infiniflow/ragflow · error · SandboxProviderConfigError

SANDBOX_LOCAL_MAX_MEMORY_MB must be greater than 0.

Error message

SANDBOX_LOCAL_MAX_MEMORY_MB must be greater than 0.

What it means

Raised by LocalProvider._validate_limits() during initialize() when 'max_memory_mb' is zero or negative. The value is applied as an address-space rlimit (RLIMIT_AS) on the spawned child process via _limit_child_process, so a non-positive limit is rejected at configuration time with SandboxProviderConfigError. Default is 512 MB.

Source

Thrown at agent/sandbox/providers/local.py:262

                "min": 0,
                "max": 100,
            },
            "max_artifact_bytes": {
                "type": "integer",
                "required": False,
                "default": 10485760,
                "label": "Max Artifact Size (bytes)",
                "description": "Maximum size of a single artifact file. Unit: bytes.",
                "min": 1024,
                "max": 104857600,
            },
        }

    def _validate_limits(self) -> None:
        if self.timeout <= 0:
            raise SandboxProviderConfigError("SANDBOX_LOCAL_TIMEOUT must be greater than 0.")
        if self.max_memory_mb <= 0:
            raise SandboxProviderConfigError("SANDBOX_LOCAL_MAX_MEMORY_MB must be greater than 0.")
        if self.max_output_bytes <= 0:
            raise SandboxProviderConfigError("SANDBOX_LOCAL_MAX_OUTPUT_BYTES must be greater than 0.")
        if self.max_artifacts < 0:
            raise SandboxProviderConfigError("SANDBOX_LOCAL_MAX_ARTIFACTS must be greater than or equal to 0.")
        if self.max_artifact_bytes <= 0:
            raise SandboxProviderConfigError("SANDBOX_LOCAL_MAX_ARTIFACT_BYTES must be greater than 0.")

    def _prepare_script(self, instance_dir: Path, language: str, code: str, args_json: str) -> tuple[list[str], Path]:
        if language == "python":
            script_path = instance_dir / "main.py"
            script_path.write_text(build_python_wrapper(code, args_json), encoding="utf-8")
            return [self.python_bin, str(script_path)], script_path
        if language in {"javascript", "nodejs"}:
            script_path = instance_dir / "main.js"
            script_path.write_text(build_javascript_wrapper(code, args_json), encoding="utf-8")
            return [self.node_bin, str(script_path)], script_path
        raise RuntimeError(f"Unsupported language for local provider: {language}")

View on GitHub (pinned to 554fb1133a)

Solutions

1. Set 'max_memory_mb' to a positive MB value (e.g. 512, the default) in the initialize() config.
2. Omit the key to use the 512 MB default.
3. Verify the upstream env/config variable that feeds max_memory_mb and ensure it is expressed in megabytes, not bytes.
4. If the workload genuinely needs no cap, still set a large positive value (e.g. 65536, the schema max) — 0 is not accepted.

Example fix

// before
provider.initialize({"max_memory_mb": 0})

// after
provider.initialize({"max_memory_mb": 512})

def valid_memory(config: dict) -> bool:
    return int(config.get("max_memory_mb", 512)) > 0

try:
    provider.initialize(config)
except SandboxProviderConfigError as e:
    if "SANDBOX_LOCAL_MAX_MEMORY_MB" in str(e):
        config["max_memory_mb"] = 512
        provider.initialize(config)
    else:
        raise

When it happens

Trigger: Calling initialize({'max_memory_mb': 0}) or a negative value; int() coercion of a config string '0'. Also setting it very low (e.g. 1) passes validation but causes child MemoryError at runtime — though that is a separate failure.

Common situations: Trying to express 'unlimited memory' with 0; copying a container memory limit of 0 from an orchestration manifest; unit confusion (passing bytes like 536870912 instead of MB, which silently passes but is absurd).

Related errors

• SANDBOX_LOCAL_TIMEOUT must be greater than 0.
• SANDBOX_LOCAL_MAX_OUTPUT_BYTES must be greater than 0.
• SANDBOX_LOCAL_MAX_ARTIFACTS must be greater than or equal to
• SANDBOX_LOCAL_MAX_ARTIFACT_BYTES must be greater than 0.
• Invalid SSH provider configuration.

AI-assisted analysis of infiniflow/ragflow@554fb1133a (2026-08-15). Data as JSON: /api/errors/e67dc2fdcaa72dbc. Report an issue: GitHub.