File size: 1,591 Bytes
ca15036 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 | ---
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**.
|