Skip to main content

​
Overview

A Knowledge Pack is a structured collection of business-specific information that your thinkrr agents can reference during live calls. This could include documents, manuals, FAQs, policies, pricing sheets, inventories, and public web pages. By attaching a Knowledge Pack to an inbound agent or outbound project, you ensure that the AI responds with accurate, context-specific information every time.
⚠️ Important Knowledge Packs are intended for reference information and factual knowledge only. They should not be used to instruct the agent on how to behave, what to say, or how to structure conversations. Behavioral guidance, tone, scripts, and decision logic must be configured using the agent’s dedicated instruction and guideline fields inside the platform. Placing instructions inside a Knowledge Pack can lead to inconsistent or unreliable agent behavior.
Key capabilities:
  • Attach to both inbound and outbound configurations
  • Upload multiple supported file types
  • Use URL scraping to pull and reference live web content
  • Enable or disable packs without losing stored content
  • Edit, refresh, and manage packs directly from the configuration panel
  • Maintain separate packs for different products, brands, or departments
Supported file types:
  • .md (Markdown) (NEW)
  • .csv, .tsv, .xlsx, .xls, .pdf, .txt, .docx
    • Up to 25 files per Knowledge Pack
    • 50MB maximum per file
  • Public web pages via URLs
Plan limits:
  • Business Professional – 1 Knowledge Pack
  • Business Growth – 2 Knowledge Packs
  • Agency Lite – 1 Knowledge Pack
  • Agency Unlimited – 2 Knowledge Packs
🚫 Note: Sub-accounts cannot access Knowledge Packs created by an agency’s main account.

​
Setting Up a Knowledge Pack

Knowledge Packs have a dedicated page in the left-hand sidebar menu, making it easier to create and manage.
1

Open the Knowledge Packs page

  • In the thinkrr platform, click Knowledge Packs in the left-hand sidebar menu.
KPSIDEBAR
2

Create a New Knowledge Pack

  • Click Create New Knowledge Pack to launch the builder.
KPSIDEBAR2
3

Add and configure your content

  • Upload supported files or enter URLs as needed.
  • Once created, you can edit, refresh, disable, or delete the pack at any time.
  • The same Knowledge Pack can be assigned to inbound agents or outbound projects from their configuration screens.
Add inbound content
Alternatively, you can create and configure when setting up your inbound agents or outbound projects.
1

Access the Knowledge Packs section

  • In your inbound agent configuration screen, click **Knowledge Packs **from the configurable tabs at the top of the screen
Inbound Knowledge Pack access
2

Create a New Knowledge Pack

  • Click Create New Knowledge Pack to launch the builder
Create inbound pack
3

Add Content

  • Upload supported files, or
  • Enter one or more URLs to let thinkrr scan public web pages
Add inbound content
4

Manage Your Pack

  • Edit, delete, or refresh your pack from the right-hand menu
  • Disable the pack temporarily instead of deleting it
  • Re-enable anytime without losing content
Edit Knowledge Pack
1

Access the Knowledge Packs section

  • From the Project Configuration screen, click Knowledge Packs in the panel
Outbound Knowledge Pack access
2

Create a New Knowledge Pack

  • Click Create New Knowledge Pack to launch the builder
Create outbound pack
3

Add Content

  • Upload files or add URLs as needed
Add outbound content
4

Manage Your Pack

  • Edit, refresh, disable, or delete your pack
  • Disabled packs can be re-enabled anytime without losing data
Edit Knowledge Pack

​
Updating and Refreshing a Knowledge Pack

Updating a Knowledge Pack is a simple two-step process:
  1. Remove the outdated files or URLs
  2. Upload or add the updated versions
There is no manual refresh, reindexing, or additional action required. Once new content is added, thinkrr automatically processes it in the background and updates the Knowledge Pack. 💡 Important: Always remove old data before adding new content to avoid overlap or conflicting information.

​
Using URLs in Knowledge Packs

Knowledge Packs allow you to add URLs as a source of information, in addition to uploaded files. This is useful when your content already lives online , such as in documentation pages or shared documents. For example:
  • If you manage product inventory in a Google Sheet, you can make that sheet visible to anyone with the link and add the Google Sheet URL to the Knowledge Pack.
  • If you maintain policies, pricing details, or internal FAQs in a Google Doc, you can make the document publicly accessible (or accessible via link) and add the Google Doc URL directly.
This approach removes the need to:
  • Export the document as a file
  • Convert it to Markdown or PDF
When you add a URL to a Knowledge Pack:
  • The system scans the page once, at the time the URL is added
  • The extracted content is saved and used by the agent during calls
  • The URL is not continuously re-scraped
If the content in the source document changes later, those updates will not automatically apply. To sync changes made to a URL-based document:
  • Remove the existing URL from the Knowledge Pack
  • Add the same URL again to trigger a fresh scan
Important considerations:
  • The document must be publicly accessible or set to “anyone with the link”
  • Limit edit access to only trusted users who are responsible for maintaining the content
  • Review sharing permissions carefully to avoid exposing sensitive information
Markdown files still provide the most controlled and predictable results, but using Google Sheets or Google Docs via URLs is often the easiest and lowest-maintenance option for frequently updated information.
⚠️ Important Some URLs cannot be scraped due to restrictions set by the website’s hosting provider or service platform. These protections are designed to prevent automated scraping and may block content access without any visible error. There is no built-in way for thinkrr to detect whether a page was partially or fully blocked during scraping. If a page cannot be scraped, the information will simply be missing from the Knowledge Pack. If your agent appears inconsistent or inaccurate when referencing URL-based content, this may be the cause. To avoid this:
  • Check with your website host or service provider to confirm scraping is allowed
  • Test the URL manually by adding it to a Knowledge Pack and verifying the agent’s responses
  • Be especially cautious when scraping your own website or protected pages
💡 Tip: Check the site’s robots.txt file Some websites block automated scraping using a file called robots.txt. You can quickly check this yourself by visiting: yourwebsite.com/robots.txt This file tells crawlers which parts of the site are allowed or disallowed. How to read it (at a high level):
  • An asterisk (*) usually means “all crawlers”
  • Allow: / generally means the entire site can be crawled
  • Disallow: / means the entire site is blocked
  • Specific paths under Disallow mean only those sections are blocked
Things to keep in mind:
  • If certain pages are disallowed, those pages may not be scraped into a Knowledge Pack
  • If the robots.txt file exists, scraping rules are explicitly defined and should be reviewed
  • If the file does not exist at all, the site may still block scraping through other protections
If your agent seems to miss or ignore information from a URL, checking robots.txt is a good first step to understand whether scraping restrictions may be the cause.

​
Plan Credits and Usage

Knowledge Pack usage is tied to your account’s plan credit allocation. Credits are consumed when an agent actively references Knowledge Pack content during calls. You can monitor remaining credits and purchase more from the Payments screen.
Payments screen
  • Credit counter: Displays remaining credits for the current billing cycle
  • Credit transfers: Agencies can transfer credits between sub-accounts to balance usage

​
Optimization

Knowledge Packs work best when the content is clean, concise, and well-structured. Poorly formatted content can reduce answer accuracy and slightly increase response time. In most cases, this adds around 100 milliseconds of latency, which is generally not noticeable.

​
Content Formatting Best Practices

  • Prefer .md (Markdown) files over .txt whenever possible (NEW)
  • Use clear and descriptive section headings
  • Keep each section focused and reasonably short
  • If a section becomes long, split it into multiple subsections
  • Write short paragraphs and lists to separate concepts
  • Avoid large walls of text
  • Be specific and explicit, include names, dates, and units
  • Avoid vague references like “it,” “this,” or “they,” since earlier context may not always be present
  • Group related information within the same section so content stays cohesive
  • Use Knowledge Packs to supply reference information only, not agent instructions or prompts
❌ Bad Example – Unstructured Content: “Our company offers product A which costs 50 dollars and product B which costs 100 dollars and we have a return policy of 30 days and our hours are 9 to 5 on weekdays.” ✅ Good Example – Structured Content (.md Format): 
## Product Pricing
- Product A: $50
- Product B: $100

## Return Policy
Returns accepted within 30 days of purchase.

## Business Hours
Monday–Friday: 9am–5pm

​
Working With CSV and Tabular Data

.csv and .tsv files are supported, but tabular data on its own can be harder for agents to retrieve accurately, especially when rows lack context. Best practices when using CSV or spreadsheet-style data:
  • Add explanatory text that describes what the table represents
  • Keep related rows grouped together
  • Avoid uploading raw tables without any surrounding explanation
  • When possible, summarize critical information in Markdown instead of relying only on rows and columns
❌ Bad Example – Raw Table Only
ProductPriceCategory
Widget A$50Electronics
Widget B$100Home
✅ Good Example – Table With Context 
## Product Pricing

Our product catalog includes:
- **Widget A**: $50 (Electronics category)
- **Widget B**: $100 (Home category)

​
Using URLs and Auto-Crawling Effectively

When adding URLs to a Knowledge Pack, it is best to use specific and targeted paths instead of broad website URLs. This improves crawl performance and helps you avoid hitting exclusion URL limits. What the tip means ❌ Bad approach (broad path + many exclusions):
Set auto-crawl on https://example.com/entire site)
Then add 100+ exclusion URLs for pages you don’t want ✅ Good approach (granular paths):
Set auto-crawl on specific paths like:
https://example.com/docs/
https://example.com/faq/
Only crawl what you actually need Why it matters
  • Better performance, crawling fewer, targeted pages is faster
  • Avoids hitting limits, there’s a max of 200 exclusion URLs per auto-crawling path and 500 exclusion URLs per Knowledge Pack
Planning your URLs carefully improves reliability, performance, and long-term maintainability.

​
FAQs & Troubleshooting

​
General Questions

Yes. A single Knowledge Pack can be linked to multiple inbound agents and outbound projects.
thinkrr can scan most publicly accessible pages but cannot access login-protected or private content.

​
Configuration

No. Knowledge Packs update automatically after new files or URLs are added. (NEW)
Disabling a pack removes access for the agent while keeping all content intact. You can re-enable it anytime.
No. Knowledge Packs created in an agency’s main account are not accessible to sub-accounts.

​
Usage and Results

Split the file into smaller parts and upload them separately. Even a few megabytes of text can represent a large amount of content. (NEW)
No. thinkrr optimizes response generation even when referencing large Knowledge Packs.
For additional questions or guidance, try using our Virtual Support Agent! Available 24/7 to help resolve most issues quickly at thinkrr.ai/support. If you still need assistance, visit our support site at help.thinkrr.ai and submit a Ticket or contact our team directly at [email protected].