`jianmu.config`¶

适用对象：框架使用者 / 节点作者 / 维护者是否必读：按需相关模块：jianmu.node, jianmu.guard

1. 模块职责¶

jianmu.config 是配置 re-export 层，汇总了框架级配置模型、节点配置模型以及少量与约束相关的公共类型。

这里适合查两类内容：

jianmu.yaml 对应的配置模型
内置节点和预制模式使用的配置结构

2. 适合查什么¶

项目级配置入口：get_config()、load_project_env()
框架配置模型：JianmuConfig、PathsConfig
节点配置模型：AgentLLMConfig、ToolExecutorConfig
预制模式配置：ReActConfig、PlanExecuteConfig

3. 使用建议¶

需要读取项目级默认配置时，优先看 get_config() 和 load_project_env()
需要为节点构造结构化配置时，优先使用这里公开的 *Config
如果只是使用现成节点，不必先理解全部配置模型

4. 注意事项¶

这里公开的是稳定配置入口，不代表所有底层字段都适合作为外部协议长期依赖
Constraints 的 canonical path 是 jianmu.config.constraints.Constraints

5. 最小示例¶

from jianmu.config import AgentLLMConfig, ReActConfig, get_config

project_config = get_config()
react_config = ReActConfig(model="gpt-4.1-mini")
llm_config = AgentLLMConfig(temperature=0.2)

6. 常见入口¶

想读项目级配置：看 get_config()
想加载 .env / 项目环境：看 load_project_env()
想给节点传结构化配置：看 AgentLLMConfig、ToolExecutorConfig、ReActConfig

7. API 参考¶

config ¶

Jianmu configuration package.

Constraints `dataclass` ¶

Constraints(
    *,
    guard: GuardConstraints | dict[str, Any] | None = None,
    execution: ExecutionConstraints
    | dict[str, Any]
    | None = None,
    timeout: float | None = None,
    max_iterations: int | None = None,
    max_messages: int | None = None,
)

High-level runtime constraints shared across guard and execution layers.

max_iterations caps the main agent/react loop rounds when a node exposes iterative planning/execution behaviour.

属性：

名称	类型	描述
`guard`	`GuardConstraints`	Guard-layer constraints applied to approvals and policy checks.
`execution`	`ExecutionConstraints`	Execution-layer constraints applied to tool/runtime sandboxes.
`timeout`	`Optional[float]`	Optional end-to-end timeout in seconds.
`max_iterations`	`Optional[int]`	Optional cap for iterative agent or ReAct loops.
`max_messages`	`Optional[int]`	Optional cap for retained message history.

源代码位于： jianmu/config/constraints.py

def __init__(
    self,
    *,
    guard: GuardConstraints | dict[str, Any] | None = None,
    execution: ExecutionConstraints | dict[str, Any] | None = None,
    timeout: float | None = None,
    max_iterations: int | None = None,
    max_messages: int | None = None,
) -> None:
    self.guard = (
        guard
        if isinstance(guard, GuardConstraints)
        else GuardConstraints.from_dict(guard)
    ) or GuardConstraints()
    self.execution = _coerce_execution_constraints(execution)
    self.timeout = _parse_float(timeout)
    self.max_iterations = _parse_int(max_iterations)
    self.max_messages = _parse_int(max_messages)

from_dict `classmethod` ¶

from_dict(data: dict[str, Any] | None) -> 'Constraints'

Build constraints from the canonical nested dictionary structure.

参数：

名称	类型	描述	默认
`data`	`dict[str, Any] \| None`	Canonical nested constraints dictionary.	必需

返回：

类型	描述
`'Constraints'`	Parsed `Constraints` instance.

源代码位于： jianmu/config/constraints.py

@classmethod
def from_dict(cls, data: dict[str, Any] | None) -> "Constraints":
    """Build constraints from the canonical nested dictionary structure.

    Args:
        data: Canonical nested constraints dictionary.

    Returns:
        Parsed ``Constraints`` instance.
    """
    payload = dict(data or {})
    return cls(
        guard=payload.get("guard"),
        execution=payload.get("execution"),
        timeout=payload.get("timeout"),
        max_iterations=payload.get("max_iterations"),
        max_messages=payload.get("max_messages"),
    )

coerce `classmethod` ¶

coerce(data: Any | None) -> 'Constraints | None'

Coerce arbitrary input into a Constraints instance.

参数：

名称	类型	描述	默认
`data`	`Any \| None`	Arbitrary value to coerce into constraints.	必需

返回：

类型	描述
`'Constraints \| None'`	Constraints instance, or `None` when coercion is not possible.

源代码位于： jianmu/config/constraints.py

@classmethod
def coerce(cls, data: Any | None) -> "Constraints | None":
    """Coerce arbitrary input into a Constraints instance.

    Args:
        data: Arbitrary value to coerce into constraints.

    Returns:
        Constraints instance, or ``None`` when coercion is not possible.
    """
    if data is None:
        return None
    if isinstance(data, cls):
        return data
    if isinstance(data, dict):
        return cls.from_dict(data)
    return None

to_dict ¶

to_dict() -> dict[str, Any]

Serialize constraints to the nested canonical structure.

返回：

类型	描述
`dict[str, Any]`	Dictionary representation of the constraints object.

源代码位于： jianmu/config/constraints.py

def to_dict(self) -> dict[str, Any]:
    """Serialize constraints to the nested canonical structure.

    Returns:
        Dictionary representation of the constraints object.
    """
    return asdict(self)

JianmuConfig ¶

Bases: BaseModel

Project-level project configuration root.

属性：

名称	类型	描述
`paths`	`PathsConfig`	File-system paths used by the project.
`models`	`ModelsConfig`	Model-name defaults for different workflows.
`llm`	`LLMDefaultsConfig`	Shared LLM sampling and timeout defaults.
`limits`	`LimitsConfig`	Cross-cutting operational limits.
`prompt`	`PromptConfig`	Prompt-template and bootstrap defaults.
`logging`	`LoggingConfig`	Logging behavior defaults.
`trace`	`TraceConfig`	Telemetry and trace emission defaults.
`provider_limits`	`ProviderLimitsConfig`	Provider concurrency defaults.
`execution`	`ExecutionConfig`	Tool execution backend defaults.
`guard`	`GuardConfig`	Project-level tool permission defaults.
`runtime`	`RuntimeConfig`	Reactive runner scheduling defaults.
`rag`	`RagConfig`	Retrieval-augmented generation defaults.

PathsConfig ¶

Bases: BaseModel

File path configuration.

属性：

名称	类型	描述
`outputs_dir`	`str`	Root output directory used for generated artifacts.
`store`	`str`	Default state/message store file path.
`knowledge_db`	`str`	Default SQLite database path for knowledge or RAG data.
`checkpoints_dir`	`str`	Directory used to persist runtime checkpoints.
`sandbox_config`	`Optional[str]`	Reserved for future use.
`agents_dir`	`Optional[str]`	Reserved for future use.
`workflows_dir`	`Optional[str]`	Reserved for future use.
`prompts_dir`	`Optional[str]`	Reserved for future use.
`skills_dir`	`Optional[str]`	Optional directory containing custom skills.

ModelsConfig ¶

Bases: BaseModel

Configuration settings for LLM models used by different components of Jianmu.

属性：

名称	类型	描述
`default`	`str`	Default model name for general-purpose generation.
`fallback`	`str`	Reserved for future use.
`react`	`str`	Default model for ReAct-based nodes.
`plan_execute`	`str`	Default model for plan-execute workflows.
`evaluate`	`str`	Default model for evaluator or reviewer nodes.

LLMDefaultsConfig ¶

Bases: BaseModel

Default hyperparameter settings for LLM interactions in the system.

属性：

名称	类型	描述
`temperature`	`float`	Default sampling temperature.
`top_p`	`float`	Default nucleus-sampling cutoff.
`top_k`	`int`	Default top-k sampling cutoff.
`timeout`	`float`	Default request timeout in seconds.
`max_tokens`	`Optional[int]`	Optional default completion-token limit.
`max_budget_tokens`	`Optional[int]`	Optional budget limit aggregated across one run.
`max_retries`	`int`	Default retry count for model calls.

LimitsConfig ¶

Bases: BaseModel

Limits and threshold configurations for execution steps and memory storage.

属性：

名称	类型	描述
`max_rounds`	`int`	Default maximum rounds for plan-execute loops.
`max_iterations`	`int`	Default maximum iterations for ReAct loops.
`history_limit`	`int`	Default maximum number of messages kept in history windows.
`memory_top_k`	`int`	Default number of memory items retrieved per query.
`chunk_size`	`int`	Default text chunk size for indexing or splitting.
`chunk_overlap`	`int`	Default overlap between adjacent chunks.
`rerank_alpha`	`float`	Default hybrid-search weighting factor.

PromptConfig ¶

Bases: BaseModel

Static prompt defaults.

属性：

名称	类型	描述
`persona_default_prompt`	`Optional[str]`	Optional default persona/base prompt shared by context presets.
`react_default_prompt`	`Optional[str]`	Legacy combined ReAct prompt override.
`react_protocol_prompt`	`Optional[str]`	Optional reusable protocol-only prompt fragment shared by ReAct-like execution paths.
`react_soft_landing_prompt`	`Optional[str]`	Optional final-answer prompt fragment used by ReAct soft-landing fallback when execution budget is exhausted.
`plan_execute_plan_prompt`	`Optional[str]`	Optional default planning prompt template.
`plan_execute_execute_prompt`	`Optional[str]`	Optional default execution prompt template.
`plan_execute_review_prompt`	`Optional[str]`	Optional default review prompt template.
`tools_prompt_template`	`Optional[str]`	Optional template wrapping the rendered tool description block.
`history_semantics_prompt`	`Optional[str]`	Optional prompt explaining normalized tool planning and observation history semantics.
`history_self_tool_plan_text`	`Optional[str]`	Optional rewritten content for prior assistant tool-plan messages.
`history_tool_observation_note`	`Optional[str]`	Optional rewritten note for prior tool observation messages.
`skill_capability_policy_prompt`	`Optional[str]`	Optional prompt describing skill capability semantics.
`skill_on_demand_policy_prompt`	`Optional[str]`	Optional prompt describing lazy skill loading behavior.
`skill_fallback_policy_prompt`	`Optional[str]`	Optional prompt describing fallback skill injection behavior.
`skill_resource_usage_policy_prompt`	`Optional[str]`	Optional prompt describing runtime skill resource usage.
`active_skills_header`	`Optional[str]`	Optional heading for inlined active skill blocks.
`skill_detail_heading_template`	`Optional[str]`	Optional template for one inlined skill detail block.
`skill_resources_heading_template`	`Optional[str]`	Optional template for one skill resource heading.
`skill_workdir_label`	`Optional[str]`	Optional label template for skill working directory display.
`skill_root_label`	`Optional[str]`	Optional label template for skill root display.
`skill_file_label`	`Optional[str]`	Optional label template for a single skill file.
`skill_files_label`	`Optional[str]`	Optional label template for multiple skill files.
`skills_index_heading`	`Optional[str]`	Optional heading template for the skill summary index block.
`skills_list_heading`	`Optional[str]`	Optional heading template for a non-indexed skill list block.
`skills_list_footer`	`Optional[str]`	Optional footer used beneath non-indexed skill lists.
`skill_instructions_heading`	`Optional[str]`	Optional heading template for fully inlined skill instructions.
`bt_skill_summary_hint_template`	`Optional[str]`	Optional template describing how to invoke a BT skill from summaries.
`prompt_skill_summary_hint`	`Optional[str]`	Optional hint describing prompt-skill behavior in summaries.
`skill_summary_line_template`	`Optional[str]`	Optional template for one skill summary line.
`bootstrap_files`	`list[str]`	Workspace files summarized into prompt bootstrap context.

LoggingConfig ¶

Bases: BaseModel

Configuration settings for logging level and terminal colorization.

属性：

名称	类型	描述
`level`	`str`	Default logger level name.
`colorize`	`bool`	Whether terminal logs should include ANSI colors.

TraceConfig ¶

Bases: BaseModel

Settings for trace logging, telemetry events, and routing protocols.

属性：

名称	类型	描述
`log_events`	`bool`	Whether trace events are mirrored to logs.
`safe_mode`	`bool`	Whether telemetry payloads should be sanitized conservatively.
`sample_rate`	`float`	Default sampling rate for trace emission.
`token_sample_rate`	`float`	Sampling rate applied specifically to token events.
`max_per_sec`	`int`	Per-trace/per-agent emission rate limit.
`event_sample_rates`	`dict[str, float]`	Per-event sampling overrides.
`always_events`	`list[str]`	Event names that bypass normal sampling.
`router`	`Optional[str]`	Optional trace router URI such as a file sink.

ProviderLimitsConfig ¶

Bases: BaseModel

Provider concurrency defaults.

属性：

名称	类型	描述
`openai_max_concurrency`	`Optional[int]`	Default max concurrent requests for OpenAI-compatible providers.
`litellm_max_concurrency`	`Optional[int]`	Default max concurrent requests for LiteLLM providers.

ExecutionDockerConfig ¶

Bases: BaseModel

Docker-backed tool execution defaults.

属性：

名称	类型	描述
`image`	`str`	Default Docker image used for sandboxed execution.
`timeout`	`float`	Default Docker job timeout in seconds.
`network`	`str`	Default Docker network mode.
`memory`	`Optional[str]`	Optional Docker memory limit.
`cpus`	`Optional[float]`	Optional Docker CPU limit.
`reuse_container`	`bool`	Whether tool discovery and execution should reuse one long-lived helper container via `docker exec`.
`env_allowlist`	`list[str]`	Explicit environment variable names passed into containers.
`env_allowprefix`	`list[str]`	Environment variable prefixes allowed into containers.

ExecutionConfig ¶

Bases: BaseModel

Configuration options for executing commands, tools, and Docker sandbox settings.

属性：

名称	类型	描述
`max_retries`	`int`	Default retry count for tool execution.
`retry_backoff`	`float`	Default delay in seconds between tool retry attempts.
`timeout_s`	`float \| None`	Default timeout in seconds for one tool execution.
`tool_runner`	`str`	Default tool execution backend name.
`confirm_default`	`str`	Default confirmation mode for guarded execution.
`tool_args_max`	`Optional[int]`	Optional maximum serialized tool-argument size.
`docker`	`ExecutionDockerConfig`	Docker-specific execution defaults.

RuntimeCheckpointConfig ¶

Bases: BaseModel

Runtime checkpoint behavior defaults.

属性：

名称	类型	描述
`enabled`	`bool`	Whether runners should use a config-derived default checkpointer when one is not passed explicitly.
`backend`	`str`	Default checkpoint backend name.
`interval`	`int`	Default tick interval between checkpoint saves.

RuntimeConfig ¶

Bases: BaseModel

Reactive runner scheduling defaults.

属性：

名称	类型	描述
`max_fps`	`float`	Maximum runner tick frequency.
`max_pending_wakeups`	`int`	Maximum queued wake-up signals for async runners.
`hot_loop_warn_factor`	`float`	Threshold multiplier used to warn about hot loops.
`checkpoint`	`'RuntimeCheckpointConfig'`	Runtime checkpoint behavior defaults.

RagEmbeddingConfig ¶

Bases: BaseModel

Configuration for RAG embedding models and API endpoint locations.

属性：

名称	类型	描述
`provider`	`str`	Embedding provider name.
`model`	`Optional[str]`	Optional embedding model identifier.
`base_url`	`Optional[str]`	Optional custom API base URL.

RagConfig ¶

Bases: BaseModel

Root configurations for Retrieval Augmented Generation (RAG) components.

属性：

名称	类型	描述
`embedding`	`RagEmbeddingConfig`	Embedding-provider defaults used by RAG components.

StateKeys ¶

Bases: BaseModel

Single source of truth for state field naming conventions.

属性：

名称	类型	描述
`messages`	`str`	State key for conversation history.
`rounds`	`str`	State key for loop round counters.
`actions`	`str`	State key for model-emitted actions or tool calls.
`done`	`str`	State key indicating workflow completion.
`text_output`	`str`	State key for latest plain-text node output.
`final_answer`	`str`	State key for final user-facing answer text.
`skill_result`	`str`	State key for structured skill outputs.
`usage`	`str`	State key for aggregated model-usage metadata.
`streaming_output`	`str`	State key for incremental streamed text.
`tool_effects`	`str`	State key for aggregated tool-side-effect counters.

EvaluationStateKeys ¶

Bases: BaseModel

State field routing for EvaluationNode outputs.

属性：

名称	类型	描述
`score`	`str \| None`	State key receiving the evaluator score.
`reflection`	`str \| None`	State key receiving the evaluator reflection text.
`highest_score`	`str \| None`	State key tracking the best score so far.
`best_answer`	`str \| None`	State key storing the best answer content.
`best_answer_source`	`str`	State key pointing to the answer candidate being scored.

AgentLLMConfig ¶

Bases: BaseModel

LLM generation parameters plus state routing keys.

system_prompt configures the persona/base prompt. Context presets may still append protocol, tools, skills, history, and other runtime blocks.

属性：

名称	类型	描述
`model`	`str`	Primary model identifier used for inference.
`temperature`	`float`	Sampling temperature for generation.
`max_tokens`	`Optional[int]`	Optional output-token cap.
`top_p`	`Optional[float]`	Optional nucleus-sampling parameter.
`top_k`	`Optional[int]`	Optional top-k sampling parameter.
`timeout`	`float`	Provider timeout in seconds.
`stream`	`bool`	Whether streaming responses are requested.
`max_budget_tokens`	`Optional[int]`	Optional budget cap shared with runtime accounting.
`resilience`	`ModelResilienceConfig`	Retry and fallback policy for provider calls.
`extra_params`	`Dict[str, Any]`	Additional provider-specific generation parameters.
`system_prompt`	`Optional[str]`	Optional persona/base prompt override.
`keys`	`StateKeys`	State-key mapping used by the node.

ModelResilienceConfig ¶

Bases: BaseModel

Framework-level retry/fallback policy for one model call path.

属性：

名称	类型	描述
`max_retries`	`int`	Maximum retry count for a single model call.
`fallback_model`	`Optional[str]`	Optional fallback model used after retries fail.
`enable_fallback`	`bool`	Whether fallback model routing is enabled.

ReActConfig ¶

Bases: BaseModel

ReAct (Reason + Act) loop configuration.

system_prompt configures the persona/base prompt. The ReAct protocol prompt and tool descriptions are added by the context builder. max_iterations is the main agent-loop round limit, not a low-level tree tick limit.

属性：

名称	类型	描述
`model`	`str`	Primary model identifier used for ReAct turns.
`temperature`	`float`	Sampling temperature for generation.
`max_tokens`	`Optional[int]`	Optional output-token cap.
`top_p`	`Optional[float]`	Optional nucleus-sampling parameter.
`top_k`	`Optional[int]`	Optional top-k sampling parameter.
`timeout`	`float`	Provider timeout in seconds.
`stream`	`bool`	Whether streaming responses are requested.
`max_budget_tokens`	`Optional[int]`	Optional budget cap shared with runtime accounting.
`max_iterations`	`int`	Main ReAct loop round limit.
`soft_landing`	`bool`	Whether soft-landing summary behavior is enabled.
`resilience`	`ModelResilienceConfig`	Retry and fallback policy for provider calls.
`system_prompt`	`Optional[str]`	Optional persona/base prompt override.
`tool_choice`	`Optional[Any]`	Provider-specific tool selection directive applied only to the first model round of each ReAct run.
`extra_params`	`Dict[str, Any]`	Additional provider-specific generation parameters.
`keys`	`StateKeys`	State-key mapping used by the preset.

SkillNodeConfig ¶

Bases: BaseModel

Configuration options specifically for skill-execution nodes in Jianmu.

属性：

名称	类型	描述
`push_to_chat`	`bool \| None`	Whether skill outputs should be appended to chat history.
`use_history`	`bool`	Whether prior message history is included in skill runs.
`history_limit`	`int \| None`	Optional cap on retained history for skill prompting.
`react_config`	`Optional[ReActConfig]`	Optional ReAct configuration for nested skill execution.
`constraints`	`Optional[Constraints]`	Optional runtime and guard constraints for the skill node.
`skill_prompt_mode`	`Literal['summary', 'full']`	Prompt rendering mode for skills.
`keys`	`StateKeys`	State-key mapping used by the node.

ToolExecutorConfig ¶

Bases: BaseModel

Tool executor configuration.

属性：

名称	类型	描述
`max_retries`	`int`	Maximum retry count for tool execution failures.
`retry_backoff`	`float`	Backoff delay between retries.
`timeout_s`	`float \| None`	Optional timeout in seconds for each tool execution.
`observation_format`	`str`	Format used for tool observations.
`execution_mode`	`Literal['auto', 'serial', 'parallel']`	Tool scheduling mode: auto, serial, or parallel.
`keys`	`StateKeys`	State-key mapping used by the executor.

PlanExecuteConfig ¶

Bases: BaseModel

Plan-Execute(-Review) loop configuration.

属性：

名称	类型	描述
`model`	`str`	Primary model identifier used across the preset.
`temperature`	`float`	Sampling temperature for generation.
`max_tokens`	`Optional[int]`	Optional output-token cap.
`top_p`	`Optional[float]`	Optional nucleus-sampling parameter.
`top_k`	`Optional[int]`	Optional top-k sampling parameter.
`timeout`	`float`	Provider timeout in seconds.
`max_budget_tokens`	`Optional[int]`	Optional budget cap shared with runtime accounting.
`stream`	`bool`	Whether streaming responses are requested.
`max_rounds`	`int`	Maximum plan/execute rounds before termination.
`enable_review`	`bool`	Whether review steps are enabled.
`review_threshold`	`float`	Score threshold for successful review.
`plan_prompt`	`Optional[str]`	Optional custom planning prompt override.
`execute_prompt`	`Optional[str]`	Optional custom execution prompt override.
`review_prompt`	`Optional[str]`	Optional custom review prompt override.
`plan_key`	`str`	State key used to store the plan.
`score_key`	`str`	State key used to store review scores.
`keys`	`StateKeys`	State-key mapping used by the preset.

bootstrap_project ¶

bootstrap_project(
    config: JianmuConfig | None = None,
    root: str | Path | None = None,
) -> JianmuConfig

Load environment variables and apply logging for a project config.

参数：

名称	类型	描述	默认
`config`	`JianmuConfig \| None`	Configuration object for the operation.	`None`
`root`	`str \| Path \| None`	Root path used for resolution.	`None`

返回：

类型	描述
`JianmuConfig`	The resulting `JianmuConfig` value.

源代码位于： jianmu/config/loader.py

def bootstrap_project(config: JianmuConfig | None = None, root: str | Path | None = None) -> JianmuConfig:
    """Load environment variables and apply logging for a project config.

    Args:
        config: Configuration object for the operation.
        root: Root path used for resolution.

    Returns:
        The resulting `JianmuConfig` value.
    """
    resolved = config or get_config(root=root)
    load_project_env(root, override=True)
    configure_logging_from_config(resolved)
    try:
        from jianmu.telemetry.bootstrap import bootstrap_telemetry
    except Exception:
        return resolved
    bootstrap_telemetry(resolved)
    return resolved

configure_logging_from_config ¶

configure_logging_from_config(config: JianmuConfig) -> None

Apply logging settings from config when telemetry is available.

参数：

名称	类型	描述	默认
`config`	`JianmuConfig`	Configuration object for the operation.	必需

源代码位于： jianmu/config/loader.py

def configure_logging_from_config(config: JianmuConfig) -> None:
    """Apply logging settings from config when telemetry is available.

    Args:
        config: Configuration object for the operation.
    """
    try:
        from jianmu.telemetry.logging import configure_logging
    except Exception:
        return
    configure_logging(
        level=config.logging.level,
        colorize=config.logging.colorize,
        force=True,
    )

get_config ¶

get_config(
    reload: bool = False, root: str | Path | None = None
) -> JianmuConfig

Load and cache the project configuration.

参数：

名称	类型	描述	默认
`reload`	`bool`	Whether to force a fresh load from disk.	`False`
`root`	`str \| Path \| None`	Optional project root or config file path.	`None`

返回：

类型	描述
`JianmuConfig`	Resolved project configuration with absolute paths applied.

引发：

类型	描述
`FileNotFoundError`	If a root path is specified but no config file is found.

源代码位于： jianmu/config/loader.py

def get_config(reload: bool = False, root: str | Path | None = None) -> JianmuConfig:
    """Load and cache the project configuration.

    Args:
        reload: Whether to force a fresh load from disk.
        root: Optional project root or config file path.

    Returns:
        Resolved project configuration with absolute paths applied.

    Raises:
        FileNotFoundError: If a root path is specified but no config file is found.
    """
    global _global_config

    if _global_config is not None and not reload:
        return _global_config

    project_root = _resolve_project_root(root)
    load_project_env(root, override=True)
    config_path = _find_config_path(root)
    if config_path is None:
        if root is not None:
            raise FileNotFoundError("No jianmu.yaml/jianmu.yml/jianmu.toml found in the config root.")
        config = _apply_env_model_defaults(JianmuConfig())
        config = _apply_env_logging_overrides(config)
        config = _resolve_paths(config, project_root)
        _global_config = config
        return _global_config

    data = _load_config_file(config_path) or {}
    config = JianmuConfig(**data)
    config = _apply_env_model_defaults(config)
    config = _apply_env_logging_overrides(config)
    config = _resolve_paths(config, config_path.parent)
    _global_config = config
    return _global_config

load_project_env ¶

load_project_env(
    root: str | Path | None = None,
    *,
    override: bool = False,
) -> Path

Load .env from the target project root into os.environ.

参数：

名称	类型	描述	默认
`root`	`str \| Path \| None`	Root path used for resolution.	`None`
`override`	`bool`	The `override` value.	`False`

返回：

类型	描述
`Path`	The resulting `Path` value.

源代码位于： jianmu/config/loader.py

def load_project_env(root: str | Path | None = None, *, override: bool = False) -> Path:
    """Load ``.env`` from the target project root into ``os.environ``.

    Args:
        root: Root path used for resolution.
        override: The `override` value.

    Returns:
        The resulting `Path` value.
    """
    project_root = _resolve_project_root(root)
    env_path = project_root / ".env"
    cache_key = str(env_path.resolve())
    if cache_key in _env_loaded_roots and not override:
        return env_path
    try:
        from dotenv import load_dotenv
    except Exception:
        return env_path
    load_dotenv(env_path, override=override)
    _env_loaded_roots.add(cache_key)
    return env_path

jianmu.config¶

1. 模块职责¶

2. 适合查什么¶

3. 使用建议¶

4. 注意事项¶

5. 最小示例¶

6. 常见入口¶

7. API 参考¶

config ¶

Constraints dataclass ¶

from_dict classmethod ¶

coerce classmethod ¶

to_dict ¶

JianmuConfig ¶

PathsConfig ¶

ModelsConfig ¶

LLMDefaultsConfig ¶

LimitsConfig ¶

PromptConfig ¶

LoggingConfig ¶

TraceConfig ¶

ProviderLimitsConfig ¶

ExecutionDockerConfig ¶

ExecutionConfig ¶

RuntimeCheckpointConfig ¶

RuntimeConfig ¶

RagEmbeddingConfig ¶

RagConfig ¶

StateKeys ¶

EvaluationStateKeys ¶

AgentLLMConfig ¶

ModelResilienceConfig ¶

ReActConfig ¶

SkillNodeConfig ¶

ToolExecutorConfig ¶

PlanExecuteConfig ¶

bootstrap_project ¶

configure_logging_from_config ¶

get_config ¶

load_project_env ¶

`jianmu.config`¶

Constraints `dataclass` ¶

from_dict `classmethod` ¶

coerce `classmethod` ¶