This guide provides recommendations for monitoring and improving the performance of data store handlers.
Track your agent's performance
You can monitor your agents conversation history and you can use the analytics tool for agent statistics.
Run self-service evaluation
You can run self-service evaluation which will assess the quality of your data store agent and recommend changes.
Improve agent responses
If responses don't meet your expectations during testing, try these options:
- Overwrite an answer by adding an FAQ entry for a specific question.
- If you experience low quality answers with a structured FAQ, try using an unstructured FAQ instead.
- Use the parse and chunk configurations to optimize responses.
- To ban certain phrases, see agent settings: banned phrases.
- To customize Responsible AI (RAI) content filtering, see agent settings: safety filters.
- To fine-tune your agent's response based on information from earlier in the conversation, configure search conditions. This feature is only available for data store tools.
Optimization of data store tools used by playbooks
This setting accelerates a data store's RAG response and improves latency by not rewriting the user query or dynamically populating tool parameters. Consequently, the user query is passed to the data store as-is, and no filters or user metadata-related parameters are populated.
You can enable this feature in the Dialogflow CX console by selecting the playbook, and then selecting the Settings tab. You can configure the playbook to continue processing any follow-up actions immediately or wait for the next end-user input.
If enabled, the following playbook instructions are still honored:
- When to answer the user query by using the data store or stores.
- When not to answer the user query by using the data store or stores and instead provide an alternate response.
- Any instructions that are defensive, for example, "don't answer questions about competitors."
If enabled, the following playbook instructions are ignored. Don't activate this setting if you have any of the following query requirements:
- Specific data store filters to be populated using the instructions.
- Instructions to apply any user personalization metadata passed by filters.
- Any other instructions to rewrite the user query before querying the data store.
Handle conversation digressions
An end-user can ask clarifying questions during a conversation. For example, during credit card information collection, they might want to clarify what a CVV is. In this case, your agent answers the question and returns to collecting the necessary credit card information. To accomplish this, you can create a data store handler with data stores that answer the question, apply that handler to the flow start page of the flow that handles credit card information collection, and set a transition target for this handler to return to the "current page."
Handle undesired intent matches
If your agent is matching intents when it should be using a data store handler, you can try the following to correct this:
- Delete or modify training phrases that are vague, so that all of your training phrases precisely handle the chosen intent and don't conflict with your data store content.
- Use negative examples to avoid intent matching.
Data store filtering
If you want only certain data stores available for queries, you can base them on session parameter values. For example, you can use unique data stores for product categories. To accomplish data store filtering for product categories:
- Set session parameters to product categories.
- Create condition routes that check the values of the session parameters and transition to a specific page that has the chosen data store handler.
- Transition back to the calling page from the data store handler so that the conversation can continue.
Data store search configuration
For more information about fine-tuning your data store search results, see Data store search configuration.