veryfront knowledge ingest is the primary CLI workflow for getting documents
into a project’s knowledge base. It finds a source file, parses it, and writes
generated markdown back into the project.
Ingest one uploaded file:
Prerequisites
Authenticate with the CLI and set the target project:veryfront knowledge ingest parses PDF, Office, EPUB, HTML, and RTF sources
through the built-in Kreuzberg document extension. Plain text, Markdown, JSON,
CSV, TSV, and common code files are converted directly by the CLI.
Single-file examples
Ingest a remote upload
Useuploads/... to reference a file from the project’s remote uploads store:
Ingest a local file
Use a local path to ingest a file that already exists on disk:Exact-file batch ingestion
To ingest a specific list of files without ingesting the entire folder:ingested array. Agent workflows
can match each output back to the original source path.
Batch ingestion
To ingest every supported file under a remote uploads prefix:--path ... --all only when you want everything under that uploads prefix
or local directory. For an exact file list, pass the file paths as positional
arguments instead.
What the JSON output looks like
With--json, the command returns a machine-readable run result with
ingested, skipped, and failed arrays:
stats shape varies by source type, but the top-level result fields
are stable.
Path rules
The source path determines how the command behaves:uploads/...means a remote project upload./uploads/...means a local file or directory relative to the current working directory- multiple explicit sources are passed as positional arguments:
veryfront knowledge ingest <source...> --json
uploads/... triggers the remote upload
download step, while local paths skip it.
Supported file types
veryfront knowledge ingest supports these source formats:
pdfcsvtsvdocxxlsxxlspptxhtmlhtmtxtjsonmdmdx
Troubleshooting
Unknown command: knowledge
Your installed CLI is older than the branch or release that added the command.
Update the CLI or run the current source tree directly with:
Missing API token
Set VERYFRONT_API_TOKEN, run veryfront login, or use a local CLI config with
a saved token.
Could not determine project slug
Set VERYFRONT_PROJECT_SLUG or pass the project explicitly:
Document extraction errors
Use a supported document type and ensure the source file is readable. Rich document formats use the built-in Kreuzberg extension, while text-like formats are converted directly by the CLI.Verify it worked
After ingesting a source, the command writes one or more markdown files under the project’sknowledge/ directory:
knowledge/<name>.md entry. Open the
generated markdown and confirm the parsed content matches the original.
For automation, capture the JSON output of the command directly:
ingested array names every file the command wrote. If the array is empty
or the command exited non-zero, check skipped, failed, and the command
output for the reason.