Upload arbitrary text data to Swarm as an immutable, content-addressed blob. Returns a Swarm reference hash that permanently identifies the uploaded bytes. Use this tool whenever the user asks to "upload data", "upload text", "store data", or similar, without mentioning a feed, topic, or memory. This is NOT a feed operation — if the user wants mutable, topic-indexed storage (i.e. mentions a feed, topic, or memory name), use update_feed instead. Only data is required. redundancyLevel and postageBatchId are optional — use their defaults and do NOT ask the user for them unless the user explicitly brings them up.

Update a mutable, topic-indexed Swarm feed with new data. Requires a memoryTopic supplied by the user. Use this tool ONLY when the user explicitly mentions a feed, topic, or memory name. If the user asks to upload data without mentioning a feed/topic/memory, use upload_data instead — do NOT prompt the user for a topic to route them here. postageBatchId is optional — do not ask the user for it unless they explicitly bring it up.

Download raw text data from a Swarm reference and return it as a string. Use this tool ONLY when the user explicitly asks for the text content, string content, or raw data behind a reference, or when the reference is known to have been uploaded via upload_data. If the user mentions "file", "files", "folder", or asks to "download" without specifying that they want the raw text content, use download_files instead. When in doubt about the reference type, prefer download_files — it handles both single files and folder manifests and can be saved to disk.

Retrieve the latest data from the feed of a given topic.

Upload a SINGLE file to Swarm. To upload a local file, pass its filesystem path as data — the server reads the file itself (stdio mode only). Alternatively, pass the raw text content directly as data. This tool handles one file only — if the path refers to a directory, or the user mentions "folder", "directory", "the contents of", or otherwise asks to upload more than one file, use upload_folder instead and pass the path as its folderPath. Never ask the user for the file content when a path is given, and never pass a Swarm reference — references are the OUTPUT of this tool, not an input. Small files upload synchronously and return a reference. Large files (over the server's deferred-upload threshold) upload in the background and immediately return the final reference (computed locally) plus a tagId for query_upload_progress; the content becomes retrievable at the reference once the upload completes. With redundancyLevel > 0 only the tagId is returned immediately. Optional options (ignore if they are not requested): redundancyLevel: redundancy level for fault tolerance. Optional, value is 0 if not requested. postageBatchId: The postage stamp batch ID which will be used to perform the upload, if it is provided.

Upload a folder (directory). Use this tool whenever the user mentions "folder", "directory", "the contents of", or asks to upload a path that refers to a directory rather than one file — including phrasings like "upload to Swarm folder ", where is the folder to upload, not a destination. Prefer this tool over upload_file when it is unclear whether a given path is a file or a directory: upload_file cannot upload a directory. folderPath is REQUIRED — pass the folder path from the user's message verbatim. Small folders upload synchronously and return the manifest reference. Large folders (over the server's deferred-upload threshold) upload in the background and return only a tagId; unlike upload_file, a folder's reference cannot be computed up front, so retrieve it by polling query_upload_progress with that tagId until processedPercentage is 100. Optional options (ignore if they are not requested): redundancyLevel: redundancy level for fault tolerance. Optional, value is 0 if not requested. postageBatchId: The postage stamp batch ID which will be used to perform the upload, if it is provided.

Download a file or folder from a Swarm reference. Handles both single files and folder manifests, saves them to disk (in stdio mode) or returns the file list. Use this tool whenever the user asks to "download" from a reference and mentions "file", "files", "folder", or does not specify the data type. Prefer this tool over download_data unless the user explicitly asks for the raw text/string content behind a reference. This is the safe default for downloads when the reference type is unknown.