Extract residual-stream activations and apply steering vectors (including activation oracles) to any vLLM model during inference. - View it on GitHub