This script enables CLI interaction with language models through LiteLLM. It supports multiple language model providers like OpenAI, Anthropic, and Ollama.
- An API key for language model providers like OpenAI or Anthropic, set in
config.yamlor as respective environment variables (e.g.,OPENAI_API_KEY). - If using Ollama ensure it is running, the model is available, and the base URL is set correctly in
config.yaml.
Run the following command in your terminal:
/bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/pahar0/cli-gpt-chat/main/install.sh)"# Clone the repo
git clone [email protected]:pahar0/cli-gpt-chat.git
# Navigate to the project directory
cd cli-gpt-chat
# Install dependencies
pip install -r requirements.txt
# Make gpt.py executable
chmod +x gpt.py
# Create a symlink
sudo ln -sf $(pwd)/gpt.py /usr/local/bin/gpt
# Refresh environment
source ~/.bashrcsystem_prompt: Define the initial system message to start conversations.conversation_file: Designate a file to save conversation history, aiding in context awareness.conversation_expiry_hours: Set the duration after which the conversation history expires and resets.default_provider: Set your default language model provider.debug: Enable or disable debug mode for additional log information.providers_map: Map provider names to their respective models, API keys and base URLs for API requests.
Run the gpt command in your terminal. You can interact with the script using the following commands:
/prompt: Show or change the prompt./conversation: Show the current conversation./provider: Show or change the provider and model./delete: Delete the current conversation./clear: Clear the terminal screen./debug: Toggle debug mode./bye: Exit the program. Same as Ctrl+C.
These settings will apply only to the current session. To change the default settings, edit the config.yaml file.
You can add any of the LiteLLM providers to the providers_map in the config.yaml file.
You can find a list of available providers here.
- You may encounter a warning about protected namespaces due to the pydantic dependency. This warning can safely be ignored. Alternatively, you can eliminate the warning by downgrading the pydantic version using the following command:
pip install pydantic==1.10.13