Agents can be equipped with extra information to help them ground their responses to users. There are two ways that extra information can be provided, on top of the content in the system message. These are resources and knowledge.
| Resources | Knowledge | |
| Summary | Uploaded resources are used as a searchable set of reference material for agents. Most of the time, most of the information in the provided resources is not used. Check out: How are resources used by agents? | Uploaded knowledge is sent to the AI in their entirety. The AI will always ‘read’ the entire contents of uploaded knowledge. |
| How to add | Resources can be added to your agent through its configuration page | Knowledge can be added to your agent through its configuration page |
| Benefits | Lower environmental impact and the search may more specifically identify relevant parts of the provided information. | More accurate understanding of provided information. |
| Limitations | Most of the uploaded information is ignored by the agent each time it responds to students. Quality of information retrieved depends largely on what Cogniti can find, which is based on what the student types into their query. If students will be asking specific questions, resources can be quite powerful. | Limited by the length of input that an AI model can take in. Longer documents cannot be added. Higher environmental impact. |
| Environmental impact | Low, because only small snippets of the provided resources are fed to the AI engine. | Higher, because the entirety of provided knowledge is fed to the AI engine every single time a student posts a query to the agent. |
| Examples | Reference material that you have permission to upload to AI, such as your own extensive notes, a problem bank, etc. | Essential material that you have permission to upload to AI, such as a course syllabus, teaching philosophy, assignment instructions, etc. |
