You can export interaction logging to BigQuery. Once configured, all live interaction logging is written to your BigQuery table. This provides you with advanced analysis tools that can help you debug and improve your agent and to discover patterns in conversation data.
Limitations
The following limitations apply:
- The Dialogflow agent and the BigQuery dataset must be in the same project.
Table description
Each row of the table contains one conversational turn with the following columns:
Column | Type | Description |
---|---|---|
project_id | STRING | The project ID. |
agent_id | STRING | The agent ID. |
conversation_name | STRING | The fully qualified resource name for the session. |
turn_position | INTEGER | The conversational turn number. |
request_time | TIMESTAMP | The time of the conversational turn. |
language_code | STRING | The language tag. |
request | JSON | The detect intent request. |
response | JSON | The detect intent response. |
partial_responses | JSON | Partial responses if applicable. |
derived_data | JSON | Additional metadata for this conversational turn. |
conversation_signals | JSON | NLU related analytics data. See ConversationSignals for the JSON schema. |
Configuration
To configure interaction logging export:
- Ensure that interaction logging is enabled.
- Follow the BigQuery dataset creation guide to create a dataset. Note the dataset name, as you will need this in the next step.
Follow the BigQuery table creation guide to create a table with a SQL schema definition. Use the following SQL statement for creation:
CREATE TABLE <your_dataset_name>.dialogflow_bigquery_export_data( project_id STRING, agent_id STRING, conversation_name STRING, turn_position INTEGER, request_time TIMESTAMP, language_code STRING, request JSON, response JSON, partial_responses JSON, derived_data JSON, conversation_signals JSON );
Configure your agent settings to enable BigQuery export, and to provide the dataset and table names created above.