Yappr AI Agent

A locally running AI agent built with Python and Ollama that can use tools, search the web, remember information, read documents, and retrieve relevant information using RAG. It combines these capabilities into a modular local assistant that can handle different tasks and interact with external information.

Python · Ollama · Qwen 2.5 · SQLite · Tavily · RAG

What is Yappr?

Yappr is a local AI agent built with Python and Ollama.

It can use external tools, access real-world information, maintain persistent memory, process documents, and retrieve relevant information using a lightweight RAG pipeline.

Current capabilities

• Calculator
• Current weather
• Time and time zones
• Web search
• Persistent SQLite memory
• Document processing
• RAG retrieval
• Multiple tool calls
• Runtime logging
• Error handling
• Automated testing

Architecture

User → Yappr → Local LLM → Tool Registry → Tools → Results → Local LLM → Response

Turing uses a dynamic tool registry to keep tools separate from the main agent, making the architecture easier to extend and maintain.

Tools

Supported documents

• .txt
• .md
• .json
• .csv
• .pdf
• .docx

Tech Stack

• Python — Agent logic and tool execution
• Ollama — Local LLM runtime
• Qwen 2.5 — Language model
• SQLite — Persistent memory
• Tavily — Web search
• Open-Meteo — Weather data
• Ollama Embeddings — Document embeddings
• PyMuPDF — PDF processing
• python-docx — DOCX processing
• pytest — Automated testing

Project Structure

turing-ai-agent/
│
├── assets/
│   ├── architecture.png
│   └── demo.gif
│
├── data/
│   ├── memory.db
│   ├── memoryexample.db
│   └── README.md
│
├── documents/
│   ├── empty.txt
│   ├── notes.csv
│   ├── notes.docx
│   ├── notes.json
│   ├── notes.md
│   ├── notes.pdf
│   ├── notes.txt
│   └── README.md
│
├── logs/
│   ├── agent.log
│   ├── agentexample.log
│   └── README.md
│
├── RAG/
│   ├── embeddings/
│   │   └── ollama_embedding.py
│   ├── chunker.py
│   ├── indexer.py
│   ├── retriever.py
│   ├── similarity.py
│   └── vector_store.py
│
├── tests/
│   ├── __init__.py
│   ├── test_calculator.py
│   ├── test_document_reader.py
│   ├── test_memory.py
│   ├── test_registry.py
│   └── test_web_search.py
│
├── tools/
│   ├── __init__.py
│   ├── calculator.py
│   ├── document_reader.py
│   ├── memory_tool.py
│   ├── rag.py
│   ├── registry.py
│   ├── time.py
│   ├── weather.py
│   └── web_search.py
│
├── .env.example
├── .gitignore
├── main.py
├── memory.py
├── requirements.txt
├── setup.sh
├── shell.sh
└── README.md

Directory overview

assets/
Contains the project architecture image and demo GIF used by the README.

data/
Contains the SQLite database used for persistent memory and an example database for reference.

documents/
Contains sample documents used for document processing and RAG testing. The project supports TXT, Markdown, JSON, CSV, PDF, and DOCX files.

logs/
Contains the runtime log generated by Turing and an example log file.

RAG/
Contains the retrieval-augmented generation pipeline, including document chunking, indexing, embeddings, similarity calculation, vector storage, and retrieval.

tests/
Contains automated tests for the project's core tools and components.

tools/
Contains Yappr's individual tools and the dynamic tool registry used by the agent.

Setup

Requirements

• Python 3
• Ollama
• Git
• Tavily API key

1. Clone the repository

git clone <your-repository-url>
cd ai_agent

2. Run the setup script

chmod +x setup.sh shell.sh
./setup.sh

The setup script:

• Checks for Python
• Creates the virtual environment
• Installs Python dependencies
• Checks for Ollama
• Pulls the Qwen model if required
• Pulls the embedding model if required

3. Configure environment variables

Create a .env file based on .env.example:

TAVILY_API_KEY=your_api_key_here

The .env file contains private credentials and should never be committed to GitHub.

4. Start Yappr

./shell.sh

The shell script activates the virtual environment and starts main.py.

Testing

Turing uses pytest for automated testing.

Run the complete test suite with:

pytest

The test suite covers components including:

• Calculator
• Document reader
• Memory
• Tool registry
• Web search

Contributions

Contributions are welcome.

You can:

• Report bugs
• Suggest improvements
• Improve documentation
• Submit pull requests
• Add useful tools or features

Please open an issue before making major architectural changes.

Author

Slyyr $