Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.
Python agentagentic-aiagentic-frameworkagentic-workflowaiautomationclaudedeepseekgenaigptllmlow-codemcpnextjsno-codeopenaipythonskillsworkflow
"""Opt-in HTTP fixtures for provider discovery without a deployed plugin daemon."""
from collections.abc import Callable, Iterator
from threading import Thread
import pytest
from flask import Flask
from werkzeug.serving import make_server
from yarl import URL
from core.plugin.impl import base as plugin_base
@pytest.fixture(scope="session")
def _empty_model_provider_daemon_url() -> Iterator[URL]:
app = Flask(__name__)
@app.get("/plugin/<tenant_id>/management/models")
def list_model_providers(tenant_id: str) -> dict[str, object]:
return {"code": 0, "message": "success", "data": []}
server = make_server("127.0.0.1", 0, app, threaded=True)
thread = Thread(target=server.serve_forever, kwargs={"poll_interval": 0.01}, daemon=True)
thread.start()
try:
yield URL(f"http://127.0.0.1:{server.server_port}")
finally:
server.shutdown()
thread.join()
server.server_close()
@pytest.fixture
def _empty_model_provider_daemon(
_empty_model_provider_daemon_url: URL,
config_overrides: Callable[..., None],
monkeypatch: pytest.MonkeyPatch,
) -> None:
"""Run real model discovery against an empty installation over loopback HTTP.
Skip the global unit-test Redis mock so a cache response cannot determine
whether malformed stored providers reach their intended resolution error.
"""
config_overrides(PLUGIN_MODEL_PROVIDERS_CACHE_ENABLED=False)
monkeypatch.setattr(plugin_base, "plugin_daemon_inner_api_baseurl", _empty_model_provider_daemon_url)