Development guidelines for Oros, a multi-LLM chat application bridging user prompts to blockchain tasks through a stateless Golang proxy and React frontend.
Development guidelines and context for building Oros (f.k.a. "kavachat"), a multi-LLM chat application that interacts with blockchain tasks.
Oros is an evolving codebase for a multi-LLM chat application focusing on bridging user prompts to GPT-based AI through a stateless Golang server with a React frontend. The long-term vision includes multi-chain agentic blockchain actions, long-term memory, user-shared context, and expansion to different user interfaces (Telegram, mobile, etc.).
**Marketing tagline:** "Oros: bring deAI to any dApp"
The current architecture consists of three main components:
1. **Golang Proxy**: Stateless server between frontend and LLM
- Exposes HTTP endpoints
- Routes user chat requests to remote LLM (e.g., GPT-4o)
- Handles prompt engineering
- Manages streaming responses
- Prevents code injection
2. **React Frontend**: Chat interface for users
- Send messages
- Connect wallets
- Real-time streaming replies from proxy
3. **Tests**: Comprehensive test coverage
- Unit tests (Go and TypeScript)
- Integration tests verifying LLM responses
- Docker-based environment for end-to-end checks
1. **Test Coverage**
- Build robust, readable tests for proxy, UI, and integration endpoints
- Achieve high coverage for critical paths
- Document behavior to ensure safe future refactors
2. **Architecture Clarity**
- Keep stateless proxy design straightforward
- Defer complex session logic to future expansions
- Maintain minimal dependencies for faster iteration
3. **Incremental Improvements**
- Prioritize small, frequent PRs with clear commit histories
- Integrate new features behind flags or environment configs
- Ensure stability for production usage
4. **User Focus**
- Present user-friendly chat UI
- Enable smooth LLM interactions
- Maintain clear docs and function signatures for developer experience
**Code Reviews:**
**Versioning:**
**CI Process:**
GitHub Actions runs:
**Platforms:**
**Feedback Loop:**
1. **Stateless Design**: Current proxy is stateless; session logic is ephemeral in UI or planned for future
2. **Core Functionality**: Focus on solidifying chat proxy + tests + minimal UI
3. **Easy Expansion**: Architecture designed to easily extend to multi-LLM or multi-chain usage
4. **Living Documentation**: Update guidelines as architecture evolves
**Q: What is the main goal of this development stage?**
A: Solidify core functionality (chat proxy + tests + minimal UI) and ensure easy expansion to multi-LLM or multi-chain usage.
**Q: Are we storing conversation data or user sessions in the backend?**
A: Currently, the proxy is stateless. Any partial memory or session logic is either ephemeral in the UI or a future enhancement.
Leave a review
No reviews yet. Be the first to review this skill!
# Download SKILL.md from killerskills.ai/api/skills/oros-multi-llm-chat-application-development/raw