Enum ClaudeError 

pub enum ClaudeError {
    ApiKeyNotFound,
    ApiRequestFailed(String),
    ApiHttpError {
        status: u16,
        body: String,
    },
    InvalidResponseFormat(String),
    AmendmentParsingFailed(String),
    PromptTooLarge {
        estimated_tokens: usize,
        max_tokens: usize,
        model: String,
    },
    RateLimitExceeded,
    NetworkError(String),
    SubprocessBinaryMissing(String),
    SubprocessSpawnFailed(String),
    SubprocessTimeout {
        secs: u64,
    },
    SubprocessOutputTooLarge {
        limit: usize,
    },
    SubprocessJsonParseFailed(String),
}

Claude API specific errors.

Variants

ApiKeyNotFound

API key not found in environment variables.

ApiRequestFailed(String)

Claude API request failed with error message.

Used where no HTTP status is available (subprocess failures, or an error the backend could not attribute to a status). Prefer ClaudeError::ApiHttpError whenever a status is known, so callers can tell a permanent failure from a retryable one.

ApiHttpError

AI API returned a non-success HTTP status.

Fields

status: u16

HTTP status code returned by the API.

body: String

Response body, used as the error detail.

InvalidResponseFormat(String)

Invalid response format from Claude API.

AmendmentParsingFailed(String)

Failed to parse amendments from Claude response.

PromptTooLarge

Prompt exceeds the model’s available input token budget.

Fields

estimated_tokens: usize

Estimated token count of the assembled prompt.

max_tokens: usize

Maximum available input tokens (context minus reserved output).

model: String

Model identifier.

RateLimitExceeded

Rate limit exceeded for Claude API.

NetworkError(String)

Network connectivity error.

SubprocessBinaryMissing(String)

Required subprocess binary is missing from PATH.

SubprocessSpawnFailed(String)

Failed to spawn a subprocess.

SubprocessTimeout

Subprocess exceeded the configured timeout.

Fields

secs: u64

Timeout that was exceeded, in seconds.

SubprocessOutputTooLarge

Subprocess produced more output than the configured cap.

Fields

limit: usize

Configured stdout cap in bytes.

SubprocessJsonParseFailed(String)

Subprocess stdout was not valid JSON.

Implementations

impl ClaudeError

pub fn is_transient(&self) -> bool

Returns true when retrying the request could plausibly succeed.

Only a non-retryable 4xx is treated as permanent: the request is malformed, unauthorised, or names something that does not exist (a misspelled model, say), so no amount of retrying or falling back will help. Everything else — 5xx, network failures, timeouts, and any error this cannot positively classify — is reported as transient, which preserves the historical fall-back-and-degrade behaviour for errors whose permanence is unproven.

pub fn is_structured_output_rejection(&self) -> bool

Returns true when the endpoint rejected the structured-output field output_config itself, rather than anything about the request’s content.

Anthropic’s Messages API takes output_config.format on models the registry flags via supports_structured_output, but a gateway named by ANTHROPIC_BEDROCK_BASE_URL may not pass the field through, and answers a strict-schema rejection — {"message": "output_config.format: Extra inputs are not permitted"} — rather than a model or content error (issue #1561). That is a property of the endpoint, which the per-model registry gate cannot know, so callers use this to drop to the YAML response path instead of failing the run.

The field-name match is the load-bearing half, keeping this narrow enough that no ordinary 400 (bad model, oversized prompt, malformed body) can trip it. 422 is accepted alongside 400 because pydantic-style gateways conventionally use it for exactly this unrecognised-field rejection.

Trait Implementations

impl Debug for ClaudeError

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Display for ClaudeError

fn fmt(&self, __formatter: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Error for ClaudeError

fn source(&self) -> Option<&(dyn Error + 'static)>

Returns the lower-level source of this error, if any.

fn description(&self) -> &str

👎Deprecated since 1.42.0:

use the Display impl or to_string()

fn cause(&self) -> Option<&dyn Error>

👎Deprecated since 1.33.0:

replaced by Error::source, which can support downcasting

fn provide<'a>(&'a self, request: &mut Request<'a>)

🔬This is a nightly-only experimental API. (error_generic_member_access)

Provides type-based access to context intended for error reports.