feat(annotate): auto_serve mode spawns and tears down inference server

Setting --vlm.auto_serve=true with --vlm.backend=openai makes the CLI
launch 'transformers serve <model_id> --port <serve_port>
--continuous-batching' as a child process, poll /v1/models until ready
(up to serve_ready_timeout_s), run the pipeline, then SIGINT the
server on process exit.

Override the spawn command with --vlm.serve_command='vllm serve ...'
or any OpenAI-compatible launcher.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
This commit is contained in:
Pepijn
2026-04-28 16:39:50 +02:00
parent cd12665e85
commit 5003989ae6
2 changed files with 87 additions and 3 deletions
@@ -83,6 +83,18 @@ class VlmConfig:
"""Base URL for the ``openai`` backend."""
api_key: str = "EMPTY"
"""API key for the ``openai`` backend; ``EMPTY`` works for local servers."""
auto_serve: bool = False
"""When True with ``backend=openai``, the CLI spawns the inference
server itself (default: ``transformers serve``), waits for it to be
ready, runs the pipeline, and tears it down on exit. Override the
command via ``serve_command``."""
serve_port: int = 8000
"""Port the auto-spawned server binds to. Sets ``api_base`` automatically."""
serve_command: str | None = None
"""Override the auto-serve command (full shell command). When ``None``,
we run ``transformers serve <model_id> --port <serve_port> --continuous-batching``."""
serve_ready_timeout_s: float = 600.0
"""Max seconds to wait for the server to start serving requests."""
max_new_tokens: int = 512
temperature: float = 0.2
json_mode: bool = True
@@ -257,8 +257,10 @@ def _make_openai_client(config: VlmConfig) -> VlmClient:
"""Backend that talks to any OpenAI-compatible server.
Compatible with ``vllm serve``, ``transformers serve``,
``ktransformers serve``, and hosted endpoints. The server is
expected to be already running and to host ``config.model_id``.
``ktransformers serve``, and hosted endpoints. By default the server
is expected to be already running. Set ``auto_serve=True`` to have
this client spawn one (default: ``transformers serve``), wait until
it's ready, and tear it down on process exit.
Image blocks ``{"type":"image", "image":<PIL.Image>}`` are
auto-converted to ``image_url`` data-URLs. Video blocks
@@ -273,7 +275,11 @@ def _make_openai_client(config: VlmConfig) -> VlmClient:
"Install with `pip install openai`."
) from exc
client = OpenAI(base_url=config.api_base, api_key=config.api_key)
api_base = config.api_base
if config.auto_serve:
api_base = _spawn_inference_server(config)
client = OpenAI(base_url=api_base, api_key=config.api_key)
def _gen(
batch: Sequence[Sequence[dict[str, Any]]], max_tok: int, temp: float
@@ -293,6 +299,72 @@ def _make_openai_client(config: VlmConfig) -> VlmClient:
return _GenericTextClient(_gen, config)
def _spawn_inference_server(config: VlmConfig) -> str:
"""Spawn ``transformers serve`` (or ``serve_command``), wait until it
accepts ``/v1/models``, and register a shutdown hook.
Returns the full ``api_base`` URL the OpenAI client should use.
"""
import atexit # noqa: PLC0415
import logging # noqa: PLC0415
import shlex # noqa: PLC0415
import signal # noqa: PLC0415
import subprocess # noqa: PLC0415
import time # noqa: PLC0415
import urllib.request # noqa: PLC0415
log = logging.getLogger(__name__)
cmd = config.serve_command
if not cmd:
cmd = (
f"transformers serve {shlex.quote(config.model_id)} "
f"--port {config.serve_port} --continuous-batching"
)
api_base = f"http://localhost:{config.serve_port}/v1"
log.info("auto_serve: launching: %s", cmd)
proc = subprocess.Popen(
shlex.split(cmd),
stdout=subprocess.PIPE,
stderr=subprocess.STDOUT,
text=True,
bufsize=1,
)
def _shutdown() -> None:
if proc.poll() is None:
log.info("auto_serve: stopping pid=%s", proc.pid)
proc.send_signal(signal.SIGINT)
try:
proc.wait(timeout=15)
except subprocess.TimeoutExpired:
proc.kill()
proc.wait(timeout=5)
atexit.register(_shutdown)
deadline = time.monotonic() + config.serve_ready_timeout_s
health_url = api_base.rstrip("/") + "/models"
while time.monotonic() < deadline:
if proc.poll() is not None:
tail = proc.stdout.read() if proc.stdout else ""
raise RuntimeError(
f"auto_serve: inference server exited (rc={proc.returncode}). "
f"Tail of output:\n{tail}"
)
try:
with urllib.request.urlopen(health_url, timeout=2) as resp:
if resp.status == 200:
log.info("auto_serve: server ready at %s", api_base)
return api_base
except Exception: # noqa: BLE001 - intentional broad except
pass
time.sleep(2)
proc.terminate()
raise RuntimeError(
f"auto_serve: server did not become ready within {config.serve_ready_timeout_s}s"
)
def _to_openai_message(message: dict[str, Any]) -> dict[str, Any]:
"""Convert an internal message dict to OpenAI chat format.