← The Quick Suite AI Layer

▶ Play from here · 109 min

Starts this lesson and continues through 14 more to the end of the course.

Spaces, knowledge bases, and the Quick Index

Why this matters

Every AI answer in Quick Suite is only as good as what it can retrieve. Spaces are the container you curate; the Quick Index is the storage and retrieval engine underneath; knowledge bases are the pipes that pull external content in. Get these three straight and every other feature in this module — chat, agents, research, automations — makes sense. Get them confused and you'll be the person asking AWS Support why Salesforce content "isn't in the index" (spoiler: it can't be).

What a space is

"A space in Amazon Quick is a collection of data and Quick resources scoped for a particular team or domain. You can use spaces to aggregate and organize files, dashboards, topics, knowledge bases, and application actions into a unified and customizable knowledge center for your team." — Working with spaces (verified 2026-08-26)

The point of a space is grounding: "Spaces allow your team to get the most relevant results from conversational agents and other AI tools inside Quick by grounding the results with only data relevant for your task or domain" (same page). A space is also "a data layer for apps in Amazon Quick applications."

Five resource types can go in (Managing spaces, verified 2026-08-26): Dashboards, Topics, Knowledge bases, Actions, File uploads.

Note which doc tree each page above lives in — working-with-spaces is in /quick/, managing-spaces is in /quicksuite/. Same product, two live trees (Q0 lesson 1).

Permissions — one rule and one exception

Two levels, from the same working-with-spaces page (verified 2026-08-26):

Sharing is to users or groups, and the creator is owner by default (Creating spaces, verified 2026-08-26).

The rule, verbatim — this is the one to memorize:

"Quick resources added to a space respect user access permissions. If a user can't access a resource outside a space, they won't be able to access it inside a space. Files uploaded to a space are always available to everyone with access to the space."

So a dashboard in a space still enforces its own sharing — but an uploaded file has no permissions of its own. ⚠️ Uploading payroll.xlsx into a widely shared space is the access grant. There is no finer-grained control on space files.

⚠️ Deletion semantics differ by content type: "Deleting a space deletes all files associated with the space. Deleting a space unlinks assets associated with the space without deleting the assets" (managing-spaces, verified 2026-08-26). Files die with the space; linked assets survive.

Files: what fits

From the "Limitations of spaces" section of working-with-spaces (verified 2026-08-26):

• Types: documents (.pdf .docx .pptx .xlsx …), text/code (.py .sql .java …), and — easy to miss — audio and video (.mp3 .wav .mp4 .mov …).
• Size: "File uploads are limited to 500 MB for PDF, Word, and PowerPoint files, and 50 MB for all other supported file types."
• Count: "Spaces support up to 10,000 files (compared to 20 files in regular chat conversations)…" — that parenthetical is the whole pitch for spaces over ad-hoc chat uploads.
• A space can't contain another space.
• Upload lifecycle: Uploading → Processing → Text ready (text queryable, images/tables still processing) → Ready. A file can answer questions before its tables can.

⚠️ The storage number contradicts itself across pages. The user guide says total space storage "less than 10 GB"; the Service Quotas table says "Maximum total size of uploaded files per space: 1 Gigabytes — not adjustable" (both verified 2026-08-26). Ten-times discrepancy, unresolved in the docs. Plan against the quota table (the conservative one) and re-verify before you promise anyone a number. Same story for resource counts: the user guide says "at most 20 resources of a specific resource type", the quota table says "Maximum number of resources per space: 100."

The Quick Index

"An index is a structured corpus of documents that allows for efficient searching, sorting, and data access. Amazon Quick uses an index to store, prepare, sync any files you upload to your Quick spaces. It then uses this indexed data to generate agent-driven responses during chat conversations." — Manage data capacity (verified 2026-08-26)

Key mechanics from that page (verified 2026-08-26):

• An index is created automatically in your home Region at signup.
• Capacity is measured as original file size at the source — "before any processing", not extracted text. A 40 MB PowerPoint costs 40 MB of index even if it contains 2 KB of words.
• Included capacity: "Professional users get 25 GB of storage per user and Enterprise users get 50 GB per user… pooled on a payer account level." Overage: $5 per GB per month.
• Auto-scaling grows in 0.25 GB increments "up to 2 TB" (support ticket to go beyond); this "replac[es] the previous limit of 60 GB of extracted text."
• Index capacity in Regions beyond your home Region bills as overage from dollar one.
• To release capacity: "you must delete all datasets and files uploaded to spaces" first.
• Space uploads are rejected when index capacity is full — the space error is really an index error.

⚠️ The Service Quotas table still lists "Maximum total data capacity of an index: 60,000 MB — not adjustable" (verified 2026-08-26) — the superseded pre-2 TB figure. Another live contradiction; the capacity page is the newer story.

Lineage worth knowing on support calls: knowledge-base sync with a CMK requires both the qbusiness.amazonaws.com and quicksight.amazonaws.com service principals (Knowledge base integrations, verified 2026-08-26) — the index is Amazon Q Business machinery under the hood. You can also attach an existing Q Business index: "When you connect an Amazon Q Business index to Amazon Quick, it becomes available as a knowledge base" (Q Business indexes, verified 2026-08-26).

Knowledge bases: the pipes in

A knowledge base syncs an external source into the index. Built-in integrations (knowledge-base-integrations, verified 2026-08-26): Google Drive, OneDrive, Confluence, SharePoint, Amazon S3, and Web Crawler.

Sync behavior, same page: refresh is daily by default; a deletion-percentage safeguard skips the delete phase of a sync that would remove too much; a sync run that exceeds 14 days fails. Per-item limits: 500 MB text docs, 2 GB audio, 10 GB video, max 30 MB of text extracted per document.

The matrix that settles arguments — which integrations can feed the index vs only act (Supported integrations, verified 2026-08-26):

⚠️ Salesforce content cannot be indexed. An agent can act on Salesforce (query it live via an action), but there is no Salesforce knowledge base. People assume otherwise constantly.

The API surface (it's still quicksight)

Spaces and knowledge bases are fully scriptable — under the old namespace, per the CLI reference (verified 2026-08-26):

aws quicksight create-space | describe-space | list-spaces | search-spaces |
    update-space | delete-space |
    describe-space-permissions | update-space-permissions |
    list-space-resources | update-space-resources
aws quicksight create-knowledge-base | describe-knowledge-base | list-knowledge-bases |
    update-knowledge-base | delete-knowledge-base | batch-delete-knowledge-base |
    describe-knowledge-base-permissions | update-knowledge-base-permissions
aws quicksight list-users-index-capacity

There is no aws quicksuite namespace. IAM actions follow the same pattern (quicksight:DescribeActionConnector appears verbatim in Action connector APIs, verified 2026-08-26).

Check yourself

1. A reader can open a space but gets "no access" on a dashboard inside it. Bug?
2. Which of these lands in the Quick Index: a PDF uploaded to a space, a SharePoint library behind a knowledge base, a Salesforce org connected as an action?
3. Finance uploads a 300 MB PDF and a 300 MB MP4 to a space. Which upload fails, and why?
4. Space uploads suddenly fail account-wide. What do you check before opening a case?
5. Two pages give two different per-space storage limits. Which do you plan against, and what do you tell a stakeholder?
6. What single fact proves the index is shared machinery with Amazon Q Business?

Answers

1. Not a bug. Space resources respect the resource's own permissions; the space grants access only to files. The dashboard needs to be shared with the reader separately.
2. The PDF and the SharePoint library. Salesforce is Actions-only in the supported-integrations matrix — its content is never indexed, only queried live.
3. The MP4. 500 MB applies only to PDF/Word/PowerPoint; everything else caps at 50 MB per file.
4. Index capacity. "Files uploaded to spaces will be rejected if your index data storage capacity is full" — check Manage Quick → Subscriptions → index capacity, and whether auto-scale hit its configured maximum.
5. The Service Quotas figure (1 GB, not adjustable) — plan against the conservative documented number, flag the contradiction, and re-verify at decision time. Don't silently pick the bigger number.
6. CMK-encrypted knowledge-base sync requires granting both qbusiness.amazonaws.com and quicksight.amazonaws.com service principals in the key policy.

Teaching this section

• Open with the permission exception. "Everything in a space respects its own permissions — except files" reframes spaces as an access-control decision, not a convenience feature.
• Demo: create a space, upload a file, watch the status walk Uploading → Text ready → Ready; then aws quicksight list-spaces and list-space-resources to show the API view.
• Good discussion: "Your legal team wants one space for all contracts. Who should be a viewer, given the file rule?"
• Watch for: learners equating "connected" with "indexed". The Salesforce row of the matrix is the corrective.
• Timing: 25 min presented; 10 more if you demo a knowledge-base sync.