What Reveto Does From Message to Reply Key Concepts
Creating Your Account First Login and the Dashboard Quick Start: Your First Client Understanding Navigation
Overview Website Crawling Document Upload Meeting Notes and Recordings Google Drive Import Telegram Chats
Generating a Knowledge Base Updating with New Sources Viewing and Navigating AI Search Editing Content Versions and History Exporting Access Control Preparing a Knowledge Base for Replies
Overview Connecting a Facebook Page Connecting Instagram The Review Queue Autopilot Reply Settings Insights The 24-Hour Window Notifications Pausing and Disconnecting
How Website Chat Works Installing the Widget Allowed Websites Widget Settings Conversations Leads Unanswered Questions Following Up on Chats
Managing Several Clients Organizing Team Access by Client
Inviting Team Members Roles and Permissions
Overview Creating a Test AI Question Generation Assigning Tests Taking a Test Grading and Results Test Analytics
Profile Settings Company Settings Notifications Claude Connector Activity Stream
Plans and Pricing Free Trial AI Usage and Billing Managing Your Subscription
Data Protection Account Security Conversation Data and Deletion
Getting Help Troubleshooting
Common Questions

Generating a Knowledge Base

Generate a knowledge base from a client's websites, documents and Telegram chats in four steps on the Content Generation page. Reveto reads the sources you pick and writes them into sections that replies are drafted from.

Open 'Content Generation' in the sidebar. The page is titled 'Generate Knowledge Base' and has four steps: 'Choose', 'Review', 'Generate' and 'Done'. Collect the client's material on the Data Collection page first. If nothing is collected yet, the page shows 'No data sources connected yet' and a 'Go to Data Collection' button.

PlanGenerate from sourcesKnowledge bases
Free trialYes, from the $10 AI testing credit once your email is verifiedUnlimited
ProfessionalYesUnlimited
BusinessYesUnlimited
EnterpriseYesUnlimited
StarterNoUp to 3, created blank and filled in by hand

You also need a role with the Content Generation permission and permission to create knowledge bases. Only one generation runs at a time for each person.

Step 1: Choose sources

The source tree lists everything you collected, grouped as 'Websites', 'Documents', 'Meetings', 'Google Drive' and 'Telegram'. Documents kept in folders are grouped by folder. An empty group shows 'Nothing here yet'.

  1. Tick a group to select everything in it, or open the group with its arrow and tick single items.
  2. Check the line at the bottom. It counts the sources you selected, for example '3 sources selected'.
  3. Click 'Continue to Review'. The button stays disabled until at least one source is ticked.

Build one knowledge base per client, from that client's material only. In Data Collection, 'Generate KB from this folder' opens this page with that folder's documents already ticked.

Step 2: Review and configure

  • 'Knowledge Base Title': optional. Without one, the knowledge base is named with the date and time it was created.
  • 'Output Language': the language the knowledge base is written in, whatever language the sources use. 25 languages are available.
  • 'Selected Data Sources': a summary of your selection, with page, document and message counts.
  • 'Sections to Generate': the 9 default sections, all on. Turn off the ones you don't need, or use 'Select all' and 'Deselect all'. At least one section must stay on.
  • 'Estimated Cost': the expected tokens and AI cost for this run, and the size of each source.

The default sections are 'Company Overview', 'Products & Services', 'Customer Communication', 'Sales & Marketing Knowledge', 'Operations & Service Delivery', 'Customer Interaction Scenarios', 'Knowledge Map', 'Roles and Authority' and 'Glossary'. Click 'Add Custom Section' to add your own. A custom section needs a title and a prompt that describes what it should contain.

For an advertiser whose ads bring in messages, add custom sections for what customers ask first, such as current offers, prices, opening hours, location and parking, and booking and cancellations. A section per topic keeps each fact easy to find and to update.

Step 3: Generate

Click 'Generate Knowledge Base'. The progress screen shows 'Overall Progress' as a percentage and a card for each section: waiting, in progress, done, or amber if it was skipped. Telegram chats are summarized first, so the screen may show 'Summarizing Telegram conversations...' for a while. Generation takes a few minutes, depending on how much data you selected.

Generation runs on the server, so you can leave the page. When you come back to Content Generation, the progress screen picks up where it was, and the notification bell shows 'Knowledge Base Ready' when it's done. If you see 'Generation Failed', click 'Try Again' or 'Start Over'. A run that makes no progress for 30 minutes is stopped and marked as failed.

'Stop & Discard' cancels the run and throws away what was generated. AI usage up to that point is still charged, and the screen shows it under 'AI Usage Before Cancellation'.

Step 4: Done

The last step shows 'Knowledge Base Generated!' with the input tokens, output tokens and 'Actual Cost'. Click 'View Knowledge Base' to open it, or 'Generate Another' to start again. The knowledge base has two levels, sections and subsections, and its text sits in blocks: paragraphs, lists, tables, notes, warnings, examples, summaries, quotes and code. If a section failed, the rest still finish and the progress screen says, for example, 'Completed with 1 section skipped'.