Resource types at a glance: formats, examples, restrictions
Resources are the sources for AI agents. Information stored in the knowledge base forms the basis for the answers that agents provide in customer communications. Various document types and sources can be added and managed here to enable efficient and accurate knowledge delivery.This article describes how to use document types as sources. The following resource types can be uploaded to the knowledge base:
Resource management and agent control are handled in the moinAI Hub (article).
1. PDF
PDF files or PDF links can be added to the knowledge base. The PDFs must contain readable text content.
Examples
- Manuals: User manuals or troubleshooting guides for a specific product model.
- Brochures: Product brochures with detailed specifications.
Potential sources of error
- Password-protected PDFs
- PDFs that only contain scanned images
- Unsupported special formats
2. Website
For website URLs, the website content is read by a scraper and added to the knowledge base.
Special features
- Support for public and non-public websites
- Automatic updates (every 7, 14 or 30 days)
- Optional setup of custom HTTP headers
Examples
- Product pages: Reading information about available products
- FAQs on company websites: Automatic transfer of existing answers
Potential sources of error
- Incorrect URL or website not accessible
- Blocking of the scraper
- Content behind JavaScript elements cannot be read directly
3. Document
A document consists of formatted text with paragraphs, lists, tables and headings. It can be created, edited and published in the moinAI Hub.
Special features
- Contains titles and structured content
- Immediately available in preview and live after editing
Examples
- Knowledge articles: Troubleshooting guide for a specific product.
- Guidelines: Return policy for a complaint.
4. Question / answer
In a question-answer pair, a specific question is linked to a structured answer.
Difference to document: The AI uses the relationship between question and answer, which leads to greater accuracy in answer generation.
Examples
- Frequently asked questions: Answer ‘How can I reset my password?’ with step-by-step instructions.
- Product questions: ‘How much does the premium subscription cost?’ answered with pricing information.
5. CSV file
CSV files can be used to map tables or a series of lists of varying lengths.
Restrictions
- Maximum of 5,000 rows per file
- Row lengths are dynamically limited
Special features
- Title columns for identifying rows
- Selective selection of relevant columns increases response quality
Examples
- Product lists: Tables with product numbers, names and prices
- Customer data: Customer information for targeted contact
Potential sources of error
- Too much information in one row
- Maximum number of rows exceeded
- Inconsistent structure
- Incorrect file format
- Incorrect separator symbol
This article provides detailed instructions on how to import a CSV file in the ideal way.
How creating sources helps to improve response generation is described in this article.
6. External articles
External articles are linked to the knowledge base via an API (article).