Group2255's picture
Update README.md
5f1dc3d verified
|
Raw History Blame Contribute Delete
2.25 kB
---
title: PromptPilot (AML-3304)
emoji: πŸ€–
colorFrom: blue
colorTo: indigo
sdk: streamlit
sdk_version: 1.20.0
app_file: app.py
pinned: false
---
# πŸ€– PromptPilot – Unified AI Chatbot (AML-3304)
This is an MVP AI chatbot developed for the **AML-3304 assignment**:
**"From Tokens to Transformers: Building AI Systems with Embeddings, Generative Models, and MLOps."**
PromptPilot combines five intelligent chat modes powered by popular Transformer-based models, enabling users to explore code generation, factual Q&A, probabilistic reasoning, and paraphrasing.
---
## πŸ”„ Chat Modes
| Mode | Description | Model Used |
|-------------------------|-------------------------------------------------------------|----------------------------------------|
| πŸ§‘β€πŸ’» Code Generator | Generate Python code from plain language | `Salesforce/codegen-350M-mono` |
| πŸ“˜ General Q&A | Factual question answering and explanations | `google/flan-t5-base` |
| πŸ“Š Bayesian Q&A | Uncertainty-based reasoning using sampling | `google/flan-t5-base` + top-k sampling |
| 🌐 Deep Coder | Technical Q&A and code completion tasks | `deepseek-ai/deepseek-coder-6.7b-instruct` |
| πŸ” Transformer Explorer | Paraphrasing and text transformation using sequence models | `google/flan-t5-base` |
---
## πŸš€ Live Demo
Try the chatbot here:
**πŸ”— [https://huggingface.co/spaces/Group2255/programming-help-chatbot](https://huggingface.co/spaces/Group2255/programming-help-chatbot)**
---
## 🧩 Features
- Multi-mode UI for interactive exploration
- Efficient model switching using `st.session_state`
- GPU-compatible on Hugging Face Spaces
- Supports both deterministic and sampling-based generation
- Lightweight models for fast response and free-tier compatibility
---
## πŸ’» Run Locally
### βœ… Option 1: Python + Streamlit
```bash
git clone https://huggingface.co/spaces/Group2255/programming-help-chatbot
cd programming-help-chatbot
pip install -r requirements.txt
streamlit run app.py