 将 _storage_key 改为存储在 clip.config 字典中，所有读取处同步修改。 - render_adapter.py: c._storage_key = sk → c.config['_storage_key'] = sk - unified_render_service.py: getattr → config.get('_storage_key') - gpu_direct_pipeline.py: getattr → config.get('_storage_key')	2026-09-28 14:46:44 +08:00
xiaoxia	d0e5ef1753	Merge pull request 'feat(worker): 全GPU直连渲染，取消mezzanine CPU中间编码' (#2086) from feature/gpu-direct-pipeline into develop	2026-09-28 14:20:48 +08:00
saas-backend-agent	6e8199581d	fix(gpu-direct): 修复单测+style E741，默认字体改为Noto Sans CJK SC ... - DEFAULT_DRAWTEXT_FONT 从 sans 改为 'Noto Sans CJK SC'（P4000已装中文字体） - _download_assets 返回 4-tuple 导致老单测解包失败，更新 test_render_adapter - 修复 E741 模糊变量名 l → _lbl/_lyr - ruff format 统一格式	2026-09-28 14:02:57 +08:00
CI Bot	1e23a3f094	style: auto-format with black + isort + ruff + prettier [skip ci-format-check]	2026-09-28 05:38:30 +00:00
业务服务器运维	b04a803655	feat(worker): 全GPU直连渲染，取消mezzanine CPU中间编码 ... - gpu_encoder 新增 render_inputs_to_output：多输入+filter_complex 直接交 P4000 一次出片 - 新增 gpu_direct_pipeline：原始素材/BGM 签名URL、TTS 上传、drawtext 标题/字幕、 filter_complex 完成 trim/scale/pad/concat/边缘crop/amix，末端 h264_nvenc 仅编码一次 - unified_render_service：render() 步骤4.8 接入直连，命中即出片； 不满足条件或 GPU 失败自动回退现有 mezzanine/CPU 链路；config gpu_direct_enabled 可灰度关闭 - render_adapter：_download_assets 额外返回 asset_storage_map， _do_render 给 clip 注入 _storage_key 供直连签名 消除 mezzanine 85s + 边缘裁剪 26s + 合成 2s + 中转 1s ≈ 114s CPU 开销	2026-09-28 13:30:31 +08:00
xiaoxia	e496f127a3	feat(p1): OSS 双endpoint分离 — 上传/下载走VPC内网，签名URL走公网 (#2085) ... Co-authored-by: xiaoxia <dev@xiaoxiajianji.com> Co-committed-by: xiaoxia <dev@xiaoxiajianji.com>	2026-09-28 11:35:13 +08:00
xiaoxia	423be1446f	Merge pull request 'fix(p0): GPU mezzanine传输切回OSS绕开Tailscale反向卡顿' (#2084) from fix/p0-gpu-mezzanine-oss into develop ... fix(p0): GPU mezzanine传输切回OSS绕开Tailscale反向卡顿 Tailscale 反向链路（P4000 从 116 relay 下载 mezzanine）间歇性 TCP stall， 导致 GPU 编码首字节 20s 超时 fallback 到 CPU 软编，渲染变慢到 ~3 分钟。 改 GPU_ENCODE_MEZZANINE_TRANSPORT=relay → oss：P4000 直接从阿里云 OSS 公网下载 mezzanine，绕开不稳定的 Tailscale 反向链路；编码结果回传仍走 relay（116→P4000 正向 POST 正常）。仅改 .env.staging，部署后 watchtower 自动拉起容器生效。	2026-09-28 08:51:37 +08:00
xiaoxia	6d9d2e8179	fix(p0): GPU mezzanine传输切回OSS绕开Tailscale反向卡顿 ... Tailscale 反向链路（P4000 从 116 relay 下载 mezzanine）间歇性 TCP stall， 导致 GPU 编码 20s 首字节超时 fallback 到 CPU 软编，渲染变慢到 ~3 分钟。 改 transport=oss：P4000 直接从阿里云 OSS 公网下载 mezzanine，绕开不稳定的 Tailscale 反向链路；编码结果回传仍走 relay（116→P4000 正向链路正常）。	2026-09-28 08:31:20 +08:00