AI Templates Repository

This repository contains a collection of code templates, snippets, and SDK docs for AI projects. It is code I create during my projects and has some reusability.

Overview

This repository serves as a central location for reusable code patterns across different AI domains, including:

• Retrieval Augmented Generation (RAG)
• Agent-based systems with BeeAI
• OpenAI API integration
• Document processing and extraction
• Vector database integration

Repository Structure

ai-templates/
├── Agents                    # Agent-based systems and frameworks
│   └── BeeAI                 # BeeAI framework examples
├── OpenAI                    # OpenAI API integration templates
│   └── OpenAI Responses API  # Responses API documentation and examples
├── Retrieval augmented generation      # RAG implementation examples
│   ├── Advanced techniques             # Enhanced RAG techniques
│   ├── Document processors             # Document parsing and chunking
│   ├── Elastic search                  # Elasticsearch integration
│   └── Milvus                          # Milvus integration
└── Watsonx                   # WatsonX AI platform integration
    └── Watsonx.ai Python SDK # WatsonX Python SDK documentation

Components

Agents & BeeAI Framework

The BeeAI framework provides tools for building agent-based systems with LLMs. Key examples include:

• Basic agents: Foundation for building AI agents (basics_notebook.py)
• ReAct agents: Implementation of the ReAct (Reasoning and Acting) pattern (agents_notebook.py)
• Travel agent example: Complete agent implementation for travel recommendations (travel-agent-notebook.py/.ipynb)
• Workflow implementation: Process orchestration with BeeAI (workflows_notebook.py)
• WatsonX integration: Connecting BeeAI with WatsonX models (watsonx_notebook.py)

OpenAI Integration

Documentation and examples for the OpenAI Responses API, covering:

• Built-in tools: Leverage OpenAI's built-in capabilities
• Conversation state: Managing multi-turn conversations
• File search: Retrieve information from uploaded files
• Function calling: Enable models to call your code
• Images and vision: Process and analyze images
• Structured outputs: Generate responses in specific formats
• Text generation and prompting: Core text generation capabilities
• Web search: Integrate web search functionality

Retrieval Augmented Generation (RAG)

Templates and implementations for RAG systems, including:

Document Processors

• MarkdownProcessor: Process and chunk markdown documents
• PDFChunker: Extract and segment text from PDFs
• PyMuPDFProcessor: Alternative PDF processing with PyMuPDF
• UnstructuredProcessor: General document processing

Vector Databases

• Elasticsearch integration: Full implementation using Elasticsearch
• Milvus integration: Alternative implementation with Milvus

Advanced Techniques

• Contextual Retrieval: Enhance retrieval with contextually relevant questions

WatsonX Integration

Documentation for the WatsonX.ai Python SDK, including:

• Base functionality
• Connection management
• Foundation models
• Extension frameworks

Getting Started

Prerequisites

This repository contains templates for different frameworks. Depending on the specific template, you'll need:

• Python 3.8+
• Required Python packages in the respective requirements.txt files

Environment Setup

Most examples require API keys or other credentials. Set these up in a .env file based on the provided env_example files.

For WatsonX examples:

WATSONX_API_KEY=your_api_key
WATSONX_PROJECT_ID=your_project_id
WATSONX_URL=your_url

For Elasticsearch examples:

ELASTIC_URL=your_elastic_url
ELASTIC_USER=your_username
ELASTIC_PASSWORD=your_password

For Milvus examples:

MILVUS_HOST=localhost
MILVUS_PORT=19530

Running Examples

Most examples are provided as Python scripts (.py) or Jupyter notebooks (.ipynb). To run:

1. Install the required dependencies
2. Set up your environment variables
3. Execute the example file:
   • For Python scripts: python example.py
   • For Jupyter notebooks: Open in Jupyter Lab/Notebook and run cells

Key Technologies

• BeeAI: Framework for building AI agents
• WatsonX.ai: IBM's AI platform with language models and embeddings
• OpenAI APIs: Various OpenAI APIs, particularly the Responses API
• Vector Databases: Elasticsearch and Milvus
• Document Processing Libraries: PyMuPDF, PDF-Chunker, Unstructured

Use Cases

These templates can be used for:

• Building conversational AI agents
• Implementing knowledge retrieval systems
• Processing and extracting information from documents
• Creating domain-specific assistants
• Developing travel recommendation systems
• Implementing question-answering systems over custom data

Some interesting links

• https://github.com/IBM/watsonx-developer-hub
• https://github.com/i-am-bee/beeai-framework
• https://github.com/IBM/watsonx-ai-samples
• https://github.com/watson-developer-cloud/watsonx-orchestrate-developer-toolkit
• https://suedbroecker.net/archive/
• https://heidloff.net
• https://github.com/IBM/watson-openscale-samples/tree/main

Contributing

Feel free to add new templates or improve existing ones. Create a pull request with your changes or additions.

License

[Specify your license information here]