sigoden/aichat

Features

• Agent Orchestration Frameworks - Provides a framework for orchestrating specialized AI agents with custom prompts and tool-calling capabilities.
• Interactive REPLs - Provides a terminal-based interactive shell for executing AI prompts with tab-completion and multi-line editing.
• Interactive Shells - Provides an interactive REPL with tab autocompletion, history search, and multi-line input for conversational AI.
• Model Provider Integrations - Offers a unified interface for connecting and switching between multiple different language model providers.
• OpenAI-Compatible APIs - Translates internal requests into the OpenAI API standard to ensure compatibility with various model providers.
• AI Agent Orchestration - Provides a framework to build specialized AI agents combining custom prompts, tool calling, and RAG.
• Conversational Session Management - Stores conversation history in sessions to maintain context across multiple interactions.
• Conversational Document Querying - Enables interactive querying of indexed documents with citations using retrieval-augmented generation.
• RAG Document Retrieval - Retrieves relevant snippets from local files and URLs to provide grounded context for model responses.
• External Tool Integration - Integrates the AI agent with external tools and data sources to automate tasks and retrieve real-time info.
• Function Calling Interfaces - Trigger external tools and organize them into aliases for faster activation.
• Persistent Chat Histories - Persists conversation history and compresses long contexts into summaries to stay within token limits.
• Retrieval Augmented Generation - Implements a RAG pipeline to ground model responses in external data for improved accuracy.
• LLM Gateways - Functions as a local HTTP server that aggregates multiple LLM providers into a single API gateway.
• LLM Provider Integrations - Provides configurations and authentication adapters to connect to various external large language model providers.
• Local RAG Implementations - Implements a local-first RAG system that indexes private files and URLs for custom model context.
• Role-Based Model Assignment - Enables combining specific prompts and model configurations to tailor AI behavior for specialized roles.
• Parallel Function Calling - Enables models to trigger external functions and API invocations to extend their operational capabilities.
• Persona Instruction Sets - Provides reusable prompt templates to set system context and standardize model behavior via personas.
• Shell Command Generation - Translates natural language descriptions into precise shell commands tailored to the current operating system.
• Tool Calling - Implements mechanisms for the model to request and execute external shell commands and API tools.
• AI Agents and Automation - Orchestrates specialized AI agents and conversation starters to automate complex workflows.
• Local Knowledge Base Indexers - Indexes local files and URLs into a vector database to create a custom knowledge base for RAG.
• Stateful Session Management - Persists conversation history to local storage and utilizes summarization to maintain context within token limits.
• Interactive AI Shells - Offers a unified terminal-based REPL for interacting with multiple LLM providers.
• Natural Language Automation - Translates natural language descriptions of tasks into system-specific shell commands and executes them.
• Tool-Call Orchestrators - Integrates external functions into the chat workflow to perform system tasks like file system operations.
• Conversational Session Managers - Organizes conversations into distinct sessions that can be saved, cleared, compressed, or switched.
• Agent Deployment - Packages custom instructions, tools, and documents into specialized agent instances for complex tasks.
• YAML-Based Workflows - Runs predefined sequences of commands using YAML definitions and variables within an isolated context.
• Cross-Model Comparators - Provides a web-based interface to send the same prompt to multiple models and compare responses side-by-side.
• Document Loaders - Extends the system to process non-text binary files through external document loader services.
• LLM Response Streaming - Delivers language model outputs incrementally to the client to reduce perceived latency.
• Model Configuration Settings - Adjusts model parameters such as temperature and token limits during an active session.
• Sampling Parameter Tuning - Allows adjusting inference-time sampling parameters like temperature to control the randomness of generated text.
• Model Switching Interfaces - Enables dynamic switching of active language models and provides details on their capabilities and pricing.
• Result Reranking - Uses a reranker model to optimize the relevance of merged vector and keyword search results.
• Prompt Augmenters - Injects relevant document snippets into prompts using embedding models and rerankers.
• Macro Automation - Combines a sequence of REPL commands into single custom macros for streamlined operation.
• Data Ingestion Sources - Accepts data from standard input, local files, directories, and remote URLs for model processing.
• Contextual Content Injections - Imports local files, remote URLs, or shell command outputs to serve as immediate context for prompts.
• Command Automation - Executes predefined sequences of commands to automate repetitive tasks within the interface.
• Command Line Input Utilities - Optimizes command entry via tab completion, history search, and external editor integration for multi-line input.
• REPL Macros - Groups series of interactive REPL commands into custom macros to streamline repetitive workflows.
• AI Personas - Provides a system for switching between predefined AI personas and custom system prompts to change response behavior.
• External Content Handlers - Attaches the contents of local files, directories, or remote URLs to prompts as context.
• Embedded Web Servers - Ships an embedded HTTP server that exposes chat completions and embeddings as a local API.
• Local API Servers - Runs a local HTTP server that exposes chat completions and embeddings for use by other applications.
• Multi-Source File Ingestion - Collects data from diverse sources including stdin, local files, and remote URLs to use as model context.
• Hosted Web Interfaces - Hosts a lightweight HTTP server providing a browser-based playground for interacting with and comparing models.
• Command Line Interfaces - Terminal interface for multiple OpenAI model versions.
• Command Line Tools - All-in-one LLM CLI featuring RAG, shell assistance, and agentic capabilities.
• AI Assistant Tools - All-in-one CLI for LLM interaction and RAG.
• AI Assistants and Tools - All-in-one CLI for LLM interaction and RAG.

Star history