.cursorrules
.llmrules
version: 1
Sign in to like and favorite skills
.llmrules
Code Style and Structure
Here are some pieces of information you need to know about this project:
version: 1
project: name: "oros" description: > Oros (f.k.a. “kavachat") is an evolving codebase for a multi-LLM chat application that interacts with blockchain tasks. Currently, it focuses on bridging user prompts to GPT-based AI, proxying requests through a stateless Golang server, and rendering a React frontend. In the long term, Oros may expand to encompass many single and multi-chain agentic blockchain actions (using one or more LLMs/deModels), long-term memory and user-shared context, and expansion to different user-interfaces /entry points (TG, mobile,etc). marketing: "Oros: bring deAI to any dApp".
repos:
context:
overview: > The project’s current architecture has three main components: 1) A Golang Proxy that sits between the frontend and the LLM. This is a stateless server exposing HTTP endpoints and routing user chat requests to a remote LLM (e.g., GPT-4o). It also handles partial application logic, such as prompt engineering, streaming responses, and code injection prevention.
2) A **React Frontend** that provides a chat interface. Users can send messages, connect wallets, and see real-time streaming replies from the proxy. 3) **Tests** that aim to cover critical behaviors, integration points, and resilience. The codebase includes unit tests (Go and TypeScript), plus integration tests verifying correct LLM responses, and a Docker-based environment for end-to-end checks.
future: - Incorporating multi-LLM support (community open models + GPT, deModels). - Adding ephemeral or partial memory for conversation context. - Expanding to multi-chain interactions. - Evolving to multi-chain AI agent that can propose on-chain transactions, orchestrate bridging, etc. - Expanding to other user-interfaces/entry points (TG, mobile,etc).
guidingPrinciples:
workflow: codeReviews: > - Create a short-lived feature branch off of main or dev. - Open a Pull Request with a succinct description, referencing the relevant issue or user story. - Tag relevant reviewers (product lead, lead engineer, or others if domain-specific knowledge is needed). - Ensure tests pass. - Merge with a squash commit if approved. versioning: > We follow semantic versioning for major releases. However, as the codebase is still early, we might do frequent minor or patch releases with
v0.x.x.
ciProcess: >
GitHub Actions / any CI pipeline runs:
- Linting
- Tests (unit, integration)
- Build steps for both Go and React
- Docker build for local e2e tests
communication:
faq:
notes:
This .cursorrules file is living documentation. Update it as the architecture evolves (e.g., multi-LLM or memory logic), so new developers grasp the purpose, flow, and higher-level context of Oros.