Agent Context

Agent Context is the foundation of oper8r's intelligence. It houses the knowledge base, uploaded files, and company identity that power every AI-generated response across all session types - Chat, Spreadsheet, Document, and more.

When logged in, click Data in the navigation bar, then select the Agent Context tab to access these settings.

Knowledge Library

The Knowledge Library is your single source of truth - a centralized repository of approved Q&A content, boilerplate text, and company information that oper8r draws on when generating responses.

Knowledge Item Types

Each knowledge item is categorized by how it was created:

Creating Knowledge Items

There are several ways to add knowledge to your library:

• Create in text editor - write new content directly in the rich markdown editor
• Upload a file - import content from documents (oper8r extracts and indexes the content)
• Auto-generated - oper8r creates knowledge items during sessions when it identifies reusable content

Rich Markdown Editor

Click on any knowledge item to open the detail modal. The left side provides a full markdown editor for viewing and editing content. The right sidebar displays metadata, version history, citations, and similarity scores.

Organized Categorization & Tagging

Keep your knowledge library organized with multiple methods:

• Tags - add user-defined tags to categorize items by topic, product, industry, or any custom taxonomy. Tags are searchable and can be added or removed inline.
• Owner assignment - assign a team member as the owner/maintainer of each knowledge item
• Type-based grouping - filter by knowledge item type (uploaded file, agent generated, etc.)
• File-based grouping - filter to see all knowledge items extracted from a specific uploaded file

Advanced Search & Retrieval

Find the right content instantly with powerful search and filtering:

• Text search - search across all knowledge items by keyword
• Multi-faceted filtering - combine filters for type, file, owner, and tag
• Sorting options - order by edited recently/least, cited most/least, or created recently/least

Version Control & Updated Content

The library holds the latest approved version of each knowledge item, with older versions archived and tracked:

• Auto-snapshots - every edit automatically creates a new version snapshot (up to 50 retained per item)
• Diff view - compare any two versions side by side to see exactly what changed
• Version history - browse numbered versions with the actor who made each change and the timestamp

Every edit to a knowledge item creates an immutable snapshot. The version history forms a linked chain where each version knows its parent, providing full traceability of how content has evolved over time. Each snapshot records the actor, timestamp, and content hash for change detection.

Citation Tracking

Monitor how your knowledge is being used across sessions with 30-day rolling citation tracking:

• Total citation count - see how many times each item has been cited
• Sparkline chart - visual trend of citations over the last 30 days
• Per-session citation logs - detailed breakdown showing which session type (Chat, Document, Spreadsheet, etc.), which team member used it, and how many times

Semantic Similarity Detection

oper8r tracks similarity scores when knowledge items are retrieved during sessions. This helps you understand:

• Which knowledge items are being matched most frequently
• The average semantic similarity between queries and your content
• Trends over time with a 30-day rolling average sparkline

Knowledge File Uploads

Upload files that provide context to oper8r; these files are automatically analyzed and broken out into chunks which contain question/answer content and relevant context. There is no manual question+answer mapping process, this is all done intelligently by the oper8r agent.

Ideal uploads include completed questionnaires, RFPs, FAQs, competitive battlecards, product documentation, and any other materials that contain answers your team frequently reuses.

Supported File Types

CSV, DOC, DOCX, MD, MDX, ODP, ODS, ODT, PDF, PPT, PPTX, TXT, XLS, XLSX

Processing Pipeline

When you upload a file, oper8r:

1. Stores the file in its original format
2. Converts the content to markdown
3. Chunks and indexes the content for retrieval augmented generation
4. Extracts knowledge items - uploaded files automatically generate individual knowledge items in your library

All uploaded files are automatically available to oper8r across all session types.

Company Identity

Company Identity stores the foundational information about your organization that oper8r uses to personalize every response.

Core Fields

These fields are set during onboarding and can be edited at any time:

• Company Name - how oper8r refers to your company
• Website URL - your company's primary website
• Documentation URL - your product documentation site (optional)

Auto-Generated Prompts

oper8r generates two key prompts from your website and documentation data:

• Company Overview - a detailed description of your company, products, and value proposition
• Ideal Customer Profile (ICP) - a profile of your target customer including industry, role, pain points, and buying behavior

Both prompts are fully editable - review and refine them to ensure accuracy.

Website Scraping Controls

oper8r scrapes your website and documentation to build a comprehensive understanding of your company. The Scrape Status Card shows real-time progress and controls for managing this process.

Scrape Status

The status card displays separately for both your website and documentation:

• Page count - how many pages have been scraped
• Scraped URLs - list of recently scraped pages
• Last scraped timestamp - when the last scrape completed
• Status indicators:
  • Completed (green) - scraping finished successfully
  • In Progress (cyan, pulsing) - scraping is actively running
  • Pending (gray) - scraping is queued

Re-Scrape

Click Re-scrape Sites to update your scraped data. This opens a modal where you can:

• Edit URLs - update your website or documentation URL before re-scraping
• Real-time validation - URLs are validated as you type, with auto-healing suggestions (e.g., adding https://, www.)
• Content verification - oper8r checks that your company name appears on the page
• Start the scrape - trigger a new scrape job once validation passes

Scraping Details

oper8r's scraping engine crawls up to 500 pages by default and unlimited pages for enterprise, per scrape, going 20 levels deep. It intelligently strips navigation noise - cookies banners, ads, sidebars, and footers - to extract only the meaningful content from each page.

Continuous Updates

Your website data stays current with automatic weekly re-scraping. The scrape status indicator on the Company Identity page shows the health and recency of your data at a glance.