|
|
--- |
|
|
license: mit |
|
|
language: |
|
|
- en |
|
|
base_model: meta-llama/Llama-3.1-8B-Instruct |
|
|
pipeline_tag: text-generation |
|
|
tags: |
|
|
- coding |
|
|
- assistant |
|
|
- instruction-tuned |
|
|
- iit-kgp |
|
|
--- |
|
|
|
|
|
# 🚀 Coding Assistant Pro |
|
|
|
|
|
**Coding Assistant Pro** is a domain-specific AI assistant designed to provide high-quality programming support. This project was developed as part of an M.Tech CSE research initiative to demonstrate strict logical filtering and technical accuracy in AI-driven development tools. |
|
|
|
|
|
## 🛠️ Project Goals |
|
|
- **Domain Strictness:** The model is constrained to respond only to Computer Science and Programming queries. |
|
|
- **Accuracy:** Provides line-by-line code explanations and identifies specific logic errors in buggy code. |
|
|
- **Efficiency:** Utilizes Meta's Llama-3.1-8B-Instruct via the Hugging Face Serverless Inference API for low-latency performance without high hardware costs. |
|
|
|
|
|
## 📝 Model Behavior Rules |
|
|
1. **Programming Queries:** Returns detailed explanations and complete code blocks. |
|
|
2. **Correct Code Analysis:** Performs a line-by-line technical walkthrough. |
|
|
3. **Bug Identification:** Identifies specific line numbers and explains the root cause of errors. |
|
|
4. **Out-of-Domain Filter:** Returns "I have no knowledge" for any non-technical or non-CS topics. |
|
|
|
|
|
## 🚀 Technical Implementation |
|
|
- **Frontend:** Streamlit 1.51.0 |
|
|
- **Backend:** Hugging Face Inference API (`huggingface_hub`) |
|
|
- **Model:** Meta-Llama-3.1-8B-Instruct |
|
|
- **Environment Management:** Python-dotenv for secure local credential storage |
|
|
|
|
|
## 📜 License |
|
|
This project is licensed under the **MIT License**. |
|
|
|
|
|
|
|
|
|
