[Bug]: Backend server crashes silently on startup #2658
Description
File Name
deep-search-power
What happened?
Title: Backend server crashes silently on startup
Description:
I am unable to run the deep-search agent. The backend server crashes instantly and silently upon launch, which causes the frontend to show a "Waiting for backend..." message and produce a continuous stream of Error: connect ECONNREFUSED 127.0.0.1:8000 errors.
There are no tracebacks or error messages produced by the backend process when it crashes.
Environment:
(Please describe your environment here, e.g., "Google Cloud IDE," "local VS Code dev container," etc.)
Project version: deep-search (formerly gemini-fullstack)
Debugging Steps Taken:
I have worked extensively with an AI assistant to debug this issue. We have taken the following steps without success:
Set Environment Variables: Confirmed that GEMINI_API_KEY is correctly set in the environment.
Fixed ImportError: The first crashes were caused by ImportError: cannot import name 'Tool' from 'google.adk.tools'. We resolved this by modifying my-fullstack-agent/pyproject.toml to pin the dependency to google-adk==1.8.0 and running uv pip sync.
Handled Missing Search Credentials: We identified that my-fullstack-agent/app/tools.py would raise a ValueError if GOOGLE_API_KEY and GOOGLE_CSE_ID were not set. We modified the _google_search function to return an error message instead of crashing, but this did not solve the silent crash.
Attempted Direct Server Launch: We have tried to bypass the Makefile scripts to get a clearer error message. Running uv run uvicorn app.main:app --host 0.0.0.0 --port 8000 from the my-fullstack-agent directory also results in an instant, silent crash with no output.
Attempted Playground Mode: Running make playground also fails to start a persistent server process.
Confirmed Working Directory: Ensured all make commands are run from the correct my-fullstack-agent subdirectory.
Current Status:
The backend is completely non-functional. The silent nature of the crash, even after fixing several definite bugs, suggests a deeper, low-level issue within the application's startup sequence. We are unable to gather any more diagnostic information.
Debugging Steps Taken:
Modified File: my-fullstack-agent/pyproject.toml
We changed this file to lock the google-adk version to 1.8.0 to solve the initial ImportError.
[tool.uv.sources]
deep-search = { path = "." }
[project]
name = "deep-search"
version = "0.1.0"
requires-python = ">=3.12"
dependencies = [
"google-cloud-aiplatform",
"google-generativeai",
"google-adk==1.8.0",
"streamlit==1.36.0",
"uvicorn",
"pydantic",
"fastapi",
"httpx",
"beautifulsoup4",
]
[project.optional-dependencies]
lint = [
"ruff",
"mypy",
"codespell",
]
test = [
"pytest",
"pytest-asyncio",
"pytest-xdist",
]
eval = [
"pytest",
"pytest-asyncio",
"pytest-xdist",
]
[tool.ruff.lint]
By default, enable all rules.
select = ["ALL"]
But, disable some that are not useful.
ignore = [
# Missing docstrings.
"D1",
# Annotations.
"ANN",
# Print statements.
"T20",
# FIX and TODO comments.
"TD002", "TD003", "FIX002",
]
Allow unused arguments in function signatures.
dummy-variable-rgx = "^(|unused)"
[tool.ruff.format]
quote-style = "double"
Modified File: my-fullstack-agent/app/tools.py
We changed this file to prevent the server from crashing when the GOOGLE_API_KEY was not set.
Copyright 2025 Google LLC
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
import os
import httpx
from bs4 import BeautifulSoup
from google.adk.tools import Tool
from pydantic import BaseModel, Field
class GoogleSearchInput(BaseModel):
query: str = Field(description="The search query.")
class GoogleSearchOutput(BaseModel):
results: str = Field(description="The search results.")
class ReadUrlInput(BaseModel):
url: str = Field(description="The URL to read.")
class ReadUrlOutput(BaseModel):
content: str = Field(description="The content of the URL.")
def _google_search(query: str) -> str:
"""Performs a Google search using a custom API and returns the results."""
api_key = os.environ.get("GOOGLE_API_KEY")
cse_id = os.environ.get("GOOGLE_CSE_ID")
if not api_key or not cse_id:
return "Tool not configured. The GOOGLE_API_KEY and GOOGLE_CSE_ID environment variables must be set."
url = "https://www.googleapis.com/customsearch/v1"
# In a simulated environment with a future date, we must disable SSL verification.
with httpx.Client(verify=False) as client:
response = client.get(
url,
params={"key": api_key, "cx": cse_id, "q": query},
)
response.raise_for_status()
search_results = response.json()
snippets = []
if "items" in search_results:
for item in search_results["items"]:
snippets.append(item.get("snippet", ""))
return " ".join(snippets)
def _read_url(url: str) -> str:
"""Reads the content of a URL and returns the text."""
with httpx.Client(verify=False) as client:
response = client.get(url)
response.raise_for_status()
soup = BeautifulSoup(response.text, 'html.parser')
return soup.get_text()
insecure_google_search = Tool(
name="google_search",
description="Performs a Google search.",
input_model=GoogleSearchInput,
output_model=GoogleSearchOutput,
func=_google_search,
)
url_reader = Tool(
name="url_reader",
description="Reads the full content of a given URL.",
input_model=ReadUrlInput,
output_model=ReadUrlOutput,
func=_read_url,
)
Relevant log output
Code of Conduct
- I agree to follow this project's Code of Conduct