希望能够支持调用嵌入模型api构建index

[bigben446]

对于几千甚至上万的pdf大型zotero库而言，仅仅依靠电脑自己的硬件速度实在是太慢了。
可以使用大模型api构建index，速度快而且价格不贵。

[bigben446]

Support Online Embedding APIs + Resumeable Full Indexing for Large Libraries

I would like to kindly request two important features that would make the tool much more practical for users with large collections (especially 10,000+ PDFs):

1. Support for Online Embedding API Providers (during indexing)
   Please add the ability to use cloud embedding APIs (e.g. OpenAI text-embedding-3-small/large, Alibaba Cloud DashScope text-embedding-v3/v4, ByteDance Doubao-embedding, etc.) for the PDF chunking + embedding phase.
   Currently the app only uses local SentenceTransformers models. For libraries with thousands or tens of thousands of PDFs, the initial full indexing is extremely slow (days or even weeks). Online APIs are dramatically faster, have excellent batch support, and are very affordable for this one-time bulk task (often just a few dollars for 15k PDFs).

2. Support resuming a Full Index after interruption
   When indexing a large Zotero database, the process can easily take many hours or days.
   It is very important to be able to pause/stop the indexing and resume later (or automatically resume after a crash, computer sleep, power outage, etc.) without having to start the entire process from scratch. Incremental/resumable indexing is critical for real-world large-library use cases.

These two features would make Zotseek the go-to solution for serious researchers and heavy Zotero users with big PDF collections.
Thank you very much for considering these suggestions.