charmbracelet/mods

Features

• Language Model Querying - Sends text or file content to an AI model and returns the generated response directly in the command line.
• LLM Query Clients - Queries large language models directly from the command line by piping text or files and streaming responses.
• Multi-turn Interaction Managers - Maintains a history of user and assistant messages so subsequent queries can reference prior context.
• Prompt Persona Definitions - Loads predefined instruction sets or character profiles that shape how the LLM interprets input.
• Provider Switching - Connect to OpenAI, Azure OpenAI, Cohere, Groq, Gemini, or any OpenAI-compatible endpoint by setting the corresponding API key.
• MCP Server Connections - Create and initialize a client for a server using stdio, SSE, or HTTP transport.
• AI Gateways - A command-line gateway that routes queries to OpenAI, Cohere, Groq, Gemini, and local endpoints.
• Direct Model Queries - Sends text or file content as a prompt to a large language model and displays the generated response directly in the command line.
• Persona and Behavioral Instructions - Defines named personas and behavioral roles that control how the AI model interprets and formats output.
• MCP Client Implementations - Implements a Model Context Protocol client that connects to external servers via stdio, SSE, or HTTP to extend model capabilities with specialized tools.
• System Prompt Management - A configuration system that defines named personas and behavioral roles to control model output.
• CLI Prompt Piping - Feeds piped input from other commands or file contents into an AI model for processing and response generation.
• Conversation History Management - Saves, lists, shows, and deletes past interactions with AI models for later review or continuation.
• Conversation Management Systems - Lists, shows, continues, and deletes saved conversations using dedicated command-line flags.
• Conversation State Persistence - Saves each interaction locally with a title and SHA-1 identifier for later review or continuation.
• External Tool Integrations - Connect to MCP servers to access additional tools and capabilities during AI interactions.
• Interactive AI Conversations - Resumes an earlier AI interaction from its saved state to extend the dialogue without starting over.
• LLM Response Streaming - Receives and displays model output token by token as it is generated, showing results before the full response completes.
• LLM-Powered Shell Response Generators - Pipes text or file content into a large language model and returns the generated response directly in the command line.
• Tool Call Executions - Call a tool on a named MCP server by passing JSON arguments and returning the result.
• System Instruction Personas - Loads named role definitions from configuration that prepend custom system prompts to every query, shaping model behavior without modifying user input.
• Model Provider Configurations - Selects which API backend and model to use for a given query from multiple providers.
• File-Based Stores - Persists multi-turn conversation state as local JSON files keyed by SHA-1 hash, enabling save, resume, and branching of dialogues.
• LLM Clients - An AI query tool that sends prompts to large language models and streams responses back to the terminal.
• Streaming AI Gateways - Routes AI model requests through a streaming JSON-RPC interface that handles authentication, rate limiting, and response parsing for multiple backends.
• LLM-As-A-Judge Scoring - Pipes text or file content as a prompt to a large language model and returns the generated response directly in the command line.
• AI Provider Routing - Send queries to OpenAI, Cohere, Groq, Gemini, or a local LocalAI endpoint based on configuration.
• Endpoint Aliases - Define endpoints and aliases for various OpenAI-compatible APIs to switch between providers seamlessly.
• Provider-Agnostic LLM Routing - Sends queries to different cloud and local AI providers and switches between them with a single configuration.
• AI Query Input Loaders - Fetch the body of an HTTP or HTTPS URL and use it as input for an AI query.
• Custom Prompt Editors - Stores reusable role definitions that set the system prompt, allowing users to apply consistent instructions across queries.
• Generation Parameter Configurations - Sets model, temperature, top-p, top-k, max tokens, retries, and proxy to control how the LLM generates responses.
• MCP Servers - Lists configured MCP servers and shows which ones are enabled.
• MCP Tool Retrievers - Connects to each enabled MCP server and prints the tools it exposes.
• Model Context Protocol Clients - Connects to external MCP servers via stdio, SSE, or HTTP to extend AI model capabilities with specialized tools.
• Model Fallbacks - Switch to a configured fallback model when the primary model is unavailable or returns a 404.
• Prompt Prefixes - Prepends a fixed text to every query to steer the model's behavior without modifying the input.
• Theme and Behavior Customizations - Adjusts temperature, token limit, and stop sequences to control AI response style and length.
• AI-Powered Formatting - Transforms raw terminal output into structured formats like Markdown or JSON using an LLM.
• LLM Response Renderers - Asks the model to return output in Markdown and optionally applies syntax highlighting or raw printing.
• Prompt Editors - Opens the current prompt in a text editor for modification before sending it to the AI model.
• Server Error Retries - Re-attempt a request after a transient server-side failure, such as a 500 or 404 for a fallback model.
• Exponential Backoff Retries - Detects rate-limit and transient server errors (404, 500) and re-attempts the request with exponential backoff before falling back to an alternate model.
• API Rate Limit Management - Automatically wait and resend a request when the API reports a rate-limit or overload error.
• Markdown - Asks the language model to structure its reply in Markdown for easier reading and further pipeline processing.
• Command Line Tools - CLI tool for piping text through AI prompts.

Star history