May 9th 2024
You can now customize your assistant's knowledge base according to your specific needs and domain by feeding it with various types of data, including raw text, PDFs, and specific web pages. This allows you to expand your assistant's knowledge base and provide more accurate and relevant responses to user queries.
Once the data is ingested, our RAG (Retrieval-Augmented Generation) feature enables your AI assistant to retrieve and utilize this information during conversations. By leveraging the ingested data, your assistant can provide more informed and context-aware responses.
Llama 3 is now available, offering lower latency (up to 3x faster) and compatibility with more powerful synthesizers.
The new synthesizer is more accurate, natural, stable, and supports multiple languages (we will introduce new languages with the next release).
Stability and Diversity: Excellent stability, language diversity, and accuracy in voice cloning and accents.
Default Synthesizer: This is now the default synthesizer for all assistants using Synthflow LLM and Llama 3.
Agencies can now resell new products - Assistants, Teams, and Flows
New Pricing Tier Editor
More Granular Subaccount Controls
LLM as judge: Our fine-tuned LLMs now generate high-quality feedback to assess the success of AI-customer interactions based on various criteria.
Why is it Useful?
  â Great feedback for prompt optimization
  â Monitor quality of customer calls
End Call Reason: The new end_call_reason parameter provides insight into why a call ended.
Echo & Noise Cancellation Improvements
Crosstalk Improvements
Speech-to-Text Improvements: Leveraging the latest Deepgram version.
Voicemail Detection Enhancements
Incomplete Sentences: Fixed the issue with assistant saying an incomplete sentence but providing the full one in transcript.
Weâve lowered our pay-as-you-go prices to just 8 cents per minute. You can now enjoy our product at a lower cost, and as an agency, you can now maximize your margins when reselling assistants. Plus, you can now add additional minutes while in trial mode.
Specify the number of seconds before the agent picks up.
Infrastructure upgrade for improved performance and scalability.
More assistant templates.
Improvements to calcom booking format
New integrations tab