Default to HF free embedding (huggingface); require OPENAI_API_KEY only when using openai

Switch to Docker SDK: Fix Gradio API issue and use Docker for stability

Simplify config: Use default Gradio and Python versions

Remove pyaudioop-lts (not needed with Python 3.12)

Fix Python 3.13 audioop issue: Add pyaudioop and set Python 3.12

Fix sdk_version: Use specific Gradio version 4.44.0

Downgrade to Gradio 4.x for HfFolder compatibility

Update Gradio to latest version to fix HfFolder compatibility

Update gradio version constraint

Remove huggingface_hub version constraint, let gradio handle it

Use huggingface_hub 0.20.0 for HfFolder compatibility

Pin gradio and huggingface_hub to compatible versions

Update gradio version constraint and fix sdk_version format

Fix dependencies: Add compatible huggingface_hub version

Fix README: Add required YAML frontmatter for HF Space

Add Weaviate index builder Gradio app

方案 A: 预构建索引 Dataset，启动时直接下载免 embedding

Add retrieve API for Clare integration (api_name=retrieve)

Use Retriever instead of QueryEngine for HF embedding (fix Settings.llm import)

Add free Hugging Face embedding option (EMBEDDING_PROVIDER=huggingface)

Use HfApi + hf_hub_download instead of snapshot_download

Improve dataset snapshot download robustness

Use dataset-backed GENAI COURSES and build vector index

initial commit