github/deer-flow
1
0
Fork 0
mirror of https://github.com/bytedance/deer-flow.git synced 2026-10-26 13:16:24 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(config): guard numeric config fields that accepted silently-broken values (#6026)

Browse Source 
* fix(config): guard numeric config fields that accepted silently-broken values

Ten numeric fields across six config models admitted zero, negatives, or
booleans (YAML/JSON true coerces to 1) and only failed at runtime — or kept
running with a nonsense value:

- CircuitBreakerConfig: failure_threshold >= 1, recovery_timeout_sec > 0
- McpServerConfig: tool_call_timeout > 0, session_init_timeout > 0
- McpOAuthConfig.refresh_skew_seconds >= 0 (0 stays valid)
- ModelConfig.stream_chunk_timeout > 0; context_window keeps gt=0 and
  gains boolean rejection (review follow-up)
- RunEventsConfig.max_trace_content >= 1
- SandboxConfig.port 1..65535, idle_timeout >= 0 (0 = documented disable)
- SummarizationConfig.trim_tokens_to_summarize >= 1

Review follow-ups: boolean rejection extended to every bounded numeric
sibling in sandbox_config (replicas, acquire_timeout, burst_limit,
health_check_skip_seconds, bash_output_max_chars, read_file_output_max_chars,
ls_output_max_chars, bash_command_timeout, temporary_grant_ttl,
renewal_interval_seconds, ttl_multiplier) so the file is uniform, and to
model_config.context_window. Tests cover allow_inf_nan=False rejection
(float('inf')/'nan') and the refresh_skew_seconds=0 boundary.

Generated-by: ZCode (GLM, coding agent)

Authored-by: liwenjie200543 <liwenjie200543@users.noreply.github.com>

* style(config): apply Ruff formatting to numeric guards

Signed-off-by: Willem Jiang <willem.jiang@gmail.com>

---------

Signed-off-by: Willem Jiang <willem.jiang@gmail.com>
Co-authored-by: liwenjie200543 <liwenjie200543@users.noreply.github.com>
Co-authored-by: Willem Jiang <willem.jiang@gmail.com>
This commit is contained in:
kuse 2026-09-30 10:48:26 +08:00 committed by GitHub
parent fcff55a075
commit f2f9e6f6fe
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
7 changed files with 278 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

21
backend/packages/harness/deerflow/config/app_config.py
View File

@ -7,7 +7,7 @@ from typing import Any, Literal, Self
import yaml
from dotenv import load_dotenv
from pydantic import BaseModel, ConfigDict, Field, PrivateAttr, model_validator
from pydantic import BaseModel, ConfigDict, Field, PrivateAttr, field_validator, model_validator
from deerflow.config.acp_config import ACPAgentConfig, load_acp_config_from_dict
from deerflow.config.agent_storage_config import AgentStorageConfig
@ -77,8 +77,23 @@ CONFIG_FILE_DATABASE_DEFAULTS = {
class CircuitBreakerConfig(BaseModel):
"""Configuration for the LLM Circuit Breaker."""
failure_threshold: int = Field(default=5, description="Number of consecutive failures before tripping the circuit")
recovery_timeout_sec: int = Field(default=60, description="Time in seconds before attempting to recover the circuit")
failure_threshold: int = Field(
default=5,
ge=1,
description="Number of consecutive failures before tripping the circuit",
)
recovery_timeout_sec: int = Field(
default=60,
gt=0,
description="Time in seconds before attempting to recover the circuit",
)
@field_validator("failure_threshold", "recovery_timeout_sec", mode="before")
@classmethod
def _reject_boolean_circuit_settings(cls, value: object) -> object:
if isinstance(value, bool):
raise ValueError("must be an integer, not a boolean")
return value
class LlmCallConfig(BaseModel):

26
backend/packages/harness/deerflow/config/extensions_config.py
View File

@ -198,7 +198,19 @@ class McpOAuthConfig(BaseModel):
token_type_field: str = Field(default="token_type", description="Field name containing token type in token response")
expires_in_field: str = Field(default="expires_in", description="Field name containing expiry (seconds) in token response")
default_token_type: str = Field(default="Bearer", description="Default token type when missing in token response")
refresh_skew_seconds: int = Field(default=60, description="Refresh token this many seconds before expiry")
refresh_skew_seconds: int = Field(
default=60,
ge=0,
description="Refresh token this many seconds before expiry",
)
@field_validator("refresh_skew_seconds", mode="before")
@classmethod
def _reject_boolean_refresh_skew(cls, value: object) -> object:
if isinstance(value, bool):
raise ValueError("must be an integer, not a boolean")
return value
extra_token_params: dict[str, str] = Field(default_factory=dict, description="Additional form params sent to token endpoint")
model_config = ConfigDict(extra="allow")
@ -232,10 +244,14 @@ class McpServerConfig(BaseModel):
)
tool_call_timeout: float | None = Field(
default=None,
gt=0,
allow_inf_nan=False,
description=("Timeout in seconds for individual stdio MCP tool calls and durable-task calls on every transport. Other HTTP/SSE tools use transport-level timeouts. None means no call-level timeout."),
)
session_init_timeout: float | None = Field(
default=DEFAULT_MCP_SESSION_INIT_TIMEOUT,
gt=0,
allow_inf_nan=False,
description=(
"Timeout in seconds for MCP server bring-up: tool discovery (subprocess spawn + initialize + tools/list) "
"and persistent stdio session initialization, plus ephemeral HTTP/SSE durable-task session "
@ -243,6 +259,14 @@ class McpServerConfig(BaseModel):
"construction or the task poller indefinitely. None means no timeout."
),
)
@field_validator("tool_call_timeout", "session_init_timeout", mode="before")
@classmethod
def _reject_boolean_mcp_timeouts(cls, value: object) -> object:
if isinstance(value, bool):
raise ValueError("must be a number, not a boolean")
return value
task_toolsets: list[McpTaskToolsetConfig] = Field(
default_factory=list,
description="Ordinary submit/status/cancel tool groups managed by the durable MCP task runtime",

10
backend/packages/harness/deerflow/config/model_config.py
View File

@ -184,6 +184,8 @@ class ModelConfig(BaseModel):
)
stream_chunk_timeout: float | None = Field(
default=None,
gt=0,
allow_inf_nan=False,
description=(
"Maximum seconds to wait between successive streaming chunks before "
"langchain-openai raises StreamChunkTimeoutError. None means use the "
@ -192,6 +194,14 @@ class ModelConfig(BaseModel):
"interactive endpoints. Has no effect on non-OpenAI-compatible providers."
),
)
@field_validator("stream_chunk_timeout", "context_window", mode="before")
@classmethod
def _reject_boolean_numeric_settings(cls, value: object) -> object:
if isinstance(value, bool):
raise ValueError("must be a number, not a boolean")
return value
thinking: dict | None = Field(
default_factory=lambda: None,
description=(

11
backend/packages/harness/deerflow/config/run_events_config.py
View File

@ -15,7 +15,7 @@ from __future__ import annotations
from typing import Literal
from pydantic import BaseModel, Field
from pydantic import BaseModel, Field, field_validator
class RunEventsConfig(BaseModel):
@ -25,8 +25,17 @@ class RunEventsConfig(BaseModel):
)
max_trace_content: int = Field(
default=10240,
ge=1,
description="Maximum trace content size in bytes before truncation (db backend only).",
)
@field_validator("max_trace_content", mode="before")
@classmethod
def _reject_boolean_max_trace_content(cls, value: object) -> object:
if isinstance(value, bool):
raise ValueError("must be an integer, not a boolean")
return value
track_token_usage: bool = Field(
default=True,
description="Whether RunJournal should accumulate token counts to RunRow.",

39
backend/packages/harness/deerflow/config/sandbox_config.py
View File

@ -36,6 +36,14 @@ class SandboxNetworkConfig(BaseModel):
le=3600,
description="Lifetime in seconds for the temporary approval choice.",
)
@field_validator("temporary_grant_ttl", mode="before")
@classmethod
def _reject_boolean_temporary_grant_ttl(cls, value: object) -> object:
if isinstance(value, bool):
raise ValueError("must be an integer, not a boolean")
return value
proxy_image: str = Field(
default="ghcr.io/bytedance/deer-flow-sandbox-network-proxy:latest",
min_length=1,
@ -112,6 +120,14 @@ class SandboxOwnershipConfig(BaseModel):
allow_inf_nan=False,
description="Lease TTL as a multiple of renewal_interval_seconds. At least 2, so a single missed renewal (slow host, brief Redis blip) cannot expire a live owner's lease. Default 4 tolerates three consecutive misses.",
)
@field_validator("renewal_interval_seconds", "ttl_multiplier", mode="before")
@classmethod
def _reject_boolean_ownership_settings(cls, value: object) -> object:
if isinstance(value, bool):
raise ValueError("must be a number, not a boolean")
return value
key_prefix: str = Field(
default="deerflow:sandbox:owner",
description="Redis key prefix for ownership leases. Only applies to the redis ownership type.",
@ -204,6 +220,8 @@ class SandboxConfig(BaseModel):
)
port: int | None = Field(
default=None,
ge=1,
le=65535,
description="Base port for sandbox containers",
)
replicas: int | None = Field(
@ -231,8 +249,29 @@ class SandboxConfig(BaseModel):
)
idle_timeout: int | None = Field(
default=None,
ge=0,
description="Idle timeout in seconds before released warm sandboxes/VMs are stopped (default: 600 = 10 minutes). Set to 0 to disable.",
)
@field_validator(
"port",
"replicas",
"acquire_timeout",
"burst_limit",
"idle_timeout",
"health_check_skip_seconds",
"bash_output_max_chars",
"read_file_output_max_chars",
"ls_output_max_chars",
"bash_command_timeout",
mode="before",
)
@classmethod
def _reject_boolean_numeric_settings(cls, value: object) -> object:
if isinstance(value, bool):
raise ValueError("must be a number, not a boolean")
return value
health_check_skip_seconds: float | None = Field(
default=None,
ge=0,

11
backend/packages/harness/deerflow/config/summarization_config.py
View File

@ -3,7 +3,7 @@
import math
from typing import Literal
from pydantic import BaseModel, Field, model_validator
from pydantic import BaseModel, Field, field_validator, model_validator
ContextSizeType = Literal["fraction", "tokens", "messages"]
DEFAULT_SKILL_FILE_READ_TOOL_NAMES: tuple[str, ...] = ("read_file", "read", "view", "cat")
@ -82,8 +82,17 @@ class SummarizationConfig(BaseModel):
)
trim_tokens_to_summarize: int | None = Field(
default=4000,
ge=1,
description="Maximum tokens to keep when preparing messages for summarization. Pass null to skip trimming.",
)
@field_validator("trim_tokens_to_summarize", mode="before")
@classmethod
def _reject_boolean_trim_tokens(cls, value: object) -> object:
if isinstance(value, bool):
raise ValueError("must be an integer, not a boolean")
return value
summary_prompt: str | None = Field(
default=None,
description="Custom prompt template for generating summaries. If not provided, uses the default LangChain prompt.",

166
backend/tests/test_numeric_config_guards.py Normal file
View File

@ -0,0 +1,166 @@
# SPDX-License-Identifier: MIT
"""Numeric config fields that previously accepted silently-broken values.
Each guarded field used to admit zero, negatives, or booleans (YAML/JSON
``true`` coerces to ``1``) and only failed later at runtime — or worse, kept
running with a nonsense value. Bounds and the boolean rejection mirror the
guards already shipped for ``stream_bridge_config`` (heartbeat interval) and
``database_config`` (pool settings).
"""
from __future__ import annotations
import pytest
from pydantic import ValidationError
from deerflow.config.app_config import CircuitBreakerConfig
from deerflow.config.extensions_config import McpOAuthConfig, McpServerConfig
from deerflow.config.model_config import ModelConfig
from deerflow.config.run_events_config import RunEventsConfig
from deerflow.config.sandbox_config import SandboxConfig
from deerflow.config.summarization_config import SummarizationConfig
class TestCircuitBreakerGuards:
def test_defaults_pass(self) -> None:
config = CircuitBreakerConfig()
assert config.failure_threshold == 5
assert config.recovery_timeout_sec == 60
@pytest.mark.parametrize("field", ["failure_threshold", "recovery_timeout_sec"])
@pytest.mark.parametrize("bad", [0, -1, True])
def test_rejects_non_positive_and_boolean(self, field: str, bad: int | bool) -> None:
with pytest.raises(ValidationError):
CircuitBreakerConfig(**{field: bad})
class TestMcpTimeoutGuards:
def test_defaults_pass(self) -> None:
config = McpServerConfig()
assert config.tool_call_timeout is None
assert config.session_init_timeout is not None
@pytest.mark.parametrize("field", ["tool_call_timeout", "session_init_timeout"])
@pytest.mark.parametrize("bad", [0, -1.0, True])
def test_rejects_non_positive_and_boolean(self, field: str, bad: float | bool) -> None:
with pytest.raises(ValidationError):
McpServerConfig(**{field: bad})
def test_zero_point_five_and_none_are_valid(self) -> None:
config = McpServerConfig(tool_call_timeout=0.5, session_init_timeout=None)
assert config.tool_call_timeout == 0.5
assert config.session_init_timeout is None
@pytest.mark.parametrize("bad", [-1, True])
def test_refresh_skew_rejects_negative_and_boolean(self, bad: int | bool) -> None:
with pytest.raises(ValidationError):
McpOAuthConfig(refresh_skew_seconds=bad)
class TestStreamChunkTimeoutGuard:
def test_none_default_passes(self) -> None:
config = ModelConfig(name="m", use="default", model="gpt-x", stream_chunk_timeout=None)
assert config.stream_chunk_timeout is None
@pytest.mark.parametrize("bad", [0, -1.0, True])
def test_rejects_non_positive_and_boolean(self, bad: float | bool) -> None:
with pytest.raises(ValidationError):
ModelConfig(name="m", use="default", model="gpt-x", stream_chunk_timeout=bad)
class TestMaxTraceContentGuard:
def test_default_passes(self) -> None:
assert RunEventsConfig().max_trace_content == 10240
@pytest.mark.parametrize("bad", [0, -1, True])
def test_rejects_non_positive_and_boolean(self, bad: int | bool) -> None:
with pytest.raises(ValidationError):
RunEventsConfig(**{"max_trace_content": bad})
class TestSandboxPortAndIdleTimeoutGuards:
def test_unset_defaults_pass(self) -> None:
config = SandboxConfig(use="local")
assert config.port is None
assert config.idle_timeout is None
def test_zero_idle_timeout_is_documented_valid(self) -> None:
assert SandboxConfig(use="local", idle_timeout=0).idle_timeout == 0
@pytest.mark.parametrize("bad", [0, 65536, -1, True])
def test_port_rejects_out_of_range_and_boolean(self, bad: int | bool) -> None:
with pytest.raises(ValidationError):
SandboxConfig(use="local", port=bad)
@pytest.mark.parametrize("bad", [-1, True])
def test_idle_timeout_rejects_negative_and_boolean(self, bad: int | bool) -> None:
with pytest.raises(ValidationError):
SandboxConfig(use="local", idle_timeout=bad)
class TestSiblingBoolRejection:
"""Same-class siblings with bounds but no boolean rejection, flagged in review."""
def test_sandbox_numeric_fields_reject_booleans(self) -> None:
import pytest
from deerflow.config.sandbox_config import SandboxOwnershipConfig
with pytest.raises(ValidationError):
SandboxConfig(use="local", replicas=True)
with pytest.raises(ValidationError):
SandboxConfig(use="local", acquire_timeout=True)
with pytest.raises(ValidationError):
SandboxConfig(use="local", burst_limit=True)
with pytest.raises(ValidationError):
SandboxConfig(use="local", health_check_skip_seconds=True)
with pytest.raises(ValidationError):
SandboxConfig(use="local", bash_output_max_chars=True)
with pytest.raises(ValidationError):
SandboxConfig(use="local", read_file_output_max_chars=True)
with pytest.raises(ValidationError):
SandboxConfig(use="local", ls_output_max_chars=True)
with pytest.raises(ValidationError):
SandboxConfig(use="local", bash_command_timeout=True)
with pytest.raises(ValidationError):
SandboxOwnershipConfig(renewal_interval_seconds=True)
with pytest.raises(ValidationError):
SandboxOwnershipConfig(ttl_multiplier=True)
def test_ownership_defaults_still_pass(self) -> None:
from deerflow.config.sandbox_config import SandboxOwnershipConfig
config = SandboxOwnershipConfig()
assert config.renewal_interval_seconds == 30.0
assert config.ttl_multiplier == 4.0
def test_context_window_rejects_booleans_and_nonpositive(self) -> None:
with pytest.raises(ValidationError):
ModelConfig(name="m", use="default", model="gpt-x", context_window=True)
with pytest.raises(ValidationError):
ModelConfig(name="m", use="default", model="gpt-x", context_window=0)
with pytest.raises(ValidationError):
ModelConfig(name="m", use="default", model="gpt-x", context_window=-1)
def test_allow_inf_nan_rejection(self) -> None:
with pytest.raises(ValidationError):
McpServerConfig(tool_call_timeout=float("inf"))
with pytest.raises(ValidationError):
ModelConfig(name="m", use="default", model="gpt-x", stream_chunk_timeout=float("inf"))
with pytest.raises(ValidationError):
ModelConfig(name="m", use="default", model="gpt-x", stream_chunk_timeout=float("nan"))
def test_refresh_skew_zero_is_a_valid_boundary(self) -> None:
config = McpOAuthConfig(token_url="https://sso.example.org/token", refresh_skew_seconds=0)
assert config.refresh_skew_seconds == 0
class TestTrimTokensToSummarizeGuard:
def test_default_and_none_pass(self) -> None:
assert SummarizationConfig().trim_tokens_to_summarize == 4000
assert SummarizationConfig(trim_tokens_to_summarize=None).trim_tokens_to_summarize is None
@pytest.mark.parametrize("bad", [0, -1, True])
def test_rejects_non_positive_and_boolean(self, bad: int | bool) -> None:
with pytest.raises(ValidationError):
SummarizationConfig(**{"trim_tokens_to_summarize": bad})