Changelog
Follow up on the latest improvements and updates.
RSS
Excited to announce that the Mistral Large 2 model, the latest from Mistral, is now live on Millis! It’s been a busy week for open-source LLMs, and we're thrilled to bring you this state-of-the-art model. Experience the new frontier of open-source LLMs on our platform!
ElevenLabs's new model, Turbo v2.5, has been released and is now live on Millis! This new model brings significant improvements.
- Hindi, French, Spanish, Mandarin, and 27 other languages are now 3x faster. English is now25% faster.
- New language support for Vietnamese, Hungarian, and Norwegian.
This update unlocks high-quality, low-latency conversational AI across even more languages!
Llama 3.1 (8B/70B/405B) models are now live on Millis! Officially launched today, it's already available on our playground. Navigate to the playground and choose Llama 3.1 from the LLM dropdown to try it out.
- Custom Phrases: Now you can set specific phrases for the agent to say when running your functions or searching from knowledge.
- Custom Vocabulary: Define a list of specialized terms or uncommon proper nouns for the agent to recognize.
- Twilio Phone Number Import: You can now import your own Twilio phone number.
Our partner, Cartesia, has released a new text-to-speech model, now fully functional on the playground and it's fast.
We’ve added more webhooks to help you sync data between Millis and your source:
- End-of-Call Webhook: At the end of each call, we send all session-related data to your webhook, including the full transcript, cost, and more.
- Prefetch Data Webhook: A webhook being called at the beginning of the session to fetch any extra data you want to include in the agent's prompt. For example: use phone number to fetch user data that agent should know.
- Run Function After Call: Set any function to execute after the call ends. When turned on, the agent will guarantee to trigger your function at the end of the call. Example use cases: capturing leads to CRM, saving records after the call.
We’ve added EU region servers! Now, you can specify your calls to be in EU regions for even faster performance. Check out the documentation here: https://millisai.mintlify.app/core-concepts/regions
First and foremost, GPT-4o-Mini was released yesterday and is now available for you to try in our playground. Just select it as the LLM model to experience its blazing-fast performance and improved response handling compared to GPT-3.5.
We have added new settings to give you more control over your agents:
- Set Greeting Line: Customize what your agent says at the beginning of a conversation.
- Voice Detection Confidence Threshold: Adjust how sensitive the agent is to detecting human voices. This can help if the agent's voice is getting cut off frequently.
- Session Timeout: Set an automatic end to conversations after a period of no voice activity or a fixed call length.
- Data Privacy Opt-Out: Choose to opt-out of storing personal data and conversation history to comply with privacy requirements like HIPAA.
We have expanded our inbound/outbound phone call support.
With Twilio, we allow you to purchase phone number and setup phone call in the US, Canada, and Australia. Now, with Vonage, we support many more countries, including France, Poland, and India. Simply buy a phone number from Vonage and link it to your Millis agents.
Load More
→