Feature Requests

EVI easily picks up background noise and sounds (chatter, music, etc) which causes a lot of disruption to the conversation. Feature request for background noise filtering and speaker isolation. I tried doing so using the prompt but it didn't help. It doesn't pause the conversation, but the EVI is "listening" to the background and takes that as part of the conversation now

Currently, when you use a CLM, EVI doesn't start speaking until all the text for its turn has been generated. There is no way for the CLM to start generating some text, have EVI start speaking, and then decide to generate more. This is a feature request to support "streaming" so EVI can start speaking sooner.

I need to add trace_id's to LLM responses for feedback annotation purposes in my web application. It is not currently possible to add any information to Custom LLM responses. See https://dev.hume.ai/docs/empathic-voice-interface-evi/guides/custom-language-model#clm-outgoing-message-data-format I propose adding metadata to the assistant_message responses that would be passed through to the client. type OutgoingCLMMessage = AssistantInputMessage | AssistantEndMessage; interface AssistantInputMessage { type: "assistant_input"; text: string; metadata?: { [key: string]: any }; // 👈🏼 Add } interface AssistantEndMessage { type: "assistant_end"; metadata?: { [key: string]: any }; // 👈🏼 Add }

I know this is a pipe dream but.. For my use case that I am building, I want users to be able to interact with Hume via text message, phone call, and in-app. The problem is syncing what was said in each different modality. If I had a way to append chats/summaries/context to the existing chat group or session that would be great. The idea is to be able to add context that doesn't actually occur during a socket open connection or happens off the converastion. That'd be useful to me!

When trying to connect to the EVI ChatSocket, I'm getting this error: { "type": "error", "code": "I0100", "slug": "uncaught", "message": "Unexpected service error occurred. Please visit https://dev.hume.ai/support for assistance.", "requestId": "0155a95c-f670-44b9-8a13-d69bba6fc5945155824", "receivedAt": "2025-11-20T12:45:29.030Z" }

ws.mjs:389 WebSocket connection to 'wss: //api.hume.ai/v0/evi/chat?config_id=9e3b3048-4f27-420c-86b6-7d84642853c…experienceLevel%5D=senior&fernSdkLanguage=JavaScript&fernSdkVersion=0.15.4 ' failed:

If hume could open up the door to VOIP providers that would be amazing. I do undrstand that Twillio is kind of a defualt in the space, but they are really a terrible company, and even though I am being biased (Twillio user for 12 years) it would be great to have flexibility. Someone recently hacked my twillio account and becuase of it I am now banned and can't use Twillio - now i am out of the game to use hume in any capacity becuase only one VOIP provider is an option. What would be even better is if we could get a SIP integration maybe and we could add any procider and not be just limited to Twillio! Also Twillio is very expensive for some and they are not even able to get the numbers that they need sometimes.

I need to access the number that was called, which is a twilio number, as well as the originating number. The caller number is supplied in the chat_started webhook, but not the <Parameter name="to_number" value="+1xxxxxxxxxxx" /> which is posted when twilio connects to Hume. Is there a way to access that parameter? This is so the tools for a configuration and prompt parameters can be dynamically assigned.

We are using Hume AI in an industry with strict data privacy protection requirements. I would like an API endpoint that allows us to delete chat sessions after use or an auto-deletion setting in Hume AI that removes chat history after 1 or 2 days.

Hi I am setting up the example here: https://github.com/HumeAI/hume-api-examples/tree/eeccd72e5a200f658354669f81cbf7b27b0a487f/evi/evi-python-phone-calling-proxy-server . It worked yesterday but today, I am getting this error: ❌ Error: 38 validation errors for union[AssistantEnd,AssistantMessage,AssistantProsody,AudioOutput,ChatMetadata,WebSocketError,UserInterruption,UserMessage,ToolCallMessage,ToolResponseMessage,ToolErrorMessage] AssistantEnd.type Input should be 'assistant_end' [type=literal_error, input_value='session_settings', input_type=str] For further information visit https://errors.pydantic.dev/2.12/v/literal_error AssistantMessage.from_text Field required [type=missing, input_value={'type': 'session_setting...es': {'name': 'Joshua'}}, input_type=dict] For further information visit https://errors.pydantic.dev/2.12/v/missing AssistantMessage.message Field required [type=missing, input_value={'type': 'session_setting...es': {'name': 'Joshua'}}, input_type=dict] For further information visit https://errors.pydantic.dev/2.12/v/missing AssistantMessage.models Field required [type=missing, input_value={'type': 'session_setting...es': {'name': ..... I don't think this should happen since this is on your official github and was working yesterday. Can someone please look into this.