xpanderaihub/brain-ingest repository overview

⁠xpander.ai brain-ingest

Parses documents with Docling (PDF, Office, HTML, Markdown, scans with OCR), chunks them by structure, and embeds them with BAAI/bge-m3 for the Company Brain's on-site knowledge. Model weights are baked into the image, so it runs air-gapped.

Ships in both self-hosted editions (hybrid and air-gapped). Pull access requires a customer access token issued by xpander.ai.

Docs: https://docs.xpander.ai/self-hosted⁠