RAGfly CLI

Operate RAGfly from the terminal and CI — login + the full cloud API surface against api.ragfly.ai. Lightweight: depends only on click, rich, httpx and keyring. No desktop / PySide6 / OCR dependencies.

pip install ragfly-cli
ragfly version
ragfly login
ragfly cloud me

This package ships the ragfly binary for scripting and automation. It is distinct from pip install ragfly (the Python SDK, import import ragfly) and from RAGfly Desktop (the DMG/exe that bundles the local file worker for ragfly local scan/sync/daemon).

Authentication

# A person (JWT, stored in the OS keyring)
ragfly login

# Non-interactive / CI / agents (API key, only reaches /v1)
export RAGFLY_API_KEY=rf_xxxxxxxxxx

cloud group and cloud api-key manage a person's session and credentials: they need ragfly login, and the server refuses them for an API key.

Opening an original file on disk (optional)

Searching and citing need no local path setup. A document result may include an fs object. For a local relative file, read fs.home_var and join that environment variable's value with fs.relative_path. Each root has its own variable; configure only the roots available on this machine.

export RAGFLY_HOME_442681="/Users/ana/Dropbox"
export RAGFLY_HOME_991203="/Volumes/Archive"

If fs.is_cloud_only is true, use the authorized cloud provider. If fs.is_public_url is true, open its public URL. If fs.is_absolute is true, use the absolute path on the machine where it exists. If fs.home_var is null or unset, do not guess a local root. See the MCP guide for the full order and examples.

Command surface

ragfly
├── login / logout / version
└── cloud                      ← the English REST /v1 contract of api.ragfly.ai
    ├── me
    ├── group          list | switch | clear      (signed-in person)
    ├── api-key        create | list | revoke     (signed-in person)
    ├── document       list | show | edges
    ├── space          list | show | compose | read | refresh | promote
    ├── queue          show | runs
    ├── skill          list | show | run
    ├── catalog
    ├── document-type  list
    ├── characteristic list
    ├── function       show
    ├── search
    ├── chat           ask
    ├── agent          context | tool
    ├── usage
    ├── conversation   list | delete
    ├── process        list | show | update
    ├── organization   show | update | draft
    └── operation      list | show | run

Every operation of the RAGfly application is available through cloud operation, with the same permissions and audit as the web app:

ragfly cloud operation list
ragfly cloud operation show document_types.update          # input/output schema
ragfly cloud operation run document_types.update --input-json '{"code": "TDOC_...", "name": "Invoices"}'

# write_confirm operations (deletes, reverts, resets) only run with --confirm
ragfly cloud operation run document_types.delete --input-json @input.json --confirm

--input-json takes inline JSON, @file or - (stdin). Never put secrets in it. Output is English end to end (--status VECTORIZED, -o json | jq).

ragfly cloud document list --status VECTORIZED --limit 20 -o id
ragfly cloud skill run SUMMARIZE_DOCUMENT --space 12
ragfly cloud search "Q1 revenue"
ragfly cloud search "statements" --filter @filter.json --space 12
ragfly cloud chat ask "How do I release the entity of my key?" --mode help

Full reference: https://api.ragfly.ai/docs.

Local operations (ragfly local scan/sync/daemon) require the local file worker and ship with RAGfly Desktop, not with this package.

Relation to RAGfly Desktop

RAGfly Desktop bundles its own ragfly binary with the local commands and a signed-in person's session. This package is the standalone cloud surface: since 1.19.0 its cloud commands use only the public /v1 contract, so they no longer mirror the Desktop copy.