OARC CLI
Advancing AI Agent Development with Local Models and Dataset Generation
Ollama Agent Roll Cage (OARC) is a Python-based framework combining the power of ollama LLMs, Coqui-TTS, Keras classifiers, LLaVA Vision, Whisper Speech Recognition, and YOLOv8 Object Detection into a unified chatbot agent API for local, custom automation.
Streamline chatbot agent design and deployment with a comprehensive API.
Integrate speech, vision, and data retrieval seamlessly.
Build tailored workflows for your unique use cases.
"An agent refers to the algorithmic logic that wraps a model and via iteration, generates the chain of thought output for the model" — Borch
OARC aims to achieve multi-modal super alignment within Agentic action spaces and inter-communication protocols.
Where LLM-generated text prompts transcend mere words and become actionable through programming logic.
Configure agent capabilities (TTS_FLAG, STT_FLAG, LLAVA_FLAG, etc.)
Define which language, vision, and speech models the agent uses
External services and functions the agent can access
OARC implements a sophisticated speech-to-speech pipeline that enables fluid voice interaction with agents.
Silence removal preprocess, smart user interrupt, wake words, and debate moderator mode.
LLM sentence chunking preprocess for TTS with real-time generation.
A comprehensive Python library for AI research, dataset generation, and conversation management using large language models.
Search and process web content, ArXiv papers, and GitHub repositories.
Create and expand high-quality conversation datasets for AI training.
Clean and validate generated datasets to ensure high quality.