Build out new target connector for uploading documents to AstraDB from supported sources (S3, Google Drive, Local).
Requires chunking + embedding for vector retrieval. Chunking can happen within Docling but embeddings are new territory --> Where/how should we do this?
This will also break from some patterns set by existing connectors (S3, Google Drive) in that it is not just simple file storage. --> How coupled/decoupled should this be from old connectors?
Build out new target connector for uploading documents to AstraDB from supported sources (S3, Google Drive, Local).
Requires chunking + embedding for vector retrieval. Chunking can happen within Docling but embeddings are new territory --> Where/how should we do this?
This will also break from some patterns set by existing connectors (S3, Google Drive) in that it is not just simple file storage. --> How coupled/decoupled should this be from old connectors?