proxoar / talk Goto Github PK

Is your feature request related to a problem? Please describe.
At present, the static file middleware of https://github.com/labstack/echo lacks cache control functionality. This omission could potentially result in protracted load times and increased network bandwidth usage.

This is how the backend responds with an image request:

Describe the solution you'd like
Given that the static files in the Talk server remain unchanged between each deployment, it seems prudent to introduce a middleware that generates a random Etag upon startup. This Etag could then be written into the HTTP response header, with an expiration time set to max.

In light of the factors mentioned, I believe it would be beneficial to consider replacing web-socket with Server Side Events (SSE).

• web-socket offers a distinct authentication approach compared to RESTful API.
• Raw web-socket lacks a convenient method for segregating event types. While socket.io excels in this aspect, it lacks a robust Go SDK.
• SSE enjoys improved browser support for reconnection.

Settings are at default

The audio volume of Safari's recorder is noticeably lower compared to Chrome, resulting in inaccurate responses from the speech-to-text server.

Are you suggesting an enhancement in relation to an issue? Please elaborate.
It appears that Google's speech-to-text and Elevenlabs' text-to-speech offer streaming capabilities. It would be splendid if Talk could utilise these features.

Could you specify your preferred resolution?
Implementing this would necessitate a slight modification in both backend and frontend. Given the recent major overhaul that shifted the communication model towards versatility, the implementation of this feature should be considerably easier.

• https://cloud.google.com/speech-to-text/docs/transcribe-streaming-audio
• https://docs.elevenlabs.io/api-reference/text-to-speech-stream

Container starts fine. http://localhost:8000 is showing home page /auth endpoint. Giving below error in the log. It is not clear what is causing this error. Executed container with LOG_LEVEL=debug , does not give details. Please assist troubleshooting the issue

2024-02-03 13:35:01 1.7069853014169762e+09 WARN [email protected]/logger.go:47 Client error {"remote_ip": "172.17.0.1", "latency": "141.202µs", "host": "localhost:8000", "request": "GET /api/health", "status": 401, "size": 45, "user_agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/121.0.0.0 Safari/537.36", "request_id": ""}

Describe the bug

A few messages were successfully sent but no reply was received.

The SSE connection terminated when the operating system entered sleep mode. It appears there was no reconnection upon wake-up.

Screenshots

Desktop (please complete the following information):

• macOS Version 14.1.1 (23B81)
• Arc Version 1.17.2 (43249)

Version: 1.2.7

console.js:213:

Could not load content for webpack://react-devtools-extensions/react-devtools-shared/src/backend/console.js (Fetch through target failed: Unsupported URL scheme; Fallback: HTTP error: status code 404, net::ERR_UNKNOWN_URL_SCHEME)

If a website is not using HTTPS, modern browsers may restrict access to the microphone. It would be more convenient if the program could operate with a generated untrusted HTTPS certificate when the user doesn't provide their own.

Talk backend can return mocked data for testing and exploration purpose.

The concept of data synchronisation across different browsers is indeed intriguing:

The client transmits data to the server
The client retrieves data from the server
Such synchronisation should be exclusive to authenticated users.

This functionality is anticipated to significantly influence user interface design and interaction. It may be prudent to initially develop a rudimentary version.

Google Gemini's precision and depth truly stand out, surpassing(likely) the ChatGPT-4 model and far eclipsing Google Bard. It's time to back Google Gemini.