[codex] voxcpm phone assistant bridge

.gitignore

@@ -5,3 +5,4 @@ voxcpm.egg-info
 .DS_Store
 ./pretrained_models/
 app_local.py
+.voxcpm-phone-profiles/

README.md

@@ -247,6 +247,29 @@ python app.py --device auto
 
 Supported values are `auto`, `cpu`, `mps`, `cuda`, and `cuda:N`. On Apple Silicon Macs, `auto` uses MPS when available.
 
+### Phone Assistant Bridge
+
+For a mobile-first assistant flow with voice input, webhook-backed replies, and VoxCPM voice cloning:
+
+```bash
+voxcpm-assistant --port 8809
+```
+
+Use a clean reference sample plus an exact transcript for the highest-quality cloned voice. If you connect a webhook, the app POSTs the user message and conversation history and expects a reply in plain text or in a `reply` field.
+If you want phone microphone dictation, install the optional ASR dependency separately with `pip install funasr`.
+You can also save a named voice enrollment once and reuse it later from the app's profile panel. Pass `--default-profile-name <name>` to auto-load a saved profile on start.
+
+To wire in an assistant backend automatically, set `ASSISTANT_BACKEND_URL` plus optional `ASSISTANT_BACKEND_TOKEN`, `ASSISTANT_CONTEXT`, `ASSISTANT_BACKEND_MODE=auto|openai|custom`, and `ASSISTANT_BACKEND_MODEL` (model name for OpenAI-compatible endpoints, default `assistant`) before launching.
+
+To stage each turn to a capture webhook, also set `CAPTURE_WEBHOOK_URL` plus optional `CAPTURE_WEBHOOK_TOKEN`. The assistant POSTs a neutral, backend-agnostic capture event:
+
+```json
+{"source": "voxcpm-phone-assistant", "type": "text", "timestamp": "...",
+ "message": "...", "reply": "...", "profile": "...", "turn": 1}
+```
+
+Mapping this onto a specific ingestion schema (field renames, classification, validation) is the receiving webhook's responsibility.
+
 ### 🚢 Production Deployment (Nano-vLLM)
 
 For high-throughput serving, use **[Nano-vLLM-VoxCPM](https://github.com/a710128/nanovllm-voxcpm)** — a dedicated inference engine built on Nano-vLLM with concurrent request support and an async API.
@@ -627,4 +650,4 @@ VoxCPM model weights and code are open-sourced under the [Apache-2.0](LICENSE) l
 
 ## ⭐ Star History
 
-[Star History Chart](https://star-history.com/#OpenBMB/VoxCPM&Date)
+[Star History Chart](https://star-history.com/#OpenBMB/VoxCPM&Date)

app.py

@@ -5,9 +5,13 @@
 import numpy as np
 import gradio as gr
 from typing import Optional, Tuple
-from funasr import AutoModel
 from pathlib import Path
 
+try:
+    from funasr import AutoModel
+except ImportError:  # Optional dependency for ASR-assisted prompt transcription
+    AutoModel = None
+
 os.environ["TOKENIZERS_PARALLELISM"] = "false"
 
 import voxcpm
@@ -245,6 +249,10 @@ def get_or_load_voxcpm(self) -> voxcpm.VoxCPM:
         return self.voxcpm_model
 
     def get_or_load_asr_model(self) -> AutoModel:
+        if AutoModel is None:
+            raise RuntimeError(
+                "funasr is not installed. Install the optional ASR dependency to enable prompt transcription."
+            )
         if self.asr_model is not None:
             return self.asr_model
         logger.info(

lora_ft_webui.py

@@ -22,7 +22,11 @@
 from voxcpm.core import VoxCPM
 from voxcpm.model.voxcpm import LoRAConfig
 import numpy as np
-from funasr import AutoModel
+
+try:
+    from funasr import AutoModel
+except ImportError:  # Optional dependency for ASR-assisted prompt transcription
+    AutoModel = None
 
 # --- Localization ---
 LANG_DICT = {
@@ -121,6 +125,10 @@ def detect_sample_rate(pretrained_path: str) -> Optional[int]:
 
 def get_or_load_asr_model():
     global asr_model
+    if AutoModel is None:
+        raise RuntimeError(
+            "funasr is not installed. Install the optional ASR dependency to enable prompt transcription."
+        )
     if asr_model is None:
         print("Loading ASR model (SenseVoiceSmall)...", file=sys.stderr)
         device = "cuda:0" if torch.cuda.is_available() else "cpu"

pyproject.toml

@@ -44,7 +44,6 @@ dependencies = [
     "soundfile",
     "librosa",
     "matplotlib",
-    "funasr",
     "spaces",
     "argbind",
     "safetensors"
@@ -62,6 +61,7 @@ dev = [
 
 [project.scripts]
 voxcpm = "voxcpm.cli:main"
+voxcpm-assistant = "voxcpm.phone_assistant:main"
 
 [project.urls]
 Homepage = "https://github.com/OpenBMB/VoxCPM"