🧠 AI Knowledge Copilot

AI-powered document intelligence assistant built with OpenAI + Streamlit.

Upload documents, ask questions in natural language, and generate context-aware insights using a lightweight Retrieval-Augmented Generation (RAG-lite) architecture.

---

🌐 Live Demo

🚀 https://ai-knowledge-copilot-qqurmsj8tcmzugptlwo245.streamlit.app/

💻 Best experience on desktop (V1 limitation).

---

🚀 Overview

AI Knowledge Copilot enables users to:

• Upload multiple document types (PDF, DOCX, PPTX, CSV, TXT)
• Extract and aggregate knowledge across files
• Ask natural language questions
• Generate AI-powered summaries and insights
• Interact through a ChatGPT-style interface

This project demonstrates how enterprise AI copilots can transform unstructured data into conversational intelligence.

---

🧠 Core Use Cases

• 📚 Knowledge synthesis
• 🔍 Insight extraction
• ⚖️ Multi-document comparison
• 🧾 Executive briefings
• 🏢 Internal enterprise copilots

---

🏗️ System Architecture

User Uploads Documents
        ↓
Document Parsing Layer
(PDF / DOCX / PPTX / CSV)
        ↓
Text Extraction
        ↓
Context Aggregation
        ↓
GPT-4o-mini Prompting
        ↓
AI Response Generation
        ↓
Chat-style UI Rendering

---

⚙️ Tech Stack

Layer	Technology
Frontend/UI	Streamlit
LLM	OpenAI GPT-4o-mini
PDF Parsing	PyPDF2
DOCX Parsing	python-docx
PPTX Parsing	python-pptx
Structured Data	pandas
Deployment	Streamlit Cloud
Version Control	GitHub

---

💡 Features

• 📂 Multi-file upload
• 💬 Chat-style AI interaction
• ⚡ Suggested questions
• 👍 👎 Feedback system
• 🧠 Context-aware responses
• ☁️ Live cloud deployment

---

⚖️ Product & Engineering Trade-offs

Decision	Reason
RAG-lite architecture	Faster MVP development
No vector database in V1	Reduced complexity and deployment overhead
Streamlit instead of React	Rapid prototyping and iteration
GPT-4o-mini	Lower inference cost and fast responses
Desktop-first UX	Stable and cleaner demo experience

---

📌 Known Limitations (V1)

• Optimized primarily for desktop usage
• Some mobile-uploaded PDFs may fail due to encoding inconsistencies
• OCR for scanned/image-based documents is not yet supported
• No semantic retrieval/vector search yet

---

🚀 Future Roadmap

V2

• Vector embeddings + semantic retrieval
• Inline citations
• Persistent memory
• Robust mobile support
• OCR fallback for scanned documents

V3

• Multi-agent orchestration
• Enterprise integrations (Drive, Slack, Notion)
• Analytics dashboard
• Human-in-the-loop review workflows

---

⚙️ Run Locally

1. Clone the repository

git clone https://github.com/skyplon/ai-knowledge-copilot.git
cd ai-knowledge-copilot

2. Install dependencies

pip install -r requirements.txt

3. Configure API key

Create:

.streamlit/secrets.toml

Add:

OPENAI_API_KEY = "your_api_key_here"

4. Run the app

streamlit run app.py

---

☁️ Deployment

This application is deployed using:

• GitHub
• Streamlit Cloud
• OpenAI API

Every push to the main branch automatically redeploys the app.

---

🧠 AI / Product Thinking Behind the Project

This project was intentionally designed as a lightweight AI copilot MVP focused on:

• Rapid experimentation
• Human-centered AI UX
• Enterprise knowledge workflows
• AI product architecture trade-offs
• Real-world deployment simplicity

Rather than optimizing for scale in V1, the focus was on validating:

• usability
• workflow value
• interaction patterns
• deployment feasibility

---

👤 Author

Juan Navarrete

Senior AI Product Manager focused on:

• Enterprise AI systems
• AI copilots
• Agentic workflows
• Knowledge intelligence platforms

🌎 GitHub: https://github.com/skyplon

---

⭐ Support

If you found this project interesting, feel free to give it a star ⭐