Skip to main content
A file bucket holds documents instead of text snippets. You upload a file, MemoryPlugin extracts its text, splits it into chunks, and embeds those chunks for semantic search. After that you can ask questions about the document and get answers that cite the exact file and page.
File buckets are still under development. Today you can upload and search files through the dashboard only (the Ask tool, plus the MCP and TypingMind search tools). They are not yet wired into the browser extension or auto-injected into your chats.

​
Memory buckets vs file buckets

A memory bucket stores short text memories you write or import. A file bucket stores whole documents that get processed for search. Both live in the same bucket selector, grouped under Memories and Files.

Persistent

Files stay available across conversations and tools

Semantic search

Matches passages by meaning, not just keywords

Cited answers

Results name the file and page they came from

Queryable

Search from Ask, the MCP server, or the TypingMind plugin

​
Supported file types

FormatExtensionsNotes
PDF.pdfText is extracted from the pages
Word.docx, .docMicrosoft Word documents
Markdown.mdPlain Markdown
Text.txtPlain text
The upload area states the limit plainly: PDF, Word, Markdown, Text (max 10MB each). Anything else is rejected with “Unsupported file type”.
Scanned or image-only PDFs are not supported. The PDF has to contain selectable text, because extraction reads the text layer, not the image.

​
Create a file bucket

1

Open the bucket selector

Click the bucket dropdown at the top of the sidebar.
2

Choose New File Bucket

Click New File Bucket. The Create New Bucket dialog opens with Bucket Type set to Files.
3

Name it

Type a name under Bucket Name (for example, “Research Papers”), then click Create Bucket.

​
Upload files

File bucket view with the upload dropzone and a table of files in Ready and Processing states
1

Select the file bucket

Pick your file bucket in the sidebar so its contents show.
2

Add files

In the dropzone (Drop files here or click to upload), drag files in or click to browse. You can add several at once; they upload one after another.
3

Wait for processing

Each file moves through Processing to Ready. A file that fails extraction shows Failed, and you can retry it. You can only search a file once it reads Ready.
The file table lists Name, Type, Size, Status, and Uploaded date. Delete a file from its row when you no longer need it.

​
Query your files

Once files are Ready, ask questions about them from any of these: Ask tool. Open Ask, switch to Files mode, and optionally scope to one file bucket. Answers cite the file name and page. MCP. Call the search_uploaded_files tool: 
Tool: search_uploaded_files
Parameters:
  - query: "your search query"
  - bucketId: (optional) a specific file bucket
  - topK: (optional) number of results
See MCP server. TypingMind. The plugin’s search_uploaded_files function searches your documents inside a TypingMind chat. See TypingMind plugin.

​
Use cases

  • Reference docs. Upload API references or style guides and query them without leaving your AI chat.
  • Research. Store papers and reports, then ask questions that pull across several of them.
  • Project context. Keep specs, notes, and requirements searchable for every related conversation.

​
Best practices

Separate buckets for separate projects let you scope a search to just the relevant documents.
File names appear in results. A clear name tells you at a glance which source an answer came from.
A tight, single-topic document usually retrieves better than one huge file that mixes everything together.
A file is not searchable until its status reads Ready. Large files take longer to get there.

​
Limits

  • File size: 10 MB per file.
  • Processing: files are queued and handled in order. Larger files take longer to become searchable.

​
Next steps

Ask

Query your files in Files mode

Memory Buckets

Store text memories alongside your files