Optional Pipeline Connection

The pipeline issues have now been resolved, and updated instructions are included below. However, due to the earlier delays, we are making this component optional. Connecting your entire pipeline will still count as an extension (See Assessment Rubric for more information).

Key Points:

• If you choose to keep your intent and slot classifier and MARBEL+Dialogflow agent separate, that is completely fine and will not negatively affect your grade.

• Connecting the entire pipeline is now optional but is considered an "easier" extension and will count toward your extensions.

If you choose not to connect your entire pipeline, you can focus on implementing your exclusion features and extensions directly in Dialogflow. For detailed guidance, refer to the Exclusion Features section and its related subpages.

However, if you decide to connect your pipeline, please ensure that these two sections are fully completed within your intent and slot classifier.

Instructions for Connecting and Running the Pipeline

Setting Up the New Pipeline Connection

If you decide to connect your pipeline, follow these updated instructions carefully. Ensure you meet the prerequisites and run the commands in the correct sequence.

Prerequisites:

1. Activate PCA25:
   Ensure that PCA25 is activated in your environment.

2. Model Checkpoints:
   Confirm you have a trained model checkpoint saved in the nlu.utils.checkpoints folder.

3. Replace File:
   Replace the existing file located at:

   social-interaction-cloud/sic_framework/services/eis/eiscomponent.py

   with the updated eiscomponent.py file provided: . To use Dialogflow with this file this variable should be changed to False on line 105.

   # If true then you use your intent and slot classifier including Whisper otherwise Dialogflow self.params.nlu = False

Steps to Run the Pipeline:

Run each of the following commands in separate terminals after running your redis server:

1. Run Whisper:
   Start the Whisper component to handle speech-to-text conversion:

   run-whisper

   Note: The first time you use Whisper, the transcription may take some time due to downloading the model. Be patient during this process. If it still takes a long time after the first time read the top of the eiscomponent.py file.

2. Run NLU:
   Start the natural language understanding (NLU) module:

3. Run Webserver:
   Launch the webserver for pipeline integration:

4. Start Framework:
   Initialize the framework:

5. Run EIS:
   Start the External Interface Service (EIS):

6. Launch Localhost:
   Open your browser and navigate to the localhost start page.

7. Run MARBEL Agent in Eclipse:
   Start the MARBEL agent within Eclipse. Ensure the agent connects successfully to the pipeline.