Context plans and limits

What your plan limits in Sanity Context, how source chunks are counted across your organization's Knowledge Bases, and what happens when you reach a limit.

Your organization's plan sets how many Knowledge Bases it can have and how many source chunks they can hold between them. Source chunks are one pool shared by every Knowledge Base in the organization, and you decide how to spread it across them. Only source chunks that a build has taken in count toward the limit. For what a source chunk is, see Sources, documents, and source chunks in Knowledge Bases.

Each plan sets the number of Knowledge Bases, the organization's source chunk limit, and whether scheduled refresh is available:

For example, on Self-serve one Knowledge Base can hold 40,000 source chunks and the other four share the remaining 10,000.

On every plan, each Knowledge Base also holds at most 5,000 documents. File size limits for uploads are listed in Knowledge Base source types. For prices, see the pricing page.

Creating a Knowledge Base beyond your plan's number is refused with this message:

The organization's plan limit for knowledge bases is reached (1 of 1 used). Upgrade the plan to continue.

The source chunk limit counts what builds have taken in, across every Knowledge Base in the organization:

• Source chunks count once a build takes them in. A Knowledge Base that has never been built counts nothing. A source added since a Knowledge Base's last build counts nothing until the next build or scheduled refresh decides whether it fits.
• Source chunks a build took in stay in. A document that grows, or a move to a smaller plan, never drops a page from a built Knowledge Base, even when that takes the organization past its limit.
• Excluded documents don't count. A document a build had no room for is kept, but it doesn't count toward the source chunk limit.
• Removing a source frees its source chunks and its documents.
• Files from an upload that failed don't count.
• Builds and refreshes don't use anything up. The limit measures what your Knowledge Bases hold, not how often they build.

Every build and scheduled refresh decides, when it starts, which of the Knowledge Base's documents it takes in:

• Documents the Knowledge Base already took in stay in.
• New documents go in the order they were added, each one if it fits in the room left. The room is the organization's source chunk limit less what every Knowledge Base in the organization has already taken in, and no more than the Knowledge Base's 5,000-document limit.
• A document that doesn't fit is excluded. A later, smaller document can still go in.

A document goes in or stays out whole, so a long file that Context split into sections goes in with all its sections.

An excluded document stays in its source and shows as excluded, but no build or refresh reads it, so nothing in the Knowledge Base comes from it. The next build or scheduled refresh with room for it takes it in, for example after you upgrade the plan or remove another source.

For example, an organization on Free has 9,800 of its 10,000 source chunks taken in and uploads a PDF that becomes about 500 source chunks. The upload succeeds, and the next build excludes the PDF. Once the organization has room for it, the next build takes it in.

Adding a source checks a separate, higher ceiling: everything the organization stores, including excluded documents and sources in Knowledge Bases that have never been built. Adding a website crawl, dataset source, or upload to any Knowledge Base is refused once the organization stores 30% more source chunks than its plan allows:

The organization's plan limit for source chunks is reached (13,010 of 10,000 used). Upgrade the plan to continue.

Context only knows a source's size once it has processed it, so one large upload or dataset added below that ceiling can take the organization past it. The next source added is then refused.

Once the organization's Knowledge Bases have taken in as many source chunks as the plan allows:

• Built Knowledge Bases keep serving everything their last build took in.
• Building a Knowledge Base is refused when the organization's other Knowledge Bases hold the whole limit:

The organization's plan limit for source chunks is reached (10,000 of 10,000 used). Upgrade the plan to continue.

• A build that can take in none of the Knowledge Base's documents fails before it starts any work:

None of this knowledge base's documents fit under the organization's plan limit for source chunks (9,950 of 10,000 used by other knowledge bases). Remove sources or upgrade the plan to build it.

• Scheduled refreshes still run. Pages and Sanity documents already in a Knowledge Base keep refreshing, even if their content grows. New documents a refresh finds, such as Sanity documents that newly match a dataset query, are excluded until there's room.
• A recrawl stops adding new pages once their source chunks fill the remaining room, and completes as partial.
• A crawl that starts with no room fails:

The organization's plan limit for source chunks is reached (10,000 of 10,000 used), so no pages can be added. Remove sources or upgrade the plan to continue.

To make room, remove sources you don't need, or narrow a dataset query or crawl scope. To raise the limit, buy Self-serve in Manage, or talk to Sanity sales about Knowledge Base add-ons.

Each Knowledge Base holds at most 5,000 documents across all its sources. Context checks this while it imports a dataset source or an upload, so the request to add the source succeeds and the import then fails. When the import would go past the limit, the message reads:

This import would add 1,200 documents, but this knowledge base only has room for 900 more (4,100 of 5,000 used). Remove sources to make room, or add a smaller import.

An upload can also go past the limit only once its files are split. That import fails with:

This knowledge base has 5,210 documents, which exceeds the 5,000-document limit. Remove sources to get under the limit.

A website crawl stops adding pages when the Knowledge Base runs out of room, and completes as partial.

In the Context app, the Sources list on each Knowledge Base shows how many documents and source chunks it holds, and each source shows its own count. A crawl's page list shows the source chunks in each page and marks pages a build left out as excluded.

• Sources, documents, and source chunks in Knowledge Bases. How material becomes source chunks, and what changes how many a source uses.
• Keep a Knowledge Base current. Remove or narrow sources to make room.