add inference api

pyproject.toml

@@ -5,6 +5,9 @@ description = ""
 authors = ["Ftps <ftpsflandre@gmail.com>"]
 readme = "README.md"
 
+[tool.poetry.urls]
+github = "https://github.com/RVC-Project/Retrieval-based-Voice-Conversion"
+
 [tool.poetry.dependencies]
 python = "3.11.2"
 torch = "^2.1.0"
@@ -20,9 +23,19 @@ python-dotenv = "^1.0.0"
 pydub = "^0.25.1"
 click = "^8.1.7"
 tensorboardx = "^2.6.2.2"
+poethepoet = "^0.24.4"
+uvicorn = {version = "^0.26.0", optional=true}
+fastapi = "^0.109.0"
+python-multipart = {version = "^0.0.6", optional=true}
+
+[tool.poetry.extras]
+api = ["uvicorn", "fastapi"]
 
 [tool.poetry.scripts]
-rvc = "rvc.cli.cli:main"
+rvc = "rvc.utils.cli.cli:main"
+
+[tool.poe.tasks]
+rvc-api = "uvicorn rvc.utils.api.api:app --reload"
 
 [build-system]
 requires = ["poetry-core"]

rvc/utils/api/api.py

@@ -0,0 +1,11 @@
+from dotenv import load_dotenv
+from fastapi import FastAPI
+import uvicorn
+
+from rvc.utils.api.endpoints import inference
+
+load_dotenv()
+
+app = FastAPI()
+
+app.include_router(inference.router)

rvc/utils/api/endpoints/inference.py

@@ -0,0 +1,50 @@
+from io import BytesIO
+from pathlib import Path
+
+from fastapi import APIRouter, UploadFile, Response, responses
+from pydantic import BaseModel
+from scipy.io import wavfile
+
+from rvc.modules.vc.modules import VC
+
+router = APIRouter()
+
+
+@router.post("/inference")
+def inference(
+    modelpath: str | UploadFile,
+    input: Path | UploadFile,
+    sid: int = 0,
+    f0_up_key: int = 0,
+    f0_method: str = "rmvpe",
+    f0_file: Path | None = None,
+    index_file: Path | None = None,
+    index_rate: float = 0.75,
+    filter_radius: int = 3,
+    resample_sr: int = 0,
+    rms_mix_rate: float = 0.25,
+    protect: float = 0.33,
+):
+    print(protect)
+    vc = VC()
+    vc.get_vc(modelpath)
+    tgt_sr, audio_opt, times, _ = vc.vc_single(
+        sid,
+        input,
+        f0_up_key,
+        f0_method,
+        f0_file,
+        index_file,
+        index_rate,
+        filter_radius,
+        resample_sr,
+        rms_mix_rate,
+        protect,
+    )
+    wavfile.write(wv := BytesIO(), tgt_sr, audio_opt)
+    print(times)
+    return responses.StreamingResponse(
+        wv,
+        media_type="audio/wav",
+        headers={"Content-Disposition": "attachment; filename=inference.wav"},
+    )