ollama-chat

ollama-chat

Ollama Chat is a conversational AI chat client that uses Ollama to interact with local large language models (LLMs).

Features

• Platform independent - tested on macOS, Windows, and Linux
• Chat with any local Ollama model
• Prompt commands to include files, images, and URL content
• Regenerate the most recent conversation response
• Delete the most recent conversation exchange
• View responses as Markdown or text
• Save conversations as Markdown text
• Multiple concurrent chats
• Conversation templates for repeating prompts with variable substitutions
• Start a converstation or template from the command line
• Download and manage models

Installation

To get up and running with Ollama Chat follows these steps:

1. Install and start Ollama

2. Install Ollama Chat

   pip install ollama-chat
   

Updating

To update Ollama Chat:

pip install -U ollama-chat

Start Ollama Chat

To start Ollama Chat, open a terminal prompt and run the Ollama Chat application:

ollama-chat

A web browser is launched and opens the Ollama Chat web application.

By default, a configuration file, "ollama-chat.json", is created in the user's home directory.

To start a conversation from the command line, use the -m argument:

ollama-chat -m "Why is the sky blue?"

To start a named template from the command line, use the -t and -v arguments:

ollama-chat -t AskAristotle -v Subject "Why is the sky blue?"

Conversation Templates

Conversation Templates allow you to repeat the same prompts with different models. Templates can define variables for use in the template title and prompt text (e.g., {{var}}).

There are two ways to create a template. Click "Add Template" from the index page, and a new template is created and opened in the template editor. The other way is to click "Template" from a conversation view's menu.

Prompt Commands

Ollama Chat supports special prompt commands that allow you to include files, images, and URL content in your prompt, among other things. The following prompt commands are available:

• /file - include a file

  /file README.md
  
  Please summarize the README file.
  

• /image - include an image

  /image image.jpeg
  
  Please summarize the image.
  

• /dir - include files from a directory

  /dir src/ollama_chat py
  
  Please provide a summary for each Ollama Chat source file.
  

• /url - include a URL resource

  /url https://craigahobbs.github.io/ollama-chat/README.md
  
  Please summarize the README file.
  

• /do - execute a conversation template by name

  /do city-report -v CityState "Seattle, WA"
  

To get prompt command help use the -h option:

/file -h

File Format and API Documentation

Ollama Chat File Format

Ollama Chat API

Development

This package is developed using python-build. It was started using python-template as follows:

template-specialize python-template/template/ ollama-chat/ -k package ollama-chat -k name 'Craig A. Hobbs' -k email '[e-mail]' -k github 'craigahobbs' -k noapi 1