Ollama Tools
Ollama Tools Ai
Ollama tools is a side project I am working on to allow ollama models to use tools. What is a tool? A function that allows the AI to do something it navitily can’t like create a folder and file on your locally machine. You can ask the AI to preform a certain task and based on the function you have created enable the AI to preform a task using tools. Some of the newer models like llama3.1 and llama3-groq-tool-use work well preforming these tasks.
The goal is to create a large tools class that the Ai is aware of and can use one or more as needed based on a chat session. I have also added a vector databse with similarity score so the AI can referance previous tasks and conversations. Some of the next steps are to add new tools, add a way to drop any document into the folder and prompt the AI a certain way and it can use a tools to create RAG knowledge set and understand your data better.
Overview
Ollama Tools AI is a work-in-progress project focused on implementing advanced AI functionalities, including function calling and tool usage. This project aims to provide a robust and flexible framework for AI-driven tasks and interactions.
Features
- Function Calling: Seamlessly integrate and call various functions to perform tasks.
- Tool Usage: Utilize different tools for enhanced AI capabilities.
- Web Search: Use searXNG or Tavily for web searches
- Create: The AI will create folders and files, read files, list files, delete files, web search and more..
- DB Similarity: Every chat and tool use is saved in DB for future context aware AI responses.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
# API Keys
TAVILY_API_KEY=tvly-
# Models
OLLAMA_MODEL=llama3-groq-tool-use
EMBED_MODEL=nomic-embed-text
# Search Configuration
SEARCH_PROVIDER=SEARXNG
SEARCH_RESULTS_LIMIT=5
# URLs
OLLAMA_URL=http://127.0.0.1:11434
SEARXNG_URL=http://192.168.70.48:4000
# Database Configuration
DB_DIR=./chromadb # database location realitive to current directory
N_CONTEXTS=3 # Number of contexts to retrieve from DB - you can adjust this for testing
SIMILARITY_THRESHOLD=0.7 # Adjust this value to control context relevance (lower is more strict)
# Debug print statements in terminal False or True
DEBUG_MODE=False
Install dependencies
pip install -r requirements.txt
Usage
1
python ol.py
screenshots
Search results
