Components, hooks, visualizers, and composable blocks for Pipecat and voice AI.
Canvas-2D audio spectrum visualizer with mel-scale bands and the shared visualizer lifecycle states (connecting, silent, speaking, thinking). Headless view for any MediaStreamTrack, plus a Pipecat-connected wrapper.
Canvas-2D radial audio visualizer: a ring of spokes growing outward with mel-scale voice bands, sharing the kit visualizer lifecycle states (connecting, silent, speaking, thinking). Headless view for any MediaStreamTrack, plus a Pipecat-connected wrapper.
Shader aura audio visualizer with optional dithering: swirling ribbons of turbulent light that animate between the shared visualizer lifecycle states (connecting, silent, speaking, thinking), with the aura swelling on live volume. Ships its own Shadertoy-compatible WebGL runtime. Headless view for any MediaStreamTrack, plus a Pipecat-connected wrapper.
Bot audio output element with a shared volume store, mute toggle, horizontal or vertical volume slider, and a popover volume control.
Client and agent connection status rows, live-updated from transport state and bot lifecycle events.
One button for the whole session lifecycle, driven by the Pipecat transport state. Every state's label, icon, style, and action can be overridden sparsely.
Live scrolling transcript of the user/agent conversation with role labels, karaoke text rendering, function-call cards, and connection-aware empty states.
Media device picker for microphones, speakers, or cameras — as a select or a dropdown menu around any trigger.
Telephone keypad that sends DTMF tones — buffered sequence entry or immediate per-key dialing, with synthesized press feedback.
Single-datapoint metric tile — label, big tabular value, optional unit — fully generic and props-driven, chart-free by design.
Metrics dashboard block with a Performance/Usage tab navigator: per-processor latency tiles and charts, plus running token/character/audio usage totals, on the stock shadcn chart (recharts).
Session metadata rows: transport type, copyable ids, RTVI client/server versions.
Message composer with inline send button that injects the message into the conversation and sends it to the bot.
Caption-style overlay for bot speech with word-by-word fade-in and turn-end fade-out.
Client bootstrap hook with an explicit transport factory, connect/disconnect lifecycle, startBot handling and error state.
Shared RTVI event log: rAF-batched capture with a rolling cap, pause, clear-on-reconnect, and per-subscriber filtering and grouping.
Shared store of RTVI metrics: per-processor ttfb/ttfa/processing/characters/stt_usage series with a rolling cap, running token totals (incl. cache-read and reasoning), and session-scoped reset.
Microphone control with toggle and push-to-talk modes, live audio visualizer, and a mic/speaker device picker in a split-button dropdown.
Screen-share toggle with a preview tile while sharing.
Camera toggle with preview tile and camera picker in a split-button dropdown.