error-handling

Error Handling Architecture Guide

Two-Tier Error Hierarchy

LiteLLM-RS uses a two-tier error architecture optimized for 66+ providers:

┌────────────────────────────────────────────────────────┐
│                    Gateway Layer                        │
│  LiteLLMError (core/types/errors/litellm.rs)          │
│  - 15 variants for gateway-level errors                │
│  - Routing, configuration, authentication errors       │
└────────────────────────────────────────────────────────┘
                          ↓
┌────────────────────────────────────────────────────────┐
│                   Provider Layer                        │
│  ProviderError (core/providers/unified_provider.rs)   │
│  - 24 variants for provider-specific errors            │
│  - Each variant includes provider: &'static str        │
│  - Rich factory methods for error creation             │
└────────────────────────────────────────────────────────┘

ProviderError (24 Variants)

// src/core/providers/unified_provider.rs

#[derive(Debug, Clone, thiserror::Error)]
pub enum ProviderError {
    // Authentication & Authorization
    #[error("[{provider}] Authentication failed: {message}")]
    Authentication {
        provider: &'static str,
        message: String,
    },

    // Rate Limiting & Quotas
    #[error("[{provider}] Rate limit exceeded: {message}")]
    RateLimit {
        provider: &'static str,
        message: String,
        retry_after: Option<u64>,
        rpm_limit: Option<u32>,
        tpm_limit: Option<u32>,
        current_usage: Option<u32>,
    },

    #[error("[{provider}] Quota exceeded: {message}")]
    QuotaExceeded {
        provider: &'static str,
        message: String,
    },

    // Model & Request Errors
    #[error("[{provider}] Model not found: {model}")]
    ModelNotFound {
        provider: &'static str,
        model: String,
    },

    #[error("[{provider}] Invalid request: {message}")]
    InvalidRequest {
        provider: &'static str,
        message: String,
    },

    // Network & Availability
    #[error("[{provider}] Network error: {message}")]
    Network {
        provider: &'static str,
        message: String,
    },

    #[error("[{provider}] Request timeout: {message}")]
    Timeout {
        provider: &'static str,
        message: String,
    },

    #[error("[{provider}] Provider unavailable: {message}")]
    ProviderUnavailable {
        provider: &'static str,
        message: String,
    },

    // Feature Support
    #[error("[{provider}] Feature not supported: {feature}")]
    NotSupported {
        provider: &'static str,
        feature: String,
    },

    #[error("[{provider}] Not implemented: {feature}")]
    NotImplemented {
        provider: &'static str,
        feature: String,
    },

    #[error("[{provider}] Feature disabled: {feature}")]
    FeatureDisabled {
        provider: &'static str,
        feature: String,
    },

    // Content & Token Limits
    #[error("[{provider}] Context length exceeded: max={max}, actual={actual}")]
    ContextLengthExceeded {
        provider: &'static str,
        max: u32,
        actual: u32,
    },

    #[error("[{provider}] Token limit exceeded: {message}")]
    TokenLimitExceeded {
        provider: &'static str,
        message: String,
    },

    #[error("[{provider}] Content filtered: {reason}")]
    ContentFiltered {
        provider: &'static str,
        reason: String,
        policy_violations: Option<Vec<String>>,
        potentially_retryable: Option<bool>,
    },

    // Configuration & Serialization
    #[error("[{provider}] Configuration error: {message}")]
    Configuration {
        provider: &'static str,
        message: String,
    },

    #[error("[{provider}] Serialization error: {message}")]
    Serialization {
        provider: &'static str,
        message: String,
    },

    // Advanced Errors
    #[error("[{provider}] API error (status={status}): {message}")]
    ApiError {
        provider: &'static str,
        status: u16,
        message: String,
    },

    #[error("[{provider}] Deployment error ({deployment}): {message}")]
    DeploymentError {
        provider: &'static str,
        deployment: String,
        message: String,
    },

    #[error("[{provider}] Response parsing error: {message}")]
    ResponseParsing {
        provider: &'static str,
        message: String,
    },

    #[error("[{provider}] Routing error: {message}")]
    RoutingError {
        provider: &'static str,
        attempted_providers: Vec<String>,
        message: String,
    },

    #[error("[{provider}] Transformation error ({from_format} -> {to_format}): {message}")]
    TransformationError {
        provider: &'static str,
        from_format: String,
        to_format: String,
        message: String,
    },

    #[error("[{provider}] Streaming error ({stream_type}): {message}")]
    Streaming {
        provider: &'static str,
        stream_type: String,
        position: Option<u64>,
        last_chunk: Option<String>,
        message: String,
    },

    #[error("[{provider}] Operation cancelled ({operation_type}): {cancellation_reason}")]
    Cancelled {
        provider: &'static str,
        operation_type: String,
        cancellation_reason: String,
    },

    #[error("[{provider}] Error: {message}")]
    Other {
        provider: &'static str,
        message: String,
    },
}

Factory Methods

Basic Factory Methods

impl ProviderError {
    pub fn authentication(provider: &'static str, message: impl Into<String>) -> Self {
        Self::Authentication {
            provider,
            message: message.into(),
        }
    }

    pub fn rate_limit(provider: &'static str, retry_after: Option<u64>) -> Self {
        Self::RateLimit {
            provider,
            message: "Rate limit exceeded".to_string(),
            retry_after,
            rpm_limit: None,
            tpm_limit: None,
            current_usage: None,
        }
    }

    pub fn model_not_found(provider: &'static str, model: impl Into<String>) -> Self {
        Self::ModelNotFound {
            provider,
            model: model.into(),
        }
    }

    pub fn invalid_request(provider: &'static str, message: impl Into<String>) -> Self {
        Self::InvalidRequest {
            provider,
            message: message.into(),
        }
    }

    pub fn network(provider: &'static str, message: impl Into<String>) -> Self {
        Self::Network {
            provider,
            message: message.into(),
        }
    }

    pub fn timeout(provider: &'static str, message: impl Into<String>) -> Self {
        Self::Timeout {
            provider,
            message: message.into(),
        }
    }

    pub fn provider_unavailable(provider: &'static str, message: impl Into<String>) -> Self {
        Self::ProviderUnavailable {
            provider,
            message: message.into(),
        }
    }

    pub fn not_supported(provider: &'static str, feature: impl Into<String>) -> Self {
        Self::NotSupported {
            provider,
            feature: feature.into(),
        }
    }

    pub fn configuration(provider: &'static str, message: impl Into<String>) -> Self {
        Self::Configuration {
            provider,
            message: message.into(),
        }
    }

    pub fn serialization(provider: &'static str, message: impl Into<String>) -> Self {
        Self::Serialization {
            provider,
            message: message.into(),
        }
    }

    pub fn api_error(provider: &'static str, status: u16, message: impl Into<String>) -> Self {
        Self::ApiError {
            provider,
            status,
            message: message.into(),
        }
    }

    pub fn response_parsing(provider: &'static str, message: impl Into<String>) -> Self {
        Self::ResponseParsing {
            provider,
            message: message.into(),
        }
    }
}

Enhanced Factory Methods

impl ProviderError {
    pub fn rate_limit_with_limits(
        provider: &'static str,
        retry_after: Option<u64>,
        rpm_limit: Option<u32>,
        tpm_limit: Option<u32>,
        current_usage: Option<u32>,
    ) -> Self {
        Self::RateLimit {
            provider,
            message: format!(
                "Rate limit exceeded. RPM: {:?}, TPM: {:?}, Current: {:?}",
                rpm_limit, tpm_limit, current_usage
            ),
            retry_after,
            rpm_limit,
            tpm_limit,
            current_usage,
        }
    }

    pub fn context_length_exceeded(
        provider: &'static str,
        max: u32,
        actual: u32,
    ) -> Self {
        Self::ContextLengthExceeded {
            provider,
            max,
            actual,
        }
    }

    pub fn content_filtered(
        provider: &'static str,
        reason: impl Into<String>,
        policy_violations: Option<Vec<String>>,
        potentially_retryable: Option<bool>,
    ) -> Self {
        Self::ContentFiltered {
            provider,
            reason: reason.into(),
            policy_violations,
            potentially_retryable,
        }
    }

    pub fn streaming_error(
        provider: &'static str,
        stream_type: impl Into<String>,
        position: Option<u64>,
        last_chunk: Option<String>,
        message: impl Into<String>,
    ) -> Self {
        Self::Streaming {
            provider,
            stream_type: stream_type.into(),
            position,
            last_chunk,
            message: message.into(),
        }
    }

    pub fn routing_error(
        provider: &'static str,
        attempted_providers: Vec<String>,
        message: impl Into<String>,
    ) -> Self {
        Self::RoutingError {
            provider,
            attempted_providers,
            message: message.into(),
        }
    }
}

HTTP Status Mapping

Standard Mapping Pattern

impl MyProvider {
    fn map_http_error(&self, status: u16, body: &str) -> ProviderError {
        match status {
            // Authentication errors
            401 => ProviderError::authentication(PROVIDER_NAME, "Invalid API key"),
            403 => ProviderError::authentication(PROVIDER_NAME, "Access forbidden"),

            // Not found errors
            404 => ProviderError::model_not_found(PROVIDER_NAME, body),

            // Client errors
            400 => ProviderError::invalid_request(PROVIDER_NAME, body),
            422 => ProviderError::invalid_request(PROVIDER_NAME, "Unprocessable entity"),

            // Rate limiting
            429 => ProviderError::rate_limit(PROVIDER_NAME, self.parse_retry_after(body)),

            // Server errors
            500 => ProviderError::provider_unavailable(PROVIDER_NAME, "Internal server error"),
            502 => ProviderError::provider_unavailable(PROVIDER_NAME, "Bad gateway"),
            503 => ProviderError::provider_unavailable(PROVIDER_NAME, "Service unavailable"),
            504 => ProviderError::timeout(PROVIDER_NAME, "Gateway timeout"),

            // Default
            _ => ProviderError::api_error(PROVIDER_NAME, status, body),
        }
    }

    fn parse_retry_after(&self, body: &str) -> Option<u64> {
        // Parse retry-after from response headers or body
        serde_json::from_str::<serde_json::Value>(body)
            .ok()
            .and_then(|v| v.get("retry_after"))
            .and_then(|v| v.as_u64())
    }
}

ErrorMapper Trait

// src/core/traits/error_mapper/trait_def.rs

pub trait ErrorMapper: Send + Sync {
    type Error;

    fn map_http_error(&self, status: u16, body: &str) -> Self::Error;
    fn map_network_error(&self, error: reqwest::Error) -> Self::Error;
    fn map_parse_error(&self, error: serde_json::Error) -> Self::Error;
}

// Generic implementation for ProviderError
pub struct GenericErrorMapper;

impl ErrorMapper for GenericErrorMapper {
    type Error = ProviderError;

    fn map_http_error(&self, status: u16, body: &str) -> Self::Error {
        match status {
            401 | 403 => ProviderError::authentication("generic", body),
            404 => ProviderError::model_not_found("generic", body),
            429 => ProviderError::rate_limit("generic", None),
            400 | 422 => ProviderError::invalid_request("generic", body),
            500..=599 => ProviderError::provider_unavailable("generic", body),
            _ => ProviderError::api_error("generic", status, body),
        }
    }

    fn map_network_error(&self, error: reqwest::Error) -> Self::Error {
        if error.is_timeout() {
            ProviderError::timeout("generic", error.to_string())
        } else if error.is_connect() {
            ProviderError::network("generic", error.to_string())
        } else {
            ProviderError::network("generic", error.to_string())
        }
    }

    fn map_parse_error(&self, error: serde_json::Error) -> Self::Error {
        ProviderError::response_parsing("generic", error.to_string())
    }
}

Retry Logic

Retryable Error Detection

impl ProviderError {
    pub fn is_retryable(&self) -> bool {
        matches!(
            self,
            Self::RateLimit { .. }
                | Self::Timeout { .. }
                | Self::Network { .. }
                | Self::ProviderUnavailable { .. }
        )
    }

    pub fn retry_after(&self) -> Option<Duration> {
        match self {
            Self::RateLimit { retry_after, .. } => {
                retry_after.map(Duration::from_secs)
            }
            Self::Timeout { .. } => Some(Duration::from_secs(1)),
            Self::Network { .. } => Some(Duration::from_millis(500)),
            Self::ProviderUnavailable { .. } => Some(Duration::from_secs(5)),
            _ => None,
        }
    }

    pub fn should_fallback(&self) -> bool {
        matches!(
            self,
            Self::ProviderUnavailable { .. }
                | Self::RateLimit { .. }
                | Self::QuotaExceeded { .. }
                | Self::ModelNotFound { .. }
        )
    }
}

Retry Implementation

pub async fn execute_with_retry<F, T, E>(
    operation: F,
    max_retries: u32,
    base_delay: Duration,
) -> Result<T, E>
where
    F: Fn() -> Pin<Box<dyn Future<Output = Result<T, E>> + Send>>,
    E: std::fmt::Debug,
{
    let mut attempts = 0;
    let mut last_error;

    loop {
        match operation().await {
            Ok(result) => return Ok(result),
            Err(e) => {
                attempts += 1;
                last_error = e;

                if attempts >= max_retries {
                    break;
                }

                // Exponential backoff
                let delay = base_delay * 2u32.pow(attempts - 1);
                tokio::time::sleep(delay).await;
            }
        }
    }

    Err(last_error)
}

LiteLLMError (Gateway Level)

// src/core/types/errors/litellm.rs

#[derive(Debug, thiserror::Error)]
pub enum LiteLLMError {
    #[error("Provider error: {0}")]
    Provider(#[from] ProviderError),

    #[error("Routing error: {0}")]
    Routing(String),

    #[error("Configuration error: {0}")]
    Configuration(String),

    #[error("Authentication error: {0}")]
    Authentication(String),

    #[error("Authorization error: {0}")]
    Authorization(String),

    #[error("Validation error: {0}")]
    Validation(String),

    #[error("Timeout error: {0}")]
    Timeout(String),

    #[error("Rate limit error: {0}")]
    RateLimit(String),

    #[error("Internal error: {0}")]
    Internal(String),

    #[error("Not found: {0}")]
    NotFound(String),

    #[error("Conflict: {0}")]
    Conflict(String),

    #[error("Service unavailable: {0}")]
    ServiceUnavailable(String),

    #[error("Bad gateway: {0}")]
    BadGateway(String),

    #[error("Cancelled: {0}")]
    Cancelled(String),

    #[error("Unknown error: {0}")]
    Unknown(String),
}

Error Context Preservation

Adding Context to Errors

// Using the anyhow pattern for context
use anyhow::Context;

async fn process_request(request: ChatRequest) -> Result<ChatResponse, LiteLLMError> {
    let provider = select_provider(&request.model)
        .context("Failed to select provider")?;

    let response = provider
        .chat_completion(request, context)
        .await
        .context("Chat completion failed")?;

    Ok(response)
}

Error Chain Display

impl ProviderError {
    pub fn display_chain(&self) -> String {
        let mut chain = vec![self.to_string()];

        // Add any source errors
        if let Some(source) = std::error::Error::source(self) {
            chain.push(format!("Caused by: {}", source));
        }

        chain.join("\n")
    }
}

Best Practices

1. Always Use Factory Methods

// Good
ProviderError::authentication(PROVIDER_NAME, "Invalid API key")

// Bad - verbose and error-prone
ProviderError::Authentication {
    provider: PROVIDER_NAME,
    message: "Invalid API key".to_string(),
}

2. Include Provider Name

// Good - error clearly identifies source
ProviderError::network("openai", "Connection refused")

// Bad - unclear which provider failed
ProviderError::network("unknown", "Connection refused")

3. Preserve Error Context

// Good - preserves original error
self.pool_manager.execute_request(&url, method, headers, body)
    .await
    .map_err(|e| ProviderError::network(PROVIDER_NAME, e.to_string()))?

// Bad - loses original error
self.pool_manager.execute_request(&url, method, headers, body)
    .await
    .map_err(|_| ProviderError::network(PROVIDER_NAME, "Request failed"))?

4. Use Specific Error Types

// Good - specific error type
if response.status() == 429 {
    return Err(ProviderError::rate_limit(PROVIDER_NAME, retry_after));
}

// Bad - generic error loses information
if !response.status().is_success() {
    return Err(ProviderError::api_error(PROVIDER_NAME, status, "Failed"));
}

5. Handle All Error Variants in Match

// Good - exhaustive handling
match error {
    ProviderError::RateLimit { retry_after, .. } => {
        if let Some(delay) = retry_after {
            tokio::time::sleep(Duration::from_secs(delay)).await;
        }
        // Retry...
    }
    ProviderError::Authentication { .. } => {
        // Don't retry, return immediately
        return Err(error);
    }
    e if e.is_retryable() => {
        // Retry with backoff
    }
    _ => return Err(error),
}

HTTP to ProviderError Mapping Reference

HTTP Status	ProviderError Variant	Retryable
400	`InvalidRequest`	No
401	`Authentication`	No
403	`Authentication`	No
404	`ModelNotFound`	No
408	`Timeout`	Yes
422	`InvalidRequest`	No
429	`RateLimit`	Yes
500	`ProviderUnavailable`	Yes
502	`ProviderUnavailable`	Yes
503	`ProviderUnavailable`	Yes
504	`Timeout`	Yes

About this skill

Two-Tier Error Hierarchy

ProviderError (24 Variants)

Factory Methods

Basic Factory Methods

Enhanced Factory Methods

HTTP Status Mapping

Standard Mapping Pattern

ErrorMapper Trait

Retry Logic

Retryable Error Detection

Retry Implementation

LiteLLMError (Gateway Level)

Error Context Preservation

Adding Context to Errors

Error Chain Display

Best Practices

1. Always Use Factory Methods

2. Include Provider Name

3. Preserve Error Context

4. Use Specific Error Types

5. Handle All Error Variants in Match

HTTP to ProviderError Mapping Reference