liter_llm 1.6.4 → 1.7.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 5e435f10f7f7a4190e12335aed8fee67caeff9da56709ef23b56c20aa77b4c64
-  data.tar.gz: 674282900ad04eeef03753f41b9c25e1a8ebe7676881fc0ae9a101f4aaa2be30
+  metadata.gz: 336f60aa43a86dbbddd88a10084f24180cc2e02dda02294754e66203b7d2698d
+  data.tar.gz: '0816afd2387ade4af23e63e05f24d970da571a7f799f41399548e49ea42fa885'
 SHA512:
-  metadata.gz: acf1c4dff2e26e01aa11f1c1c2bba1b66d170cebf8ef11dafe69bf5b4b6c8b7cbe7083a451df2b215221fd539c8cdbc2b0fce65067516930c7f23a21848d0262
-  data.tar.gz: a2451cff7040de44bb9ff738811d103b3b54eaf397da918e4bde59cc044ec0cdcf6624d0f38bb2165963bae3dd9dc0e21fb46ee89311c4506094a7b39c90f0f6
+  metadata.gz: 777e16f5ebecd0e90576f0714fc1b7b8f6bf41e8b7ccb6f23734d581bd2909a0c99c2279a9edcd45c56f742116915c7b226cbfdda0054980cab2fbb3842c1f6a
+  data.tar.gz: 2008f0e64537acc3fa8578667dff3cc10c5fd45a008d40785dc74aa526113a285523e7eddc8b0a1c0839b7deceb904f7379d6a6accea4cdc1cb9ebdbe629b676

data/ext/liter_llm_rb/native/Cargo.lock

@@ -1875,9 +1875,9 @@ checksum = "92daf443525c4cce67b150400bc2316076100ce0b3686209eb8cf3c31612e6f0"
 
 [[package]]
 name = "liter-llm"
-version = "1.6.4"
+version = "1.7.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "82177d82ab01638c1567473be387c83492e6201bc59b399c12a254bde133d4ee"
+checksum = "0552ad2a855bd57c422933227a1d664ad66b1e773dd2244a14194500880ef544"
 dependencies = [
  "ahash 0.8.12",
  "async-trait",
@@ -1917,7 +1917,7 @@ dependencies = [
 
 [[package]]
 name = "liter-llm-rb"
-version = "1.6.4"
+version = "1.7.1"
 dependencies = [
  "futures",
  "liter-llm",

data/ext/liter_llm_rb/native/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "liter-llm-rb"
-version = "1.6.4"
+version = "1.7.1"
 edition = "2024"
 license = "MIT"
 description = "Universal LLM API client with Rust-powered polyglot bindings."
@@ -24,7 +24,7 @@ wasm-http = ["liter-llm/wasm-http"]
 
 [dependencies]
 futures = "0.3"
-liter-llm = { version = "1.6.4", features = ["native-http", "full"] }
+liter-llm = { version = "1.7.1", features = ["native-http", "full"] }
 magnus = "0.8"
 rb-sys = ">=0.9, <0.9.128"
 serde = { version = "1", features = ["derive"] }

data/ext/liter_llm_rb/src/lib.rs

@@ -1,5 +1,5 @@
 // This file is auto-generated by alef. DO NOT EDIT.
-// alef:hash:52effca2513e1c5f8a9a5b90a372d645388de4b92e4f822acc38d97502c322b6
+// alef:hash:fec9b44a56d9714b7b12931bb48e1e8f6d57ef2d0ccaad1f9ac6046ce01a89af
 // Re-generate with: alef generate
 #![allow(dead_code, unused_imports, unused_variables)]
 #![allow(
@@ -67,7 +67,7 @@ fn json_to_ruby(handle: &Ruby, val: serde_json::Value) -> magnus::Value {
 #[serde(default)]
 #[magnus::wrap(class = "LiterLlm::SystemMessage")]
 pub struct SystemMessage {
-    content: String,
+    content: UserContent,
     name: Option<String>,
 }
 
@@ -101,7 +101,10 @@ unsafe impl TryConvertOwned for SystemMessage {}
 
 impl Default for SystemMessage {
     fn default() -> Self {
-        liter_llm::types::SystemMessage::default().into()
+        Self {
+            content: Default::default(),
+            name: None,
+        }
     }
 }
 
@@ -114,7 +117,7 @@ impl SystemMessage {
         Ok(Self {
             content: kwargs
                 .get(ruby.to_symbol("content"))
-                .and_then(|v| String::try_convert(v).ok())
+                .and_then(|v| UserContent::try_convert(v).ok())
                 .unwrap_or_default(),
             name: kwargs
                 .get(ruby.to_symbol("name"))
@@ -122,18 +125,13 @@ impl SystemMessage {
         })
     }
 
-    fn content(&self) -> String {
+    fn content(&self) -> UserContent {
         self.content.clone()
     }
 
     fn name(&self) -> Option<String> {
         self.name.clone()
     }
-
-    #[allow(clippy::should_implement_trait)]
-    fn to_s(&self) -> String {
-        self.content.clone()
-    }
 }
 
 #[derive(Clone, Debug, serde::Serialize, serde::Deserialize)]
@@ -420,7 +418,7 @@ impl AudioContent {
 #[serde(default)]
 #[magnus::wrap(class = "LiterLlm::AssistantMessage")]
 pub struct AssistantMessage {
-    content: Option<String>,
+    content: Option<AssistantContent>,
     name: Option<String>,
     tool_calls: Option<Vec<ToolCall>>,
     refusal: Option<String>,
@@ -476,7 +474,7 @@ impl AssistantMessage {
         Ok(Self {
             content: kwargs
                 .get(ruby.to_symbol("content"))
-                .and_then(|v| String::try_convert(v).ok()),
+                .and_then(|v| AssistantContent::try_convert(v).ok()),
             name: kwargs
                 .get(ruby.to_symbol("name"))
                 .and_then(|v| String::try_convert(v).ok()),
@@ -492,7 +490,7 @@ impl AssistantMessage {
         })
     }
 
-    fn content(&self) -> Option<String> {
+    fn content(&self) -> Option<AssistantContent> {
         self.content.clone()
     }
 
@@ -512,9 +510,64 @@ impl AssistantMessage {
         self.function_call.clone()
     }
 
-    #[allow(clippy::should_implement_trait)]
-    fn to_s(&self) -> String {
-        self.content.clone().unwrap_or_default()
+    fn text(&self) -> Option<String> {
+        let core_self = liter_llm::types::AssistantMessage {
+            content: self.content.clone().map(Into::into),
+
+            name: self.name.clone(),
+
+            tool_calls: self.tool_calls.clone().map(|v| v.into_iter().map(Into::into).collect()),
+
+            refusal: self.refusal.clone(),
+
+            function_call: self.function_call.clone().map(Into::into),
+        };
+        core_self.text()
+    }
+
+    fn refusal_text(&self) -> Option<String> {
+        let core_self = liter_llm::types::AssistantMessage {
+            content: self.content.clone().map(Into::into),
+
+            name: self.name.clone(),
+
+            tool_calls: self.tool_calls.clone().map(|v| v.into_iter().map(Into::into).collect()),
+
+            refusal: self.refusal.clone(),
+
+            function_call: self.function_call.clone().map(Into::into),
+        };
+        core_self.refusal_text().map(|v| v.to_owned())
+    }
+
+    fn output_images(&self) -> Vec<ImageUrl> {
+        let core_self = liter_llm::types::AssistantMessage {
+            content: self.content.clone().map(Into::into),
+
+            name: self.name.clone(),
+
+            tool_calls: self.tool_calls.clone().map(|v| v.into_iter().map(Into::into).collect()),
+
+            refusal: self.refusal.clone(),
+
+            function_call: self.function_call.clone().map(Into::into),
+        };
+        core_self.output_images().into_iter().map(Into::into).collect()
+    }
+
+    fn output_audio(&self) -> Vec<AudioContent> {
+        let core_self = liter_llm::types::AssistantMessage {
+            content: self.content.clone().map(Into::into),
+
+            name: self.name.clone(),
+
+            tool_calls: self.tool_calls.clone().map(|v| v.into_iter().map(Into::into).collect()),
+
+            refusal: self.refusal.clone(),
+
+            function_call: self.function_call.clone().map(Into::into),
+        };
+        core_self.output_audio().into_iter().map(Into::into).collect()
     }
 }
 
@@ -1445,6 +1498,7 @@ pub struct ChatCompletionRequest {
     stream_options: Option<StreamOptions>,
     seed: Option<i64>,
     reasoning_effort: Option<ReasoningEffort>,
+    modalities: Option<Vec<Modality>>,
     extra_body: Option<String>,
 }
 
@@ -1498,6 +1552,7 @@ impl Default for ChatCompletionRequest {
             stream_options: None,
             seed: None,
             reasoning_effort: None,
+            modalities: None,
             extra_body: None,
         }
     }
@@ -1567,6 +1622,9 @@ impl ChatCompletionRequest {
             reasoning_effort: kwargs
                 .get(ruby.to_symbol("reasoning_effort"))
                 .and_then(|v| ReasoningEffort::try_convert(v).ok()),
+            modalities: kwargs
+                .get(ruby.to_symbol("modalities"))
+                .and_then(|v| <Vec<Modality>>::try_convert(v).ok()),
             extra_body: kwargs
                 .get(ruby.to_symbol("extra_body"))
                 .and_then(|v| String::try_convert(v).ok()),
@@ -1649,6 +1707,10 @@ impl ChatCompletionRequest {
         self.reasoning_effort.clone()
     }
 
+    fn modalities(&self) -> Option<Vec<Modality>> {
+        self.modalities.clone()
+    }
+
     fn extra_body(&self) -> Option<String> {
         self.extra_body.clone()
     }
@@ -2885,6 +2947,75 @@ impl Image {
     }
 }
 
+#[derive(Clone, Debug, serde::Serialize, serde::Deserialize)]
+#[serde(default)]
+#[magnus::wrap(class = "LiterLlm::DecodedDataUrl")]
+pub struct DecodedDataUrl {
+    mime: String,
+    data: Vec<u8>,
+}
+
+unsafe impl IntoValueFromNative for DecodedDataUrl {}
+
+impl magnus::TryConvert for DecodedDataUrl {
+    fn try_convert(val: magnus::Value) -> Result<Self, magnus::Error> {
+        if let Ok(r) = <&DecodedDataUrl as magnus::TryConvert>::try_convert(val) {
+            return Ok(r.clone());
+        }
+        let json_str: String = if let Ok(s) = <String as magnus::TryConvert>::try_convert(val) {
+            s
+        } else {
+            val.funcall::<_, _, String>("to_json", ()).map_err(|e| {
+                magnus::Error::new(
+                    unsafe { magnus::Ruby::get_unchecked() }.exception_type_error(),
+                    format!("no implicit conversion into DecodedDataUrl: {}", e),
+                )
+            })?
+        };
+        serde_json::from_str::<DecodedDataUrl>(&json_str).map_err(|e| {
+            magnus::Error::new(
+                unsafe { magnus::Ruby::get_unchecked() }.exception_type_error(),
+                format!("failed to deserialize DecodedDataUrl: {}", e),
+            )
+        })
+    }
+}
+
+unsafe impl TryConvertOwned for DecodedDataUrl {}
+
+impl Default for DecodedDataUrl {
+    fn default() -> Self {
+        liter_llm::image::DecodedDataUrl::default().into()
+    }
+}
+
+impl DecodedDataUrl {
+    fn new(args: &[magnus::Value]) -> Result<Self, magnus::Error> {
+        let ruby = unsafe { magnus::Ruby::get_unchecked() };
+        let args = magnus::scan_args::scan_args::<(), (Option<magnus::RHash>,), (), (), (), ()>(args)?;
+        let (kwargs_opt,) = args.optional;
+        let kwargs = kwargs_opt.unwrap_or_else(|| ruby.hash_new());
+        Ok(Self {
+            mime: kwargs
+                .get(ruby.to_symbol("mime"))
+                .and_then(|v| String::try_convert(v).ok())
+                .unwrap_or_default(),
+            data: kwargs
+                .get(ruby.to_symbol("data"))
+                .and_then(|v| <Vec<u8>>::try_convert(v).ok())
+                .unwrap_or_default(),
+        })
+    }
+
+    fn mime(&self) -> String {
+        self.mime.clone()
+    }
+
+    fn data(&self) -> Vec<u8> {
+        self.data.clone()
+    }
+}
+
 #[derive(Clone, Debug, serde::Serialize, serde::Deserialize)]
 #[serde(default)]
 #[magnus::wrap(class = "LiterLlm::CreateSpeechRequest")]
@@ -7710,6 +7841,122 @@ impl magnus::TryConvert for ImageDetail {
 unsafe impl IntoValueFromNative for ImageDetail {}
 unsafe impl TryConvertOwned for ImageDetail {}
 
+#[derive(Clone, Debug, serde::Serialize, serde::Deserialize)]
+#[serde(untagged)]
+pub enum AssistantContent {
+    Text(String),
+    Parts(Vec<AssistantPart>),
+}
+
+impl Default for AssistantContent {
+    fn default() -> Self {
+        Self::Text(Default::default())
+    }
+}
+
+impl magnus::IntoValue for AssistantContent {
+    fn into_value_with(self, handle: &Ruby) -> magnus::Value {
+        match serde_json::to_value(&self) {
+            Ok(v) => json_to_ruby(handle, v),
+            Err(_) => handle.qnil().into_value_with(handle),
+        }
+    }
+}
+
+impl magnus::TryConvert for AssistantContent {
+    fn try_convert(val: magnus::Value) -> Result<Self, magnus::Error> {
+        // For data enums with fields (e.g., PageAction), try to deserialize from JSON first.
+        // For unit enums or when passed as a string, fall back to string-based conversion.
+        let json_str: String = if let Ok(s) = <String as magnus::TryConvert>::try_convert(val) {
+            s
+        } else {
+            val.funcall::<_, _, String>("to_json", ()).map_err(|e| {
+                magnus::Error::new(
+                    unsafe { Ruby::get_unchecked() }.exception_type_error(),
+                    format!("no implicit conversion into AssistantContent: {}", e),
+                )
+            })?
+        };
+        // Try deserializing as JSON first (handles JSON strings like "\"markdown\"" or "{\"click\":{\"selector\":\"...\"}}\"")
+        // If that fails, try treating it as a plain string value and wrap in quotes
+        // If both fail, try as Custom variant (for untagged enum support)
+        serde_json::from_str(&json_str)
+            .or_else(|_| serde_json::from_str(&format!("\"{json_str}\"")))
+            .or_else(|_| {
+                // Try as a JSON string for Custom variant (untagged enums accept any remaining value)
+                match serde_json::to_value(&json_str) {
+                    Ok(val) => serde_json::from_value(val),
+                    Err(e) => Err(e),
+                }
+            })
+            .map_err(|e| magnus::Error::new(unsafe { Ruby::get_unchecked() }.exception_type_error(), e.to_string()))
+    }
+}
+
+unsafe impl IntoValueFromNative for AssistantContent {}
+unsafe impl TryConvertOwned for AssistantContent {}
+
+#[derive(Clone, Debug, serde::Serialize, serde::Deserialize)]
+#[serde(tag = "type")]
+#[serde(rename_all = "snake_case")]
+pub enum AssistantPart {
+    Text { text: String },
+    Refusal { refusal: String },
+    OutputImage { image_url: ImageUrl },
+    OutputAudio { audio: AudioContent },
+}
+
+impl Default for AssistantPart {
+    fn default() -> Self {
+        Self::Text {
+            text: Default::default(),
+        }
+    }
+}
+
+impl magnus::IntoValue for AssistantPart {
+    fn into_value_with(self, handle: &Ruby) -> magnus::Value {
+        match serde_json::to_value(&self) {
+            Ok(v) => json_to_ruby(handle, v),
+            Err(_) => handle.qnil().into_value_with(handle),
+        }
+    }
+}
+
+impl magnus::TryConvert for AssistantPart {
+    fn try_convert(val: magnus::Value) -> Result<Self, magnus::Error> {
+        // For data enums with fields (e.g., PageAction), try to deserialize from JSON first.
+        // For unit enums or when passed as a string, fall back to string-based conversion.
+        let json_str: String = if let Ok(s) = <String as magnus::TryConvert>::try_convert(val) {
+            s
+        } else {
+            val.funcall::<_, _, String>("to_json", ()).map_err(|e| {
+                magnus::Error::new(
+                    unsafe { Ruby::get_unchecked() }.exception_type_error(),
+                    format!("no implicit conversion into AssistantPart: {}", e),
+                )
+            })?
+        };
+        // Try deserializing as JSON first (handles JSON strings like "\"markdown\"" or "{\"click\":{\"selector\":\"...\"}}\"")
+        // If that fails, try treating it as a plain string value and wrap in quotes
+        // If both fail, try as Custom variant (for untagged enum support)
+        serde_json::from_str(&json_str)
+            .or_else(|_| serde_json::from_str(&format!("\"{json_str}\"")))
+            .or_else(|_| {
+                // Try as a JSON string for Custom variant (untagged enums accept any remaining value)
+                match serde_json::to_value(&json_str) {
+                    Ok(val) => serde_json::from_value(val),
+                    Err(e) => Err(e),
+                }
+            })
+            .map_err(|e| magnus::Error::new(unsafe { Ruby::get_unchecked() }.exception_type_error(), e.to_string()))
+    }
+}
+
+unsafe impl IntoValueFromNative for AssistantPart {}
+unsafe impl TryConvertOwned for AssistantPart {}
+impl AssistantPart {}
+
 #[derive(Clone, Copy, PartialEq, Eq, Debug, serde::Serialize, serde::Deserialize)]
 pub enum ToolType {
     #[serde(rename = "function")]
@@ -7964,6 +8211,51 @@ impl magnus::TryConvert for StopSequence {
 unsafe impl IntoValueFromNative for StopSequence {}
 unsafe impl TryConvertOwned for StopSequence {}
 
+#[derive(Clone, Copy, PartialEq, Eq, Debug, serde::Serialize, serde::Deserialize)]
+#[serde(rename_all = "snake_case")]
+pub enum Modality {
+    Text,
+    Audio,
+    Image,
+}
+
+impl Default for Modality {
+    fn default() -> Self {
+        Self::Text
+    }
+}
+
+impl magnus::IntoValue for Modality {
+    fn into_value_with(self, handle: &Ruby) -> magnus::Value {
+        let sym = match self {
+            Modality::Text => "text",
+            Modality::Audio => "audio",
+            Modality::Image => "image",
+        };
+        handle.to_symbol(sym).into_value_with(handle)
+    }
+}
+
+impl magnus::TryConvert for Modality {
+    fn try_convert(val: magnus::Value) -> Result<Self, magnus::Error> {
+        let s: String = magnus::TryConvert::try_convert(val)?;
+        // Accept the serde wire name (snake_case), the PascalCase Rust variant name,
+        // and a lowercase fallback so fixtures written in any of those styles work.
+        match s.as_str() {
+            "text" | "Text" => Ok(Modality::Text),
+            "audio" | "Audio" => Ok(Modality::Audio),
+            "image" | "Image" => Ok(Modality::Image),
+            other => Err(magnus::Error::new(
+                unsafe { Ruby::get_unchecked() }.exception_arg_error(),
+                format!("invalid Modality value: {other}"),
+            )),
+        }
+    }
+}
+
+unsafe impl IntoValueFromNative for Modality {}
+unsafe impl TryConvertOwned for Modality {}
+
 #[derive(Clone, Copy, PartialEq, Eq, Debug, serde::Serialize, serde::Deserialize)]
 #[serde(rename_all = "snake_case")]
 pub enum FinishReason {
@@ -8762,6 +9054,27 @@ fn create_client_from_json(json: String) -> Result<DefaultClient, Error> {
     })
 }
 
+fn encode_data_url(args: &[magnus::Value]) -> Result<String, Error> {
+    let args = magnus::scan_args::scan_args::<(Vec<u8>,), (Option<magnus::Value>,), (), (), (), ()>(args)?;
+    let (bytes,) = args.required;
+
+    let (mime,) = args.optional;
+
+    let mime: Option<String> = mime.and_then(|v| {
+        if v.is_nil() {
+            None
+        } else {
+            Some(String::try_convert(v).unwrap_or_default())
+        }
+    });
+
+    Ok(liter_llm::image::encode_data_url(&bytes, mime.as_deref()))
+}
+
+fn decode_data_url(url: String) -> Option<DecodedDataUrl> {
+    liter_llm::image::decode_data_url(&url).map(Into::into)
+}
+
 fn register_custom_provider(config: magnus::Value) -> Result<(), Error> {
     let config: liter_llm::CustomProviderConfig = {
         let binding_val: CustomProviderConfig = CustomProviderConfig::try_convert(config)
@@ -8880,6 +9193,11 @@ fn ensure_crypto_provider() -> () {
     liter_llm::ensure_crypto_provider()
 }
 
+#[cfg(all(feature = "native-http", target_os = "windows"))]
+fn ensure_crypto_provider() -> () {
+    liter_llm::ensure_crypto_provider()
+}
+
 fn chat_stream(engine: DefaultClient, req: ChatCompletionRequest) -> Result<magnus::Value, Error> {
     engine.chat_stream(req)
 }
@@ -8888,7 +9206,7 @@ fn chat_stream(engine: DefaultClient, req: ChatCompletionRequest) -> Result<magn
 impl From<SystemMessage> for liter_llm::types::SystemMessage {
     fn from(val: SystemMessage) -> Self {
         Self {
-            content: val.content,
+            content: val.content.into(),
             name: val.name,
         }
     }
@@ -8898,7 +9216,7 @@ impl From<SystemMessage> for liter_llm::types::SystemMessage {
 impl From<liter_llm::types::SystemMessage> for SystemMessage {
     fn from(val: liter_llm::types::SystemMessage) -> Self {
         Self {
-            content: val.content.to_string(),
+            content: val.content.into(),
             name: val.name.map(|v| v.to_string()),
         }
     }
@@ -8988,7 +9306,7 @@ impl From<liter_llm::types::AudioContent> for AudioContent {
 impl From<AssistantMessage> for liter_llm::types::AssistantMessage {
     fn from(val: AssistantMessage) -> Self {
         Self {
-            content: val.content,
+            content: val.content.map(Into::into),
             name: val.name,
             tool_calls: val.tool_calls.map(|v| v.into_iter().map(Into::into).collect()),
             refusal: val.refusal,
@@ -9001,7 +9319,7 @@ impl From<AssistantMessage> for liter_llm::types::AssistantMessage {
 impl From<liter_llm::types::AssistantMessage> for AssistantMessage {
     fn from(val: liter_llm::types::AssistantMessage) -> Self {
         Self {
-            content: val.content.map(|v| v.to_string()),
+            content: val.content.map(Into::into),
             name: val.name.map(|v| v.to_string()),
             tool_calls: val.tool_calls.map(|v| v.into_iter().map(Into::into).collect()),
             refusal: val.refusal.map(|v| v.to_string()),
@@ -9285,6 +9603,7 @@ impl From<ChatCompletionRequest> for liter_llm::types::ChatCompletionRequest {
             stream_options: val.stream_options.map(Into::into),
             seed: val.seed,
             reasoning_effort: val.reasoning_effort.map(Into::into),
+            modalities: val.modalities.map(|v| v.into_iter().map(Into::into).collect()),
             extra_body: val.extra_body.as_ref().and_then(|s| serde_json::from_str(s).ok()),
         }
     }
@@ -9313,6 +9632,7 @@ impl From<liter_llm::types::ChatCompletionRequest> for ChatCompletionRequest {
             stream_options: val.stream_options.map(Into::into),
             seed: val.seed,
             reasoning_effort: val.reasoning_effort.map(Into::into),
+            modalities: val.modalities.map(|v| v.into_iter().map(Into::into).collect()),
             extra_body: val.extra_body.as_ref().map(ToString::to_string),
         }
     }
@@ -9598,6 +9918,26 @@ impl From<liter_llm::types::Image> for Image {
     }
 }
 
+#[allow(clippy::redundant_closure, clippy::useless_conversion)]
+impl From<DecodedDataUrl> for liter_llm::image::DecodedDataUrl {
+    fn from(val: DecodedDataUrl) -> Self {
+        Self {
+            mime: val.mime,
+            data: val.data.to_vec().into(),
+        }
+    }
+}
+
+#[allow(clippy::redundant_closure, clippy::useless_conversion)]
+impl From<liter_llm::image::DecodedDataUrl> for DecodedDataUrl {
+    fn from(val: liter_llm::image::DecodedDataUrl) -> Self {
+        Self {
+            mime: val.mime.to_string(),
+            data: val.data.to_vec().into(),
+        }
+    }
+}
+
 #[allow(clippy::redundant_closure, clippy::useless_conversion)]
 impl From<CreateSpeechRequest> for liter_llm::types::CreateSpeechRequest {
     fn from(val: CreateSpeechRequest) -> Self {
@@ -10813,6 +11153,52 @@ impl From<liter_llm::types::ImageDetail> for ImageDetail {
     }
 }
 
+impl From<AssistantContent> for liter_llm::types::AssistantContent {
+    fn from(val: AssistantContent) -> Self {
+        match val {
+            AssistantContent::Text(_0) => Self::Text(_0),
+            AssistantContent::Parts(_0) => Self::Parts(_0.into_iter().map(Into::into).collect()),
+        }
+    }
+}
+
+impl From<liter_llm::types::AssistantContent> for AssistantContent {
+    fn from(val: liter_llm::types::AssistantContent) -> Self {
+        match val {
+            liter_llm::types::AssistantContent::Text(_0) => Self::Text(_0),
+            liter_llm::types::AssistantContent::Parts(_0) => Self::Parts(_0.into_iter().map(Into::into).collect()),
+        }
+    }
+}
+
+impl From<AssistantPart> for liter_llm::types::AssistantPart {
+    fn from(val: AssistantPart) -> Self {
+        match val {
+            AssistantPart::Text { text } => Self::Text { text },
+            AssistantPart::Refusal { refusal } => Self::Refusal { refusal },
+            AssistantPart::OutputImage { image_url } => Self::OutputImage {
+                image_url: image_url.into(),
+            },
+            AssistantPart::OutputAudio { audio } => Self::OutputAudio { audio: audio.into() },
+        }
+    }
+}
+
+impl From<liter_llm::types::AssistantPart> for AssistantPart {
+    fn from(val: liter_llm::types::AssistantPart) -> Self {
+        match val {
+            liter_llm::types::AssistantPart::Text { text } => Self::Text { text: text.to_string() },
+            liter_llm::types::AssistantPart::Refusal { refusal } => Self::Refusal {
+                refusal: refusal.to_string(),
+            },
+            liter_llm::types::AssistantPart::OutputImage { image_url } => Self::OutputImage {
+                image_url: image_url.into(),
+            },
+            liter_llm::types::AssistantPart::OutputAudio { audio } => Self::OutputAudio { audio: audio.into() },
+        }
+    }
+}
+
 impl From<ToolType> for liter_llm::types::ToolType {
     fn from(val: ToolType) -> Self {
         match val {
@@ -10909,6 +11295,26 @@ impl From<liter_llm::types::StopSequence> for StopSequence {
     }
 }
 
+impl From<Modality> for liter_llm::types::Modality {
+    fn from(val: Modality) -> Self {
+        match val {
+            Modality::Text => Self::Text,
+            Modality::Audio => Self::Audio,
+            Modality::Image => Self::Image,
+        }
+    }
+}
+
+impl From<liter_llm::types::Modality> for Modality {
+    fn from(val: liter_llm::types::Modality) -> Self {
+        match val {
+            liter_llm::types::Modality::Text => Self::Text,
+            liter_llm::types::Modality::Audio => Self::Audio,
+            liter_llm::types::Modality::Image => Self::Image,
+        }
+    }
+}
+
 impl From<FinishReason> for liter_llm::types::FinishReason {
     fn from(val: FinishReason) -> Self {
         match val {
@@ -11328,8 +11734,6 @@ fn ruby_init(ruby: &Ruby) -> Result<(), Error> {
 
     class.define_method("name", method!(SystemMessage::name, 0))?;
 
-    class.define_method("to_s", method!(SystemMessage::to_s, 0))?;
-
     let class = module.define_class("UserMessage", ruby.class_object())?;
 
     class.define_singleton_method("new", function!(UserMessage::new, -1))?;
@@ -11376,7 +11780,13 @@ fn ruby_init(ruby: &Ruby) -> Result<(), Error> {
 
     class.define_method("function_call", method!(AssistantMessage::function_call, 0))?;
 
-    class.define_method("to_s", method!(AssistantMessage::to_s, 0))?;
+    class.define_method("text", method!(AssistantMessage::text, 0))?;
+
+    class.define_method("refusal_text", method!(AssistantMessage::refusal_text, 0))?;
+
+    class.define_method("output_images", method!(AssistantMessage::output_images, 0))?;
+
+    class.define_method("output_audio", method!(AssistantMessage::output_audio, 0))?;
 
     let class = module.define_class("ToolMessage", ruby.class_object())?;
 
@@ -11542,6 +11952,8 @@ fn ruby_init(ruby: &Ruby) -> Result<(), Error> {
 
     class.define_method("reasoning_effort", method!(ChatCompletionRequest::reasoning_effort, 0))?;
 
+    class.define_method("modalities", method!(ChatCompletionRequest::modalities, 0))?;
+
     class.define_method("extra_body", method!(ChatCompletionRequest::extra_body, 0))?;
 
     let class = module.define_class("StreamOptions", ruby.class_object())?;
@@ -11726,6 +12138,14 @@ fn ruby_init(ruby: &Ruby) -> Result<(), Error> {
 
     class.define_method("revised_prompt", method!(Image::revised_prompt, 0))?;
 
+    let class = module.define_class("DecodedDataUrl", ruby.class_object())?;
+
+    class.define_singleton_method("new", function!(DecodedDataUrl::new, -1))?;
+
+    class.define_method("mime", method!(DecodedDataUrl::mime, 0))?;
+
+    class.define_method("data", method!(DecodedDataUrl::data, 0))?;
+
     let class = module.define_class("CreateSpeechRequest", ruby.class_object())?;
 
     class.define_singleton_method("new", function!(CreateSpeechRequest::new, -1))?;
@@ -12405,6 +12825,10 @@ fn ruby_init(ruby: &Ruby) -> Result<(), Error> {
 
     module.define_module_function("create_client_from_json", function!(create_client_from_json, 1))?;
 
+    module.define_module_function("encode_data_url", function!(encode_data_url, -1))?;
+
+    module.define_module_function("decode_data_url", function!(decode_data_url, 1))?;
+
     module.define_module_function("register_custom_provider", function!(register_custom_provider, 1))?;
 
     module.define_module_function("unregister_custom_provider", function!(unregister_custom_provider, 1))?;
@@ -12429,6 +12853,8 @@ fn ruby_init(ruby: &Ruby) -> Result<(), Error> {
 
     module.define_module_function("ensure_crypto_provider", function!(ensure_crypto_provider, 0))?;
 
+    module.define_module_function("ensure_crypto_provider", function!(ensure_crypto_provider, 0))?;
+
     module.define_module_function("chat_stream", function!(chat_stream, 2))?;
 
     let liter_llm_error_info_class = module.define_class("LiterLlmErrorInfo", ruby.class_object())?;

data/lib/liter_llm/native.rb

@@ -1,5 +1,5 @@
 # This file is auto-generated by alef — DO NOT EDIT.
-# alef:hash:52effca2513e1c5f8a9a5b90a372d645388de4b92e4f822acc38d97502c322b6
+# alef:hash:fec9b44a56d9714b7b12931bb48e1e8f6d57ef2d0ccaad1f9ac6046ce01a89af
 # To regenerate: alef generate
 # To verify freshness: alef verify --exit-code
 # frozen_string_literal: true
@@ -314,7 +314,142 @@ module LiterLlm
 end
 
 module LiterLlm
-  # Response format constraint.
+  # One part of a structured assistant response.
+  #
+  # `#[serde(tag = "type", rename_all = "snake_case")]` matches OpenAI's
+  # parts-spec discriminator (`"type": "text"`, `"type": "output_image"`, …).
+  module AssistantPart
+    extend T::Helpers
+    extend T::Sig
+
+    interface!
+
+    # Dispatch from a Hash to the appropriate variant constructor.
+    # @param hash [Hash] with discriminator field and variant-specific fields
+    # @return [variant_class] an instance of the appropriate variant
+    sig { params(hash: T::Hash[T.untyped, T.untyped]).returns(T.untyped) }
+    def self.from_hash(hash)
+      discriminator = hash[:type] || hash["type"]
+      case discriminator
+      when "text" then AssistantPartText.from_hash(hash)
+      when "refusal" then AssistantPartRefusal.from_hash(hash)
+      when "output_image" then AssistantPartOutputImage.from_hash(hash)
+      when "output_audio" then AssistantPartOutputAudio.from_hash(hash)
+      else raise "Unknown discriminator: #{discriminator}"
+      end
+    end
+  end
+  ## A text segment of the response.
+  AssistantPartText = Data.define(:text) do
+    include AssistantPart
+    extend T::Sig
+
+    # The text content of this part.
+    sig { returns(String) }
+    def text = super # rubocop:disable Lint/UselessMethodDefinition
+    sig { returns(T::Boolean) }
+    def text? = true
+    sig { returns(T::Boolean) }
+    def refusal? = false
+    sig { returns(T::Boolean) }
+    def output_image? = false
+    sig { returns(T::Boolean) }
+    def output_audio? = false
+    # @param hash [Hash] deserialized from the native extension
+    # @return [self]
+    sig { params(hash: T::Hash[T.untyped, T.untyped]).returns(T.attached_class) }
+    def self.from_hash(hash)
+      new(text: hash[:text] || hash["text"])
+    end
+  end
+  ## A refusal — the model declined to respond.
+  AssistantPartRefusal = Data.define(:refusal) do
+    include AssistantPart
+    extend T::Sig
+
+    # The refusal reason.
+    sig { returns(String) }
+    def refusal = super # rubocop:disable Lint/UselessMethodDefinition
+    sig { returns(T::Boolean) }
+    def text? = false
+    sig { returns(T::Boolean) }
+    def refusal? = true
+    sig { returns(T::Boolean) }
+    def output_image? = false
+    sig { returns(T::Boolean) }
+    def output_audio? = false
+    # @param hash [Hash] deserialized from the native extension
+    # @return [self]
+    sig { params(hash: T::Hash[T.untyped, T.untyped]).returns(T.attached_class) }
+    def self.from_hash(hash)
+      new(refusal: hash[:refusal] || hash["refusal"])
+    end
+  end
+  ## An image produced by the model (e.g. `gpt-image-1`, Gemini Imagen).
+  AssistantPartOutputImage = Data.define(:image_url) do
+    include AssistantPart
+    extend T::Sig
+
+    # Image URL or data URI referencing the generated image.
+    sig { returns(ImageUrl) }
+    def image_url = super # rubocop:disable Lint/UselessMethodDefinition
+    sig { returns(T::Boolean) }
+    def text? = false
+    sig { returns(T::Boolean) }
+    def refusal? = false
+    sig { returns(T::Boolean) }
+    def output_image? = true
+    sig { returns(T::Boolean) }
+    def output_audio? = false
+    # @param hash [Hash] deserialized from the native extension
+    # @return [self]
+    sig { params(hash: T::Hash[T.untyped, T.untyped]).returns(T.attached_class) }
+    def self.from_hash(hash)
+      new(image_url: hash[:image_url] || hash["image_url"])
+    end
+  end
+  ## Audio produced by the model (e.g. `gpt-4o-audio-preview`).
+  AssistantPartOutputAudio = Data.define(:audio) do
+    include AssistantPart
+    extend T::Sig
+
+    # Base64-encoded audio data and its format.
+    sig { returns(AudioContent) }
+    def audio = super # rubocop:disable Lint/UselessMethodDefinition
+    sig { returns(T::Boolean) }
+    def text? = false
+    sig { returns(T::Boolean) }
+    def refusal? = false
+    sig { returns(T::Boolean) }
+    def output_image? = false
+    sig { returns(T::Boolean) }
+    def output_audio? = true
+    # @param hash [Hash] deserialized from the native extension
+    # @return [self]
+    sig { params(hash: T::Hash[T.untyped, T.untyped]).returns(T.attached_class) }
+    def self.from_hash(hash)
+      new(audio: hash[:audio] || hash["audio"])
+    end
+  end
+end
+
+module LiterLlm
+  # Wire format for the chat completions `response_format` field.
+  #
+  # # Provider mapping
+  #
+  # - **OpenAI** (and OpenAI-compatible providers): emitted verbatim as
+  #   `{"type": "json_schema", "json_schema": {...}}` per the
+  #   chat-completions spec.
+  # - **Gemini / Vertex AI**: translated to
+  #   `generationConfig.responseMimeType = "application/json"` and
+  #   `generationConfig.responseSchema = <schema>`. The `name`,
+  #   `description`, and `strict` fields are dropped — Gemini's
+  #   structured-output API does not consume them.
+  # - **Anthropic**: no native JSON mode. A system instruction is
+  #   prepended asking the model to respond with valid JSON.
+  #   `strict` is advisory only; callers should still validate the
+  #   returned JSON if the schema is load-bearing.
   module ResponseFormat
     extend T::Helpers
     extend T::Sig

data/lib/liter_llm/version.rb

@@ -1,10 +1,10 @@
 # This file is auto-generated by alef — DO NOT EDIT.
-# alef:hash:52effca2513e1c5f8a9a5b90a372d645388de4b92e4f822acc38d97502c322b6
+# alef:hash:fec9b44a56d9714b7b12931bb48e1e8f6d57ef2d0ccaad1f9ac6046ce01a89af
 # To regenerate: alef generate
 # To verify freshness: alef verify --exit-code
 # frozen_string_literal: true
 
 module LiterLlm
   ## The version string for this package.
-  VERSION = "1.6.4"
+  VERSION = "1.7.1"
 end

data/lib/liter_llm.rb

@@ -1,5 +1,5 @@
 # This file is auto-generated by alef — DO NOT EDIT.
-# alef:hash:52effca2513e1c5f8a9a5b90a372d645388de4b92e4f822acc38d97502c322b6
+# alef:hash:fec9b44a56d9714b7b12931bb48e1e8f6d57ef2d0ccaad1f9ac6046ce01a89af
 # To regenerate: alef generate
 # To verify freshness: alef verify --exit-code
 # frozen_string_literal: true

data/sig/types.rbs

@@ -1,5 +1,5 @@
 # This file is auto-generated by alef — DO NOT EDIT.
-# alef:hash:52effca2513e1c5f8a9a5b90a372d645388de4b92e4f822acc38d97502c322b6
+# alef:hash:fec9b44a56d9714b7b12931bb48e1e8f6d57ef2d0ccaad1f9ac6046ce01a89af
 # To regenerate: alef generate
 # To verify freshness: alef verify --exit-code
 
@@ -10,10 +10,10 @@ module LiterLlm
   type json_value = Hash[String, untyped] | Array[untyped] | String | Integer | Float | bool | nil
 
   class SystemMessage
-    attr_accessor content: String?
+    attr_accessor content: UserContent?
     attr_accessor name: String?
 
-    def initialize: (?content: String, ?name: String) -> void
+    def initialize: (?content: UserContent, ?name: String) -> void
   end
 
   class UserMessage
@@ -45,13 +45,17 @@ module LiterLlm
   end
 
   class AssistantMessage
-    attr_accessor content: String?
+    attr_accessor content: AssistantContent?
     attr_accessor name: String?
     attr_accessor tool_calls: Array[ToolCall]?
     attr_accessor refusal: String?
     attr_accessor function_call: FunctionCall?
 
-    def initialize: (?content: String, ?name: String, ?tool_calls: Array[ToolCall], ?refusal: String, ?function_call: FunctionCall) -> void
+    def initialize: (?content: AssistantContent, ?name: String, ?tool_calls: Array[ToolCall], ?refusal: String, ?function_call: FunctionCall) -> void
+    def text: () -> String?
+    def refusal_text: () -> String?
+    def output_images: () -> Array[ImageUrl]
+    def output_audio: () -> Array[AudioContent]
   end
 
   class ToolMessage
@@ -165,9 +169,10 @@ module LiterLlm
     attr_accessor stream_options: StreamOptions?
     attr_accessor seed: Integer?
     attr_accessor reasoning_effort: ReasoningEffort?
+    attr_accessor modalities: Array[Modality]?
     attr_accessor extra_body: json_value?
 
-    def initialize: (?model: String, ?messages: Array[Message], ?temperature: Float, ?top_p: Float, ?n: Integer, ?stream: bool, ?stop: StopSequence, ?max_tokens: Integer, ?presence_penalty: Float, ?frequency_penalty: Float, ?logit_bias: Hash[String, Float], ?user: String, ?tools: Array[ChatCompletionTool], ?tool_choice: ToolChoice, ?parallel_tool_calls: bool, ?response_format: ResponseFormat, ?stream_options: StreamOptions, ?seed: Integer, ?reasoning_effort: ReasoningEffort, ?extra_body: json_value) -> void
+    def initialize: (?model: String, ?messages: Array[Message], ?temperature: Float, ?top_p: Float, ?n: Integer, ?stream: bool, ?stop: StopSequence, ?max_tokens: Integer, ?presence_penalty: Float, ?frequency_penalty: Float, ?logit_bias: Hash[String, Float], ?user: String, ?tools: Array[ChatCompletionTool], ?tool_choice: ToolChoice, ?parallel_tool_calls: bool, ?response_format: ResponseFormat, ?stream_options: StreamOptions, ?seed: Integer, ?reasoning_effort: ReasoningEffort, ?modalities: Array[Modality], ?extra_body: json_value) -> void
   end
 
   class StreamOptions
@@ -299,6 +304,13 @@ module LiterLlm
     def initialize: (?url: String, ?b64_json: String, ?revised_prompt: String) -> void
   end
 
+  class DecodedDataUrl
+    attr_accessor mime: String?
+    attr_accessor data: String?
+
+    def initialize: (?mime: String, ?data: String) -> void
+  end
+
   class CreateSpeechRequest
     attr_accessor model: String?
     attr_accessor input: String?
@@ -780,6 +792,12 @@ module LiterLlm
     type value = :low | :high | :auto
   end
 
+  class AssistantContent
+  end
+
+  class AssistantPart
+  end
+
   class ToolType
     type value = :function
   end
@@ -797,6 +815,10 @@ module LiterLlm
   class StopSequence
   end
 
+  class Modality
+    type value = :text | :audio | :image
+  end
+
   class FinishReason
     type value = :stop | :length | :tool_calls | :content_filter | :function_call | :other
   end
@@ -859,6 +881,10 @@ module LiterLlm
 
   def self.create_client_from_json: (String json) -> DefaultClient
 
+  def self.encode_data_url: (String bytes, ?String mime) -> String
+
+  def self.decode_data_url: (String url) -> DecodedDataUrl?
+
   def self.register_custom_provider: (CustomProviderConfig config) -> void
 
   def self.unregister_custom_provider: (String name) -> bool

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: liter_llm
 version: !ruby/object:Gem::Version
-  version: 1.6.4
+  version: 1.7.1
 platform: ruby
 authors:
 - Na'aman Hirschfeld
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-06-17 00:00:00.000000000 Z
+date: 2026-06-18 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: rb_sys