Skip to main content
veryfront knowledge ingest is the primary path for adding documents to a project’s knowledge base. Use it when an agent needs to turn uploads/... or a local file into knowledge/*.md.

​
Ingest one document

Use a remote upload path:
Terminal
veryfront knowledge ingest uploads/contracts/q1.pdf --json
Use a local path when the file already exists on disk:
Terminal
veryfront knowledge ingest ./contracts/q1.pdf --json
Inside a sandbox, /workspace/uploads/... is a local path:
Terminal
veryfront knowledge ingest /workspace/uploads/contracts/q1.pdf --json
The CLI resolves the source, downloads remote uploads into the workspace when needed, parses the document, and writes the result back to knowledge/<slug>.md.

​
Set auth and project context

Set the API token and project slug in your environment:
.env
VERYFRONT_API_TOKEN=vf_your_api_key
VERYFRONT_PROJECT_SLUG=my-project
Or log in interactively:
Terminal
veryfront login
Pass --project when you want to override the current project:
Terminal
veryfront knowledge ingest uploads/contracts/q1.pdf --project my-project --json

​
Choose the source path

PathMeaningBehavior
uploads/...Remote project uploadDownloads from the uploads store before parsing
./... or /absolute/path/...Local file or directoryParses directly from disk
/workspace/uploads/...Local sandbox fileParses directly from the sandbox workspace
./uploads/...Local path named uploads/Parses directly from disk
Use uploads/... for project uploads. Use ./uploads/... when the file is already in the current workspace.

​
Ingest a batch

Ingest every supported upload under a remote prefix:
Terminal
veryfront knowledge ingest --path uploads/contracts --all --json
Ingest every supported file under a local directory:
Terminal
veryfront knowledge ingest --path ./contracts --all --recursive --json
Each source document becomes its own markdown file in the project knowledge tree.

​
Verify the result

Read the generated knowledge file with veryfront files get:
Terminal
veryfront files get knowledge/demo-notes.md
For machine-readable output, keep --json enabled during ingestion:
Result
[
  {
    "source": "uploads/demo/notes.txt",
    "remotePath": "knowledge/demo-notes.md",
    "slug": "demo-notes",
    "sourceType": "txt",
    "summary": "Converted document to markdown (87 chars).",
    "stats": {
      "characters": 87,
      "lines": 3
    },
    "warnings": []
  }
]

​
Use low-level commands when you need manual control

veryfront knowledge ingest is the default path. Use the lower-level CRUD commands when you need to inspect or move files yourself:
Terminal
veryfront uploads list
veryfront uploads pull uploads/contracts/q1.pdf --output /tmp/q1.pdf
veryfront files list --path knowledge/ --json
veryfront files get knowledge/demo-notes.md

​
Requirements

Install python3 before running veryfront knowledge ingest. Outside the Veryfront sandbox image, install the parser dependencies for non-text formats:
Terminal
pip install pandas openpyxl xlrd pdfplumber python-docx python-pptx beautifulsoup4 lxml
The command supports pdf, csv, tsv, docx, xlsx, xls, pptx, html, htm, txt, json, md, and mdx.

​
Next

  • Agents: Build agents that call CLI and tool workflows.
  • Tools: Define the tools that wrap or complement ingestion flows.
  • Workflows: Orchestrate repeatable ingestion and follow-up processing.