diff --git a/litellm-rust/Cargo.lock b/litellm-rust/Cargo.lock index 500a12035cb..f00afe04465 100644 --- a/litellm-rust/Cargo.lock +++ b/litellm-rust/Cargo.lock @@ -1838,7 +1838,41 @@ source = "registry+https://github.com/rust-lang/crates.io-index" checksum = "68ab91017fe16c622486840e4c83c9a37afeff978bd239b5293d61ece587de66" [[package]] -name = "litellm-core" +name = "litellm-ai-gateway" +version = "0.1.0" +dependencies = [ + "axum", + "base64 0.22.1", + "futures-channel", + "futures-util", + "litellm-config", + "litellm-core", + "reqwest 0.12.28", + "rustls 0.23.42", + "rustls-native-certs", + "serde", + "serde_json", + "sha2 0.10.9", + "subtle", + "tokio", + "tokio-tungstenite", + "tower", + "tracing", +] + +[[package]] +name = "litellm-auth" +version = "0.1.0" +dependencies = [ + "serde", + "subtle", + "thiserror 2.0.19", + "tokio", + "veil", +] + +[[package]] +name = "litellm-auth-aws" version = "0.1.0" dependencies = [ "aws-config", @@ -1847,13 +1881,64 @@ dependencies = [ "aws-sigv4", "aws-smithy-runtime-api", "aws-types", + "litellm-auth", + "moka", + "reqwest 0.12.28", + "serde_json", + "sha2 0.10.9", + "tokio", +] + +[[package]] +name = "litellm-auth-azure" +version = "0.1.0" +dependencies = [ "azure_core", "azure_identity", + "litellm-auth", + "moka", + "serde_json", + "sha2 0.10.9", + "strum", + "tokio", + "url", +] + +[[package]] +name = "litellm-auth-gcp" +version = "0.1.0" +dependencies = [ + "gcp_auth", + "litellm-auth", + "moka", + "serde_json", + "sha2 0.10.9", + "tokio", + "tracing", +] + +[[package]] +name = "litellm-config" +version = "0.1.0" +dependencies = [ + "litellm-core", + "pyo3", + "serde_json", + "thiserror 2.0.19", +] + +[[package]] +name = "litellm-core" +version = "0.1.0" +dependencies = [ "base64 0.22.1", "bytes", "data-url", "futures-util", - "gcp_auth", + "litellm-auth", + "litellm-auth-aws", + "litellm-auth-azure", + "litellm-auth-gcp", "mime_guess", "moka", "rand 0.8.7", @@ -1865,7 +1950,6 @@ dependencies = [ "serde_json", "serde_path_to_error", "sha2 0.10.9", - "strum", "subtle", "thiserror 2.0.19", "tokio", diff --git a/litellm-rust/Cargo.toml b/litellm-rust/Cargo.toml index 2a0a7687e0e..36ffc519552 100644 --- a/litellm-rust/Cargo.toml +++ b/litellm-rust/Cargo.toml @@ -1,5 +1,16 @@ [workspace] -members = ["crates/*"] +members = [ + "crates/auth", + "crates/auth-aws", + "crates/auth-azure", + "crates/auth-gcp", + "crates/core", + "crates/token-counter", + "crates/config", + "crates/ai-gateway", + "crates/python-interop", + "crates/python-bridge", +] resolver = "2" [workspace.package] @@ -11,6 +22,10 @@ repository = "https://github.com/BerriAI/litellm" [workspace.dependencies] bytes = "1" litellm-core = { path = "crates/core" } +litellm-auth = { path = "crates/auth" } +litellm-auth-aws = { path = "crates/auth-aws" } +litellm-auth-azure = { path = "crates/auth-azure" } +litellm-auth-gcp = { path = "crates/auth-gcp" } litellm-token-counter = { path = "crates/token-counter" } litellm-python-interop = { path = "crates/python-interop" } pyo3 = "0.29.2" diff --git a/litellm-rust/crates/auth-aws/Cargo.toml b/litellm-rust/crates/auth-aws/Cargo.toml new file mode 100644 index 00000000000..a4fdbc4f754 --- /dev/null +++ b/litellm-rust/crates/auth-aws/Cargo.toml @@ -0,0 +1,24 @@ +[package] +name = "litellm-auth-aws" +version = "0.1.0" +edition.workspace = true +license.workspace = true +repository.workspace = true + +[dependencies] +litellm-auth.workspace = true + +moka = { workspace = true, features = ["sync"] } +serde_json.workspace = true +sha2.workspace = true + +aws-config = { version = "1.9.0", default-features = false, features = ["rustls", "rt-tokio"] } +aws-credential-types = { version = "1.3.0", features = ["hardcoded-credentials"] } +aws-sdk-sts = { version = "1.108.0", default-features = false, features = ["rustls", "rt-tokio"] } +aws-sigv4 = "1.5.1" +aws-types = "1.4.0" +aws-smithy-runtime-api = "1.13.0" + +[dev-dependencies] +reqwest.workspace = true +tokio.workspace = true diff --git a/litellm-rust/crates/auth-aws/src/aws.rs b/litellm-rust/crates/auth-aws/src/aws.rs new file mode 100644 index 00000000000..32035c572df --- /dev/null +++ b/litellm-rust/crates/auth-aws/src/aws.rs @@ -0,0 +1,950 @@ +use std::collections::BTreeMap; +use std::sync::OnceLock; +use std::time::Duration; +use std::time::{SystemTime, UNIX_EPOCH}; + +use moka::sync::Cache; +use serde_json::{Map, Value}; +use sha2::{Digest, Sha256}; + +use aws_credential_types::Credentials; +use aws_credential_types::provider::ProvideCredentials; +use aws_sigv4::http_request::{ + SignableBody, SignableRequest, SigningParams, SigningSettings, sign, +}; +use aws_sigv4::sign::v4; +use aws_smithy_runtime_api::client::identity::Identity; + +use litellm_auth::Error; + +use super::constants::{ + AWS_ACCESS_KEY_ID, AWS_EXTERNAL_ID, AWS_PROFILE_NAME, AWS_REGION, AWS_REGION_NAME, + AWS_ROLE_ARN, AWS_ROLE_NAME, AWS_SECRET_ACCESS_KEY, AWS_SESSION_NAME, AWS_SESSION_TOKEN, + AWS_SIGNED_HEADER_NAMES, AWS_STS_ENDPOINT, AWS_WEB_IDENTITY_TOKEN, AWS_WEB_IDENTITY_TOKEN_FILE, + BEDROCK_SERVICE, DEFAULT_BEDROCK_REGION, DEFAULT_SESSION_NAME_PREFIX, + SIGV4_COMPUTED_HEADER_NAMES, +}; + +const STATIC_CREDENTIALS_TTL: Duration = Duration::from_secs(3600 - 60); +const AMBIENT_CREDENTIALS_TTL: Duration = Duration::from_secs(600); + +static STATIC_CREDENTIALS_CACHE: OnceLock> = OnceLock::new(); +static AMBIENT_CREDENTIALS_CACHE: OnceLock> = OnceLock::new(); + +fn credential_cache_ttl(flow: &AwsAuthFlow) -> Option { + match flow { + AwsAuthFlow::StaticKeys { .. } => Some(STATIC_CREDENTIALS_TTL), + AwsAuthFlow::DefaultChain => Some(AMBIENT_CREDENTIALS_TTL), + AwsAuthFlow::WebIdentity { .. } + | AwsAuthFlow::AssumeRole { .. } + | AwsAuthFlow::Profile { .. } + | AwsAuthFlow::SessionToken { .. } => None, + } +} + +#[derive(Clone, Debug, Default, PartialEq, Eq)] +pub struct AwsAuthConfig { + pub access_key_id: Option, + pub secret_access_key: Option, + pub session_token: Option, + pub region_name: Option, + pub session_name: Option, + pub profile_name: Option, + pub role_name: Option, + pub web_identity_token: Option, + pub sts_endpoint: Option, + pub external_id: Option, +} + +impl AwsAuthConfig { + fn with_environment(self, env_lookup: &(dyn Fn(&str) -> Option + Sync)) -> Self { + Self { + access_key_id: self.access_key_id.or_else(|| env_lookup(AWS_ACCESS_KEY_ID)), + secret_access_key: self + .secret_access_key + .or_else(|| env_lookup(AWS_SECRET_ACCESS_KEY)), + session_token: self.session_token.or_else(|| env_lookup(AWS_SESSION_TOKEN)), + region_name: self.region_name.or_else(|| env_lookup(AWS_REGION_NAME)), + session_name: self.session_name.or_else(|| env_lookup(AWS_SESSION_NAME)), + profile_name: self.profile_name.or_else(|| env_lookup(AWS_PROFILE_NAME)), + role_name: self.role_name.or_else(|| env_lookup(AWS_ROLE_NAME)), + web_identity_token: self + .web_identity_token + .or_else(|| env_lookup(AWS_WEB_IDENTITY_TOKEN)), + sts_endpoint: self.sts_endpoint.or_else(|| env_lookup(AWS_STS_ENDPOINT)), + external_id: self.external_id.or_else(|| env_lookup(AWS_EXTERNAL_ID)), + } + } +} + +#[derive(Clone, Debug, PartialEq, Eq)] +pub enum AwsAuthFlow { + WebIdentity { + token: String, + role: String, + session_name: String, + }, + AssumeRole { + role: String, + session_name: Option, + }, + Profile { + name: String, + }, + SessionToken { + access_key_id: String, + secret_access_key: String, + session_token: String, + }, + StaticKeys { + access_key_id: String, + secret_access_key: String, + region_name: String, + }, + DefaultChain, +} + +fn cache_key(config: &AwsAuthConfig, flow: &AwsAuthFlow) -> String { + let mut hasher = Sha256::new(); + hasher.update(format!("{config:?}:{flow:?}")); + format!("{:x}", hasher.finalize()) +} + +fn static_credentials_cache() -> &'static Cache { + STATIC_CREDENTIALS_CACHE.get_or_init(|| { + Cache::builder() + .max_capacity(200) + .time_to_live(STATIC_CREDENTIALS_TTL) + .build() + }) +} + +fn ambient_credentials_cache() -> &'static Cache { + AMBIENT_CREDENTIALS_CACHE.get_or_init(|| { + Cache::builder() + .max_capacity(200) + .time_to_live(AMBIENT_CREDENTIALS_TTL) + .build() + }) +} + +fn get_cached_credentials(key: &str) -> Option { + static_credentials_cache() + .get(key) + .or_else(|| ambient_credentials_cache().get(key)) +} + +fn set_cached_credentials(key: String, credentials: Credentials, ttl: Duration) { + if ttl == STATIC_CREDENTIALS_TTL { + static_credentials_cache().insert(key, credentials); + } else { + ambient_credentials_cache().insert(key, credentials); + } +} + +fn role_identity(arn: &str) -> Option<(&str, &str, &str)> { + let mut parts = arn.splitn(6, ':'); + let ("arn", partition, _, _, account, resource) = ( + parts.next()?, + parts.next()?, + parts.next()?, + parts.next()?, + parts.next()?, + parts.next()?, + ) else { + return None; + }; + let role = if let Some(role) = resource.strip_prefix("role/") { + role.rsplit('/').next()? + } else { + resource.strip_prefix("assumed-role/")?.split('/').next()? + }; + Some((partition, account, role)) +} + +fn same_role_arns(target: &str, caller: &str) -> bool { + role_identity(target) == role_identity(caller) +} + +pub fn classify_auth( + config: AwsAuthConfig, + env_lookup: &(dyn Fn(&str) -> Option + Sync), +) -> AwsAuthFlow { + let config = config.with_environment(env_lookup); + if let (Some(token), Some(role), Some(session_name)) = ( + config.web_identity_token.clone(), + config.role_name.clone(), + config.session_name.clone(), + ) { + return AwsAuthFlow::WebIdentity { + token, + role, + session_name, + }; + } + if let Some(role) = config.role_name.clone() { + return AwsAuthFlow::AssumeRole { + role, + session_name: config.session_name.clone(), + }; + } + if let Some(name) = config.profile_name { + return AwsAuthFlow::Profile { name }; + } + if let (Some(access_key_id), Some(secret_access_key), Some(session_token)) = ( + config.access_key_id.clone(), + config.secret_access_key.clone(), + config.session_token, + ) { + return AwsAuthFlow::SessionToken { + access_key_id, + secret_access_key, + session_token, + }; + } + if let (Some(access_key_id), Some(secret_access_key), Some(region_name)) = ( + config.access_key_id, + config.secret_access_key, + config.region_name, + ) { + return AwsAuthFlow::StaticKeys { + access_key_id, + secret_access_key, + region_name, + }; + } + AwsAuthFlow::DefaultChain +} + +pub async fn resolve_credentials( + config: AwsAuthConfig, + env_lookup: &(dyn Fn(&str) -> Option + Sync), +) -> Result { + let resolved = config.clone().with_environment(env_lookup); + let flow = classify_auth(config, env_lookup); + match flow { + AwsAuthFlow::SessionToken { + access_key_id, + secret_access_key, + session_token, + } => Ok(Credentials::new( + access_key_id, + secret_access_key, + Some(session_token), + None, + "litellm-static-session", + )), + AwsAuthFlow::StaticKeys { + access_key_id, + secret_access_key, + region_name, + } => { + let flow = AwsAuthFlow::StaticKeys { + access_key_id: access_key_id.clone(), + secret_access_key: secret_access_key.clone(), + region_name, + }; + let key = cache_key(&resolved, &flow); + if let Some(credentials) = get_cached_credentials(&key) { + return Ok(credentials); + } + let credentials = Credentials::new( + access_key_id, + secret_access_key, + None, + None, + "litellm-static", + ); + set_cached_credentials( + key, + credentials.clone(), + credential_cache_ttl(&flow).unwrap_or(STATIC_CREDENTIALS_TTL), + ); + Ok(credentials) + } + AwsAuthFlow::Profile { name } => { + let provider = aws_config::profile::ProfileFileCredentialsProvider::builder() + .profile_name(name) + .build(); + provider + .provide_credentials() + .await + .map_err(|error| Error::AwsProfile(error.to_string())) + } + AwsAuthFlow::AssumeRole { role, session_name } => { + if is_already_running_as_role(&role, &resolved).await? { + let ambient_flow = AwsAuthFlow::DefaultChain; + let key = cache_key(&resolved, &ambient_flow); + if let Some(credentials) = get_cached_credentials(&key) { + return Ok(credentials); + } + let provider = + aws_config::default_provider::credentials::DefaultCredentialsChain::builder() + .build() + .await; + let credentials = provider + .provide_credentials() + .await + .map_err(|error| Error::AwsDefaultChain(error.to_string()))?; + set_cached_credentials( + key, + credentials.clone(), + credential_cache_ttl(&ambient_flow).unwrap_or(AMBIENT_CREDENTIALS_TTL), + ); + return Ok(credentials); + } + let mut loader = aws_config::defaults(aws_config::BehaviorVersion::latest()); + if let Some(region) = resolved.region_name.clone() { + loader = loader.region(aws_types::region::Region::new(region)); + } + if let Some(endpoint) = resolved.sts_endpoint.clone() { + loader = loader.endpoint_url(endpoint); + } + if let (Some(access_key_id), Some(secret_access_key)) = + (resolved.access_key_id, resolved.secret_access_key) + { + loader = loader.credentials_provider(Credentials::new( + access_key_id, + secret_access_key, + resolved.session_token, + None, + "litellm-role-source", + )); + } + let sdk_config = loader.load().await; + let builder = aws_config::sts::AssumeRoleProvider::builder(role); + let builder = match session_name { + Some(name) => builder.session_name(name), + None => builder.session_name(default_session_name()), + }; + let builder = match resolved.external_id { + Some(id) => builder.external_id(id), + None => builder, + }; + let provider = builder.configure(&sdk_config).build().await; + provider + .provide_credentials() + .await + .map_err(|error| Error::AwsAssumeRole(error.to_string())) + } + AwsAuthFlow::WebIdentity { + token, + role, + session_name, + } => { + let mut loader = aws_config::defaults(aws_config::BehaviorVersion::latest()); + if let Some(region) = resolved.region_name { + loader = loader.region(aws_types::region::Region::new(region)); + } + if let Some(endpoint) = resolved.sts_endpoint { + loader = loader.endpoint_url(endpoint); + } + let sdk_config = loader.load().await; + let client = aws_sdk_sts::Client::new(&sdk_config); + let response = client + .assume_role_with_web_identity() + .role_arn(role) + .role_session_name(session_name) + .web_identity_token(token) + .send() + .await + .map_err(|error| Error::AwsWebIdentity(error.to_string()))?; + let credentials = response + .credentials() + .ok_or(Error::AwsMissingWebIdentityCredentials)?; + let expiration = SystemTime::try_from(*credentials.expiration()) + .map_err(|error| Error::AwsWebIdentityExpiration(error.to_string()))?; + Ok(Credentials::new( + credentials.access_key_id(), + credentials.secret_access_key(), + Some(credentials.session_token().to_string()), + Some(expiration), + "litellm-web-identity", + )) + } + AwsAuthFlow::DefaultChain => { + let key = cache_key(&resolved, &AwsAuthFlow::DefaultChain); + if let Some(credentials) = get_cached_credentials(&key) { + return Ok(credentials); + } + let provider = + aws_config::default_provider::credentials::DefaultCredentialsChain::builder() + .build() + .await; + let credentials = provider + .provide_credentials() + .await + .map_err(|error| Error::AwsDefaultChain(error.to_string()))?; + set_cached_credentials( + key, + credentials.clone(), + credential_cache_ttl(&AwsAuthFlow::DefaultChain).unwrap_or(AMBIENT_CREDENTIALS_TTL), + ); + Ok(credentials) + } + } +} + +async fn is_already_running_as_role(role: &str, config: &AwsAuthConfig) -> Result { + if role_identity(role).is_none() { + return Ok(false); + } + if let (Ok(current_role), Ok(token_file)) = ( + std::env::var(AWS_ROLE_ARN), + std::env::var(AWS_WEB_IDENTITY_TOKEN_FILE), + ) && !token_file.is_empty() + { + return Ok(same_role_arns(role, ¤t_role)); + } + + let mut loader = aws_config::defaults(aws_config::BehaviorVersion::latest()); + if let Some(region) = config.region_name.clone() { + loader = loader.region(aws_types::region::Region::new(region)); + } + if let Some(endpoint) = config.sts_endpoint.clone() { + loader = loader.endpoint_url(endpoint); + } + let sdk_config = loader.load().await; + let response = match aws_sdk_sts::Client::new(&sdk_config) + .get_caller_identity() + .send() + .await + { + Ok(response) => response, + Err(_) => return Ok(false), + }; + Ok(response + .arn() + .is_some_and(|caller| same_role_arns(role, caller))) +} + +fn default_session_name() -> String { + let seconds = SystemTime::now() + .duration_since(UNIX_EPOCH) + .map_or(0, |duration| duration.as_secs()); + format!("{DEFAULT_SESSION_NAME_PREFIX}-{seconds}") +} + +/// The subset of `headers` SigV4 should cover. +/// +/// Python signs only these and reattaches the rest afterwards, so a forwarded +/// client header cannot change the canonical request and invalidate the +/// signature. Signing everything instead makes the request 403 on a header the +/// caller supplied, on a deployment that works on the Python path. +pub fn aws_signature_headers(headers: &BTreeMap) -> BTreeMap { + headers + .iter() + .filter(|(name, _)| { + let name = name.to_ascii_lowercase(); + AWS_SIGNED_HEADER_NAMES.contains(&name.as_str()) + || name.starts_with("x-amz-") + || name.starts_with("x-amzn-") + }) + .map(|(name, value)| (name.clone(), value.clone())) + .collect() +} + +/// Whether the signer produces `name` itself. +/// +/// Python's reattach loop skips these, so a caller-supplied copy never reaches +/// the wire next to the computed one. +pub fn is_sigv4_computed_header(name: &str) -> bool { + SIGV4_COMPUTED_HEADER_NAMES.contains(&name.to_ascii_lowercase().as_str()) +} + +pub fn sign_bedrock_post( + url: &str, + body: &[u8], + headers: &BTreeMap, + region: &str, + credentials: &Credentials, + signing_time: SystemTime, +) -> Result, Error> { + let identity: Identity = credentials.clone().into(); + let params = v4::SigningParams::builder() + .identity(&identity) + .region(region) + .name(BEDROCK_SERVICE) + .time(signing_time) + .settings(SigningSettings::default()) + .build() + .map(SigningParams::from) + .map_err(|error| Error::AwsSigningParameters(error.to_string()))?; + let header_refs = headers + .iter() + .map(|(name, value)| (name.as_str(), value.as_str())); + let request = SignableRequest::new("POST", url, header_refs, SignableBody::Bytes(body)) + .map_err(|error| Error::AwsSignableRequest(error.to_string()))?; + let (instructions, _) = sign(request, ¶ms) + .map_err(|error| Error::AwsSigning(error.to_string()))? + .into_parts(); + Ok(instructions + .headers() + .map(|(name, value)| { + let normalized_name = match name { + "authorization" => "Authorization", + "x-amz-date" => "X-Amz-Date", + "x-amz-security-token" => "X-Amz-Security-Token", + _ => name, + }; + (normalized_name.to_string(), value.to_string()) + }) + .collect()) +} + +/// Model-id and region parsing shared by every Bedrock route. +pub fn bedrock_model_id_and_region(model: &str) -> (String, Option) { + let mut stripped = model; + for prefix in ["bedrock/converse/", "bedrock/", "converse/"] { + if let Some(value) = stripped.strip_prefix(prefix) { + stripped = value; + break; + } + } + let mut region = None; + if let Some((candidate, remainder)) = stripped.split_once('/') + && is_bedrock_region(candidate) + { + region = Some(candidate.to_string()); + stripped = remainder; + } + for prefix in ["nova-2/", "nova/"] { + if let Some(value) = stripped.strip_prefix(prefix) { + stripped = value; + break; + } + } + if region.is_none() { + // Python splits the whole ARN and takes field 3, the region. Stripping + // `arn:` first shifts every field down one, so the region is field 2 + // here; field 3 is the account id. + region = stripped + .strip_prefix("arn:") + .and_then(|value| value.split(':').nth(2)) + .filter(|value| !value.is_empty()) + .map(str::to_string); + } + (stripped.to_string(), region) +} + +fn is_bedrock_region(value: &str) -> bool { + value.len() > 3 + && value.contains('-') + && value + .chars() + .all(|char| char.is_ascii_alphanumeric() || char == '-') +} + +pub fn resolve_bedrock_region( + model_region: Option<&str>, + optional_params: &Map, + env_lookup: &dyn Fn(&str) -> Option, +) -> String { + if let Some(region) = optional_params + .get("aws_region_name") + .and_then(Value::as_str) + { + return region.to_string(); + } + if let Some(region) = model_region { + return region.to_string(); + } + env_lookup(AWS_REGION_NAME) + .or_else(|| env_lookup(AWS_REGION)) + .unwrap_or_else(|| DEFAULT_BEDROCK_REGION.to_string()) +} + +pub fn aws_auth_config( + optional_params: &Map, + env_lookup: &dyn Fn(&str) -> Option, +) -> AwsAuthConfig { + let value = |key: &str| { + optional_params + .get(key) + .and_then(Value::as_str) + .map(str::to_string) + }; + let env = |key: &str| env_lookup(key); + AwsAuthConfig { + access_key_id: value("aws_access_key_id").or_else(|| env("AWS_ACCESS_KEY_ID")), + secret_access_key: value("aws_secret_access_key").or_else(|| env("AWS_SECRET_ACCESS_KEY")), + session_token: value("aws_session_token").or_else(|| env("AWS_SESSION_TOKEN")), + region_name: value("aws_region_name").or_else(|| env(AWS_REGION_NAME)), + session_name: value("aws_session_name").or_else(|| env("AWS_SESSION_NAME")), + profile_name: value("aws_profile_name").or_else(|| env("AWS_PROFILE_NAME")), + role_name: value("aws_role_name").or_else(|| env("AWS_ROLE_NAME")), + web_identity_token: value("aws_web_identity_token") + .or_else(|| env("AWS_WEB_IDENTITY_TOKEN")), + sts_endpoint: value("aws_sts_endpoint").or_else(|| env("AWS_STS_ENDPOINT")), + external_id: value("aws_external_id").or_else(|| env("AWS_EXTERNAL_ID")), + } +} + +/// Credentials a host resolved through its own chain and handed down verbatim. +/// +/// A host with its own resolution (LiteLLM's Python `BaseAWSLLM`, which reads +/// profiles, STS and boto sessions) passes the result here so the core signs +/// with exactly those. Without this the core would re-derive from ambient +/// state, where an unrelated `AWS_ROLE_NAME` or `AWS_PROFILE_NAME` in the +/// environment outranks explicit keys in [`classify_auth`] and the two sides +/// would sign as different principals. +pub fn host_supplied_credentials(optional_params: &Map) -> Option { + let value = |key: &str| { + optional_params + .get(key) + .and_then(Value::as_str) + .map(str::trim) + .filter(|value| !value.is_empty()) + }; + let access_key_id = value("aws_access_key_id")?; + let secret_access_key = value("aws_secret_access_key")?; + Some(Credentials::new( + access_key_id, + secret_access_key, + value("aws_session_token").map(str::to_string), + None, + "litellm-host-supplied", + )) +} + +#[cfg(test)] +mod tests { + use super::*; + + fn no_env(_: &str) -> Option { + None + } + + fn parity_inputs() -> (String, Vec, BTreeMap) { + ( + "https://bedrock-runtime.us-east-1.amazonaws.com/model/amazon.titan-text-express-v1/invoke" + .to_string(), + br#"{"input":"hello"}"#.to_vec(), + BTreeMap::from([("Content-Type".to_string(), "application/json".to_string())]), + ) + } + + #[test] + fn reads_the_region_field_of_a_model_arn_not_the_account_id() { + // Python's `_get_aws_region_from_model_arn` splits the whole ARN and + // takes field 3. Stripping `arn:` first shifts every field down one, so + // the region is field 2 here. Taking field 3 after the strip returns + // the account id, which is not a region at all. + let (_, region) = bedrock_model_id_and_region( + "bedrock/arn:aws:bedrock:us-west-2:123456789012:foundation-model/anthropic.claude-v2", + ); + assert_eq!(region.as_deref(), Some("us-west-2")); + } + + #[test] + fn classification_preserves_python_precedence() { + let config = AwsAuthConfig { + access_key_id: Some("ak".into()), + secret_access_key: Some("sk".into()), + session_token: Some("token".into()), + region_name: Some("us-east-1".into()), + session_name: Some("session".into()), + profile_name: Some("profile".into()), + role_name: Some("role".into()), + web_identity_token: Some("oidc".into()), + ..Default::default() + }; + assert!(matches!( + classify_auth(config, &no_env), + AwsAuthFlow::WebIdentity { .. } + )); + } + + #[test] + fn classification_covers_fallthroughs() { + let env = |key: &str| match key { + AWS_PROFILE_NAME => Some("profile".into()), + _ => None, + }; + assert!(matches!( + classify_auth(AwsAuthConfig::default(), &env), + AwsAuthFlow::Profile { .. } + )); + assert!(matches!( + classify_auth( + AwsAuthConfig { + access_key_id: Some("ak".into()), + secret_access_key: Some("sk".into()), + session_token: Some("token".into()), + ..Default::default() + }, + &no_env + ), + AwsAuthFlow::SessionToken { .. } + )); + assert!(matches!( + classify_auth( + AwsAuthConfig { + access_key_id: Some("ak".into()), + secret_access_key: Some("sk".into()), + region_name: Some("us-east-1".into()), + ..Default::default() + }, + &no_env + ), + AwsAuthFlow::StaticKeys { .. } + )); + assert_eq!( + classify_auth(AwsAuthConfig::default(), &no_env), + AwsAuthFlow::DefaultChain + ); + } + + #[tokio::test] + async fn static_credentials_do_not_use_network() { + let credentials = resolve_credentials( + AwsAuthConfig { + access_key_id: Some("ak".into()), + secret_access_key: Some("sk".into()), + region_name: Some("us-east-1".into()), + ..Default::default() + }, + &no_env, + ) + .await + .expect("static credentials"); + assert_eq!(credentials.access_key_id(), "ak"); + assert_eq!(credentials.session_token(), None); + } + + #[test] + fn cache_policy_matches_python_flows() { + assert_eq!( + credential_cache_ttl(&AwsAuthFlow::StaticKeys { + access_key_id: "ak".into(), + secret_access_key: "sk".into(), + region_name: "us-east-1".into(), + }), + Some(STATIC_CREDENTIALS_TTL) + ); + assert_eq!( + credential_cache_ttl(&AwsAuthFlow::DefaultChain), + Some(AMBIENT_CREDENTIALS_TTL) + ); + assert_eq!( + credential_cache_ttl(&AwsAuthFlow::SessionToken { + access_key_id: "ak".into(), + secret_access_key: "sk".into(), + session_token: "token".into(), + }), + None + ); + assert_eq!( + credential_cache_ttl(&AwsAuthFlow::Profile { + name: "profile".into() + }), + None + ); + assert_eq!( + credential_cache_ttl(&AwsAuthFlow::AssumeRole { + role: "arn:aws:iam::123456789012:role/demo".into(), + session_name: None, + }), + None + ); + assert_eq!( + credential_cache_ttl(&AwsAuthFlow::WebIdentity { + token: "token".into(), + role: "arn:aws:iam::123456789012:role/demo".into(), + session_name: "session".into(), + }), + None + ); + } + + #[test] + fn cache_round_trip_preserves_credentials() { + let key = format!("cache-test-{}", std::process::id()); + let credentials = Credentials::new("cache-ak", "cache-sk", None, None, "test"); + set_cached_credentials(key.clone(), credentials.clone(), STATIC_CREDENTIALS_TTL); + assert_eq!( + get_cached_credentials(&key).map(|value| value.access_key_id().to_string()), + Some("cache-ak".to_string()) + ); + } + + #[test] + fn same_role_comparison_matches_partition_account_and_role() { + assert!(same_role_arns( + "arn:aws:iam::123456789012:role/path/demo", + "arn:aws:sts::123456789012:assumed-role/demo/session" + )); + assert!(!same_role_arns( + "arn:aws:iam::123456789012:role/demo", + "arn:aws:iam::999999999999:role/demo" + )); + assert!(!same_role_arns( + "arn:aws:iam::123456789012:role/demo", + "arn:aws-cn:iam::123456789012:role/demo" + )); + assert!(!same_role_arns( + "arn:aws:iam::123456789012:user/demo", + "arn:aws:iam::123456789012:role/demo" + )); + } + + #[test] + fn a_forwarded_client_header_is_not_folded_into_the_signature() { + // Python signs only the AWS header set, so a header a caller forwarded + // cannot change the canonical request. Signing it instead makes the + // request 403 the moment anything on the wire rewrites or drops it. + let (url, body, mut headers) = parity_inputs(); + headers.insert("x-request-id".to_string(), "abc-123".to_string()); + headers.insert("Accept-Encoding".to_string(), "gzip".to_string()); + headers.insert("x-amzn-trace-id".to_string(), "Root=1-abc".to_string()); + let signable = aws_signature_headers(&headers); + + assert!(!signable.contains_key("x-request-id")); + assert!(!signable.contains_key("Accept-Encoding")); + // The AWS-prefixed one is genuinely part of the signature. + assert!(signable.contains_key("x-amzn-trace-id")); + assert!(signable.contains_key("Content-Type")); + + let credentials = Credentials::new( + "AKIDEXAMPLE", + "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY", + None, + None, + "test", + ); + let signed = sign_bedrock_post( + &url, + &body, + &signable, + "us-east-1", + &credentials, + SystemTime::UNIX_EPOCH, + ) + .expect("signs"); + let authorization = signed + .get("Authorization") + .expect("carries an authorization header"); + assert!( + !authorization.contains("x-request-id"), + "forwarded header reached SignedHeaders: {authorization}" + ); + assert!( + !authorization.contains("accept-encoding"), + "forwarded header reached SignedHeaders: {authorization}" + ); + } + + #[test] + fn signing_matches_botocore_golden_vector() { + let (url, body, headers) = parity_inputs(); + let credentials = Credentials::new( + "AKIDEXAMPLE", + "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY", + Some("session-token".to_string()), + None, + "test", + ); + let signed = sign_bedrock_post( + &url, + &body, + &headers, + "us-east-1", + &credentials, + UNIX_EPOCH + std::time::Duration::from_secs(1_704_164_645), + ) + .expect("golden signature"); + assert_eq!( + signed.get("X-Amz-Date").map(String::as_str), + Some("20240102T030405Z") + ); + assert_eq!( + signed.get("X-Amz-Security-Token").map(String::as_str), + Some("session-token") + ); + assert_eq!( + signed.get("Authorization").map(String::as_str), + Some( + "AWS4-HMAC-SHA256 Credential=AKIDEXAMPLE/20240102/us-east-1/bedrock/aws4_request, SignedHeaders=content-type;host;x-amz-date;x-amz-security-token, Signature=55c027ef47527d3ad63f1735f9d099efdbc99f296ff914bd94e727e24ec0e464" + ) + ); + } + + #[test] + fn signing_without_session_token_omits_security_header() { + let (url, body, headers) = parity_inputs(); + let credentials = Credentials::new( + "AKIDEXAMPLE", + "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY", + None, + None, + "test", + ); + let signed = sign_bedrock_post( + &url, + &body, + &headers, + "us-east-1", + &credentials, + UNIX_EPOCH + std::time::Duration::from_secs(1_704_164_645), + ) + .expect("signature"); + assert!(!signed.contains_key("X-Amz-Security-Token")); + } + + #[ignore] + #[tokio::test] + async fn live_bedrock_invoke_model_returns_200() -> Result<(), Box> { + let access_key_id = std::env::var("AWS_BEDROCK_TEST_ACCESS_KEY_ID")?; + let secret_access_key = std::env::var("AWS_BEDROCK_TEST_SECRET_ACCESS_KEY")?; + let body = br#"{"anthropic_version":"bedrock-2023-05-31","max_tokens":1,"messages":[{"role":"user","content":[{"type":"text","text":"ping"}]}]}"#.to_vec(); + let headers = + BTreeMap::from([("Content-Type".to_string(), "application/json".to_string())]); + let credentials = resolve_credentials( + AwsAuthConfig { + access_key_id: Some(access_key_id), + secret_access_key: Some(secret_access_key), + region_name: Some("us-west-2".to_string()), + ..Default::default() + }, + &no_env, + ) + .await?; + let client = reqwest::Client::new(); + let mut failures = Vec::new(); + + for region in ["us-west-2", "us-east-1"] { + let url = format!( + "https://bedrock-runtime.{region}.amazonaws.com/model/us.anthropic.claude-opus-4-8/invoke" + ); + let signed_headers = sign_bedrock_post( + &url, + &body, + &headers, + region, + &credentials, + SystemTime::now(), + )?; + let mut request = client.post(&url).body(body.clone()); + for (name, value) in &headers { + request = request.header(name, value); + } + for (name, value) in signed_headers { + request = request.header(name, value); + } + let response = request.send().await?; + let status = response.status(); + let response_body = response.text().await?; + let snippet: String = response_body.chars().take(240).collect(); + println!("region={region} status={status} response={snippet}"); + if status == reqwest::StatusCode::OK { + return Ok(()); + } + failures.push(format!("{region}: {status} {snippet}")); + } + + panic!( + "no Bedrock region returned HTTP 200: {}", + failures.join("; ") + ); + } +} diff --git a/litellm-rust/crates/auth-aws/src/constants.rs b/litellm-rust/crates/auth-aws/src/constants.rs new file mode 100644 index 00000000000..be215cc9016 --- /dev/null +++ b/litellm-rust/crates/auth-aws/src/constants.rs @@ -0,0 +1,43 @@ +pub const AWS_ACCESS_KEY_ID: &str = "AWS_ACCESS_KEY_ID"; +pub const AWS_SECRET_ACCESS_KEY: &str = "AWS_SECRET_ACCESS_KEY"; +pub const AWS_SESSION_TOKEN: &str = "AWS_SESSION_TOKEN"; +pub const AWS_REGION_NAME: &str = "AWS_REGION_NAME"; +pub const AWS_REGION: &str = "AWS_REGION"; +pub const AWS_SESSION_NAME: &str = "AWS_SESSION_NAME"; +pub const AWS_PROFILE_NAME: &str = "AWS_PROFILE_NAME"; +pub const AWS_ROLE_NAME: &str = "AWS_ROLE_NAME"; +pub const AWS_WEB_IDENTITY_TOKEN: &str = "AWS_WEB_IDENTITY_TOKEN"; +pub const AWS_ROLE_ARN: &str = "AWS_ROLE_ARN"; +pub const AWS_WEB_IDENTITY_TOKEN_FILE: &str = "AWS_WEB_IDENTITY_TOKEN_FILE"; +pub const AWS_STS_ENDPOINT: &str = "AWS_STS_ENDPOINT"; +pub const AWS_EXTERNAL_ID: &str = "AWS_EXTERNAL_ID"; +pub const AWS_BEARER_TOKEN_BEDROCK: &str = "AWS_BEARER_TOKEN_BEDROCK"; + +/// Headers SigV4 covers, beyond the `x-amz-` / `x-amzn-` prefixes. Mirrors +/// Python's `_filter_headers_for_aws_signature` allowlist. +pub const AWS_SIGNED_HEADER_NAMES: &[&str] = &[ + "host", + "content-type", + "date", + "x-amz-date", + "x-amz-security-token", + "x-amz-content-sha256", + "x-amz-algorithm", + "x-amz-credential", + "x-amz-signedheaders", + "x-amz-signature", +]; +/// Headers the signer emits itself. Mirrors Python's `SIGV4_COMPUTED_HEADERS`, +/// which the reattach loop skips so a caller's copy cannot ride alongside the +/// computed one. +pub const SIGV4_COMPUTED_HEADER_NAMES: &[&str] = &[ + "authorization", + "x-amz-date", + "x-amz-security-token", + "date", +]; +pub const BEDROCK_SERVICE: &str = "bedrock"; +pub const DEFAULT_SESSION_NAME_PREFIX: &str = "litellm-session"; +pub const DEFAULT_BEDROCK_REGION: &str = "us-west-2"; +pub const BEDROCK_RUNTIME_ENDPOINT_TEMPLATE: &str = + "https://bedrock-runtime.{region}.amazonaws.com"; diff --git a/litellm-rust/crates/auth-aws/src/lib.rs b/litellm-rust/crates/auth-aws/src/lib.rs new file mode 100644 index 00000000000..aaf1c713067 --- /dev/null +++ b/litellm-rust/crates/auth-aws/src/lib.rs @@ -0,0 +1,4 @@ +mod aws; +pub mod constants; + +pub use aws::*; diff --git a/litellm-rust/crates/auth-azure/Cargo.toml b/litellm-rust/crates/auth-azure/Cargo.toml new file mode 100644 index 00000000000..16f18a3a789 --- /dev/null +++ b/litellm-rust/crates/auth-azure/Cargo.toml @@ -0,0 +1,19 @@ +[package] +name = "litellm-auth-azure" +version = "0.1.0" +edition.workspace = true +license.workspace = true +repository.workspace = true + +[dependencies] +azure_core.workspace = true +azure_identity.workspace = true +litellm-auth.workspace = true +moka.workspace = true +serde_json.workspace = true +sha2.workspace = true +strum.workspace = true +url.workspace = true + +[dev-dependencies] +tokio.workspace = true diff --git a/litellm-rust/crates/core/src/providers/azure_ai/auth/credential_provider_cache.rs b/litellm-rust/crates/auth-azure/src/credential_provider_cache.rs similarity index 87% rename from litellm-rust/crates/core/src/providers/azure_ai/auth/credential_provider_cache.rs rename to litellm-rust/crates/auth-azure/src/credential_provider_cache.rs index 297e4cc6502..ab9ffc719df 100644 --- a/litellm-rust/crates/core/src/providers/azure_ai/auth/credential_provider_cache.rs +++ b/litellm-rust/crates/auth-azure/src/credential_provider_cache.rs @@ -4,7 +4,7 @@ use std::sync::Arc; use azure_core::credentials::TokenCredential; use moka::future::Cache; -use crate::AuthError; +use litellm_auth::Error; #[derive(Clone, Debug, PartialEq, Eq, Hash)] pub(crate) struct AzureCredentialProviderCacheKey { @@ -31,9 +31,9 @@ impl AzureCredentialProviderCache { &self, key: AzureCredentialProviderCacheKey, create: F, - ) -> Result, AuthError> + ) -> Result, Error> where - F: Future, AuthError>>, + F: Future, Error>>, { self.entries .try_get_with(key, create) diff --git a/litellm-rust/crates/auth-azure/src/lib.rs b/litellm-rust/crates/auth-azure/src/lib.rs new file mode 100644 index 00000000000..e76227d6aa2 --- /dev/null +++ b/litellm-rust/crates/auth-azure/src/lib.rs @@ -0,0 +1,7 @@ +mod credential_provider_cache; +mod native; +mod resolve; +mod types; + +pub use resolve::AzureAuthService; +pub use types::AzureAuthInputs; diff --git a/litellm-rust/crates/core/src/providers/azure_ai/auth/native.rs b/litellm-rust/crates/auth-azure/src/native.rs similarity index 94% rename from litellm-rust/crates/core/src/providers/azure_ai/auth/native.rs rename to litellm-rust/crates/auth-azure/src/native.rs index b8f19818d16..5f913a8ad01 100644 --- a/litellm-rust/crates/core/src/providers/azure_ai/auth/native.rs +++ b/litellm-rust/crates/auth-azure/src/native.rs @@ -1,4 +1,3 @@ -use crate::auth::error::AuthConfigurationError; use std::sync::Arc; use std::time::{Duration, UNIX_EPOCH}; @@ -13,8 +12,8 @@ use azure_identity::{ }; use sha2::{Digest, Sha256}; -use crate::AuthError; -use crate::auth::{InputSource, ResolvedCredential, SecretValue, Sourced}; +use litellm_auth::Error; +use litellm_auth::{InputSource, ResolvedCredential, SecretValue, Sourced}; use super::credential_provider_cache::{ AzureCredentialProviderCache, AzureCredentialProviderCacheKey, @@ -62,7 +61,7 @@ pub(crate) struct ValidatedAzureRequest { } impl ValidatedAzureRequest { - pub(crate) fn new(request: NativeAzureRequest) -> Result { + pub(crate) fn new(request: NativeAzureRequest) -> Result { validate_authority(&request)?; let credential_source = validate_sources(&request)?; Ok(Self { @@ -120,7 +119,7 @@ impl NativeAzureTokenAcquirer { pub(crate) async fn acquire( &self, request: ValidatedAzureRequest, - ) -> Result { + ) -> Result { let scope = request.request.scope().to_string(); let key = request.request.cache_key(); let transport = self.transport.clone(); @@ -134,7 +133,7 @@ impl NativeAzureTokenAcquirer { let token = credential .get_token(&[scope.as_str()], None) .await - .map_err(|error| AuthError::AzureTokenAcquisition(error.to_string()))?; + .map_err(|error| Error::AzureTokenAcquisition(error.to_string()))?; let expires_on = u64::try_from(token.expires_on.unix_timestamp()) .ok() .map(|seconds| UNIX_EPOCH + Duration::from_secs(seconds)); @@ -239,7 +238,7 @@ impl NativeAzureRequest { } } -fn validate_authority(request: &NativeAzureRequest) -> Result<(), AuthError> { +fn validate_authority(request: &NativeAzureRequest) -> Result<(), Error> { let authority = match request { NativeAzureRequest::ClientSecret { authority, .. } | NativeAzureRequest::ClientAssertion { authority, .. } @@ -251,8 +250,7 @@ fn validate_authority(request: &NativeAzureRequest) -> Result<(), AuthError> { let Some(authority) = authority else { return Ok(()); }; - let url = url::Url::parse(authority.value()) - .map_err(|_| AuthError::Configuration(AuthConfigurationError::InvalidAzureAuthority))?; + let url = url::Url::parse(authority.value()).map_err(|_| Error::InvalidAzureAuthority)?; if url.scheme() != "https" || url.host_str().is_none() || !url.username().is_empty() @@ -261,14 +259,12 @@ fn validate_authority(request: &NativeAzureRequest) -> Result<(), AuthError> { || url.fragment().is_some() || !matches!(url.path(), "" | "/") { - return Err(AuthError::Configuration( - AuthConfigurationError::InvalidAzureAuthority, - )); + return Err(Error::InvalidAzureAuthority); } Ok(()) } -fn validate_sources(request: &NativeAzureRequest) -> Result { +fn validate_sources(request: &NativeAzureRequest) -> Result { match request { NativeAzureRequest::ClientSecret { tenant_id, @@ -356,7 +352,7 @@ fn is_request_controlled(value: &Sourced, optional: Option<&Sourced Result { +fn trusted_only(sources: &[InputSource]) -> Result { if sources.contains(&InputSource::Request) { return mixed_sources(); } @@ -371,16 +367,14 @@ fn trusted_source(sources: &[InputSource]) -> InputSource { } } -fn mixed_sources() -> Result { - Err(AuthError::Configuration( - AuthConfigurationError::MixedAzureCredentialSources, - )) +fn mixed_sources() -> Result { + Err(Error::MixedAzureCredentialSources) } fn build_credential( request: NativeAzureRequest, transport: Option, -) -> Result, AuthError> { +) -> Result, Error> { match request { NativeAzureRequest::ClientSecret { tenant_id, @@ -439,11 +433,7 @@ fn build_credential( NativeAzureRequest::DeveloperTools { .. } => DeveloperToolsCredential::new(None) .map(|credential| credential as Arc), } - .map_err(|error| { - AuthError::Configuration(AuthConfigurationError::AzureCredentialInitialization( - error.to_string(), - )) - }) + .map_err(|error| Error::AzureCredentialInitialization(error.to_string())) } fn client_options( @@ -494,7 +484,7 @@ mod tests { use azure_core::{Bytes, Result}; use super::{NativeAzureRequest, NativeAzureTokenAcquirer, ValidatedAzureRequest}; - use crate::auth::{InputSource, SecretValue, Sourced}; + use litellm_auth::{InputSource, SecretValue, Sourced}; fn deployment(value: T) -> Sourced { Sourced::new(value, InputSource::Deployment) @@ -659,9 +649,7 @@ mod tests { assert!(matches!( error, - crate::AuthError::Configuration( - crate::auth::error::AuthConfigurationError::MixedAzureCredentialSources - ) + litellm_auth::Error::MixedAzureCredentialSources )); } @@ -691,12 +679,7 @@ mod tests { authority, )) .unwrap_err(); - assert!(matches!( - error, - crate::AuthError::Configuration( - crate::auth::error::AuthConfigurationError::InvalidAzureAuthority - ) - )); + assert!(matches!(error, litellm_auth::Error::InvalidAzureAuthority)); } } } diff --git a/litellm-rust/crates/core/src/providers/azure_ai/auth/resolve.rs b/litellm-rust/crates/auth-azure/src/resolve.rs similarity index 89% rename from litellm-rust/crates/core/src/providers/azure_ai/auth/resolve.rs rename to litellm-rust/crates/auth-azure/src/resolve.rs index 025dd4f8740..660a95b79d8 100644 --- a/litellm-rust/crates/core/src/providers/azure_ai/auth/resolve.rs +++ b/litellm-rust/crates/auth-azure/src/resolve.rs @@ -1,6 +1,5 @@ -use crate::AuthError; -use crate::auth::error::AuthConfigurationError; -use crate::auth::{ +use litellm_auth::Error; +use litellm_auth::{ CredentialFileRef, CredentialLookup, CredentialRef, InputSource, ResolvedCredential, SecretValue, Sourced, TokenProviderHandle, }; @@ -37,7 +36,7 @@ pub(crate) enum AzureCredentialPlan { } /// Rust counterpart to Python's `get_azure_ad_token`, not `BaseAzureLLM`. -pub(crate) struct AzureAuthService { +pub struct AzureAuthService { native: Arc, } @@ -45,14 +44,14 @@ trait AzureTokenAcquirer: Send + Sync { fn acquire( &self, request: ValidatedAzureRequest, - ) -> Pin> + Send + '_>>; + ) -> Pin> + Send + '_>>; } impl AzureTokenAcquirer for NativeAzureTokenAcquirer { fn acquire( &self, request: ValidatedAzureRequest, - ) -> Pin> + Send + '_>> { + ) -> Pin> + Send + '_>> { Box::pin(NativeAzureTokenAcquirer::acquire(self, request)) } } @@ -71,17 +70,17 @@ impl AzureAuthService { Self { native } } - pub(crate) async fn get_azure_ad_token( + pub async fn get_azure_ad_token( &self, inputs: &AzureAuthInputs, env_lookup: &(dyn Fn(&str) -> Option + Sync), - ) -> Result>, AuthError> { + ) -> Result>, Error> { match select_auth_plan(inputs, env_lookup)? { AzureCredentialPlan::Supplied(credential) => Ok(Some(credential)), AzureCredentialPlan::Caller(caller) => { let credential = caller.acquire().await?; if credential.secret().expose().is_empty() { - return Err(AuthError::EmptyAzureToken); + return Err(Error::EmptyAzureToken); } Ok(Some(Sourced::new(credential, InputSource::Deployment))) } @@ -94,7 +93,7 @@ impl AzureAuthService { } => { let assertion = resolve_reference(inputs, env_lookup, reference.value()) .await? - .ok_or(AuthError::UnresolvedOidcReference)?; + .ok_or(Error::UnresolvedOidcReference)?; let request = ValidatedAzureRequest::new(NativeAzureRequest::ClientAssertion { tenant_id, client_id, @@ -126,7 +125,7 @@ impl AzureAuthService { Err(error) => failures.push(error), } } - Err(AuthError::CredentialChain(failures)) + Err(Error::CredentialChain(failures)) } AzureCredentialPlan::Missing => Ok(None), } @@ -136,7 +135,7 @@ impl AzureAuthService { pub(crate) fn select_auth_plan( inputs: &AzureAuthInputs, env_lookup: &dyn Fn(&str) -> Option, -) -> Result { +) -> Result { let token = configured_secret(&inputs.azure_ad_token, AZURE_AD_TOKEN_ENV, env_lookup); let tenant_id = configured_string(&inputs.tenant_id, AZURE_TENANT_ID_ENV, env_lookup); let client_id = configured_string(&inputs.client_id, AZURE_CLIENT_ID_ENV, env_lookup); @@ -157,7 +156,7 @@ pub(crate) fn select_auth_plan( .map(|selector| Sourced::new(selector, value.source())) }) .transpose() - .map_err(|_| AuthError::Configuration(AuthConfigurationError::InvalidAzureSelector))?; + .map_err(|_| Error::InvalidAzureSelector)?; let federated_token_file = configured_string( &inputs.federated_token_file, AZURE_FEDERATED_TOKEN_FILE_ENV, @@ -229,7 +228,7 @@ fn select_native_plan( scope: Sourced, authority: Option>, refresh_source: InputSource, -) -> Result { +) -> Result { let selected = selector.unwrap_or_else(|| { Sourced::new( { @@ -247,9 +246,7 @@ fn select_native_plan( let selection_source = selected.source(); match selected.into_value() { - AzureCredentialType::ClientSecretCredential => Err(AuthError::Configuration( - AuthConfigurationError::MissingClientSecretFields, - )), + AzureCredentialType::ClientSecretCredential => Err(Error::MissingClientSecretFields), AzureCredentialType::WorkloadIdentityCredential => { Ok(AzureCredentialPlan::Native(ValidatedAzureRequest::new( workload_request(tenant_id, client_id, federated_token_file, scope, authority)?, @@ -331,17 +328,11 @@ fn workload_request( token_file_path: Option>, scope: Sourced, authority: Option>, -) -> Result { +) -> Result { Ok(NativeAzureRequest::WorkloadIdentity { - tenant_id: tenant_id.ok_or(AuthError::Configuration( - AuthConfigurationError::MissingWorkloadTenant, - ))?, - client_id: client_id.ok_or(AuthError::Configuration( - AuthConfigurationError::MissingWorkloadClient, - ))?, - token_file_path: token_file_path.ok_or(AuthError::Configuration( - AuthConfigurationError::MissingWorkloadTokenFile, - ))?, + tenant_id: tenant_id.ok_or(Error::MissingWorkloadTenant)?, + client_id: client_id.ok_or(Error::MissingWorkloadClient)?, + token_file_path: token_file_path.ok_or(Error::MissingWorkloadTokenFile)?, scope, authority, }) @@ -383,7 +374,7 @@ async fn resolve_reference( inputs: &AzureAuthInputs, env_lookup: &(dyn Fn(&str) -> Option + Sync), reference: &CredentialRef, -) -> Result, AuthError> { +) -> Result, Error> { let lookup = match reference { CredentialRef::Explicit(secret) => return Ok(Some(secret.clone())), CredentialRef::Env(name) => env_lookup(name) @@ -395,9 +386,7 @@ async fn resolve_reference( let resolver = inputs .credential_resolver .as_ref() - .ok_or(AuthError::Configuration( - AuthConfigurationError::MissingHostResolver, - ))?; + .ok_or(Error::MissingHostResolver)?; resolver.resolve(reference).await? } }; @@ -409,15 +398,13 @@ async fn resolve_reference( fn oidc_reference( token: &Option>, -) -> Result>, AuthError> { +) -> Result>, Error> { let Some(token) = token.as_ref() else { return Ok(None); }; let value = token.value().expose(); if token.source() == InputSource::Request && value.starts_with("oidc/") { - return Err(AuthError::Configuration( - AuthConfigurationError::RequestAzureCredentialReference, - )); + return Err(Error::RequestAzureCredentialReference); } if let Some(name) = value.strip_prefix("oidc/env/") { return non_empty_reference(name, "OIDC environment reference") @@ -439,18 +426,14 @@ fn oidc_reference( ))); } if value.starts_with("oidc/") { - return Err(AuthError::Configuration( - AuthConfigurationError::UnsupportedOidcReference, - )); + return Err(Error::UnsupportedOidcReference); } Ok(None) } -fn non_empty_reference(value: &str, kind: &str) -> Result { +fn non_empty_reference(value: &str, kind: &str) -> Result { if value.is_empty() { - return Err(AuthError::Configuration( - AuthConfigurationError::EmptyReference(kind.to_string()), - )); + return Err(Error::EmptyReference(kind.to_string())); } Ok(value.to_string()) } @@ -466,14 +449,14 @@ mod tests { AzureAuthService, AzureCredentialPlan, AzureTokenAcquirer, oidc_reference, resolve_reference, select_auth_plan, }; - use crate::AuthError; - use crate::auth::ResolvedCredential; - use crate::auth::{ + use crate::native::ValidatedAzureRequest; + use crate::types::AzureAuthInputs; + use litellm_auth::Error; + use litellm_auth::ResolvedCredential; + use litellm_auth::{ CredentialFileRef, CredentialLookup, CredentialLookupFuture, CredentialRef, CredentialResolver, CredentialResolverHandle, InputSource, SecretValue, Sourced, }; - use crate::providers::azure_ai::auth::native::ValidatedAzureRequest; - use crate::providers::azure_ai::auth::types::AzureAuthInputs; #[derive(Debug)] struct FileResolver; @@ -487,9 +470,8 @@ mod tests { fn acquire( &self, request: ValidatedAzureRequest, - ) -> std::pin::Pin< - Box> + Send + '_>, - > { + ) -> std::pin::Pin> + Send + '_>> + { let kind = request.kind(); self.requests.lock().unwrap().push(kind); Box::pin(async move { @@ -499,7 +481,7 @@ mod tests { expires_on: None, }) } else { - Err(AuthError::AzureTokenAcquisition(format!("{kind} failed"))) + Err(Error::AzureTokenAcquisition(format!("{kind} failed"))) } }) } @@ -612,12 +594,7 @@ mod tests { }) .unwrap_err(); - assert!(matches!( - error, - AuthError::Configuration( - crate::auth::error::AuthConfigurationError::RequestAzureCredentialReference - ) - )); + assert!(matches!(error, Error::RequestAzureCredentialReference)); } #[tokio::test] @@ -678,6 +655,6 @@ mod tests { .await .unwrap_err(); - assert!(matches!(error, AuthError::CredentialChain(errors) if errors.len() == 2)); + assert!(matches!(error, Error::CredentialChain(errors) if errors.len() == 2)); } } diff --git a/litellm-rust/crates/core/src/providers/azure_ai/auth/types.rs b/litellm-rust/crates/auth-azure/src/types.rs similarity index 93% rename from litellm-rust/crates/core/src/providers/azure_ai/auth/types.rs rename to litellm-rust/crates/auth-azure/src/types.rs index f15d526d945..2a510de1f43 100644 --- a/litellm-rust/crates/core/src/providers/azure_ai/auth/types.rs +++ b/litellm-rust/crates/auth-azure/src/types.rs @@ -1,10 +1,9 @@ -use crate::auth::error::AuthConfigurationError; use serde_json::{Map, Value}; use std::collections::BTreeMap; use strum::EnumString; -use crate::AuthError; -use crate::auth::{ +use litellm_auth::Error; +use litellm_auth::{ CredentialResolverHandle, InputSource, SecretValue, Sourced, TokenProviderHandle, }; @@ -54,14 +53,14 @@ pub struct AzureAuthInputs { impl AzureAuthInputs { #[cfg(test)] - pub fn from_optional_params(params: &Map) -> Result { + pub fn from_optional_params(params: &Map) -> Result { Self::from_sourced_optional_params(params, &BTreeMap::new()) } pub fn from_sourced_optional_params( params: &Map, sources: &BTreeMap, - ) -> Result { + ) -> Result { Ok(Self { azure_ad_token: secret_config(params, sources, "azure_ad_token")?, azure_ad_token_provider: None, @@ -88,15 +87,13 @@ fn string_config( params: &Map, sources: &BTreeMap, name: &str, -) -> Result, AuthError> { +) -> Result, Error> { let source = source_for(sources, name); match params.get(name) { None => Ok(ConfigValue::Absent), Some(Value::Null) => Ok(ConfigValue::ExplicitNone(source)), Some(Value::String(value)) => Ok(ConfigValue::Value(Sourced::new(value.clone(), source))), - Some(_) => Err(AuthError::Configuration( - AuthConfigurationError::InvalidFieldType(name.to_string()), - )), + Some(_) => Err(Error::InvalidFieldType(name.to_string())), } } @@ -104,7 +101,7 @@ fn secret_config( params: &Map, sources: &BTreeMap, name: &str, -) -> Result, AuthError> { +) -> Result, Error> { Ok(match string_config(params, sources, name)? { ConfigValue::Absent => ConfigValue::Absent, ConfigValue::ExplicitNone(source) => ConfigValue::ExplicitNone(source), @@ -123,7 +120,7 @@ mod tests { use std::collections::BTreeMap; use super::{AzureAuthInputs, AzureCredentialType, ConfigValue}; - use crate::auth::{InputSource, Sourced}; + use litellm_auth::{InputSource, Sourced}; #[test] fn selector_parsing_is_exact() { diff --git a/litellm-rust/crates/auth-gcp/Cargo.toml b/litellm-rust/crates/auth-gcp/Cargo.toml new file mode 100644 index 00000000000..aaa6fbb2cda --- /dev/null +++ b/litellm-rust/crates/auth-gcp/Cargo.toml @@ -0,0 +1,15 @@ +[package] +name = "litellm-auth-gcp" +version = "0.1.0" +edition.workspace = true +license.workspace = true +repository.workspace = true + +[dependencies] +gcp_auth.workspace = true +litellm-auth.workspace = true +moka.workspace = true +serde_json.workspace = true +sha2.workspace = true +tokio.workspace = true +tracing.workspace = true diff --git a/litellm-rust/crates/core/src/auth/vertex.rs b/litellm-rust/crates/auth-gcp/src/lib.rs similarity index 90% rename from litellm-rust/crates/core/src/auth/vertex.rs rename to litellm-rust/crates/auth-gcp/src/lib.rs index e13b57abdfe..331dc12488e 100644 --- a/litellm-rust/crates/core/src/auth/vertex.rs +++ b/litellm-rust/crates/auth-gcp/src/lib.rs @@ -9,9 +9,8 @@ use moka::future::Cache; use serde_json::{Map, Value}; use sha2::{Digest, Sha256}; -use crate::auth::error::AuthConfigurationError; -use crate::auth::http::apply_credential; -use crate::auth::{AuthError, CredentialPlacement, InputSource, SecretValue, Sourced}; +use litellm_auth::http::apply_credential; +use litellm_auth::{CredentialPlacement, Error, InputSource, SecretValue, Sourced}; const CLOUD_PLATFORM_SCOPE: &str = "https://www.googleapis.com/auth/cloud-platform"; const GOOGLE_OAUTH_TOKEN_ENDPOINT: &str = "https://oauth2.googleapis.com/token"; @@ -24,17 +23,17 @@ const VERTEXAI_LOCATION_ENV: &str = "VERTEXAI_LOCATION"; const VERTEX_LOCATION_ENV: &str = "VERTEX_LOCATION"; #[derive(Clone, Debug, Default)] -pub(crate) struct VertexConfig { +pub struct VertexConfig { credentials: Option>, project_id: Option, location: Option, } impl VertexConfig { - pub(crate) fn from_sourced_optional_params( + pub fn from_sourced_optional_params( params: &Map, sources: &BTreeMap, - ) -> Result { + ) -> Result { Ok(Self { credentials: optional_credentials( params, @@ -46,16 +45,16 @@ impl VertexConfig { }) } - pub(crate) fn project_id(&self) -> Option<&str> { + pub fn project_id(&self) -> Option<&str> { self.project_id.as_deref() } - pub(crate) fn location(&self) -> Option<&str> { + pub fn location(&self) -> Option<&str> { self.location.as_deref() } } -pub(crate) struct VertexEnvironment { +pub struct VertexEnvironment { pub headers: Vec<(String, String)>, pub project_id: String, } @@ -65,7 +64,7 @@ struct VertexAccessToken { project_id: String, } -pub(crate) fn get_vertex_ai_project( +pub fn get_vertex_ai_project( config: &VertexConfig, env_lookup: &dyn Fn(&str) -> Option, ) -> Option { @@ -75,7 +74,7 @@ pub(crate) fn get_vertex_ai_project( .or_else(|| non_empty_env(env_lookup, VERTEXAI_PROJECT_ENV)) } -pub(crate) fn get_vertex_ai_location( +pub fn get_vertex_ai_location( config: &VertexConfig, env_lookup: &dyn Fn(&str) -> Option, ) -> Option { @@ -87,7 +86,7 @@ pub(crate) fn get_vertex_ai_location( } #[derive(Clone)] -pub(crate) struct VertexAuth { +pub struct VertexAuth { providers: Cache>, loader: Arc, } @@ -106,13 +105,14 @@ impl VertexAuth { } } - pub(crate) async fn validate_environment( + #[tracing::instrument(target = "litellm::function_trace", level = "trace", skip_all)] + pub async fn validate_environment( &self, headers: Vec<(String, String)>, api_key: Option<&str>, config: &VertexConfig, env_lookup: &(dyn Fn(&str) -> Option + Sync), - ) -> Result { + ) -> Result { let has_authorization = headers .iter() .any(|(name, _)| name.eq_ignore_ascii_case("Authorization")); @@ -160,7 +160,7 @@ impl VertexAuth { &self, config: &VertexConfig, env_lookup: &(dyn Fn(&str) -> Option + Sync), - ) -> Result { + ) -> Result { let provider = self.load_provider(config, env_lookup).await?; let (token, project_id) = tokio::try_join!(provider.token(), provider.project_id())?; Ok(VertexAccessToken { token, project_id }) @@ -170,7 +170,7 @@ impl VertexAuth { &self, config: &VertexConfig, env_lookup: &(dyn Fn(&str) -> Option + Sync), - ) -> Result, AuthError> { + ) -> Result, Error> { let source = credential_source(config, env_lookup); let key = source.cache_key(); self.providers @@ -189,7 +189,7 @@ trait VertexProviderLoader: Send + Sync { fn load(&self, source: CredentialSource) -> VertexAuthFuture<'_, Arc>; } -type VertexAuthFuture<'a, T> = Pin> + Send + 'a>>; +type VertexAuthFuture<'a, T> = Pin> + Send + 'a>>; struct GcpTokenSource(Arc); @@ -249,7 +249,7 @@ impl VertexProviderLoader for GcpProviderLoader { } } -fn validate_request_credentials(configured: &str) -> Result<&str, AuthError> { +fn validate_request_credentials(configured: &str) -> Result<&str, Error> { let token_uri = serde_json::from_str::(configured) .ok() .and_then(|credentials| { @@ -259,7 +259,7 @@ fn validate_request_credentials(configured: &str) -> Result<&str, AuthError> { .map(str::to_string) }); if token_uri.as_deref() != Some(GOOGLE_OAUTH_TOKEN_ENDPOINT) { - return Err(AuthConfigurationError::RequestVertexTokenEndpoint.into()); + return Err(Error::RequestVertexTokenEndpoint); } Ok(configured) } @@ -321,7 +321,7 @@ fn optional_credentials( params: &Map, sources: &BTreeMap, names: &[&str], -) -> Result>, AuthError> { +) -> Result>, Error> { for name in names { let source = source_for(sources, name); match params.get(*name) { @@ -336,17 +336,10 @@ fn optional_credentials( .map(SecretValue::new) .map(|value| Sourced::new(value, source)) .map(Some) - .map_err(|error| { - AuthError::Configuration(AuthConfigurationError::InvalidFieldType(format!( - "{}: {error}", - names[0] - ))) - }); + .map_err(|error| Error::InvalidFieldType(format!("{}: {error}", names[0]))); } Some(_) => { - return Err(AuthError::Configuration( - AuthConfigurationError::InvalidFieldType(names[0].to_string()), - )); + return Err(Error::InvalidFieldType(names[0].to_string())); } } } @@ -357,19 +350,14 @@ fn source_for(sources: &BTreeMap, name: &str) -> InputSourc sources.get(name).copied().unwrap_or_default() } -fn optional_string( - params: &Map, - names: &[&str], -) -> Result, AuthError> { +fn optional_string(params: &Map, names: &[&str]) -> Result, Error> { for name in names { match params.get(*name) { None | Some(Value::Null) => continue, Some(Value::String(value)) if value.trim().is_empty() => continue, Some(Value::String(value)) => return Ok(Some(value.clone())), Some(_) => { - return Err(AuthError::Configuration( - AuthConfigurationError::InvalidFieldType(names[0].to_string()), - )); + return Err(Error::InvalidFieldType(names[0].to_string())); } } } @@ -382,8 +370,8 @@ fn non_empty_env(env_lookup: &dyn Fn(&str) -> Option, name: &str) -> Opt .filter(|value| !value.is_empty()) } -fn auth_acquisition_error(error: gcp_auth::Error) -> AuthError { - AuthError::VertexTokenAcquisition(error.to_string()) +fn auth_acquisition_error(error: gcp_auth::Error) -> Error { + Error::VertexTokenAcquisition(error.to_string()) } #[cfg(test)] @@ -537,15 +525,11 @@ mod tests { ); assert!(matches!( validate_request_credentials(r#"{"token_uri":"http://127.0.0.1/token"}"#), - Err(AuthError::Configuration( - AuthConfigurationError::RequestVertexTokenEndpoint - )) + Err(Error::RequestVertexTokenEndpoint) )); assert!(matches!( validate_request_credentials("{}"), - Err(AuthError::Configuration( - AuthConfigurationError::RequestVertexTokenEndpoint - )) + Err(Error::RequestVertexTokenEndpoint) )); } diff --git a/litellm-rust/crates/auth/Cargo.toml b/litellm-rust/crates/auth/Cargo.toml new file mode 100644 index 00000000000..128a05c1a25 --- /dev/null +++ b/litellm-rust/crates/auth/Cargo.toml @@ -0,0 +1,15 @@ +[package] +name = "litellm-auth" +version = "0.1.0" +edition.workspace = true +license.workspace = true +repository.workspace = true + +[dependencies] +serde.workspace = true +subtle.workspace = true +thiserror.workspace = true +veil.workspace = true + +[dev-dependencies] +tokio.workspace = true diff --git a/litellm-rust/crates/core/src/auth/credential.rs b/litellm-rust/crates/auth/src/credential.rs similarity index 91% rename from litellm-rust/crates/core/src/auth/credential.rs rename to litellm-rust/crates/auth/src/credential.rs index c64d331b877..6721eb67a35 100644 --- a/litellm-rust/crates/core/src/auth/credential.rs +++ b/litellm-rust/crates/auth/src/credential.rs @@ -5,7 +5,7 @@ use std::sync::Arc; use veil::Redact; -use crate::AuthError; +use crate::Error; use super::{ResolvedCredential, SecretValue, TokenProviderHandle}; @@ -48,7 +48,7 @@ pub enum CredentialLookup { } pub type CredentialLookupFuture<'a> = - Pin> + Send + 'a>>; + Pin> + Send + 'a>>; pub trait CredentialResolver: std::fmt::Debug + Send + Sync { fn resolve<'a>(&'a self, reference: &'a CredentialRef) -> CredentialLookupFuture<'a>; @@ -62,7 +62,7 @@ impl CredentialResolverHandle { Self(resolver) } - pub async fn resolve(&self, reference: &CredentialRef) -> Result { + pub async fn resolve(&self, reference: &CredentialRef) -> Result { self.0.resolve(reference).await } } @@ -84,7 +84,7 @@ impl CredentialPlan { pub async fn resolve( &self, resolver: &CredentialResolverHandle, - ) -> Result { + ) -> Result { match self { Self::Static(CredentialRef::Explicit(secret)) => Ok( CredentialPlanResolution::Resolved(ResolvedCredential::Static(secret.clone())), @@ -103,7 +103,7 @@ impl CredentialPlan { Self::Caller(caller) => { let credential = caller.acquire().await?; if credential.secret().expose().is_empty() { - return Err(AuthError::EmptyCallerCredential); + return Err(Error::EmptyCallerCredential); } Ok(CredentialPlanResolution::Resolved(credential)) } @@ -119,8 +119,8 @@ mod tests { CredentialLookup, CredentialLookupFuture, CredentialPlan, CredentialPlanResolution, CredentialRef, CredentialResolver, CredentialResolverHandle, }; - use crate::AuthError; - use crate::auth::SecretValue; + use crate::Error; + use crate::SecretValue; #[derive(Debug)] struct HostResolver; @@ -164,7 +164,7 @@ mod tests { impl CredentialResolver for FailingResolver { fn resolve<'a>(&'a self, _reference: &'a CredentialRef) -> CredentialLookupFuture<'a> { - Box::pin(async { Err(AuthError::UnresolvedOidcReference) }) + Box::pin(async { Err(Error::UnresolvedOidcReference) }) } } @@ -178,6 +178,6 @@ mod tests { .await .expect_err("acquisition errors cannot become fallback"); - assert_eq!(error, AuthError::UnresolvedOidcReference); + assert_eq!(error, Error::UnresolvedOidcReference); } } diff --git a/litellm-rust/crates/core/src/auth/error.rs b/litellm-rust/crates/auth/src/error.rs similarity index 51% rename from litellm-rust/crates/core/src/auth/error.rs rename to litellm-rust/crates/auth/src/error.rs index e7027c0df10..9189eb44676 100644 --- a/litellm-rust/crates/core/src/auth/error.rs +++ b/litellm-rust/crates/auth/src/error.rs @@ -1,15 +1,77 @@ -use thiserror::Error; +use thiserror::Error as ThisError; -#[derive(Clone, Debug, Error, PartialEq, Eq)] -pub enum AuthError { - #[error("invalid authentication configuration: {0}")] - Configuration(#[from] AuthConfigurationError), +#[derive(Clone, Debug, ThisError, PartialEq, Eq)] +pub enum Error { + #[error("invalid authentication configuration: credential header already exists")] + ExistingCredentialHeader, + #[error( + "invalid authentication configuration: credential plan is not allowed by the provider auth policy" + )] + DisallowedCredentialPlan, + #[error("invalid authentication configuration: credential cannot be empty")] + EmptyCredential, + #[error("invalid authentication configuration: invalid Azure credential selector")] + InvalidAzureSelector, + #[error( + "invalid authentication configuration: ClientSecretCredential requires tenant_id, client_id, and client_secret" + )] + MissingClientSecretFields, + #[error("invalid authentication configuration: WorkloadIdentityCredential requires tenant_id")] + MissingWorkloadTenant, + #[error("invalid authentication configuration: WorkloadIdentityCredential requires client_id")] + MissingWorkloadClient, + #[error( + "invalid authentication configuration: WorkloadIdentityCredential requires azure_federated_token_file" + )] + MissingWorkloadTokenFile, + #[error( + "invalid authentication configuration: credential reference requires a host credential resolver" + )] + MissingHostResolver, + #[error( + "invalid authentication configuration: caller credential plan requires provider-specific inputs" + )] + MissingCallerInputs, + #[error("invalid authentication configuration: credential header {0} already exists")] + DuplicateHeader(&'static str), + #[error("invalid authentication configuration: {0} must be a string or null")] + InvalidFieldType(String), + #[error("invalid authentication configuration: unsupported OIDC reference")] + UnsupportedOidcReference, + #[error("invalid authentication configuration: {0} cannot be empty")] + EmptyReference(String), + #[error("invalid authentication configuration: Azure credential initialization failed: {0}")] + AzureCredentialInitialization(String), + #[error( + "invalid authentication configuration: Azure authority must be an HTTPS origin without credentials, query, or fragment" + )] + InvalidAzureAuthority, + #[error( + "invalid authentication configuration: request-controlled Azure auth inputs cannot be combined with host credentials" + )] + MixedAzureCredentialSources, + #[error( + "invalid authentication configuration: request-controlled Azure credential references are not allowed" + )] + RequestAzureCredentialReference, + #[error( + "invalid authentication configuration: host credentials cannot be sent to a request-controlled Azure endpoint" + )] + RequestAzureCredentialDestination, + #[error( + "invalid authentication configuration: credentials cannot be sent to a request-controlled Vertex AI endpoint" + )] + RequestVertexCredentialDestination, + #[error( + "invalid authentication configuration: request-controlled Vertex credentials must use the canonical Google OAuth token endpoint" + )] + RequestVertexTokenEndpoint, #[error("credential acquisition failed: {0}")] AzureTokenAcquisition(String), #[error("credential acquisition failed: Vertex AI credentials: {0}")] VertexTokenAcquisition(String), #[error("credential acquisition failed: {}", .0.iter().map(ToString::to_string).collect::>().join("; "))] - CredentialChain(Vec), + CredentialChain(Vec), #[error("credential caller failed: credential caller returned an empty credential")] EmptyCallerCredential, #[error("credential caller failed: Azure AD token provider returned an empty token")] @@ -27,102 +89,42 @@ pub enum AuthError { provider: &'static str, environment_variable: &'static str, }, - #[error("{0}")] - MissingCredential(#[from] MissingCredential), - #[error("{0}")] - Aws(#[from] AwsAuthError), - #[error("invalid authentication header")] - InvalidHeader, -} - -#[derive(Clone, Debug, Error, PartialEq, Eq)] -pub enum AuthConfigurationError { - #[error("credential header already exists")] - ExistingCredentialHeader, - #[error("credential plan is not allowed by the provider auth policy")] - DisallowedCredentialPlan, - #[error("credential cannot be empty")] - EmptyCredential, - #[error("invalid Azure credential selector")] - InvalidAzureSelector, - #[error("ClientSecretCredential requires tenant_id, client_id, and client_secret")] - MissingClientSecretFields, - #[error("WorkloadIdentityCredential requires tenant_id")] - MissingWorkloadTenant, - #[error("WorkloadIdentityCredential requires client_id")] - MissingWorkloadClient, - #[error("WorkloadIdentityCredential requires azure_federated_token_file")] - MissingWorkloadTokenFile, - #[error("credential reference requires a host credential resolver")] - MissingHostResolver, - #[error("caller credential plan requires provider-specific inputs")] - MissingCallerInputs, - #[error("credential header {0} already exists")] - DuplicateHeader(&'static str), - #[error("{0} must be a string or null")] - InvalidFieldType(String), - #[error("unsupported OIDC reference")] - UnsupportedOidcReference, - #[error("{0} cannot be empty")] - EmptyReference(String), - #[error("Azure credential initialization failed: {0}")] - AzureCredentialInitialization(String), - #[error("Azure authority must be an HTTPS origin without credentials, query, or fragment")] - InvalidAzureAuthority, - #[error("request-controlled Azure auth inputs cannot be combined with host credentials")] - MixedAzureCredentialSources, - #[error("request-controlled Azure credential references are not allowed")] - RequestAzureCredentialReference, - #[error("host credentials cannot be sent to a request-controlled Azure endpoint")] - RequestAzureCredentialDestination, - #[error("credentials cannot be sent to a request-controlled Vertex AI endpoint")] - RequestVertexCredentialDestination, - #[error( - "request-controlled Vertex credentials must use the canonical Google OAuth token endpoint" - )] - RequestVertexTokenEndpoint, -} - -#[derive(Clone, Debug, Error, PartialEq, Eq)] -pub enum MissingCredential { #[error( "Missing Anthropic API Key - Set `api_key` or the ANTHROPIC_API_KEY environment variable" )] - AnthropicApiKey, + MissingAnthropicApiKey, #[error("Missing Azure API Key - Set `api_key` or the AZURE_API_KEY environment variable")] - AzureApiKey, + MissingAzureApiKey, #[error( "Missing Azure API Base - Set `api_base` or the AZURE_API_BASE environment variable. Expected format: https://.services.ai.azure.com/anthropic" )] - AzureApiBase, + MissingAzureApiBase, #[error( "Missing OpenAI API Key - a realtime call is being made but no key was passed via params or the OPENAI_API_KEY environment variable" )] - OpenAiRealtimeApiKey, + MissingOpenAiRealtimeApiKey, #[error( "Missing OpenAI API Key - a Responses WebSocket call is being made but no key was passed via params or the OPENAI_API_KEY environment variable" )] - OpenAiResponsesApiKey, -} - -#[derive(Clone, Debug, Error, PartialEq, Eq)] -pub enum AwsAuthError { + MissingOpenAiResponsesApiKey, #[error("AWS profile credentials failed: {0}")] - Profile(String), + AwsProfile(String), #[error("AWS default credentials failed: {0}")] - DefaultChain(String), + AwsDefaultChain(String), #[error("AWS role credentials failed: {0}")] - AssumeRole(String), + AwsAssumeRole(String), #[error("AWS web identity credentials failed: {0}")] - WebIdentity(String), + AwsWebIdentity(String), #[error("AWS web identity expiration was invalid: {0}")] - WebIdentityExpiration(String), + AwsWebIdentityExpiration(String), #[error("AWS signing parameters failed: {0}")] - SigningParameters(String), + AwsSigningParameters(String), #[error("AWS signable request failed: {0}")] - SignableRequest(String), + AwsSignableRequest(String), #[error("AWS request signing failed: {0}")] - Signing(String), + AwsSigning(String), #[error("AWS web identity response had no credentials")] - MissingWebIdentityCredentials, + AwsMissingWebIdentityCredentials, + #[error("invalid authentication header")] + InvalidHeader, } diff --git a/litellm-rust/crates/core/src/auth/http.rs b/litellm-rust/crates/auth/src/http.rs similarity index 84% rename from litellm-rust/crates/core/src/auth/http.rs rename to litellm-rust/crates/auth/src/http.rs index 83931311550..7d20991d838 100644 --- a/litellm-rust/crates/core/src/auth/http.rs +++ b/litellm-rust/crates/auth/src/http.rs @@ -1,5 +1,4 @@ -use crate::AuthError; -use crate::auth::error::AuthConfigurationError; +use crate::Error; #[derive(Clone, Copy, Debug, PartialEq, Eq)] pub enum CredentialPlacement { @@ -16,23 +15,19 @@ impl CredentialPlacement { } } -pub(crate) fn apply_credential( +pub fn apply_credential( headers: Vec<(String, String)>, credential: &str, placement: CredentialPlacement, -) -> Result, AuthError> { +) -> Result, Error> { if credential.trim().is_empty() { - return Err(AuthError::Configuration( - AuthConfigurationError::EmptyCredential, - )); + return Err(Error::EmptyCredential); } if headers .iter() .any(|(name, _)| name.eq_ignore_ascii_case(placement.header_name())) { - return Err(AuthError::Configuration( - AuthConfigurationError::DuplicateHeader(placement.header_name()), - )); + return Err(Error::DuplicateHeader(placement.header_name())); } let value = match placement { CredentialPlacement::Bearer => format!("Bearer {credential}"), diff --git a/litellm-rust/crates/auth/src/lib.rs b/litellm-rust/crates/auth/src/lib.rs new file mode 100644 index 00000000000..7a24d2acf70 --- /dev/null +++ b/litellm-rust/crates/auth/src/lib.rs @@ -0,0 +1,56 @@ +mod credential; +mod error; +pub mod http; +mod policy; +mod secret; +mod token; + +use serde::{Deserialize, Serialize}; + +#[derive(Clone, Copy, Debug, Default, Deserialize, Eq, Hash, PartialEq, Serialize)] +#[serde(rename_all = "snake_case")] +pub enum InputSource { + Request, + #[default] + Deployment, + Environment, +} + +#[derive(Clone, Debug, Default, Eq, PartialEq)] +pub struct Sourced { + value: T, + source: InputSource, +} + +impl Sourced { + pub fn new(value: T, source: InputSource) -> Self { + Self { value, source } + } + + pub fn value(&self) -> &T { + &self.value + } + + pub fn source(&self) -> InputSource { + self.source + } + + pub fn into_value(self) -> T { + self.value + } + + pub fn map(self, map: impl FnOnce(T) -> U) -> Sourced { + Sourced::new(map(self.value), self.source) + } +} + +pub use credential::{ + CredentialFileRef, CredentialLookup, CredentialLookupFuture, CredentialPlan, + CredentialPlanResolution, CredentialRef, CredentialResolver, CredentialResolverHandle, + credential_default_fields, credential_index, +}; +pub use error::Error; +pub use http::{CredentialPlacement, RequestAuth}; +pub use policy::{CredentialPlanKind, CredentialRule, ExistingHeaderBehavior, ProviderAuthPolicy}; +pub use secret::SecretValue; +pub use token::{ResolvedCredential, TokenFuture, TokenProvider, TokenProviderHandle}; diff --git a/litellm-rust/crates/core/src/auth/policy.rs b/litellm-rust/crates/auth/src/policy.rs similarity index 82% rename from litellm-rust/crates/core/src/auth/policy.rs rename to litellm-rust/crates/auth/src/policy.rs index b796dedf0d8..4a1f5eeecf9 100644 --- a/litellm-rust/crates/core/src/auth/policy.rs +++ b/litellm-rust/crates/auth/src/policy.rs @@ -1,5 +1,4 @@ -use crate::AuthError; -use crate::auth::error::AuthConfigurationError; +use crate::Error; use super::http::apply_credential; use super::{CredentialPlacement, ResolvedCredential}; @@ -46,22 +45,18 @@ impl ProviderAuthPolicy { headers: Vec<(String, String)>, kind: CredentialPlanKind, credential: &ResolvedCredential, - ) -> Result, AuthError> { + ) -> Result, Error> { if self.has_existing_credential(&headers) { return match self.existing_header_behavior { ExistingHeaderBehavior::Preserve => Ok(headers), - ExistingHeaderBehavior::Reject => Err(AuthError::Configuration( - AuthConfigurationError::ExistingCredentialHeader, - )), + ExistingHeaderBehavior::Reject => Err(Error::ExistingCredentialHeader), }; } - let rule = - self.rules - .iter() - .find(|rule| rule.kind == kind) - .ok_or(AuthError::Configuration( - AuthConfigurationError::DisallowedCredentialPlan, - ))?; + let rule = self + .rules + .iter() + .find(|rule| rule.kind == kind) + .ok_or(Error::DisallowedCredentialPlan)?; apply_credential(headers, credential.secret().expose(), rule.placement) } } @@ -69,7 +64,7 @@ impl ProviderAuthPolicy { #[cfg(test)] mod tests { use super::{CredentialPlanKind, CredentialRule, ExistingHeaderBehavior, ProviderAuthPolicy}; - use crate::auth::{CredentialPlacement, ResolvedCredential, SecretValue}; + use crate::{CredentialPlacement, ResolvedCredential, SecretValue}; const RULES: &[CredentialRule] = &[CredentialRule { kind: CredentialPlanKind::Static, diff --git a/litellm-rust/crates/core/src/auth/secret.rs b/litellm-rust/crates/auth/src/secret.rs similarity index 100% rename from litellm-rust/crates/core/src/auth/secret.rs rename to litellm-rust/crates/auth/src/secret.rs diff --git a/litellm-rust/crates/core/src/auth/token.rs b/litellm-rust/crates/auth/src/token.rs similarity index 83% rename from litellm-rust/crates/core/src/auth/token.rs rename to litellm-rust/crates/auth/src/token.rs index cfc6b8f0d6b..94da5f259fb 100644 --- a/litellm-rust/crates/core/src/auth/token.rs +++ b/litellm-rust/crates/auth/src/token.rs @@ -5,7 +5,7 @@ use std::time::SystemTime; use veil::Redact; -use crate::AuthError; +use crate::Error; use super::secret::SecretValue; @@ -27,7 +27,7 @@ impl ResolvedCredential { } pub type TokenFuture<'a> = - Pin> + Send + 'a>>; + Pin> + Send + 'a>>; pub trait TokenProvider: std::fmt::Debug + Send + Sync { fn acquire(&self) -> TokenFuture<'_>; @@ -41,7 +41,7 @@ impl TokenProviderHandle { Self(caller) } - pub async fn acquire(&self) -> Result { + pub async fn acquire(&self) -> Result { self.0.acquire().await } } diff --git a/litellm-rust/crates/core/Cargo.toml b/litellm-rust/crates/core/Cargo.toml index 2b3b4c6a03f..19f76f85df7 100644 --- a/litellm-rust/crates/core/Cargo.toml +++ b/litellm-rust/crates/core/Cargo.toml @@ -10,10 +10,11 @@ autotests = false bytes.workspace = true futures-util.workspace = true base64.workspace = true -azure_core.workspace = true -azure_identity.workspace = true data-url = "0.3.2" -gcp_auth.workspace = true +litellm-auth.workspace = true +litellm-auth-aws = { workspace = true, optional = true } +litellm-auth-azure.workspace = true +litellm-auth-gcp.workspace = true moka.workspace = true mime_guess = "2.0.5" rand.workspace = true @@ -23,7 +24,6 @@ rustls-native-certs.workspace = true serde.workspace = true serde_json.workspace = true serde_path_to_error = "0.1" -strum.workspace = true subtle.workspace = true tokio = { workspace = true, features = ["sync"] } tokio-tungstenite.workspace = true @@ -31,23 +31,10 @@ thiserror.workspace = true sha2.workspace = true url.workspace = true veil.workspace = true -aws-config = { version = "1.9.0", default-features = false, features = ["rustls", "rt-tokio"], optional = true } -aws-credential-types = { version = "1.3.0", features = ["hardcoded-credentials"], optional = true } -aws-sdk-sts = { version = "1.108.0", default-features = false, features = ["rustls", "rt-tokio"], optional = true } -aws-sigv4 = { version = "1.5.1", optional = true } -aws-types = { version = "1.4.0", optional = true } -aws-smithy-runtime-api = { version = "1.13.0", optional = true } - [features] default = [] -bedrock-auth = [ - "dep:aws-config", - "dep:aws-credential-types", - "dep:aws-sdk-sts", - "dep:aws-sigv4", - "dep:aws-types", - "dep:aws-smithy-runtime-api", -] +bedrock-auth = ["dep:litellm-auth-aws"] +observability = ["dep:tracing-subscriber"] [dev-dependencies] rstest.workspace = true diff --git a/litellm-rust/crates/core/src/auth/mod.rs b/litellm-rust/crates/core/src/auth/mod.rs index 2940a983fb9..390928dffd7 100644 --- a/litellm-rust/crates/core/src/auth/mod.rs +++ b/litellm-rust/crates/core/src/auth/mod.rs @@ -1,57 +1 @@ -mod credential; -pub mod error; -pub(crate) mod vertex; -pub use error::AuthError; -pub(crate) mod http; -mod policy; -mod secret; -mod token; - -use serde::{Deserialize, Serialize}; - -#[derive(Clone, Copy, Debug, Default, Deserialize, Eq, Hash, PartialEq, Serialize)] -#[serde(rename_all = "snake_case")] -pub enum InputSource { - Request, - #[default] - Deployment, - Environment, -} - -#[derive(Clone, Debug, Default, Eq, PartialEq)] -pub struct Sourced { - value: T, - source: InputSource, -} - -impl Sourced { - pub fn new(value: T, source: InputSource) -> Self { - Self { value, source } - } - - pub fn value(&self) -> &T { - &self.value - } - - pub fn source(&self) -> InputSource { - self.source - } - - pub fn into_value(self) -> T { - self.value - } - - pub fn map(self, map: impl FnOnce(T) -> U) -> Sourced { - Sourced::new(map(self.value), self.source) - } -} - -pub use credential::{ - CredentialFileRef, CredentialLookup, CredentialLookupFuture, CredentialPlan, - CredentialPlanResolution, CredentialRef, CredentialResolver, CredentialResolverHandle, - credential_default_fields, credential_index, -}; -pub use http::{CredentialPlacement, RequestAuth}; -pub use policy::{CredentialPlanKind, CredentialRule, ExistingHeaderBehavior, ProviderAuthPolicy}; -pub use secret::SecretValue; -pub use token::{ResolvedCredential, TokenFuture, TokenProvider, TokenProviderHandle}; +pub use litellm_auth::*; diff --git a/litellm-rust/crates/core/src/lib.rs b/litellm-rust/crates/core/src/lib.rs index 89c80b4551a..ab5955ac255 100644 --- a/litellm-rust/crates/core/src/lib.rs +++ b/litellm-rust/crates/core/src/lib.rs @@ -15,5 +15,5 @@ pub mod providers; pub mod responses; mod url_utils; -pub use auth::AuthError; +pub use auth::Error as AuthError; pub use error::Error; diff --git a/litellm-rust/crates/core/src/llms/azure_ai/ocr/cohere_parse_transformation.rs b/litellm-rust/crates/core/src/llms/azure_ai/ocr/cohere_parse_transformation.rs index 7481ce18e00..262ba31d246 100644 --- a/litellm-rust/crates/core/src/llms/azure_ai/ocr/cohere_parse_transformation.rs +++ b/litellm-rust/crates/core/src/llms/azure_ai/ocr/cohere_parse_transformation.rs @@ -7,8 +7,8 @@ use crate::ocr::document::{inline_remote_document, validate_inline_document}; use crate::ocr::error::{OcrError, OcrRequestError, OcrResponseError}; use crate::ocr::prepare::{credential_env, transform_request_body}; use crate::ocr::types::{LiteLLMOcrRequest, LiteLLMOcrResponse}; -use crate::providers::azure_ai::auth::AzureAuthInputs; use crate::url_utils::ApiUrl; +use litellm_auth_azure::AzureAuthInputs; const AZURE_AI_API_BASE_ENV: &str = "AZURE_AI_API_BASE"; diff --git a/litellm-rust/crates/core/src/llms/azure_ai/ocr/common_utils.rs b/litellm-rust/crates/core/src/llms/azure_ai/ocr/common_utils.rs index ef11d35da7f..f95593baed3 100644 --- a/litellm-rust/crates/core/src/llms/azure_ai/ocr/common_utils.rs +++ b/litellm-rust/crates/core/src/llms/azure_ai/ocr/common_utils.rs @@ -1,11 +1,10 @@ use std::sync::OnceLock; use crate::Error; -use crate::auth::error::AuthConfigurationError; use crate::auth::{InputSource, Sourced}; use crate::ocr::error::OcrError; use crate::ocr::types::OcrConnection; -use crate::providers::azure_ai::auth::{AzureAuthInputs, AzureAuthService}; +use litellm_auth_azure::{AzureAuthInputs, AzureAuthService}; pub(super) async fn resolve_entra( config: &AzureAuthInputs, @@ -38,10 +37,7 @@ pub(super) fn validate_destination( && connection.api_base_source == InputSource::Request && credential_source != InputSource::Request { - return Err(Error::from(crate::AuthError::Configuration( - AuthConfigurationError::RequestAzureCredentialDestination, - )) - .into()); + return Err(Error::from(crate::AuthError::RequestAzureCredentialDestination).into()); } Ok(()) } diff --git a/litellm-rust/crates/core/src/llms/azure_ai/ocr/document_intelligence/transformation.rs b/litellm-rust/crates/core/src/llms/azure_ai/ocr/document_intelligence/transformation.rs index e708c81ef9c..ec26cda0771 100644 --- a/litellm-rust/crates/core/src/llms/azure_ai/ocr/document_intelligence/transformation.rs +++ b/litellm-rust/crates/core/src/llms/azure_ai/ocr/document_intelligence/transformation.rs @@ -621,8 +621,8 @@ use crate::ocr::OcrClient; use crate::ocr::error::{OcrError, OcrRequestError, OcrResponseError}; use crate::ocr::prepare::{credential_env, transform_request_body}; use crate::ocr::types::{LiteLLMOcrRequest, LiteLLMOcrResponse, OcrConnection, OcrResponseFormat}; -use crate::providers::azure_ai::auth::AzureAuthInputs; use crate::url_utils::ApiUrl; +use litellm_auth_azure::AzureAuthInputs; const AZURE_DI_API_KEY_ENV: &str = "AZURE_DOCUMENT_INTELLIGENCE_API_KEY"; const AZURE_DI_ENDPOINT_ENV: &str = "AZURE_DOCUMENT_INTELLIGENCE_ENDPOINT"; diff --git a/litellm-rust/crates/core/src/llms/azure_ai/ocr/transformation.rs b/litellm-rust/crates/core/src/llms/azure_ai/ocr/transformation.rs index 47481905946..df2be7ddb33 100644 --- a/litellm-rust/crates/core/src/llms/azure_ai/ocr/transformation.rs +++ b/litellm-rust/crates/core/src/llms/azure_ai/ocr/transformation.rs @@ -9,8 +9,8 @@ use crate::ocr::document::{inline_remote_document, validate_inline_document}; use crate::ocr::error::{OcrError, OcrRequestError, OcrResponseError}; use crate::ocr::prepare::{credential_env, transform_request_body}; use crate::ocr::types::{LiteLLMOcrRequest, LiteLLMOcrResponse, OcrConnection}; -use crate::providers::azure_ai::auth::AzureAuthInputs; use crate::url_utils::ApiUrl; +use litellm_auth_azure::AzureAuthInputs; const AZURE_AI_API_KEY_ENV: &str = "AZURE_AI_API_KEY"; const AZURE_AI_API_BASE_ENV: &str = "AZURE_AI_API_BASE"; diff --git a/litellm-rust/crates/core/src/llms/vertex_ai/ocr/common_utils.rs b/litellm-rust/crates/core/src/llms/vertex_ai/ocr/common_utils.rs index 9daa28a1e5e..d2e499232b1 100644 --- a/litellm-rust/crates/core/src/llms/vertex_ai/ocr/common_utils.rs +++ b/litellm-rust/crates/core/src/llms/vertex_ai/ocr/common_utils.rs @@ -1,15 +1,11 @@ use crate::Error; use crate::auth::InputSource; -use crate::auth::error::AuthConfigurationError; use crate::ocr::error::OcrError; use crate::ocr::types::OcrConnection; pub(super) fn validate_destination(connection: &OcrConnection) -> Result<(), OcrError> { if connection.api_base.is_some() && connection.api_base_source == InputSource::Request { - return Err(Error::from(crate::AuthError::Configuration( - AuthConfigurationError::RequestVertexCredentialDestination, - )) - .into()); + return Err(Error::from(crate::AuthError::RequestVertexCredentialDestination).into()); } Ok(()) } diff --git a/litellm-rust/crates/core/src/llms/vertex_ai/ocr/deepseek_transformation.rs b/litellm-rust/crates/core/src/llms/vertex_ai/ocr/deepseek_transformation.rs index 348fbb9d178..e186a609f7b 100644 --- a/litellm-rust/crates/core/src/llms/vertex_ai/ocr/deepseek_transformation.rs +++ b/litellm-rust/crates/core/src/llms/vertex_ai/ocr/deepseek_transformation.rs @@ -213,7 +213,6 @@ pub(crate) use mapping::{transform_ocr_request, transform_ocr_response}; use super::common_utils::validate_destination; use crate::Error; -use crate::auth::vertex::{self, VertexConfig}; use crate::llms::base_llm::ocr::transformation::BaseOcrConfig; use crate::ocr::OcrClient; use crate::ocr::error::{OcrError, OcrRequestError, OcrResponseError}; @@ -222,6 +221,7 @@ use crate::ocr::prepare::{ }; use crate::ocr::types::{LiteLLMOcrRequest, LiteLLMOcrResponse}; use crate::url_utils::ApiUrl; +use litellm_auth_gcp::{self as vertex, VertexConfig}; const DEFAULT_API_BASE: &str = "https://aiplatform.googleapis.com"; const MODEL_NAMESPACE: &str = "deepseek-ai"; const DEFAULT_LOCATION: &str = "us-central1"; diff --git a/litellm-rust/crates/core/src/llms/vertex_ai/ocr/transformation.rs b/litellm-rust/crates/core/src/llms/vertex_ai/ocr/transformation.rs index 8487eb1daa7..711ba9b887c 100644 --- a/litellm-rust/crates/core/src/llms/vertex_ai/ocr/transformation.rs +++ b/litellm-rust/crates/core/src/llms/vertex_ai/ocr/transformation.rs @@ -1,6 +1,5 @@ use super::common_utils::validate_destination; use crate::Error; -use crate::auth::vertex::{self, VertexConfig}; use crate::llms::base_llm::ocr::transformation::BaseOcrConfig; use crate::llms::mistral::ocr::MistralOcrResponse; use crate::llms::mistral::ocr::transformation::MistralOCRConfig; @@ -10,6 +9,7 @@ use crate::ocr::error::{OcrError, OcrRequestError, OcrResponseError}; use crate::ocr::prepare::{credential_env, transform_request_body}; use crate::ocr::types::{LiteLLMOcrRequest, LiteLLMOcrResponse}; use crate::url_utils::ApiUrl; +use litellm_auth_gcp::{self as vertex, VertexConfig}; const DEFAULT_LOCATION: &str = "us-central1"; #[derive(Clone, Debug, Default)] diff --git a/litellm-rust/crates/core/src/ocr/client.rs b/litellm-rust/crates/core/src/ocr/client.rs index 041c5c6fd21..8dca960a271 100644 --- a/litellm-rust/crates/core/src/ocr/client.rs +++ b/litellm-rust/crates/core/src/ocr/client.rs @@ -8,10 +8,10 @@ use super::error::{OcrError, OcrResponseError}; use super::types::{LiteLLMOcrRequest, LiteLLMOcrResponse}; use super::wire::{DecodedOcrResponse, decode_response}; use crate::Error; -use crate::auth::vertex::VertexAuth; use crate::constants::OCR_CONNECT_TIMEOUT_SECS; use crate::error::TransportError; use crate::media::MediaFetcher; +use litellm_auth_gcp::VertexAuth; #[derive(Clone)] pub struct OcrClient { diff --git a/litellm-rust/crates/core/src/ocr/handler.rs b/litellm-rust/crates/core/src/ocr/handler.rs index 96dfa19d7cc..64baecd5473 100644 --- a/litellm-rust/crates/core/src/ocr/handler.rs +++ b/litellm-rust/crates/core/src/ocr/handler.rs @@ -66,13 +66,9 @@ impl PreparedOcrCall { let http = match request.config { OcrConfigKind::Cohere => CohereParseConfig.prepare_request(&request, &client).await?, OcrConfigKind::Mistral => MistralOCRConfig.prepare_request(&request, &client).await?, - OcrConfigKind::AzureAi => { - AzureAIOCRConfig::default() - .prepare_request(&request, &client) - .await? - } + OcrConfigKind::AzureAi => AzureAIOCRConfig.prepare_request(&request, &client).await?, OcrConfigKind::AzureCohere => { - AzureAICohereParseConfig::default() + AzureAICohereParseConfig .prepare_request(&request, &client) .await? } @@ -91,11 +87,7 @@ impl PreparedOcrCall { .prepare_request(&request, &client) .await? } - OcrConfigKind::VertexAi => { - VertexAIOCRConfig::default() - .prepare_request(&request, &client) - .await? - } + OcrConfigKind::VertexAi => VertexAIOCRConfig.prepare_request(&request, &client).await?, OcrConfigKind::VertexDeepSeek => { VertexAIDeepSeekOCRConfig .prepare_request(&request, &client) @@ -130,12 +122,12 @@ impl PreparedOcrCall { .await?, ), OcrConfigKind::AzureAi => OcrProviderData::AzureAi( - AzureAIOCRConfig::default() + AzureAIOCRConfig .read_response(&self.client, response, &url, &headers, &self.request) .await?, ), OcrConfigKind::AzureCohere => OcrProviderData::AzureCohere( - AzureAICohereParseConfig::default() + AzureAICohereParseConfig .read_response(&self.client, response, &url, &headers, &self.request) .await?, ), @@ -155,7 +147,7 @@ impl PreparedOcrCall { .await?, ), OcrConfigKind::VertexAi => OcrProviderData::VertexAi( - VertexAIOCRConfig::default() + VertexAIOCRConfig .read_response(&self.client, response, &url, &headers, &self.request) .await?, ), @@ -217,12 +209,11 @@ impl OcrProviderResponse { decoded.native, ), OcrProviderData::AzureAi(decoded) => ( - AzureAIOCRConfig::default().transform_ocr_response(&self.request, decoded.data)?, + AzureAIOCRConfig.transform_ocr_response(&self.request, decoded.data)?, decoded.native, ), OcrProviderData::AzureCohere(decoded) => ( - AzureAICohereParseConfig::default() - .transform_ocr_response(&self.request, decoded.data)?, + AzureAICohereParseConfig.transform_ocr_response(&self.request, decoded.data)?, decoded.native, ), OcrProviderData::AzureDocumentIntelligence(decoded) => ( @@ -239,7 +230,7 @@ impl OcrProviderResponse { decoded.native, ), OcrProviderData::VertexAi(decoded) => ( - VertexAIOCRConfig::default().transform_ocr_response(&self.request, decoded.data)?, + VertexAIOCRConfig.transform_ocr_response(&self.request, decoded.data)?, decoded.native, ), OcrProviderData::VertexDeepSeek(decoded) => ( diff --git a/litellm-rust/crates/core/src/ocr/provider_config.rs b/litellm-rust/crates/core/src/ocr/provider_config.rs index bc43f38826b..bd87c8ab1f3 100644 --- a/litellm-rust/crates/core/src/ocr/provider_config.rs +++ b/litellm-rust/crates/core/src/ocr/provider_config.rs @@ -40,16 +40,14 @@ impl OcrConfigKind { match self { Self::Cohere => CohereParseConfig.get_supported_ocr_params(model), Self::Mistral => MistralOCRConfig.get_supported_ocr_params(model), - Self::AzureAi => AzureAIOCRConfig::default().get_supported_ocr_params(model), - Self::AzureCohere => { - AzureAICohereParseConfig::default().get_supported_ocr_params(model) - } + Self::AzureAi => AzureAIOCRConfig.get_supported_ocr_params(model), + Self::AzureCohere => AzureAICohereParseConfig.get_supported_ocr_params(model), Self::AzureDocumentIntelligence => { AzureDocumentIntelligenceOCRConfig.get_supported_ocr_params(model) } Self::ReductoLegacy => ReductoParseLegacyConfig.get_supported_ocr_params(model), Self::ReductoV3 => ReductoParseV3Config.get_supported_ocr_params(model), - Self::VertexAi => VertexAIOCRConfig::default().get_supported_ocr_params(model), + Self::VertexAi => VertexAIOCRConfig.get_supported_ocr_params(model), Self::VertexDeepSeek => VertexAIDeepSeekOCRConfig.get_supported_ocr_params(model), } } diff --git a/litellm-rust/crates/core/src/providers/anthropic/messages/transformation.rs b/litellm-rust/crates/core/src/providers/anthropic/messages/transformation.rs index fe96155bbbb..9e8cfd3df5c 100644 --- a/litellm-rust/crates/core/src/providers/anthropic/messages/transformation.rs +++ b/litellm-rust/crates/core/src/providers/anthropic/messages/transformation.rs @@ -1,4 +1,3 @@ -use crate::auth::error::MissingCredential; use crate::error::Error; use crate::messages::transformation::{AnthropicMessagesProviderConfig, MessagesAuthStrategy}; @@ -22,7 +21,7 @@ pub fn resolve_anthropic_api_key( non_empty(api_key) .map(str::to_string) .or_else(|| env_lookup(ANTHROPIC_API_KEY_ENV).filter(|value| !value.trim().is_empty())) - .ok_or_else(|| Error::from(crate::AuthError::from(MissingCredential::AnthropicApiKey))) + .ok_or_else(|| Error::from(crate::AuthError::MissingAnthropicApiKey)) } pub fn complete_anthropic_url( diff --git a/litellm-rust/crates/core/src/providers/azure_ai/auth/mod.rs b/litellm-rust/crates/core/src/providers/azure_ai/auth/mod.rs deleted file mode 100644 index 33d007c1945..00000000000 --- a/litellm-rust/crates/core/src/providers/azure_ai/auth/mod.rs +++ /dev/null @@ -1,7 +0,0 @@ -mod credential_provider_cache; -mod native; -mod resolve; -mod types; - -pub(crate) use resolve::AzureAuthService; -pub(crate) use types::AzureAuthInputs; diff --git a/litellm-rust/crates/core/src/providers/azure_ai/messages/transformation.rs b/litellm-rust/crates/core/src/providers/azure_ai/messages/transformation.rs index 175f3d20d94..c534e20d9f7 100644 --- a/litellm-rust/crates/core/src/providers/azure_ai/messages/transformation.rs +++ b/litellm-rust/crates/core/src/providers/azure_ai/messages/transformation.rs @@ -1,4 +1,3 @@ -use crate::auth::error::MissingCredential; use crate::error::Error; use crate::messages::transformation::{AnthropicMessagesProviderConfig, MessagesAuthStrategy}; use crate::messages::types::{ @@ -33,7 +32,7 @@ pub fn resolve_azure_api_key( non_empty(api_key) .map(str::to_string) .or_else(|| env_lookup(AZURE_API_KEY_ENV).filter(|value| !value.trim().is_empty())) - .ok_or_else(|| Error::from(crate::AuthError::from(MissingCredential::AzureApiKey))) + .ok_or_else(|| Error::from(crate::AuthError::MissingAzureApiKey)) } pub fn complete_azure_anthropic_url( @@ -43,7 +42,7 @@ pub fn complete_azure_anthropic_url( let api_base = non_empty(api_base) .map(str::to_string) .or_else(|| env_lookup(AZURE_API_BASE_ENV).filter(|value| !value.trim().is_empty())) - .ok_or_else(|| Error::from(crate::AuthError::from(MissingCredential::AzureApiBase)))?; + .ok_or_else(|| Error::from(crate::AuthError::MissingAzureApiBase))?; let api_base = api_base.trim_end_matches('/'); diff --git a/litellm-rust/crates/core/src/providers/azure_ai/mod.rs b/litellm-rust/crates/core/src/providers/azure_ai/mod.rs index 4f41d1d6abb..ba63992f3cb 100644 --- a/litellm-rust/crates/core/src/providers/azure_ai/mod.rs +++ b/litellm-rust/crates/core/src/providers/azure_ai/mod.rs @@ -1,2 +1 @@ -pub(crate) mod auth; pub mod messages; diff --git a/litellm-rust/crates/core/src/providers/bedrock/aws_base.rs b/litellm-rust/crates/core/src/providers/bedrock/aws_base.rs index e5e52bfce95..b51cef7545c 100644 --- a/litellm-rust/crates/core/src/providers/bedrock/aws_base.rs +++ b/litellm-rust/crates/core/src/providers/bedrock/aws_base.rs @@ -1,930 +1 @@ -use std::collections::BTreeMap; -use std::sync::{Mutex, OnceLock}; -use std::time::Duration; -use std::time::{SystemTime, UNIX_EPOCH}; - -use crate::caching::in_memory_cache::InMemoryCache; -use crate::error::Error; -use aws_credential_types::Credentials; -use aws_credential_types::provider::ProvideCredentials; -use aws_sigv4::http_request::{ - SignableBody, SignableRequest, SigningParams, SigningSettings, sign, -}; -use aws_sigv4::sign::v4; -use aws_smithy_runtime_api::client::identity::Identity; -use serde_json::{Map, Value}; -use sha2::{Digest, Sha256}; - -use super::constants::{ - AWS_ACCESS_KEY_ID, AWS_EXTERNAL_ID, AWS_PROFILE_NAME, AWS_REGION, AWS_REGION_NAME, - AWS_ROLE_ARN, AWS_ROLE_NAME, AWS_SECRET_ACCESS_KEY, AWS_SESSION_NAME, AWS_SESSION_TOKEN, - AWS_SIGNED_HEADER_NAMES, AWS_STS_ENDPOINT, AWS_WEB_IDENTITY_TOKEN, AWS_WEB_IDENTITY_TOKEN_FILE, - BEDROCK_SERVICE, DEFAULT_BEDROCK_REGION, DEFAULT_SESSION_NAME_PREFIX, - SIGV4_COMPUTED_HEADER_NAMES, -}; - -const STATIC_CREDENTIALS_TTL: Duration = Duration::from_secs(3600 - 60); -const AMBIENT_CREDENTIALS_TTL: Duration = Duration::from_secs(600); - -static IAM_CREDENTIALS_CACHE: OnceLock>> = OnceLock::new(); - -fn credential_cache_ttl(flow: &AwsAuthFlow) -> Option { - match flow { - AwsAuthFlow::StaticKeys { .. } => Some(STATIC_CREDENTIALS_TTL), - AwsAuthFlow::DefaultChain => Some(AMBIENT_CREDENTIALS_TTL), - AwsAuthFlow::WebIdentity { .. } - | AwsAuthFlow::AssumeRole { .. } - | AwsAuthFlow::Profile { .. } - | AwsAuthFlow::SessionToken { .. } => None, - } -} - -#[derive(Clone, Debug, Default, PartialEq, Eq)] -pub struct AwsAuthConfig { - pub access_key_id: Option, - pub secret_access_key: Option, - pub session_token: Option, - pub region_name: Option, - pub session_name: Option, - pub profile_name: Option, - pub role_name: Option, - pub web_identity_token: Option, - pub sts_endpoint: Option, - pub external_id: Option, -} - -impl AwsAuthConfig { - fn with_environment(self, env_lookup: &(dyn Fn(&str) -> Option + Sync)) -> Self { - Self { - access_key_id: self.access_key_id.or_else(|| env_lookup(AWS_ACCESS_KEY_ID)), - secret_access_key: self - .secret_access_key - .or_else(|| env_lookup(AWS_SECRET_ACCESS_KEY)), - session_token: self.session_token.or_else(|| env_lookup(AWS_SESSION_TOKEN)), - region_name: self.region_name.or_else(|| env_lookup(AWS_REGION_NAME)), - session_name: self.session_name.or_else(|| env_lookup(AWS_SESSION_NAME)), - profile_name: self.profile_name.or_else(|| env_lookup(AWS_PROFILE_NAME)), - role_name: self.role_name.or_else(|| env_lookup(AWS_ROLE_NAME)), - web_identity_token: self - .web_identity_token - .or_else(|| env_lookup(AWS_WEB_IDENTITY_TOKEN)), - sts_endpoint: self.sts_endpoint.or_else(|| env_lookup(AWS_STS_ENDPOINT)), - external_id: self.external_id.or_else(|| env_lookup(AWS_EXTERNAL_ID)), - } - } -} - -#[derive(Clone, Debug, PartialEq, Eq)] -pub enum AwsAuthFlow { - WebIdentity { - token: String, - role: String, - session_name: String, - }, - AssumeRole { - role: String, - session_name: Option, - }, - Profile { - name: String, - }, - SessionToken { - access_key_id: String, - secret_access_key: String, - session_token: String, - }, - StaticKeys { - access_key_id: String, - secret_access_key: String, - region_name: String, - }, - DefaultChain, -} - -fn cache_key(config: &AwsAuthConfig, flow: &AwsAuthFlow) -> String { - let mut hasher = Sha256::new(); - hasher.update(format!("{config:?}:{flow:?}")); - format!("{:x}", hasher.finalize()) -} - -fn get_cached_credentials(key: &str) -> Option { - let cache = IAM_CREDENTIALS_CACHE.get_or_init(|| Mutex::new(InMemoryCache::default())); - let mut entries = cache.lock().ok()?; - entries.get_cache(key) -} - -fn set_cached_credentials(key: String, credentials: Credentials, ttl: Duration) { - let cache = IAM_CREDENTIALS_CACHE.get_or_init(|| Mutex::new(InMemoryCache::default())); - if let Ok(mut entries) = cache.lock() { - entries.set_cache(key, credentials, Some(ttl)); - } -} - -fn role_identity(arn: &str) -> Option<(&str, &str, &str)> { - let mut parts = arn.splitn(6, ':'); - let ("arn", partition, _, _, account, resource) = ( - parts.next()?, - parts.next()?, - parts.next()?, - parts.next()?, - parts.next()?, - parts.next()?, - ) else { - return None; - }; - let role = if let Some(role) = resource.strip_prefix("role/") { - role.rsplit('/').next()? - } else { - resource.strip_prefix("assumed-role/")?.split('/').next()? - }; - Some((partition, account, role)) -} - -fn same_role_arns(target: &str, caller: &str) -> bool { - role_identity(target) == role_identity(caller) -} - -pub fn classify_auth( - config: AwsAuthConfig, - env_lookup: &(dyn Fn(&str) -> Option + Sync), -) -> AwsAuthFlow { - let config = config.with_environment(env_lookup); - if let (Some(token), Some(role), Some(session_name)) = ( - config.web_identity_token.clone(), - config.role_name.clone(), - config.session_name.clone(), - ) { - return AwsAuthFlow::WebIdentity { - token, - role, - session_name, - }; - } - if let Some(role) = config.role_name.clone() { - return AwsAuthFlow::AssumeRole { - role, - session_name: config.session_name.clone(), - }; - } - if let Some(name) = config.profile_name { - return AwsAuthFlow::Profile { name }; - } - if let (Some(access_key_id), Some(secret_access_key), Some(session_token)) = ( - config.access_key_id.clone(), - config.secret_access_key.clone(), - config.session_token, - ) { - return AwsAuthFlow::SessionToken { - access_key_id, - secret_access_key, - session_token, - }; - } - if let (Some(access_key_id), Some(secret_access_key), Some(region_name)) = ( - config.access_key_id, - config.secret_access_key, - config.region_name, - ) { - return AwsAuthFlow::StaticKeys { - access_key_id, - secret_access_key, - region_name, - }; - } - AwsAuthFlow::DefaultChain -} - -pub async fn resolve_credentials( - config: AwsAuthConfig, - env_lookup: &(dyn Fn(&str) -> Option + Sync), -) -> Result { - let resolved = config.clone().with_environment(env_lookup); - let flow = classify_auth(config, env_lookup); - match flow { - AwsAuthFlow::SessionToken { - access_key_id, - secret_access_key, - session_token, - } => Ok(Credentials::new( - access_key_id, - secret_access_key, - Some(session_token), - None, - "litellm-static-session", - )), - AwsAuthFlow::StaticKeys { - access_key_id, - secret_access_key, - region_name, - } => { - let flow = AwsAuthFlow::StaticKeys { - access_key_id: access_key_id.clone(), - secret_access_key: secret_access_key.clone(), - region_name, - }; - let key = cache_key(&resolved, &flow); - if let Some(credentials) = get_cached_credentials(&key) { - return Ok(credentials); - } - let credentials = Credentials::new( - access_key_id, - secret_access_key, - None, - None, - "litellm-static", - ); - set_cached_credentials( - key, - credentials.clone(), - credential_cache_ttl(&flow).unwrap_or(STATIC_CREDENTIALS_TTL), - ); - Ok(credentials) - } - AwsAuthFlow::Profile { name } => { - let provider = aws_config::profile::ProfileFileCredentialsProvider::builder() - .profile_name(name) - .build(); - provider - .provide_credentials() - .await - .map_err(|error| Error::Auth(format!("AWS profile credentials failed: {error}"))) - } - AwsAuthFlow::AssumeRole { role, session_name } => { - if is_already_running_as_role(&role, &resolved).await? { - let ambient_flow = AwsAuthFlow::DefaultChain; - let key = cache_key(&resolved, &ambient_flow); - if let Some(credentials) = get_cached_credentials(&key) { - return Ok(credentials); - } - let provider = - aws_config::default_provider::credentials::DefaultCredentialsChain::builder() - .build() - .await; - let credentials = provider.provide_credentials().await.map_err(|error| { - Error::Auth(format!("AWS default credentials failed: {error}")) - })?; - set_cached_credentials( - key, - credentials.clone(), - credential_cache_ttl(&ambient_flow).unwrap_or(AMBIENT_CREDENTIALS_TTL), - ); - return Ok(credentials); - } - let mut loader = aws_config::defaults(aws_config::BehaviorVersion::latest()); - if let Some(region) = resolved.region_name.clone() { - loader = loader.region(aws_types::region::Region::new(region)); - } - if let Some(endpoint) = resolved.sts_endpoint.clone() { - loader = loader.endpoint_url(endpoint); - } - if let (Some(access_key_id), Some(secret_access_key)) = - (resolved.access_key_id, resolved.secret_access_key) - { - loader = loader.credentials_provider(Credentials::new( - access_key_id, - secret_access_key, - resolved.session_token, - None, - "litellm-role-source", - )); - } - let sdk_config = loader.load().await; - let builder = aws_config::sts::AssumeRoleProvider::builder(role); - let builder = match session_name { - Some(name) => builder.session_name(name), - None => builder.session_name(default_session_name()), - }; - let builder = match resolved.external_id { - Some(id) => builder.external_id(id), - None => builder, - }; - let provider = builder.configure(&sdk_config).build().await; - provider - .provide_credentials() - .await - .map_err(|error| Error::Auth(format!("AWS role credentials failed: {error}"))) - } - AwsAuthFlow::WebIdentity { - token, - role, - session_name, - } => { - let mut loader = aws_config::defaults(aws_config::BehaviorVersion::latest()); - if let Some(region) = resolved.region_name { - loader = loader.region(aws_types::region::Region::new(region)); - } - if let Some(endpoint) = resolved.sts_endpoint { - loader = loader.endpoint_url(endpoint); - } - let sdk_config = loader.load().await; - let client = aws_sdk_sts::Client::new(&sdk_config); - let response = client - .assume_role_with_web_identity() - .role_arn(role) - .role_session_name(session_name) - .web_identity_token(token) - .send() - .await - .map_err(|error| { - Error::Auth(format!("AWS web identity credentials failed: {error}")) - })?; - let credentials = response.credentials().ok_or_else(|| { - Error::Auth("AWS web identity response had no credentials".to_string()) - })?; - let expiration = SystemTime::try_from(*credentials.expiration()).map_err(|error| { - Error::Auth(format!("AWS web identity expiration was invalid: {error}")) - })?; - Ok(Credentials::new( - credentials.access_key_id(), - credentials.secret_access_key(), - Some(credentials.session_token().to_string()), - Some(expiration), - "litellm-web-identity", - )) - } - AwsAuthFlow::DefaultChain => { - let key = cache_key(&resolved, &AwsAuthFlow::DefaultChain); - if let Some(credentials) = get_cached_credentials(&key) { - return Ok(credentials); - } - let provider = - aws_config::default_provider::credentials::DefaultCredentialsChain::builder() - .build() - .await; - let credentials = provider - .provide_credentials() - .await - .map_err(|error| Error::Auth(format!("AWS default credentials failed: {error}")))?; - set_cached_credentials( - key, - credentials.clone(), - credential_cache_ttl(&AwsAuthFlow::DefaultChain).unwrap_or(AMBIENT_CREDENTIALS_TTL), - ); - Ok(credentials) - } - } -} - -async fn is_already_running_as_role(role: &str, config: &AwsAuthConfig) -> Result { - if role_identity(role).is_none() { - return Ok(false); - } - if let (Ok(current_role), Ok(token_file)) = ( - std::env::var(AWS_ROLE_ARN), - std::env::var(AWS_WEB_IDENTITY_TOKEN_FILE), - ) && !token_file.is_empty() - { - return Ok(same_role_arns(role, ¤t_role)); - } - - let mut loader = aws_config::defaults(aws_config::BehaviorVersion::latest()); - if let Some(region) = config.region_name.clone() { - loader = loader.region(aws_types::region::Region::new(region)); - } - if let Some(endpoint) = config.sts_endpoint.clone() { - loader = loader.endpoint_url(endpoint); - } - let sdk_config = loader.load().await; - let response = match aws_sdk_sts::Client::new(&sdk_config) - .get_caller_identity() - .send() - .await - { - Ok(response) => response, - Err(_) => return Ok(false), - }; - Ok(response - .arn() - .is_some_and(|caller| same_role_arns(role, caller))) -} - -fn default_session_name() -> String { - let seconds = SystemTime::now() - .duration_since(UNIX_EPOCH) - .map_or(0, |duration| duration.as_secs()); - format!("{DEFAULT_SESSION_NAME_PREFIX}-{seconds}") -} - -/// The subset of `headers` SigV4 should cover. -/// -/// Python signs only these and reattaches the rest afterwards, so a forwarded -/// client header cannot change the canonical request and invalidate the -/// signature. Signing everything instead makes the request 403 on a header the -/// caller supplied, on a deployment that works on the Python path. -pub fn aws_signature_headers(headers: &BTreeMap) -> BTreeMap { - headers - .iter() - .filter(|(name, _)| { - let name = name.to_ascii_lowercase(); - AWS_SIGNED_HEADER_NAMES.contains(&name.as_str()) - || name.starts_with("x-amz-") - || name.starts_with("x-amzn-") - }) - .map(|(name, value)| (name.clone(), value.clone())) - .collect() -} - -/// Whether the signer produces `name` itself. -/// -/// Python's reattach loop skips these, so a caller-supplied copy never reaches -/// the wire next to the computed one. -pub fn is_sigv4_computed_header(name: &str) -> bool { - SIGV4_COMPUTED_HEADER_NAMES.contains(&name.to_ascii_lowercase().as_str()) -} - -pub fn sign_bedrock_post( - url: &str, - body: &[u8], - headers: &BTreeMap, - region: &str, - credentials: &Credentials, - signing_time: SystemTime, -) -> Result, Error> { - let identity: Identity = credentials.clone().into(); - let params = v4::SigningParams::builder() - .identity(&identity) - .region(region) - .name(BEDROCK_SERVICE) - .time(signing_time) - .settings(SigningSettings::default()) - .build() - .map(SigningParams::from) - .map_err(|error| Error::Auth(format!("AWS signing parameters failed: {error}")))?; - let header_refs = headers - .iter() - .map(|(name, value)| (name.as_str(), value.as_str())); - let request = SignableRequest::new("POST", url, header_refs, SignableBody::Bytes(body)) - .map_err(|error| Error::Auth(format!("AWS signable request failed: {error}")))?; - let (instructions, _) = sign(request, ¶ms) - .map_err(|error| Error::Auth(format!("AWS request signing failed: {error}")))? - .into_parts(); - Ok(instructions - .headers() - .map(|(name, value)| { - let normalized_name = match name { - "authorization" => "Authorization", - "x-amz-date" => "X-Amz-Date", - "x-amz-security-token" => "X-Amz-Security-Token", - _ => name, - }; - (normalized_name.to_string(), value.to_string()) - }) - .collect()) -} - -/// Model-id and region parsing shared by every Bedrock route. -pub fn bedrock_model_id_and_region(model: &str) -> (String, Option) { - let mut stripped = model; - for prefix in ["bedrock/converse/", "bedrock/", "converse/"] { - if let Some(value) = stripped.strip_prefix(prefix) { - stripped = value; - break; - } - } - let mut region = None; - if let Some((candidate, remainder)) = stripped.split_once('/') - && is_bedrock_region(candidate) - { - region = Some(candidate.to_string()); - stripped = remainder; - } - for prefix in ["nova-2/", "nova/"] { - if let Some(value) = stripped.strip_prefix(prefix) { - stripped = value; - break; - } - } - if region.is_none() { - // Python splits the whole ARN and takes field 3, the region. Stripping - // `arn:` first shifts every field down one, so the region is field 2 - // here; field 3 is the account id. - region = stripped - .strip_prefix("arn:") - .and_then(|value| value.split(':').nth(2)) - .filter(|value| !value.is_empty()) - .map(str::to_string); - } - (stripped.to_string(), region) -} - -fn is_bedrock_region(value: &str) -> bool { - value.len() > 3 - && value.contains('-') - && value - .chars() - .all(|char| char.is_ascii_alphanumeric() || char == '-') -} - -pub fn resolve_bedrock_region( - model_region: Option<&str>, - optional_params: &Map, - env_lookup: &dyn Fn(&str) -> Option, -) -> String { - if let Some(region) = optional_params - .get("aws_region_name") - .and_then(Value::as_str) - { - return region.to_string(); - } - if let Some(region) = model_region { - return region.to_string(); - } - env_lookup(AWS_REGION_NAME) - .or_else(|| env_lookup(AWS_REGION)) - .unwrap_or_else(|| DEFAULT_BEDROCK_REGION.to_string()) -} - -pub fn aws_auth_config( - optional_params: &Map, - env_lookup: &dyn Fn(&str) -> Option, -) -> AwsAuthConfig { - let value = |key: &str| { - optional_params - .get(key) - .and_then(Value::as_str) - .map(str::to_string) - }; - let env = |key: &str| env_lookup(key); - AwsAuthConfig { - access_key_id: value("aws_access_key_id").or_else(|| env("AWS_ACCESS_KEY_ID")), - secret_access_key: value("aws_secret_access_key").or_else(|| env("AWS_SECRET_ACCESS_KEY")), - session_token: value("aws_session_token").or_else(|| env("AWS_SESSION_TOKEN")), - region_name: value("aws_region_name").or_else(|| env(AWS_REGION_NAME)), - session_name: value("aws_session_name").or_else(|| env("AWS_SESSION_NAME")), - profile_name: value("aws_profile_name").or_else(|| env("AWS_PROFILE_NAME")), - role_name: value("aws_role_name").or_else(|| env("AWS_ROLE_NAME")), - web_identity_token: value("aws_web_identity_token") - .or_else(|| env("AWS_WEB_IDENTITY_TOKEN")), - sts_endpoint: value("aws_sts_endpoint").or_else(|| env("AWS_STS_ENDPOINT")), - external_id: value("aws_external_id").or_else(|| env("AWS_EXTERNAL_ID")), - } -} - -/// Credentials a host resolved through its own chain and handed down verbatim. -/// -/// A host with its own resolution (LiteLLM's Python `BaseAWSLLM`, which reads -/// profiles, STS and boto sessions) passes the result here so the core signs -/// with exactly those. Without this the core would re-derive from ambient -/// state, where an unrelated `AWS_ROLE_NAME` or `AWS_PROFILE_NAME` in the -/// environment outranks explicit keys in [`classify_auth`] and the two sides -/// would sign as different principals. -pub fn host_supplied_credentials(optional_params: &Map) -> Option { - let value = |key: &str| { - optional_params - .get(key) - .and_then(Value::as_str) - .map(str::trim) - .filter(|value| !value.is_empty()) - }; - let access_key_id = value("aws_access_key_id")?; - let secret_access_key = value("aws_secret_access_key")?; - Some(Credentials::new( - access_key_id, - secret_access_key, - value("aws_session_token").map(str::to_string), - None, - "litellm-host-supplied", - )) -} - -#[cfg(test)] -mod tests { - use super::*; - - fn no_env(_: &str) -> Option { - None - } - - fn parity_inputs() -> (String, Vec, BTreeMap) { - ( - "https://bedrock-runtime.us-east-1.amazonaws.com/model/amazon.titan-text-express-v1/invoke" - .to_string(), - br#"{"input":"hello"}"#.to_vec(), - BTreeMap::from([("Content-Type".to_string(), "application/json".to_string())]), - ) - } - - #[test] - fn reads_the_region_field_of_a_model_arn_not_the_account_id() { - // Python's `_get_aws_region_from_model_arn` splits the whole ARN and - // takes field 3. Stripping `arn:` first shifts every field down one, so - // the region is field 2 here. Taking field 3 after the strip returns - // the account id, which is not a region at all. - let (_, region) = bedrock_model_id_and_region( - "bedrock/arn:aws:bedrock:us-west-2:123456789012:foundation-model/anthropic.claude-v2", - ); - assert_eq!(region.as_deref(), Some("us-west-2")); - } - - #[test] - fn classification_preserves_python_precedence() { - let config = AwsAuthConfig { - access_key_id: Some("ak".into()), - secret_access_key: Some("sk".into()), - session_token: Some("token".into()), - region_name: Some("us-east-1".into()), - session_name: Some("session".into()), - profile_name: Some("profile".into()), - role_name: Some("role".into()), - web_identity_token: Some("oidc".into()), - ..Default::default() - }; - assert!(matches!( - classify_auth(config, &no_env), - AwsAuthFlow::WebIdentity { .. } - )); - } - - #[test] - fn classification_covers_fallthroughs() { - let env = |key: &str| match key { - AWS_PROFILE_NAME => Some("profile".into()), - _ => None, - }; - assert!(matches!( - classify_auth(AwsAuthConfig::default(), &env), - AwsAuthFlow::Profile { .. } - )); - assert!(matches!( - classify_auth( - AwsAuthConfig { - access_key_id: Some("ak".into()), - secret_access_key: Some("sk".into()), - session_token: Some("token".into()), - ..Default::default() - }, - &no_env - ), - AwsAuthFlow::SessionToken { .. } - )); - assert!(matches!( - classify_auth( - AwsAuthConfig { - access_key_id: Some("ak".into()), - secret_access_key: Some("sk".into()), - region_name: Some("us-east-1".into()), - ..Default::default() - }, - &no_env - ), - AwsAuthFlow::StaticKeys { .. } - )); - assert_eq!( - classify_auth(AwsAuthConfig::default(), &no_env), - AwsAuthFlow::DefaultChain - ); - } - - #[tokio::test] - async fn static_credentials_do_not_use_network() { - let credentials = resolve_credentials( - AwsAuthConfig { - access_key_id: Some("ak".into()), - secret_access_key: Some("sk".into()), - region_name: Some("us-east-1".into()), - ..Default::default() - }, - &no_env, - ) - .await - .expect("static credentials"); - assert_eq!(credentials.access_key_id(), "ak"); - assert_eq!(credentials.session_token(), None); - } - - #[test] - fn cache_policy_matches_python_flows() { - assert_eq!( - credential_cache_ttl(&AwsAuthFlow::StaticKeys { - access_key_id: "ak".into(), - secret_access_key: "sk".into(), - region_name: "us-east-1".into(), - }), - Some(STATIC_CREDENTIALS_TTL) - ); - assert_eq!( - credential_cache_ttl(&AwsAuthFlow::DefaultChain), - Some(AMBIENT_CREDENTIALS_TTL) - ); - assert_eq!( - credential_cache_ttl(&AwsAuthFlow::SessionToken { - access_key_id: "ak".into(), - secret_access_key: "sk".into(), - session_token: "token".into(), - }), - None - ); - assert_eq!( - credential_cache_ttl(&AwsAuthFlow::Profile { - name: "profile".into() - }), - None - ); - assert_eq!( - credential_cache_ttl(&AwsAuthFlow::AssumeRole { - role: "arn:aws:iam::123456789012:role/demo".into(), - session_name: None, - }), - None - ); - assert_eq!( - credential_cache_ttl(&AwsAuthFlow::WebIdentity { - token: "token".into(), - role: "arn:aws:iam::123456789012:role/demo".into(), - session_name: "session".into(), - }), - None - ); - } - - #[test] - fn cache_round_trip_preserves_credentials() { - let key = format!("cache-test-{}", std::process::id()); - let credentials = Credentials::new("cache-ak", "cache-sk", None, None, "test"); - set_cached_credentials(key.clone(), credentials.clone(), STATIC_CREDENTIALS_TTL); - assert_eq!( - get_cached_credentials(&key).map(|value| value.access_key_id().to_string()), - Some("cache-ak".to_string()) - ); - } - - #[test] - fn same_role_comparison_matches_partition_account_and_role() { - assert!(same_role_arns( - "arn:aws:iam::123456789012:role/path/demo", - "arn:aws:sts::123456789012:assumed-role/demo/session" - )); - assert!(!same_role_arns( - "arn:aws:iam::123456789012:role/demo", - "arn:aws:iam::999999999999:role/demo" - )); - assert!(!same_role_arns( - "arn:aws:iam::123456789012:role/demo", - "arn:aws-cn:iam::123456789012:role/demo" - )); - assert!(!same_role_arns( - "arn:aws:iam::123456789012:user/demo", - "arn:aws:iam::123456789012:role/demo" - )); - } - - #[test] - fn a_forwarded_client_header_is_not_folded_into_the_signature() { - // Python signs only the AWS header set, so a header a caller forwarded - // cannot change the canonical request. Signing it instead makes the - // request 403 the moment anything on the wire rewrites or drops it. - let (url, body, mut headers) = parity_inputs(); - headers.insert("x-request-id".to_string(), "abc-123".to_string()); - headers.insert("Accept-Encoding".to_string(), "gzip".to_string()); - headers.insert("x-amzn-trace-id".to_string(), "Root=1-abc".to_string()); - let signable = aws_signature_headers(&headers); - - assert!(!signable.contains_key("x-request-id")); - assert!(!signable.contains_key("Accept-Encoding")); - // The AWS-prefixed one is genuinely part of the signature. - assert!(signable.contains_key("x-amzn-trace-id")); - assert!(signable.contains_key("Content-Type")); - - let credentials = Credentials::new( - "AKIDEXAMPLE", - "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY", - None, - None, - "test", - ); - let signed = sign_bedrock_post( - &url, - &body, - &signable, - "us-east-1", - &credentials, - SystemTime::UNIX_EPOCH, - ) - .expect("signs"); - let authorization = signed - .get("Authorization") - .expect("carries an authorization header"); - assert!( - !authorization.contains("x-request-id"), - "forwarded header reached SignedHeaders: {authorization}" - ); - assert!( - !authorization.contains("accept-encoding"), - "forwarded header reached SignedHeaders: {authorization}" - ); - } - - #[test] - fn signing_matches_botocore_golden_vector() { - let (url, body, headers) = parity_inputs(); - let credentials = Credentials::new( - "AKIDEXAMPLE", - "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY", - Some("session-token".to_string()), - None, - "test", - ); - let signed = sign_bedrock_post( - &url, - &body, - &headers, - "us-east-1", - &credentials, - UNIX_EPOCH + std::time::Duration::from_secs(1_704_164_645), - ) - .expect("golden signature"); - assert_eq!( - signed.get("X-Amz-Date").map(String::as_str), - Some("20240102T030405Z") - ); - assert_eq!( - signed.get("X-Amz-Security-Token").map(String::as_str), - Some("session-token") - ); - assert_eq!( - signed.get("Authorization").map(String::as_str), - Some( - "AWS4-HMAC-SHA256 Credential=AKIDEXAMPLE/20240102/us-east-1/bedrock/aws4_request, SignedHeaders=content-type;host;x-amz-date;x-amz-security-token, Signature=55c027ef47527d3ad63f1735f9d099efdbc99f296ff914bd94e727e24ec0e464" - ) - ); - } - - #[test] - fn signing_without_session_token_omits_security_header() { - let (url, body, headers) = parity_inputs(); - let credentials = Credentials::new( - "AKIDEXAMPLE", - "wJalrXUtnFEMI/K7MDENG+bPxRfiCYEXAMPLEKEY", - None, - None, - "test", - ); - let signed = sign_bedrock_post( - &url, - &body, - &headers, - "us-east-1", - &credentials, - UNIX_EPOCH + std::time::Duration::from_secs(1_704_164_645), - ) - .expect("signature"); - assert!(!signed.contains_key("X-Amz-Security-Token")); - } - - #[ignore] - #[tokio::test] - async fn live_bedrock_invoke_model_returns_200() -> Result<(), Box> { - let access_key_id = std::env::var("AWS_BEDROCK_TEST_ACCESS_KEY_ID")?; - let secret_access_key = std::env::var("AWS_BEDROCK_TEST_SECRET_ACCESS_KEY")?; - let body = br#"{"anthropic_version":"bedrock-2023-05-31","max_tokens":1,"messages":[{"role":"user","content":[{"type":"text","text":"ping"}]}]}"#.to_vec(); - let headers = - BTreeMap::from([("Content-Type".to_string(), "application/json".to_string())]); - let credentials = resolve_credentials( - AwsAuthConfig { - access_key_id: Some(access_key_id), - secret_access_key: Some(secret_access_key), - region_name: Some("us-west-2".to_string()), - ..Default::default() - }, - &no_env, - ) - .await?; - let client = reqwest::Client::new(); - let mut failures = Vec::new(); - - for region in ["us-west-2", "us-east-1"] { - let url = format!( - "https://bedrock-runtime.{region}.amazonaws.com/model/us.anthropic.claude-opus-4-8/invoke" - ); - let signed_headers = sign_bedrock_post( - &url, - &body, - &headers, - region, - &credentials, - SystemTime::now(), - )?; - let mut request = client.post(&url).body(body.clone()); - for (name, value) in &headers { - request = request.header(name, value); - } - for (name, value) in signed_headers { - request = request.header(name, value); - } - let response = request.send().await?; - let status = response.status(); - let response_body = response.text().await?; - let snippet: String = response_body.chars().take(240).collect(); - println!("region={region} status={status} response={snippet}"); - if status == reqwest::StatusCode::OK { - return Ok(()); - } - failures.push(format!("{region}: {status} {snippet}")); - } - - panic!( - "no Bedrock region returned HTTP 200: {}", - failures.join("; ") - ); - } -} +pub use litellm_auth_aws::*; diff --git a/litellm-rust/crates/core/src/providers/bedrock/constants.rs b/litellm-rust/crates/core/src/providers/bedrock/constants.rs index be215cc9016..663f887c1fd 100644 --- a/litellm-rust/crates/core/src/providers/bedrock/constants.rs +++ b/litellm-rust/crates/core/src/providers/bedrock/constants.rs @@ -1,43 +1 @@ -pub const AWS_ACCESS_KEY_ID: &str = "AWS_ACCESS_KEY_ID"; -pub const AWS_SECRET_ACCESS_KEY: &str = "AWS_SECRET_ACCESS_KEY"; -pub const AWS_SESSION_TOKEN: &str = "AWS_SESSION_TOKEN"; -pub const AWS_REGION_NAME: &str = "AWS_REGION_NAME"; -pub const AWS_REGION: &str = "AWS_REGION"; -pub const AWS_SESSION_NAME: &str = "AWS_SESSION_NAME"; -pub const AWS_PROFILE_NAME: &str = "AWS_PROFILE_NAME"; -pub const AWS_ROLE_NAME: &str = "AWS_ROLE_NAME"; -pub const AWS_WEB_IDENTITY_TOKEN: &str = "AWS_WEB_IDENTITY_TOKEN"; -pub const AWS_ROLE_ARN: &str = "AWS_ROLE_ARN"; -pub const AWS_WEB_IDENTITY_TOKEN_FILE: &str = "AWS_WEB_IDENTITY_TOKEN_FILE"; -pub const AWS_STS_ENDPOINT: &str = "AWS_STS_ENDPOINT"; -pub const AWS_EXTERNAL_ID: &str = "AWS_EXTERNAL_ID"; -pub const AWS_BEARER_TOKEN_BEDROCK: &str = "AWS_BEARER_TOKEN_BEDROCK"; - -/// Headers SigV4 covers, beyond the `x-amz-` / `x-amzn-` prefixes. Mirrors -/// Python's `_filter_headers_for_aws_signature` allowlist. -pub const AWS_SIGNED_HEADER_NAMES: &[&str] = &[ - "host", - "content-type", - "date", - "x-amz-date", - "x-amz-security-token", - "x-amz-content-sha256", - "x-amz-algorithm", - "x-amz-credential", - "x-amz-signedheaders", - "x-amz-signature", -]; -/// Headers the signer emits itself. Mirrors Python's `SIGV4_COMPUTED_HEADERS`, -/// which the reattach loop skips so a caller's copy cannot ride alongside the -/// computed one. -pub const SIGV4_COMPUTED_HEADER_NAMES: &[&str] = &[ - "authorization", - "x-amz-date", - "x-amz-security-token", - "date", -]; -pub const BEDROCK_SERVICE: &str = "bedrock"; -pub const DEFAULT_SESSION_NAME_PREFIX: &str = "litellm-session"; -pub const DEFAULT_BEDROCK_REGION: &str = "us-west-2"; -pub const BEDROCK_RUNTIME_ENDPOINT_TEMPLATE: &str = - "https://bedrock-runtime.{region}.amazonaws.com"; +pub use litellm_auth_aws::constants::*; diff --git a/litellm-rust/crates/core/tests/vertex_ai_ocr.rs b/litellm-rust/crates/core/tests/vertex_ai_ocr.rs index 64ceb9942f1..4a37c40b472 100644 --- a/litellm-rust/crates/core/tests/vertex_ai_ocr.rs +++ b/litellm-rust/crates/core/tests/vertex_ai_ocr.rs @@ -122,7 +122,7 @@ async fn configs_build_complete_requests_and_share_mistral_normalization() { .prepare_request(&direct, &client) .await .unwrap(); - let vertex_http = VertexAIOCRConfig::default() + let vertex_http = VertexAIOCRConfig .prepare_request(&vertex, &client) .await .unwrap(); @@ -153,7 +153,7 @@ async fn configs_build_complete_requests_and_share_mistral_normalization() { .transform_ocr_response(&direct, serde_json::from_value(payload.clone()).unwrap()) .unwrap() .into_json(); - let vertex_response = VertexAIOCRConfig::default() + let vertex_response = VertexAIOCRConfig .transform_ocr_response(&vertex, serde_json::from_value(payload).unwrap()) .unwrap() .into_json();