Project Title

Rust Coding Assistant using Qwen Coder 2.5 and LlamaEdge

Motivation

The project aims to assist Rust developers by automating project generation, debugging errors, and improving development workflows. By integrating an LLM-powered assistant, we enable a seamless experience for Rust programming, reducing debugging time and enhancing productivity.

Expected Outcome

• A tool that generates a complete Rust Cargo project from a prompt.

• A QA system for common Rust projects and compiler errors stored in a vector database.

• A web API that takes project descriptions and iterates until the compilation succeeds.

• Integration with Qdrant for efficient vector search and retrieval.

• Deployment of the LLM and embedding model on LlamaEdge for seamless performance.

Details

• Uses Qwen Coder 2.5 on LlamaEdge to generate Rust projects based on a structured prompt.

• Parses LLM responses into multiple files and ensures compilation.

• Maintains a QA system with common Rust projects and compiler errors using Qdrant.

• Implements a feedback loop where failed compilations trigger searches in the QA database to refine the generated code.

• Provides an OpenAI-style API for seamless integration.

Milestones

• M1 [1 week] ✅

  • Discuss timeline with mentors and refine the task breakdown.
  • Set up development environment and ensure access to LlamaEdge and qdrant.

• M2 [2 weeks] ✅

  • Develop and test prompt engineering for generating complete Cargo projects.
  • Implement tools to parse LLM responses into structured files.
  • Ensure generated projects can be built and run successfully.

• M3 [2 weeks] ✅

  • Create a dataset of QAs where Q represents a common app idea and A is a generated Cargo project.
  • Store the dataset in qdrant for efficient retrieval.

• M4 [2 weeks] ✅

  • Create a dataset of QAs where Q represents Rust compiler error messages and A provides fixes.
  • Store the dataset in qdrant for efficient retrieval.
  • Build a MCP server.

• M5 [3 weeks] ✅

  • Develop the web API that processes project descriptions and matches them with examples.
  • Implement an automated compilation process and error handling.
  • Ensure the system iterates through errors and corrections until a successful build.

• M6 [2 weeks] ⏳

  • Optimize model inference speed and retrieval accuracy.
  • Implement API integrations for seamless deployment.

• M7 [2 weeks] ⏳

  • Perform extensive testing and debugging.
  • Finalize documentation and prepare the project for public release.
  • Add Github Actions.

• Reference links:

• LFX mentorship (2025/term1): Improve the WasmEdge-based Rust coding assistant for inference-time scaling #3985

• LlamaEdge API: https://llamaedge.com/docs/user-guide/llm/full-openai

• Qwen Coder 2.5: https://huggingface.co/second-state/Qwen2.5-Coder-32B-Instruct-GGUF

• Qdrant Database: https://qdrant.tech/

• Check the work progress: HERE