Overview

A corpus is a central library of your organization's knowledge (documents, web pages, and text) that Ejento AI assistants search to answer questions. Instead of relying only on what a language model already knows, an assistant connected to a corpus looks up the most relevant passages from your content and uses them to write its answer.

Why corpora matter

Generic AI answers are rarely good enough for real work. Employees need the current travel policy, customers need your return window, and analysts need figures from this quarter's report. A corpus closes that gap:

Accurate, grounded answers. Assistants answer from your approved content, which reduces guesswork and made-up details.

Always up to date. Update a document once and every assistant that uses the corpus reflects the change. Web content can even refresh on a schedule.

Build once, reuse everywhere. A single corpus can be connected to many assistants, so teams share one source of truth instead of maintaining copies.

Controlled access. You decide who can view, edit, or manage each corpus, and sensitive personal data can be redacted automatically.

How a corpus works

You add content from files, websites, text, or connected sources such as SharePoint and Google Drive.

Ejento AI indexes it, breaking the content into searchable pieces. The indexing mode you choose (Basic, Moderate, or Exhaustive) balances search quality against cost.

You connect the corpus to an assistant. When a user asks a question, the assistant retrieves the most relevant passages and grounds its answer in them.

Indexing modes: Basic, Moderate, and Exhaustive

Every corpus has an indexing mode that decides how thoroughly its content is processed for search:

Basic: The most cost-effective option, for when indexing quality isn't the highest priority. A good starting point for FAQs, short policies, and small corpora.

Moderate: Balances performance and cost, with efficient indexing when you don't need maximum accuracy. Suits most everyday business content.

Exhaustive: The most comprehensive indexing, for the best search accuracy and relevance, at the highest cost. Choose it for large or high-stakes content, such as contracts or technical manuals. It's also the only mode that supports video files.

You choose the mode when you create a corpus and can change it later. For guidance on choosing, see Assistant Corpus Settings.

Where your content can come from

Files: PDFs, Word documents, spreadsheets, presentations, email files, archives, and many more formats.

Website: Individual web pages, entered one per line or uploaded as a list.

Text Blob: Text you type or paste directly, such as FAQs or video transcripts.

SharePoint, Google Drive, Sitemaps, and Azure NetApp Files: Sources you sync so the corpus stays in step with where your content already lives.

Use cases

Customer support: Upload product manuals, return policies, and help articles, and connect the corpus to a support assistant, so customers get answers that quote your current policy.

Internal help desks: Add the employee handbook, benefits guides, and IT procedures, so staff get instant answers to HR, IT, and finance questions.

Sales and marketing: Combine pricing sheets, case studies, and brand guidelines, so reps prepare for calls with the right figures.

Research and compliance: Index contracts, regulations, and reports, and choose Exhaustive indexing where precise answers matter most.

Key terms

Corpus: A collection of indexed content that assistants can search. The plural is corpora.

Knowledge source: Where content comes from, such as uploaded files, a website, or a synced SharePoint library.

Indexing mode: How thoroughly content is processed for search: Basic, Moderate, or Exhaustive.

Scheduled Refresh: Automatic re-indexing of web links on a schedule, so web content stays current.

Tags: Labels on a corpus or document that improve search relevance and help organize content.

PII Redaction: An option that masks personal information in documents before assistants use them.

FAQs

Do I need to create a corpus for every assistant?

No. Every assistant gets a default corpus when it's created, and you can connect other corpora to it as well. See Corpus Connections.

Who can create a corpus?

Anyone with a Permission Set that has the corpora Create toggle turned on. Without it, you can't create a corpus, even if you can view and edit existing ones.

Can a corpus use video content?

Only corpora on Exhaustive indexing support video files. Video links can't be indexed, so paste a video's transcript into a Text Blob instead.

To create your first corpus, see Assistant Corpus Setup.

Why corpora matter#

How a corpus works#

Indexing modes: Basic, Moderate, and Exhaustive#

Where your content can come from#

Use cases#

Key terms#

FAQs#

Do I need to create a corpus for every assistant?#

Who can create a corpus?#

Can a corpus use video content?#