6. Appendix

6.1. Frontend

DialBB comes with two sample frontends for accessing the Web API.

6.1.1. Simple Frontend

You can access it at:

http://<host>:<port>

This frontend displays system and user utterances in speech bubbles.

It does not allow sending aux_data.

Information other than the system utterance included in the response is not displayed.

6.1.2. Debug Frontend

You can access it at:

http://<host>:<port>/test

This frontend displays system and user utterances in a list format.

It allows sending aux_data.

The aux_data included in the response is also displayed.

6.2. How to Use DialBB without Installing via pip

First, clone the GitHub repository. The cloned directory will be referred to as <DialBB Directory>.

git clone git@github.com:c4a-ri/dialbb.git <DialBB Directory>

Set the environment variable PYTHONPATH:

export PYTHONPATH=<DialBB Directory>:$PYTHONPATH

6.2.1. Using the Class API

If you want to use DialBB via the class API, start Python and import the necessary modules or classes from dialbb:

from dialbb.main import DialogueProcessor

6.2.2. Using the Web API

To use DialBB as a Web API, specify the configuration file and start the server:

$ python <DialBB Directory>/run_server.py [--port <port>] <config file>

The default port (port number) is 8080.

6.3. Tester Using a User Simulator

A tester using a user simulator that uses an LLM (ChatGPT) is included.

6.3.1. How to Run the Sample

The following explains how to run it using Bash. If you are using Windows Command Prompt, please adjust accordingly.

Install DialBB and download and extract the sample application.
Set the OpenAI API key to the environment variable OPENAI_API_KEY.
```
export OPENAI_KEY=<Your OpenAI API Key>
```

In the directory where the sample application is extracted (sample_apps), run the following command:

dialbb-sim-tester --app_config lab_app_en/config.yml --test_config lab_app_en/simulation/config.yml --output _output.txt

The result will be written to _output.txt.

To launch the tester from within a program, use the following code:

from dialbb.sim_tester.main import test_by_simulation

test_by_simulation("lab_app_en/config.yml", 
                   "lab_app_en/simulation/config.yml", output="_output.txt")

6.3.2. Specifications

Startup options

dialbb-sim-tester --app_config <DialBB application config file> --test_config <test config file> --output <output file>

Test configuration file

A YAML file containing the following keys:
- model: (string, required) Name of the OpenAI GPT model, such as gpt-4o and gpt-4o-mini. gpt-5 cannot be used.
- user_name: (string, optional) The name representing the user in the dialogue history. Default is "User".
- system_name: (string, optional) The name representing the system in the dialogue history. Default is "System".
- settings: (list of objects, required) A list of settings. Each can contain the following:
  - prompt_templates: (list of strings, required) Paths to text files containing prompt templates. The paths are relative to the configuration file.
  - initial_aux_data: (string, optional) Path to a JSON file containing content to be included in aux_data at the beginning of the dialogue when accessing the DialBB application. Path is relative to the configuration file.
- temperatures: (list of floats, optional) A list of temperature parameters for GPT. Default is a single-element list [0.7]. Sessions will be conducted for all combinations of the length of prompt_templates × this list.
- max_turns: (integer, optional) Maximum number of turns per session. Default is 15.
Function Specification
- dialbb.sim_test.main.test_by_simulation(test_config_file: str, app_config_file: str, output_file: str=None, json_output: bool=False, prompt_params: Dict[str, str])
  
  Parameters:
  - test_config_file: Path to the test configuration file.
  - app_config_file: Path to the DialBB application configuration file.
  - output_file: File to write the dialogue logs to.
  - json_output: Whether the output file should be in JSON format. If False, it will be a text file.
  - prompt_params: Dictionary of parameters to embed into the prompt. If the prompt template contains {<key>}, it will be replaced with <value>.

6.4. Discontinued Features

6.4.1. Snips Understander Built-in Block

As Snips has become challenging to install with Python 3.9 and above, it was discontinued in version 0.9. Please use the LR-CRF Understander built-in block as an alternative.

6.4.2. Whitespace Tokenizer and Sudachi Tokenizer Built-in Blocks

These blocks were discontinued in version 0.9. If you use LR-CRF Understander or ChatGPT Understander, there is no need for the Tokenizer blocks.

6.4.3. Snips+STN Sample Application

This sample application was discontinued in version 0.9.