hlcaptain / butler Goto Github PK

Implement a Customer Support Service which later can be accessed in-app via a chat. Consider technologies like rocker.chat and Zammad.

Android Jetpack library Room became multiplatform enabling a more reactive and better overall experience for local data handling.

In order for the client to show rich messages and not rely on servers to provide localized error messages, microservices respond with an Int error code which describes the issue to the client. Create a schema and implement it on server and client side.

Refine UI and error handling of Login/Registration flow.

• Add proper error handling and messages during unexpected events.
• Fix UI bugs and upgrade UI to fit each platform properly.

Needs fix:

• UI problems, wrong flows, disappearing error screens.
• Data sync problems.

Providing localized messages should ultimately be the server's responsibility. Use a LocalizationService microservice to connect with a Weblate instance in the cloud.

In case of enum or other static values, like Int which stands for an error code or static key, which is not likely to be changed, consider using enum classes or other.

Host an API for LLM inference via vLLM.

This can be done multiple ways, but some notes:

• Must run on Linux
• Load and run inference with open-sourced model (probably form HuggingFace)
• Expose the OpenAI-like API in the local network
• Connect the client app to the server and chat with the LLM

Implement a microservice, which checks if a user can initiate a chat with an AI bot (like a custom model from HuggingFace) and answer those messages, running inference on an LLM, then sending messages to the user via the chat microservice.

Create proper onboarding screens for the client app.

Enhance documentation with pictures, how-to-use docs etc.

Gitlive's firebase kotlin sdk is very buggy on the desktop platform and not reliable on other platforms, so try to find a solution with a multiplatform library for:

• Authentication (JWT?)
• Storage/Database access with Reactive data flows.

Implement an Analytics platform (preferably open-source), like Countly into Butler backend services and client.

Develop features around image generation and speech recognition.

Examples:

• User prompts AI to generate images.
• User talks to AI using ASR.
• User sends sound files and LLM replies back with sound (text-to-speech)

Implement system monitoring with Prometheus and Grafana and to get an insight on system performance.

Redesign Home and Chat screens to fit several size of screens and decide on a layout composition.

Fix websocket handling to provide reactive data stream between client and server.

Enable user authentication and chatting on the Android platform for now (then change client APIs to support proper multiplatform on a later pull request).

Implement:

• Data handling via Firebase Firestore SDK
• User authentication via Firebase Authentication
• Chatting via Firestore SDK

Libres img resource loading is crashing due to resource is not found. Include resource path.