Multiple YAML files to one Python implementation

In practice, you may want to have multiple plugins that share the same implementation. For example, you may want to have two plugins that both pull data from a database, but one pulls data from database A and the other pulls data from database B. In this case, the plugin implementation code is the same, but the plugin configuration is different. It would be cumbersome to copy and paste the same implementation code to two different files.

To solve this problem, TaskWeaver allows you to have multiple plugin configurations that share the same implementation. Here is an example of the plugin configuration for the two plugins that pull data from database A and B respectively:

The configuration for the plugin that pulls data from database A:

name: sql_pull_data_from_A
code: sql_pull_data
...
description: >-
  Pull data from a SQL database A. The database A contains information about merchandise sales.
examples: |-
  df, description = sql_pull_data_from_A("pull data from time_series table")
parameters:
  ...
returns:
  ...
configurations:
  ...
  sqlite_db_path: /path/to/sqlite_A.db

The configuration for the plugin that pulls data from database B:

name: sql_pull_data_from_B
code: sql_pull_data
...
description: >-
  Pull data from a SQL database B. The database B contains information about customer information.
examples: |-
  df, description = sql_pull_data_from_B("pull data from time_series table")
parameters:
  ...
returns:
  ...
configurations:
  ...
  sqlite_db_path: /path/to/sqlite_B.db

Let's discuss the differences between the two configurations.

First, you can see that the name field is different, and the names are different from the python file name (without extension) which is sql_pull_data.py. This name is used in CodeInterpreter for code generation. So, you can see that in the examples field, the function name is sql_pull_data_from_A and sql_pull_data_from_B, respectively.

Second, you can see that the code field is the same, and the code file name is sql_pull_data.py. This means that the two plugins share the same implementation code. The code field is optional, and if you don't specify it, the plugin name will be used as the code file name without the extension.

Third, you can see that the configurations field is different, and the sqlite_db_path is different. This means that the two plugins have different configurations. This is typically the key reason why you want to have multiple plugin configurations that share the same implementation.

Finally, you can see that the description field is different, and the descriptions are different. This is important because the Planner and the CodeInterpreter will use the description to make decisions and generate code. The two descriptions should be explicit enough to distinguish the two plugins. Otherwise, the Planner and the CodeInterpreter may not be able to make the right decisions.

Conclusion

When you meet the situation where you want to have multiple plugin configurations that share the same implementation, you can use the code field to specify the code file name, and use the configurations field to specify the configurations. The name field is used to distinguish the plugins, and the description field is used to help the Planner and the CodeInterpreter make right decisions.

Conclusion​

Conclusion