Advanced ARI Application for LLM Conversations
Budget: ₹750 – ₹1,250 INR
Requirement:
We need to create an ARI Application to drive the conversation between a user and LLM Apps. The LLM App is available via API Only. A Typical scenario includes
When the call comes it should greet the user dynamically
It should take users speech and send the text to LLM API and say back the API response
This should continue till the conversation is over
Key Challenges:
Custom Patch Limitations: Our current ARI setup relies on prebuilt patches, which hinder seamless installation via standard pip install.
Call Transfer Workflow Limitations: The existing implementation is tailored for call transfer workflows, inadequate for driving conversations with LLM Apps.
Conversation Flow Uncertainty: Difficulty detecting conversation start and stop times, which vary unpredictably (e.g., 20 seconds, 10 seconds).
Latency Issues: Current implementation introduces significant lag due to text-to-audio conversion at every conversation stage.
Responsibilities:
Redesign ARI Architecture: Refactor the existing ARI setup to accommodate LLM App integration, ensuring efficient conversation flow.
Implement Real-time Conversation Handling: Design and implement mechanisms to detect conversation start/stop times dynamically.
Optimize Text-to-Audio Conversion: Minimize latency by optimizing text-to-audio conversion processes.
Integrate with LLM Apps: Ensure seamless integration with LLM Apps, enabling efficient conversation workflows.
Performance Tuning: Conduct thorough performance testing and optimization to ensure a responsive user experience.
We need to create an ARI Application to drive the conversation between a user and LLM Apps. The LLM App is available via API Only. A Typical scenario includes
When the call comes it should greet the user dynamically
It should take users speech and send the text to LLM API and say back the API response
This should continue till the conversation is over
Key Challenges:
Custom Patch Limitations: Our current ARI setup relies on prebuilt patches, which hinder seamless installation via standard pip install.
Call Transfer Workflow Limitations: The existing implementation is tailored for call transfer workflows, inadequate for driving conversations with LLM Apps.
Conversation Flow Uncertainty: Difficulty detecting conversation start and stop times, which vary unpredictably (e.g., 20 seconds, 10 seconds).
Latency Issues: Current implementation introduces significant lag due to text-to-audio conversion at every conversation stage.
Responsibilities:
Redesign ARI Architecture: Refactor the existing ARI setup to accommodate LLM App integration, ensuring efficient conversation flow.
Implement Real-time Conversation Handling: Design and implement mechanisms to detect conversation start/stop times dynamically.
Optimize Text-to-Audio Conversion: Minimize latency by optimizing text-to-audio conversion processes.
Integrate with LLM Apps: Ensure seamless integration with LLM Apps, enabling efficient conversation workflows.
Performance Tuning: Conduct thorough performance testing and optimization to ensure a responsive user experience.