| title: PhoneLLM Pipecat XKCD Voice Agent | |
| emoji: 🎙️ | |
| colorFrom: indigo | |
| colorTo: blue | |
| sdk: gradio | |
| sdk_version: 5.20.0 | |
| app_file: app.py | |
| pinned: false | |
| license: mit | |
| short_description: PhoneLLM Pipecat voice agent with live XKCD tools | |
| models: | |
| - pipecat-ai/phonellm-alpha-1 | |
| # 🎙️ PhoneLLM + Pipecat XKCD Voice Agent | |
| This Space loads and serves **[`pipecat-ai/phonellm-alpha-1`](https://huggingface.co/pipecat-ai/phonellm-alpha-1)** (NVIDIA Nemotron 3 Nano 30B architecture) integrated with a live **Pipecat Voice Agent** pipeline. | |
| The model autonomously reasons over voice turns with `temperature=0` (and thinking disabled), making live tool calls to the CORS-enabled XKCD API at **`https://xkcd.hemanth.deno.net/`** to fetch comics and speak back the punchline. | |
| ## 🛠️ Architecture | |
| 1. **LLM Engine:** Direct in-process loading of `pipecat-ai/phonellm-alpha-1` with 4-bit / 8-bit quantization & bfloat16. | |
| 2. **Tool Execution:** Real-time API integration with `https://xkcd.hemanth.deno.net/`. | |
| 3. **Voice UI:** Interactive comic-styled voice interface with live XKCD strip visualizer. | |