Step 3: Search skill
Now, we'll create superpowers for our assistant, also known as skills. These abilities significantly expand the functionalities our assistant can perform.
Skills are like distinct abilities we give our assistant to handle a variety of tasks. This includes everything from answering specific questions about the documents we've uploaded to searching for information on the internet or even interacting with users through guided processes or integrations with other systems, all through natural conversations.
With these skills, our assistant becomes a versatile and powerful tool, capable of adapting to a wide range of needs and providing a more comprehensive and useful experience to users.
For our example, let's create a skill that allows our assistant to answer questions about our museum information.
Create search skill
Let's head to the skills section in the sidebar menu and click on the button.
We will be presented with a window to fill in the information for our skill. We will enter the following information:
Field | Description | Value |
---|---|---|
Title | Name of the skill. | museum_search |
Description | Brief description of the skill. Should be entered in each available language. | Allows answering questions about museums. |
Type | Type of the skill. Search or dialogue. | Search |
Tags | List of tags related to the skill. | |
Access Mode | Selects the access mode to the skill. Public access or members only. | Public access |
Members | List of members who will have access to the skill and their associated privilege. |
Once all the necessary information is provided, click on the Create button.
Below is the detail of our newly created skill. Let's make a couple more changes:
- We select the collection where our documents about museums are located. This indicates to the skill which collection(s) to search for the information being requested. We'll choose the collection
Museums
. - We select the languages we want to use to perform the searches. This indicates to the skill which languages it can use to search for and respond to questions. We'll mark the language as
Spanish
.
The other parameters will be left as default. Click on Save.
Other fields to consider in our search skill are as follows:
- Maximum number of results: We can specify a limit to the number of results returned by our skill.
- Include personal collections: We can include documents from users' personal collections in the search results.
- Scoring algorithm: We can change the algorithm used to score and rank document relevance, along with associated parameters.
Train skill
The last crucial step is to train the skill to fully prepare it for deployment. Training is a fundamental process as it allows the skill to adjust and adapt based on the parameters and specifications we've provided.
During training, the skill processes information, recognizes patterns, and familiarizes itself with the various scenarios in which it will be used. This ensures it's ready to respond optimally to user requests, providing accurate and relevant results.
To initiate the operation, click on the Train button and wait for the operation to conclude.
Test skill
If you wish, you can conduct some search tests on the newly created skill in the Test Skill tab.
You can adjust different parameters about your skill while conducting your tests:
- Versions: You can test a specific version to ensure everything works correctly before switching to the new version.
- Identities: By testing with different identities, we can assess how the skill behaves across different users.
- Languages: The ability to test in different languages is essential to ensure the skill is effective and understandable for users who speak different languages.
- Audiences: Testing with different audiences allows us to evaluate how the skill responds and adapts to the needs and preferences of specific user groups.
The capability to review development information associated with each message is crucial for understanding how the skill processes information, identifying potential errors or improvements, and thus refining its performance.
We have now set up the ability to search for information about museums that we'll use in our assistant. Next, let's create it.