organize-files

Organize Files

Use moviepilot_api for every MoviePilot business operation. Retired file, recognition, transfer, and history tools are not available.

Workflow

1. Establish scope. If the user provides a path, use it. If they identify a downloader task, use downloader-operation and its fixed scripts/mp-downloader.py helper to discover the instance and call tasks.list. If they only name a configured root, call storage.settings. Use storage.list to inspect the selected directory. Do not process a broad shared root without an explicit, bounded scope.
2. Classify files into movie, TV, one music recording, one complete album, subtitle/sidecar, or unrelated content. Do not group unrelated media merely because they share a directory.
3. Call media.recognize with the representative title or path. If uncertain, call media.search; if several exact candidates remain, use ask_user_choice. Never invent or translate an ID.
4. Preserve the exact media_source + media_id. For TV, verify season detail with media.detail when numbering is ambiguous. For music, a recording is one track, an album is one multi-track directory, and an artist is browse-only.
5. When duplicate risk matters, call library.exists. If an existing transfer record affects reorganization, inspect transfer.history.
6. Before a state-changing transfer, summarize the source, target identity, media type, season/music entity, storage, and mode. Continue only when the user's request already authorizes that exact action or after confirmation.
7. Call transfer.file once per verified unit. For an album, transfer the album directory once only after its supported audio-file count is consistent with the selected album detail.
8. If requested, call media.scrape after a successful transfer. Report actual tag, cover, and lyrics counts; never assume all lyrics were found.

Structured Calls

• Directory listing: storage.list with storage/path/paging/sort in body.
• Recognition: media.recognize with title/path in query.
• Search: media.search with title/type/source constraints in query.
• Detail: media.detail with path_params.media_id and identity/type in query.
• Library check: library.exists with the exact identity in query.
• Transfer: transfer.file with the manual-transfer request in body.
• Actual transfer responses include data.items even when the batch reports failure. Inspect every item's state: accepted and retry_wait mean that background work remains; only completed confirms execution and settlement. manual_review requires resolving the task in the transfer queue first; failed and skipped do not mean imported. Do not resubmit an entire batch that already contains accepted or completed items. Preview responses remain planning data and do not contain execution states.
• Scrape: media.scrape with path_params.storage, file item in body, and exact identity/type fields in query.

Stop and report instead of transferring when the source is missing, directory configuration is absent, identity remains ambiguous, an album appears mixed or incomplete, or the requested target would overwrite unrelated media.

整理恢复与历史上下文

保留 storage.list 返回的完整文件项，包括可选的 modify_time_ns 和 change_time_ns；不要自行生成本地版本凭据。恢复发现源或目标版本变化时， 应报告人工复核状态并保留现场，不通过重新提交整个批次绕过核验。 复用历史整理电视剧时保留 episode_group；不同剧集组分批处理。 目录整理的 overwrite_skipped 同样表示未导入，不能报告为成功。