Collections
Find specific information quickly by organizing and grouping knowledge sources.
Collections enhance how the agents interact with organizational knowledge. By grouping related sources into topic-based repositories, Collections enable more precise, context-aware information retrieval and intelligent knowledge management.
Feature overview
Better information retrieval: Organize knowledge sources into topic-specific repositories so the system searches within relevant content clusters instead of the entire knowledge base. This reduces the search space and improves retrieval speed.
More accurate answers: By organizing related sources together, collections create focused knowledge domains that reduce noise and interference from unrelated content.
Contextual understanding: Go beyond simple keyword matching. The system can better understand nuance and user intent by analyzing the current question, previous interactions, and configurable contextual parameters. This enables more precise responses to underlying needs.
Search types: Define how the system processes queries across your sources. Choose between Semantic, Full-text, or Hybrid approaches. Your selection applies whenever the system accesses this collection to retrieve information, ensuring consistent search behavior.
Each Collection is a standalone entity that contains knowledge sources relevant to its theme. The search functionality lets you find information within a specific collection, limiting results to its sources only. Each collection must be trained individually, and the system keeps previous versions for reference.
Creating a new collection
To create a new collection, go to the Collections section and select New Collection.
Add the collection name and the topics it covers in the description so the agent can understand when to use it. Without this field, the agent can’t activate the skill. The agent only “knows” what is defined in this description, so write it clearly, precisely, and with enough detail. The Collection description provides enough context for the agent to determine when it should be used, based on user questions or inputs. It must also indicate the topics covered by the documents in the collection. This helps the agent understand the scope of the available information.
When a user’s input relates to any of the described topics, the agent can trigger the corresponding knowledge collection, retrieve relevant information from its documents, and generate an appropriate response.
You can add up to 200 Knowledge Sources and up to 200 Collections. Knowledge Sources can be organized into collections according to the needs of the project, with no limit per collection.
Search types
When configuring the Collection, you can select the search type that best fits your needs through a dropdown menu. Collections offers three distinct search approaches, each with its own strengths: Semantic, Full-text, and Hybrid.
Semantic search
Retrieves results based on meaning rather than exact word matches. It understands user intent and can find relevant information even when terminology differs.
Example: If a user asks, “How do I reset my password?”, Semantic search may return sources with phrases such as “password restoration procedure” or “how to change forgotten login credentials,” because it recognizes that these concepts are related.
Full text search
Focuses on exact word matches within sources. It’s ideal for finding specific terminology, product names, or unique identifiers.
Example: If you search for “Model X500 error code 3021,” Full-text search looks for sources containing those exact terms, ensuring technical precision.
Hybrid search
Combines Semantic and Full-text search. This approach often produces more relevant results, especially when a single method isn’t sufficient.
If you choose “Hybrid,” you’ll be prompted to enter percentage values to define how much each search type contributes to the final results.
Example: For a query such as “smartphone battery draining quickly” (for example, semantic: 60%, full-text: 40%), a Hybrid search might:
Use Semantic search (60%) to understand concepts related to battery conservation and power management
Use Full-text search (40%) to ensure that specific terms such as “smartphone” and “battery” appear in the results
This combination ensures you get sources that both specifically mention the key terms and understand the underlying issue of power consumption, even if they don't use the exact phrase "draining quickly."
When to use each search type
Choose Semantic search when: Your content includes different ways of expressing similar concepts, or when users may use terminology that differs from what appears in your sources.
Choose Full-text search when: Precise terminology is critical, such as product codes, specific error messages, or technical documentation where exact wording matters.
Choose Hybrid search when: Your knowledge base includes both technical specifications and conceptual information, or when you want to balance exact matches with understanding user intent.
For better results, especially in complex scenarios, consider using Hybrid search to combine semantic understanding with exact word matching.
Advanced settings
You can now customize each collection with advanced settings to fine-tune how information is retrieved using the following parameters:
Top K: Defines how many paragraphs the system considers when searching for information. With a lower K value, the model selects from the most likely options, resulting in more focused and relevant results and improving perceived accuracy.
Similarity Threshold: Sets the minimum relevance level required for results to be considered relevant. Lower values may return broader but less predictable results. Higher values produce more precise matches.
Previous user inputs: Customizes search based on earlier user messages. The default value is 0, meaning only the current message is considered. Adjust the slider to include previous messages. This helps the system understand context when a previously discussed topic is revisited. For example, if a user asks about “unlock new credit card” and then follows up with “how do I do it?”, the system understands that the follow-up refers to “unlock new credit card.”
Training
Now, you have to make sure that they are now part of the knowledge base. You do that by using the training process. Click on the button Training on the top right corner. Train each collection individually to keep the agent up to date with the latest information. After making changes, retrain the affected collections to ensure the knowledge base stays up-to-date.
You can train using the icon 
available in the list or via Training section.
Retrain the agent every time you add or edit a source, create or edit a question, change the document linked to the question.
The training process of a PDF may take a little longer than training TXT files. If you face any issues with specific files in cases of very large trainings (which may involve too many documents or PDFs close to the 100-page limit), wait a few moments and try again.
Delete Collection
When deleting collections, the system carefully handles associated content. Questions and sources within the deleted collection remain accessible, ensuring that no critical information is lost.
Before deleting a collection, note that this action permanently removes all knowledge sources within it. Back up any important information or transfer essential sources to another collection if you need to preserve them..
Last updated
Was this helpful?