Discussion with My AI Clones is a Streamlit-based web application that enables dynamic, context-aware conversations using three specialized AI assistants. The assistants—Rag_York, Wiki_York, and Search_York—leverage advanced Retrieval-Augmented Generation (RAG), Wikipedia search, and web search capabilities to provide meaningful insights tailored to user queries.
This application is ideal for professionals, researchers, and enthusiasts who want to explore cutting-edge conversational AI while interacting with their uploaded documents or external information sources.
-
Three Specialized AI Clones:
- Rag_York: Uses RAG to retrieve and summarize content from uploaded PDF documents.
- Wiki_York: Searches Wikipedia for relevant information using inferred keywords from user input.
- Search_York: Performs web searches to enrich responses with real-time information.
-
Editable System Prompt: Tailor the clones' behavior with a prefilled prompt editable via the sidebar.
-
PDF Document Support: Upload multiple PDF files for processing and vectorized retrieval.
-
Responsive Chat Interface:
- Incremental chatbot responses.
- Expandable "Sources" section for each response, showing the context and references used.
-
Keyword Inference: AI-powered extraction of meaningful keywords for precise Wikipedia and web searches.
- PDF Upload: Load one or more PDF documents for context-aware retrieval.
- System Prompt: Edit the behavior of AI clones using a prefilled system prompt.
- Dynamic Responses: View responses from each AI clone, displayed incrementally.
- Expandable Sources: Check the context used for generating replies, including document excerpts, Wikipedia summaries, and web links.
- Chat History: Review previous conversations in a scrollable section.
AGENTIC RAG/
├── __pycache__/ # Compiled Python files
├── agent clone archive/ # Archived versions of agent clone implementations
│ ├── app_v4.py
│ ├── app_v5.py
├── agentic rag app archive/ # Archived versions of main app implementations
│ ├── app_v1.py
│ ├── app_v2.py
│ ├── app_v3.py
├── static/ # Static assets for UI
│ ├── York_AI_1.png
│ ├── York_AI_2.png
│ ├── York_AI_3.png
│ ├── York.jpg
├── .gitignore # Git ignore file for sensitive or unnecessary files
├── app_v6.py # Main Streamlit app file
├── htmlTemplate.py # CSS and HTML templates for chat layout
├── requirements.txt # Python dependencies
- Python 3.10 or higher
- Streamlit installed (
pip install streamlit) - API keys for Groq (https://console.groq.com/login) and Brave Search (https://api.search.brave.com/)
-
📥 Clone the repository
git clone https://github.com/yYorky/AgenticRAG.git cd AgenticRAG -
🐍 Set up a virtual environment
conda create -n <ENV_NAME> python=3.10 conda activate <ENV_NAME>
-
📦 Install dependencies
pip install -r requirements.txt
-
🛠️ Set up environment variables
- Create a
.envfile in the root directory. - Add your API keys:
GROQ_API_KEY=your_groq_api_key BRAVE_API_KEY=your_brave_api_key
- Create a
-
🏃 Run the Streamlit app
streamlit run app_v6.py
- Open the app in your browser.
- Upload one or more PDF documents via the sidebar.
- Enter your query in the input box.
- Review responses incrementally from:
- Rag_York: Contextual retrieval from uploaded PDFs.
- Wiki_York: Wikipedia-based insights using inferred keywords.
- Search_York: Web search results using keyword inference.
- Expand the "Sources" section under each response for detailed references.
-
PDF Processing:
- Uploaded PDFs are split into manageable chunks.
- FAISS vector embeddings enable efficient similarity searches.
-
Keyword Inference:
- For Wiki_York and Search_York, an LLM generates meaningful keywords based on user input.
-
Dynamic Retrieval:
- Rag_York fetches document excerpts.
- Wiki_York queries Wikipedia for the inferred keyword.
- Search_York performs web searches with Brave Search.
-
Chat Interface:
- Responses are displayed incrementally.
- Sources and search results are shown in an expandable section.