Corpus targeting and prioritizing

Most AI assistant projects will include multiple data corpuses. Alan AI gives you control over which corpus to query when a user asks a specific sort of question. You can:

• Set a corpus priority

• Use the query transform to define how to handle specific queries

Example of use

Assume you have two corpuses in the AI assistant project:

• Dynamic corpus for retrieving information about VMs in the infrastructure

• Static corpus containing generic information about VM resources and families from documentation

When the user asks a generic question, such as What types of VMs are available?, you want Alan AI to query the static corpus and return information about available machine resources and families as indicated in the documentation. To do this:

1. Set up the dynamic corpus as specified in the example of use.

2. Add a static corpus to the dialog script:

   Dialog script

   corpus({
       title: `Dynamic corpus for infrastructure requests`,
       input: project.objects,
       query: transforms.vms_queries,
       output: project.cleanObjects,
       transforms: transforms.vms_answer,
       priority: 1
   });
   
   corpus({
       title: `Static corpus for docs requests`,
       urls: [
           `https://cloud.google.com/compute/docs/machine-resource`
       ],
       maxPages: 10,
       priority: 2
   });
   

3. In the Debugging Chat, ask the following question: What types of VMs are available?

   

4. Under Transforms, select vm_queries and add a new example instructing Alan AI to return null when a generic question is asked:

   

Now, if you ask the same question, Alan AI will return an answer from the static corpus, not the dynamic one: