Skip to main content

Capacity and limits

Capacity limits and usage restrictions of the Alibaba Cloud Model Studio knowledge base across workspaces, file uploads, chunking, retrieval and Q&A services, datasets, and API calls.

Workspaces and knowledge bases

A workspace is the top-level container for resource isolation. Each workspace can host multiple knowledge bases.
ItemLimitNotes
Workspaces per account30The primary account's workspace is marked as default
Knowledge bases per workspace50Quota increase available via support ticket
Knowledge base name length1–20 charactersMust be unique within the workspace
Knowledge base description length200 characters—
Documents per knowledge base100,000Consider splitting the knowledge base if exceeding this

File uploads

Knowledge bases support various file types—including documents, spreadsheets, images, audio, and video—each subject to distinct format and size constraints.

Documents and text

TypeFormatPer-file limit
Documents.doc / .docx / .ppt / .pptx / .pdf150 MB or 1,000 pages
Plain text.md / .txt / .htmlRecommended ≤ 10 MB
Spreadsheets.xls / .xlsxRecommended ≤ 10 MB and ≤ 100,000 rows
PDFs undergo direct text extraction; scanned PDFs are automatically processed with OCR. Spreadsheets are chunked row by row, with the first row used as field names. Markdown files are split by heading; HTML files have script and style tags automatically stripped.

Images

ItemLimit
Formats.png / .jpg / .jpeg / .bmp / .gif
Single image size≤ 20 MB
Dimension constraintsShortest side > 15 px; longest side < 8192 px; aspect ratio < 50

Audio and video

ItemLimit
Audio formats.aac / .amr / .flac / .flv / .m4a / .mp3 / .mpeg / .ogg / .opus / .wav / .webm / .wma
Video formats.mp4 / .mkv / .avi / .mov / .wmv
Single file size≤ 512 MB
Files per import50 (combined total of audio and video files)

General upload limits

ItemLimit
Files per upload50

Chunking

Chunks serve as the retrieval units of a knowledge base. The chunking method and maximum segment length are configured during data import.
ItemRangeDefault
Maximum segment length10–6000600 (tokens)
Chunk content length0–6000 charactersApplies when manually adding chunks
Chunk title length0–50 charactersUsed for display in chunk lists
The chunking method (intelligent splitting / by length / by page / by heading / by regular expression / by symbol) and the maximum segment length are configured during data import and cannot be modified after creation.

Knowledge retrieval and Q&A services

Retrieval and Q&A services are the external invocation entry points of knowledge bases. The following limits are configured when creating a service.
ItemRangeDefault
Knowledge bases per service15—
Service nameMax 40 charactersRequired
Service descriptionMax 200 charactersOptional
Q&A promptMax 500 charactersOptional
Initial vector retrieval TopK1–100100
Initial keyword retrieval TopK1–100100
Similarity threshold0.01–1.0—
Max recall count1–205
Tag lengthMax 32 characters per tagPress Enter to confirm input

Datasets

Datasets are used to import files from external data sources into knowledge bases.
ItemLimitNotes
File dataset storage (platform-hosted)100,000 files / 1 TBFree for a limited time
Files per upload50—
Dataset name length20 characters—
Dataset description length200 characters—

API limits

Batch operations

OperationPer-request limitDocumentation
Batch delete chunks10 chunksDelete chunks
Batch query files20 file IDsQuery file details

Request rate limits

Each API has a request rate limit (QPS). Exceeding the limit returns HTTP 429. For complete throttling rules, see Rate limits.
API categoryMax QPS
Retrieval2,000
Task status query20
Document / file query5–15
Knowledge base management10
Chunk management10
Dataset management2–5
Monitoring data1
If the current quotas cannot meet your business needs, submit a support ticket to request a quota increase.
Agent Development
Agent Context
Agent Governance & Evolution
Store
Application Marketplace