Reduce avoidable evidence loss before a document enters Knowledge.
Practical procedure
- Remove unrelated repeated navigation and preserve the section labels needed to understand the source. For an authorized export, include its effective date and original document reference.
- Test representative tables and scanned material before importing a full collection. Compare an extracted passage with the original rather than checking only that a document identifier was returned.
- Keep a record of transformations made before import, such as selecting columns or converting to text. Use a separate source when the scope or meaning of an export changes substantially.
Import evidence with its provenance
Create a Knowledge source for the selected project before adding documents. The upload endpoint accepts JSON with base64-encoded file bytes, a title, filename, classification and retention choice. GoCux extracts text and stores the text with a checksum; the original uploaded bytes are not the retained Knowledge document. Keep your authorized original separately when your workflow needs it.
Contract example
Developer examples below describe routes on an authorized, configured GoCux control-plane base URL. They are request shapes or operation references, not claims that a particular public hostname exposes the API. Use the workspace for the corresponding human workflow; availability and permissions depend on the actual deployment.
POST /knowledge/sources/{source_id}/documents — example request body
{
"title": "Renewal note",
"filename": "renewal.txt",
"contentBase64": "UmVuZXdhbCB0ZXJtczogMzAgZGF5cy4=",
"dataClassification": "public"
}Inspect the result
After a successful import, record the document identifier and review the resulting text. Search the project for a known phrase to check that useful content was extracted. Embeddings may be filled later, so immediate verification should include a lexical query. Preserve the source name and extraction context when referring to the document elsewhere.
What to keep explicit
An extracted sentence without its table header or unit can support the wrong conclusion. Preserve enough surrounding context for an independent reviewer to understand the quoted fact.
Handle incomplete and refused work
Invalid base64, an unsupported classification, or extracted text above the configured cap is refused. The request bounds contentBase64 to 40,000,000 characters; this is not a promise that every file of that size or format can be meaningfully extracted. Rich document support depends on deployed extraction capabilities. Missing source and unavailable storage are separate errors.
Continue with related guides
Bring this workflow into your workspace.
Set the scope. Review the evidence. Approve the steps that need it.