myagents-anydoc

myagents-anydoc — local document to Markdown

This is the fully offline document conversion capability shipped with the MyAgents App. The underlying tasks are owned by the Desktop App; CLI exit, the current session ending, or switching Runtime will not transfer task ownership.

Usage order

1. First run myagents anydoc --help to choose a command.
2. Before invoking, read the exact leaf help, for example myagents anydoc convert --help; parameter and format support is subject to the help of the currently installed version, do not add parameters from memory.
3. Add --jsonwhen results need to be read by a machine. Do not add --dry-run。

Core commands

myagents anydoc convert --file <input> [--output <directory>] [--password <password>] [--wait] [--json]
myagents anydoc status <job-id> [--json]
myagents anydoc wait <job-id> [--json]
myagents anydoc cancel <job-id> [--json]
myagents anydoc list [--limit <1..100>] [--json]

convert accepts only one local file at a time.--output is always the output root directory, not the file address of document.md ; when omitted, the default conversion directory of the current Workspace is used. When there is no confirmable Workspace, explicitly pass a directory as prompted by the error.

All conversions are asynchronous jobs. By default they return immediately after submission; use --wait or a separate waitonly when the user or subsequent work must obtain the result. Ctrl-C only stops this wait and will not cancel the job held by the App; to truly stop processing you must explicitly call cancel。

Successful artifacts are located in <output-root>/<job-id>/document.md, and may include assets/referenced relatively by the Markdown. When reading later, use the actual documentPathreturned, do not guess the job directory.

Security and boundaries

• Only pass local regular files; do not pass URLs, directories, stdin, globs, or special devices.
• Passwords are only passed via --password when explicitly provided by the user; do not repeat the real password in replies, logs, or recovery commands.
• No GPT, API key, Hugging Face token, system Python/Node, Docker, GPU, or runtime network access is required.
• On failure, keep code、suggestion and the recovery command; continue as prompted, and do not treat a partial directory as a successful artifact.
• When web scraping, batch directory ingestion, knowledge base indexing, or generative visual understanding is needed, choose the corresponding tool; AnyDoc does not provide these capabilities.

Common flows

Submit immediately and let the Agent query later:

myagents anydoc convert --file ./proposal.docx --json
myagents anydoc status <job-id> --json

Current work must wait for the result:

myagents anydoc convert --file ./scan.pdf --output ./converted --wait --json

Recover a forgotten task ID:

myagents anydoc list --limit 20 --json

For any uncertain item, return to the exact help:myagents anydoc <command> --help。