| 1 | # SPDX-License-Identifier: Apache-2.0 |
| 2 | # SPDX-FileCopyrightText: Copyright contributors to the vLLM project |
| 3 | from argparse import Namespace |
| 4 | |
| 5 | from fastapi import FastAPI |
| 6 | |
| 7 | from vllm import envs |
| 8 | from vllm.config import ModelConfig |
| 9 | from vllm.tasks import POOLING_TASKS, SupportedTask |
| 10 | |
| 11 | |
| 12 | def register_api_routers( |
| 13 | args: Namespace, |
| 14 | app: FastAPI, |
| 15 | supported_tasks: tuple["SupportedTask", ...], |
| 16 | model_config: ModelConfig | None = None, |
| 17 | ): |
| 18 | from vllm.entrypoints.serve import register_vllm_serve_api_routers |
| 19 | |
| 20 | register_vllm_serve_api_routers(app) |
| 21 | |
| 22 | from vllm.entrypoints.openai.models.api_router import ( |
| 23 | attach_router as register_models_api_router, |
| 24 | ) |
| 25 | |
| 26 | register_models_api_router(app) |
| 27 | |
| 28 | from vllm.entrypoints.serve.sagemaker.api_router import ( |
| 29 | attach_router as register_sagemaker_api_router, |
| 30 | ) |
| 31 | |
| 32 | register_sagemaker_api_router(app, supported_tasks, model_config) |
| 33 | |
| 34 | if envs.VLLM_SERVER_DEV_MODE: |
| 35 | from vllm.entrypoints.serve import register_vllm_dev_api_routers |
| 36 | |
| 37 | register_vllm_dev_api_routers(app) |
| 38 | |
| 39 | if "generate" in supported_tasks: |
| 40 | from vllm.entrypoints.generate.api_router import ( |
| 41 | register_generate_api_routers, |
| 42 | ) |
| 43 | |
| 44 | register_generate_api_routers(app) |
| 45 | |
| 46 | from vllm.entrypoints.serve.elastic_ep.api_router import ( |
| 47 | attach_router as elastic_ep_attach_router, |
| 48 | ) |
| 49 | |
| 50 | elastic_ep_attach_router(app) |
| 51 | |
| 52 | if "generate" in supported_tasks or "render" in supported_tasks: |
| 53 | from vllm.entrypoints.scale_out.factories import register_scale_out_api_routers |
| 54 | |
| 55 | register_scale_out_api_routers(app, supported_tasks) |
| 56 | |
| 57 | if "transcription" in supported_tasks or "realtime" in supported_tasks: |
| 58 | from vllm.entrypoints.speech_to_text.factories import ( |
| 59 | register_speech_to_text_api_routers, |
| 60 | ) |
| 61 | |
| 62 | register_speech_to_text_api_routers(app, supported_tasks) |
| 63 | |
| 64 | if any(task in POOLING_TASKS for task in supported_tasks): |
| 65 | from vllm.entrypoints.pooling.factories import register_pooling_api_routers |
| 66 | |
| 67 | register_pooling_api_routers(app, supported_tasks, model_config) |
| 68 | |
| 69 | if getattr(args, "enable_fault_tolerance", False): |
| 70 | from vllm.entrypoints.serve.fault_tolerance.api_router import ( |
| 71 | register_fault_tolerance_api_router, |
| 72 | ) |
| 73 | |
| 74 | register_fault_tolerance_api_router(app) |