Skip to content

A curated collection of practical AI projects implementing OCR systems, RAG, AI agents, and other AI use cases.

feat(agentic_rag): Finalize OpenAI RAG app and update README

Sumanth077b4227abbc9da
diff --git a/agentic_rag/README.md b/agentic_rag/README.md
new file mode 100644
index 0000000..4948d43
--- /dev/null
+++ b/agentic_rag/README.md
@@ -0,0 +1,68 @@
+# PDF RAG Agent with Milvus, Agno and OpenAI
+
+A Streamlit application demonstrating a Retrieval-Augmented Generation (RAG) workflow using OpenAI (`gpt-4o-mini`), Agno, Milvus, and DuckDuckGo for web search.
+
+## Features
+
+- Upload & process PDF documents.
+- OpenAI embeddings & chat completion (`gpt-4o-mini`).
+- Milvus vector store for semantic search.
+- DuckDuckGo web search fallback.
+- Interactive Streamlit chat interface.
+
+## Prerequisites
+
+- Python 3.8+
+- OpenAI API Key
+- Docker (for Milvus)
+- Running Milvus Instance
+
+## Setup
+
+1. **Clone & Navigate:**
+ ```bash
+ # Clone this repo if you haven't already
+ cd agentic_rag
+ ```
+
+2. **Environment & Dependencies:**
+ ```bash
+ python3 -m venv venv
+ source venv/bin/activate # Or . env\Scripts\activate on Windows
+ pip install -r requirements.txt
+ ```
+
+3. **OpenAI API Key:**
+ Set the `OPENAI_API_KEY` environment variable or add it to your Streamlit secrets (`.streamlit/secrets.toml`).
+
+4. **Start Milvus:**
+ Follow the official Milvus guide to start a standalone instance using Docker:
+ [https://milvus.io/docs/install_standalone-docker.md](https://milvus.io/docs/install_standalone-docker.md)
+ *Ensure it's accessible (usually `http://localhost:19530`).*
+
+## Usage
+
+1. **Ensure Milvus is running.**
+2. **Run the app:**
+ ```bash
+ streamlit run app.py
+ ```
+3. Open the provided URL (e.g., `http://localhost:8501`).
+4. Upload a PDF, click "Process Document", and start chatting!
+
+## How it Works
+
+- PDF text is chunked and embedded using OpenAI.
+- Embeddings stored in Milvus.
+- User query triggers the Agno Agent:
+ - Searches Milvus knowledge base first.
+ - Uses DuckDuckGo if knowledge base results are insufficient.
+ - `gpt-4o-mini` synthesizes information to generate the final response.
+
+## Contributing
+
+Contributions, issues, and feature requests are welcome. Please feel free to submit a Pull Request or open an issue.
+
+## License
+
+This project is licensed under the MIT License. See the LICENSE file for details (if one exists).
\ No newline at end of file
diff --git a/agentic_rag/app.py b/agentic_rag/app.py
new file mode 100644
index 0000000..096c92f
--- /dev/null
+++ b/agentic_rag/app.py
@@ -0,0 +1,251 @@
+import streamlit as st
+# Import OpenAI model and embedder classes from Agno
+from agno.models.openai import OpenAIChat
+from agno.embedder.openai import OpenAIEmbedder
+# Removed Ollama imports
+from agno.agent import Agent
+from agno.knowledge.pdf import PDFKnowledgeBase
+from agno.vectordb.milvus import Milvus
+from agno.tools.duckduckgo import DuckDuckGoTools
+import os
+from dotenv import load_dotenv
+from typing import Optional
+import time
+import base64
+
+# Custom CSS for styling
+st.set_page_config(
+ page_title="PDF RAG Agent (OpenAI)",
+ page_icon="πŸ“š",
+ layout="wide"
+)
+
+# Apply custom CSS styling (Reverted to previous style + minor tweaks)
+def apply_custom_css():
+ st.markdown("""
+ <style>
+ .stApp {
+ /* Remove specific app background to allow default Streamlit theme or user settings */
+ /* background-color: #1E1E2E; */
+ color: #FAFAFA; /* Keep light text for dark theme */
+ }
+ .stSidebar > div:first-child {
+ background-image: linear-gradient(to bottom, #262936, #1e202a); /* Dark sidebar */
+ }
+ /* Ensure chat messages stand out slightly */
+ [data-testid="stChatMessage"] {
+ background-color: rgba(74, 74, 106, 0.4);
+ border-radius: 0.5rem;
+ padding: 1rem;
+ margin-bottom: 0.5rem;
+ }
+ [data-testid="stChatMessage"] p {
+ color: inherit; /* Inherit text color */
+ }
+ /* PDF Preview container styling */
+ .pdf-preview-container {
+ border: 1px solid #4A4A6A;
+ border-radius: 0.5rem;
+ padding: 0.5rem;
+ background-color: #262936; /* Match sidebar bg */
+ margin-bottom: 1rem; /* Add margin below preview */
+ }
+ </style>
+ """, unsafe_allow_html=True)
+
+apply_custom_css()
+
+# Load environment variables
+load_dotenv()
+
+# Initialize session state
+if "messages" not in st.session_state:
+ st.session_state.messages = []
+if "document_loaded" not in st.session_state:
+ st.session_state.document_loaded = False
+if "agent" not in st.session_state:
+ st.session_state.agent = None
+if "knowledge_base" not in st.session_state:
+ st.session_state.knowledge_base = None
+
+# Define the agent creation function using OpenAI with updated instructions
+def get_rag_agent(
+ knowledge_base: PDFKnowledgeBase,
+ model_id: str = "gpt-4o-mini",
+ debug_mode: bool = True,
+) -> Agent:
+ """Creates and configures an Agentic RAG Agent for PDF interaction using OpenAI."""
+
+ model = OpenAIChat(id=model_id)
+ # Updated detailed instructions
+ instructions = [
+ "1. Knowledge Base Search:",
+ " - ALWAYS start by searching the knowledge base using search_knowledge_base tool",
+ " - Analyze ALL returned documents thoroughly before responding",
+ " - If multiple documents are returned, synthesize the information coherently",
+ "2. External Search:",
+ " - If knowledge base search yields insufficient results, use duckduckgo_search",
+ " - Focus on reputable sources and recent information",
+ " - Cross-reference information from multiple sources when possible",
+ "3. Citation Precision:",
+ " - Reference page numbers and section headers",
+ " - Distinguish between main content and appendices",
+ "4. Response Quality:",
+ " - Provide specific citations and sources for claims",
+ " - Structure responses with clear sections and bullet points when appropriate",
+ " - Include relevant quotes from source materials",
+ " - Avoid hedging phrases like 'based on my knowledge' or 'depending on the information'",
+ "5. Response Structure:",
+ " - Use markdown for formatting technical content",
+ " - Create bullet points for lists found in documents",
+ " - Preserve important formatting from original PDF",
+ "6. User Interaction:",
+ " - Ask for clarification if the query is ambiguous",
+ " - Break down complex questions into manageable parts",
+ " - Proactively suggest related topics or follow-up questions",
+ "7. Error Handling:",
+ " - If no relevant information is found, clearly state this",
+ " - Suggest alternative approaches or questions",
+ " - Be transparent about limitations in available information",
+ ]
+ pdf_rag_agent: Agent = Agent(
+ model=model,
+ knowledge=knowledge_base,
+ description="You are a helpful Agent called 'Agentic RAG' assisting with questions about a PDF document.", # Updated description
+ instructions=instructions,
+ search_knowledge=True,
+ markdown=True,
+ tools=[DuckDuckGoTools()],
+ show_tool_calls=True,
+ add_datetime_to_instructions=False,
+ debug_mode=debug_mode,
+ )
+ return pdf_rag_agent
+
+# Configure OpenAI API Key
+OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
+if not OPENAI_API_KEY:
+ try:
+ OPENAI_API_KEY = st.secrets["OPENAI_API_KEY"]
+ except (FileNotFoundError, KeyError):
+ st.error("OpenAI API Key not found. Please set the OPENAI_API_KEY environment variable or add it to Streamlit secrets.")
+ st.stop()
+
+# Configure Milvus with OpenAI Embedder
+MILVUS_URI = os.getenv("MILVUS_URI", "http://localhost:19530")
+COLLECTION_NAME = "rag_documents_openai"
+vector_db = Milvus(
+ collection=COLLECTION_NAME,
+ uri=MILVUS_URI,
+ embedder=OpenAIEmbedder()
+)
+
+# Function to display a PDF preview (Using st.markdown)
+def display_pdf_preview(pdf_path):
+ try:
+ with open(pdf_path, "rb") as f:
+ base64_pdf = base64.b64encode(f.read()).decode('utf-8')
+ pdf_display = f'<iframe src="data:application/pdf;base64,{base64_pdf}" width="100%" height="300px" type="application/pdf"></iframe>'
+ st.markdown(pdf_display, unsafe_allow_html=True)
+ except Exception as e:
+ st.error(f"Error displaying PDF: {e}")
+
+# Main layout
+st.markdown("# PDF RAG Agent with Milvus and Agno πŸ“„")
+
+# Sidebar for document upload
+with st.sidebar:
+ st.markdown("## Add your documents!")
+ uploaded_file = st.file_uploader("Choose your .pdf file", type=['pdf'], key="pdf_uploader")
+
+ if uploaded_file is not None:
+ temp_pdf_path = "temp_uploaded.pdf"
+ # Check if the file content is different from the one potentially already processed
+ # This requires storing the name/hash of the last processed file if needed for robust check
+ # Simple check based on file uploader state for now
+ if not st.session_state.document_loaded or st.session_state.get("processed_file_name") != uploaded_file.name:
+ st.session_state.document_loaded = False # Reset if new file uploaded
+ st.session_state.messages = []
+ st.session_state.agent = None
+
+ with open(temp_pdf_path, "wb") as f:
+ f.write(uploaded_file.getvalue())
+
+ file_details = f"{uploaded_file.name} ({round(len(uploaded_file.getvalue())/1024/1024, 2)} MB)"
+ st.write(file_details)
+
+ st.markdown("## PDF Preview")
+ st.markdown('<div class="pdf-preview-container">', unsafe_allow_html=True)
+ display_pdf_preview(temp_pdf_path)
+ st.markdown('</div>', unsafe_allow_html=True)
+
+ # Show process button only if not already processed
+ if not st.session_state.document_loaded:
+ if st.button("Process Document"):
+ with st.spinner("Indexing your document..."):
+ try:
+ knowledge_base = PDFKnowledgeBase(
+ path=temp_pdf_path,
+ vector_db=vector_db,
+ )
+ knowledge_base.load(recreate=True)
+ agent = get_rag_agent(knowledge_base=knowledge_base)
+
+ st.session_state.knowledge_base = knowledge_base
+ st.session_state.agent = agent
+ st.session_state.document_loaded = True
+ st.session_state.processed_file_name = uploaded_file.name # Store processed file name
+ st.session_state.messages = []
+ st.rerun() # Rerun to update UI immediately
+ except Exception as e:
+ st.error(f"An error occurred during processing: {e}")
+ st.error("Ensure your OpenAI API Key is valid and has embedding permissions.")
+ import traceback
+ st.error(traceback.format_exc())
+ else:
+ # Show processed indicator instead of the button
+ st.markdown("βœ… Document Processed Successfully!")
+
+ else:
+ # Clear state if file is removed
+ if st.session_state.document_loaded:
+ st.session_state.document_loaded = False
+ st.session_state.messages = []
+ st.session_state.agent = None
+ st.session_state.pop("processed_file_name", None) # Remove processed file name
+ st.rerun() # Rerun to update main area
+
+# Main chat area
+st.markdown("## Chat")
+
+if not st.session_state.document_loaded:
+ st.info("πŸ‘ˆ Please upload and process a PDF document in the sidebar to start chatting.")
+else:
+ # Display chat messages from history using st.chat_message
+ for message in st.session_state.messages:
+ with st.chat_message(message["role"]):
+ st.markdown(message["content"])
+
+ # Accept user input using st.chat_input
+ if prompt := st.chat_input("Ask a question about the document..."):
+ st.session_state.messages.append({"role": "user", "content": prompt})
+ with st.chat_message("user"):
+ st.markdown(prompt)
+
+ with st.chat_message("assistant"):
+ message_placeholder = st.empty()
+ full_response_content = ""
+ with st.spinner("Thinking..."):
+ try:
+ response_obj = st.session_state.agent.run(prompt)
+ full_response_content = response_obj.content
+ message_placeholder.markdown(full_response_content)
+ except Exception as e:
+ error_message = f"Error generating response: {e}"
+ st.error(error_message)
+ full_response_content = error_message
+ import traceback
+ st.error(traceback.format_exc())
+
+ st.session_state.messages.append({"role": "assistant", "content": full_response_content})
+ st.rerun() # Rerun after getting response to update display immediately
\ No newline at end of file
diff --git a/agentic_rag/requirements.txt b/agentic_rag/requirements.txt
new file mode 100644
index 0000000..43f92fc
--- /dev/null
+++ b/agentic_rag/requirements.txt
@@ -0,0 +1,7 @@
+duckduckgo-search
+google-genai
+pymilvus
+agno>=1.4.2
+streamlit>=1.44.1
+pypdf>=5.4.0
+pydantic>=2.8.2
\ No newline at end of file