Promptly is a simple online service that helps you quickly find answers in your documents whether those are PDFs, text files, or other kinds of files. Imagine having a virtual assistant who has read all your paperwork and can instantly tell you what’s inside each document.
- Upload Your Files: You can upload documents (like PDFs or text files) through Promptly.
- Ask Questions: Whenever you want information, just type in your question.
- Get Accurate Answers: Promptly will search through the documents you’ve uploaded and give you a direct answer, almost like an expert who has read everything.
🚫 Manual Document Analysis – Professionals spend hours searching and summarizing documents.
🚫 Keyword Search Limitations – Basic search tools fail to provide context-aware answers.
🚫 Scalability Issues – Many solutions struggle with large document volumes.
✅ Persistent Multi-Document Memory – Users can create chatbot instances that retain knowledge.
✅ Cross-Document Referencing – AI synthesizes insights across multiple files.
✅ Role-Based Customization – Chatbot adapts to different roles (Legal, Finance, IT, HR).
✅ On-Premise & Cloud Deployment – Enterprise-friendly with security and compliance in mind
🏢 For Teams & Enterprises – A centralized AI knowledge assistant for fast and reliable document retrieval.
📖 For Individuals – A personal AI assistant for organizing and querying private notes and study materials.
├── assets/ # Diagrams and visual assets
│ ├── process_user_queries_dag.png # User Query Pipeline Workflow Diagram
│ ├── rag_data_pipeline_dag.png # Data Pipeline Workflow Diagram
│ ├── Data Pipeline Architecture.png
│ ├── Model Training & Deployment Architecture.png
│ ├── Drift Detection.png
│ ├── pipeline_optimization.png
│
├── data_pipeline/ # Data processing and ingestion pipeline
│ ├── dags/ # Airflow DAGs for pipeline orchestration
│ │ ├── dataPipeline.py # User Queries DAG
│ │ ├── rag_data_pipeline.py # Document Processing DAG
│ │ ├── scripts/ # Scripts for data preprocessing and ingestion
│ │ │ ├── email_utils.py # Email notifications
│ │ │ ├── upload_data_GCS.py # GCS Uploading
│ │ │ ├── data_preprocessing/
│ │ │ │ ├── check_pii_data.py # PII Detection
│ │ │ │ ├── validate_schema.py # Schema Validation
│ │ │ │ ├── data_utils.py # Query Cleaning Functions
│ │ │ ├── supadb/
│ │ │ │ ├── supabase_utils.py # Supabase Integration
│ │ │ ├── rag/
│ │ │ │ ├── validate_schema.py # Schema Validation
│ │ │ │ ├── rag_utils.py # Chunking & Embeddings
│ │ ├── tests/
│ │ │ ├── test_data_pii_redact.py # Unit tests for PII detection and redaction
│ │ │ ├── test_rag_pipeline.py # Unit tests for the RAG document chunking pipeline
│ │ │ ├── test_user_queries.py # Unit tests for the user queries processing pipeline
│ ├── config.py # API Keys & Configurations
│ ├── README.md # Data Pipeline Documentation
│
├── model/ # Model serving, testing, and deployment
│ ├── serve.py # FastAPI server for serving the model
│ ├── deploy_server.py # Deployment script for Google Cloud Vertex AI
│ ├── Dockerfile # Docker configuration for containerizing the model server
│ ├── requirements.txt # Python dependencies for the model
│ ├── tests/ # Unit tests for the model and APIs
│ │ ├── api_test.py # Test script for the REST API
│ │ └── sdk_test.py # Test script for SDK integration
│ ├── README.md # Model Directory Documentation
│
├── model_pipeline/ # Model training and fine-tuning pipeline
│ ├── training/ # Training scripts and notebooks
│ │ ├── promptly-finetuning.ipynb # Model training Jupyter notebook
│ │ ├── README.md # Training-specific documentation
│ ├── scripts/
│ │ ├── bias_detection.py # Script for detecting bias while training the model
│ │ ├── load_data.py # Load data from Supabase
│ │ ├── streamlit_ui.py # Streamlit user interface for the app
│ ├── mlflow/
│ │ ├── Dockerfile # Docker file for setting up MLflow in GCP Instance
│ ├── README.md # Model Pipeline Documentation
│
├── data/ # Data storage and processing
│ ├── rag_documents/ # Original PDFs & Text Files
│ ├── preprocessed_docs_chunks.csv # Cleaned & Chunked Data
│ ├── preprocessed_user_data.csv # Processed User Queries
│
├── .github/workflows/ # CI/CD workflows
│ ├── README.md # GitHub Actions Documentation
│
├── .dvc/ # DVC Configuration
├── .gitignore # Git ignore file
├── .dvcignore # DVC ignore file
├── requirements.txt # Python dependencies
├── README.md # Global project documentation
- Source: Retrieved from the conversations table in Supabase.
- Description: This table contains user-generated queries, which we have pre-filled with custom data to simulate various interaction scenarios.
- Source: Focused on IT specifications, we have curated data from publicly available requirements documents.
- Description: We have selectively gathered documents that provide detailed IT specifications, particularly from the PURE dataset, which comprises 79 publicly available natural language requirements documents collected from the web.
- Reference: https://zenodo.org/records/5195084
- The data pipeline automates document ingestion, preprocessing, and storage in Supabase.
- For detailed documentation, refer to the Data Pipeline README.
- The model training pipeline handles fine-tuning and evaluation of the model. For more details, refer to the Model Training README.
- The model serving is the deployment of model to Vertex AI. For more details, refer to the Model Serving README.
- User dominance is verified to check if the drift is caused by a single user or skewed usage pattern.
- If drift is detected, it triggers downstream actions:
- Drift trend analysis to monitor changes over time.
- Model retraining is triggered automatically if needed.
- The entire process is automated via an Apache Airflow DAG, which runs once a day.
- Upon completion or detection, an email notification is sent to alert stakeholders.
- Cross-Document Retrieval: Enables querying across multiple documents for context-aware answers.
- Fine-Tuned Model: Uses a fine-tuned version of
Qwen/Qwen2.5-0.5B-Instructfor text generation tasks. - Cloud Deployment: Supports deployment to Google Cloud Vertex AI for scalable inference.
- CI/CD Integration: Automates model training and deployment using GitHub Actions.
Project proposal View Here
For any questions or issues, reach out to the Promptly team:
- Ronak Vadhaiya - [email protected]
- Sagar Bilwal - [email protected]
- Rajiv Shah - [email protected]
- Kushal Shankar - [email protected]