🤡🤡

Updated the API reference to fully match the Swagger API Documentation.

People who are using custom llm and disabling this property urlRequestMetadataEnabled, then we changed it to model.metadataSendMode, and you need to add the "off" parameter. https://docs.vapi.ai/api-reference/assistants/create-assistant https://cdn.discordapp.com/attachments/1251889150869766225/1253633538603876362/image.png?ex=667690b6&is=66753f36&hm=dea9fe4de11d309fefca00649b81374a5c36027b51ed7f39f962e20d3289bb2d&

A couple of cool things: - assistant-speaks-first-with-model-generated-message: You can use this in the squad essentially. When transferring the call to another assistant, ideally you don't want your assistant to respond with the transfer message. Instead, it should perform the task based on the previous message. So, you can achieve this by using it. - You can use the multilingual Cartesian model through the API. You just need to add the voice model as "sonic-multilingual." - You can now pass the voicemail tool, and we will add an additional check to ensure that it is a voicemail. https://cdn.discordapp.com/attachments/1251889150869766225/1256809637072863262/image.png?ex=66821eaf&is=6680cd2f&hm=3235c9c7fb9bad7764d5fdc849db510a2ddc48963c14df7d82d14727ae04e166& https://cdn.discordapp.com/attachments/1251889150869766225/1256809637404217344/image.png?ex=66821eaf&is=6680cd2f&hm=4c977d4bbfa33491e2055c3958b38e43012dfe9f8de7abebd611f1b80773a54c&

New Transcriber is added (Gladia)

Wake-up call feature. If user doesn't respond for x time then assistant will say something https://cdn.discordapp.com/attachments/1251889150869766225/1259855791494664202/image.png?ex=668d33a3&is=668be223&hm=384682e2888e035c4ba273e87e26695026abfbf4063c705bc849c30d0498cf6b&

Squad is supported via Dashboard now.

Function calling on claude is fixed.

@here, we will be updating the assistant.maxDurationSeconds=600 as the new default; previously, it was 1800. So, if you guys want to have the call for more than 10 minutes, make these changes.

Hey guys, couple of updates we launched couple of new things which can help you alot. 1. Added the

smartTurnDetectionEnabled

,

startSpeakingPlan

and

stopSpeakingPlan

through these pararmeter you can improve the interruption issue if you guys are faicing anything related to it. 2. Added the

ChunkPlan

through this you can have more control about how data to be sent to the voice provider. We haven't written the documentation about it but you can search about it https://api.vapi.ai/api and you can search for these terms, you will get the brief idea about it. Also, do note that all these features can be used only via API. @here

🚀 API Update Announcement 🚀 Hello @here, we’ve rolled out some important updates! Here’s a quick summary of the key changes: 1. New Control and Listen Features for assistant.monitorPlan: assistant.monitorPlan.controlEnabled & assistant.monitorPlan.listenEnabled are now added (default: true). While these defaults don't impact call execution, requests sent to call.monitor.controlUrl or call.monitor.listenUrl will now be processed. 2. Language Detection in Transcriber: We’ve introduced assistant.transcriber.languageDetectionEnabled (default: false). This feature can now be toggled to enable real-time language detection in transcriptions. 3. End-of-Call Reports Enhanced: More data is now sent in the end-of-call reports. Also, the issue related to costbreakdown wasn't sent properly is also fixed. 4. Admin/Editor Role Enhancements for Organizations: New admin/editor role feature. While the default flow remains unchanged, we encourage testing to spot any 2nd-order effects. You can refer to https://api.vapi.ai/api for more detailed information about these parameters.

📢 Important Update: Upcoming Change to maxTokens Default Value 📢 Hello, @everyone! 👋 Starting this week, we will be reducing the default maxTokens from 250 to 100 to help improve response times and reduce latency across our services. 💨 This change means that around 80 words = 1 minute of speech, so responses will be shorter and quicker, ensuring faster processing of your requests. OpenAI is prioritizing smaller requests, so this update is crucial for optimizing performance. ⚡ Thank you for your understanding! If you have any questions or concerns, feel free to reach out. 😊 Happy coding! 💻

Adding to previous update Q. Will it affect you? => No, it doesn’t affect anyone. We will be applying this setting when you create a new assistant. Q. Can I increase the max tokens? => It is completely up to you. You can increase the max tokens; we just set a default, but you still have full control over it. Q. What would be the effect of it? => Lower latency, as OpenAI prioritizes smaller requests.

We still haven't figured out a daily changelog system but getting closer. For Sept 9-11 1️⃣ Save call recordings in GCP. Add credentials in the dashboard. 2️⃣ Improved cost breakdown in

call.costs

. 3️⃣ Configure caller ID for call transfers

assistant.model.tools[type=transferCall].destinations[type=number].callerId

. 4️⃣

assistant.voice.language

supported for

elevenlabs_turbo_v2_5

. 5️⃣ Store recordings at a path in S3 or GCP

assistant.artifactPlan.recordingPath

. 6️⃣ Write longer prompts in

assistant.analysisPlan

. 7️⃣

end-of-call-report

has cost, costs, startedAt, and endedAt properties now. 8️⃣ New

language-changed

webhook. Add in

assistant.serverMessages

and

assistant.clientMessages

.

Changelog Announcement We are excited to introduce two significant updates: 1. **Warm Transfer Call Support**: Now, with warm transfer capabilities, you can seamlessly transfer calls. We'll dial the transfer destination number, provide a summary or the entire conversation to your agent, and connect them with your customer for a smooth handover. For more details, check out TransferDestinationNumber in our API documentation. 2. **Custom TTS Integration**: You can now integrate your own Text-to-Speech (TTS) provider. Tailor your voice experience by selecting a custom TTS for even more flexibility and personalization. Learn more by reviewing CustomVoice in our API documentation. For detailed API implementation, visit [Vapi AI API Documentation](https://api.vapi.ai/api).