Files And Knowledge
Upload files and build searchable workspace knowledge collections.
Files and knowledge collections are workspace-scoped app features that can support richer chat workflows.
Files
File uploads use multipart form data and return 202 Accepted because scanning and maintenance are asynchronous. The file service supports listing recent files, reading metadata, creating signed URLs, and deleting files.
Knowledge Collections
Knowledge collections organize documents for retrieval. Documents can be added from inline text or workspace files.
During ingestion, Nodal parses supported file types, chunks content, requests embeddings through the billing/gateway path, and stores searchable chunk metadata.
Supported Knowledge File Types
Supported parsing includes plain text, Markdown, CSV, TSV, HTML, DOCX, and PDF text extraction. PDF parsing can use an optional vision path when configured, with cached parsed output to avoid repeated billing for the same parse.
Standalone image OCR is not currently part of knowledge ingestion.
Search
Collection search embeds the query, runs vector search over stored chunks, bills the workspace for the query embedding, and returns ranked chunks with scores.