HyxiaoGe/ai-audio-assistant-web
AI 音视频助手后端：多厂商 LLM/ASR 统一接入，支持转写、结构化摘要、成本优化 与健康监控。
What's novel
AI 音视频助手后端：多厂商 LLM/ASR 统一接入，支持转写、结构化摘要、成本优化 与健康监控。
Code Analysis
8 files read · 3 roundsA production FastAPI + Celery backend that transcribes long audio/video (multi-cloud ASR), polishes transcripts, generates structured multi-style summaries with images, and integrates YouTube, with quota/cost management and content moderation.
Strengths
Genuinely deep implementation: a real 6-dimension weighted ASR scheduler, multi-strategy SmartFactory service selection with health/cost/performance scoring, careful security (SSRF, upload-key ownership, provider validation), and nuanced error handling that distinguishes media-stream vs. envelope responses with i18n and trace-id propagation. The code is well-organized with clear separation across API/services/core/worker and thoughtful, well-commented edge-case handling.
Weaknesses
The core problem (audio/video summarization) is well-trodden, so novelty is moderate; the SmartFactory's fault-tolerance wrapper is a documented no-op stub, and some logic (e.g. task_service.py at 56KB, process_audio.py at 58KB) is concentrated in large files that could be decomposed further. I could not verify test depth directly, though a 202-file test suite and coverage config exist.
Score Breakdown
Signal breakdown
Innovation
Craft
Traction
Scope
Evidence
Commits
131
Contributors
1
Files
218
Active weeks
11
Repository
Language
Python
Stars
1
Forks
0
License
—