Enable ALPN feature in reqwest crate

.devcontainer/Dockerfile

@@ -4,6 +4,7 @@ FROM ubuntu:24.04
 RUN apt update && DEBIAN_FRONTEND=noninteractive apt install -y \
     build-essential \
     clang \
+    pkg-config \
     zlib1g \
     zlib1g-dev \
     libssl-dev \
@@ -29,7 +30,7 @@ RUN curl https://mise.run | sh \
     && ~/.local/bin/mise --verbose use -g --yes [email protected] [email protected] [email protected] \
     && echo 'export PATH="$HOME/.local/bin:$HOME/.local/share/mise/shims:$PATH"' >> ~/.bash_profile \
     && echo 'export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$HOME/.local/share/mise/installs/ruby/3.1/lib' >> ~/.bash_profile
- 
+
 # Install Rust
 RUN curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh -s -- -y
 

engine/Cargo.toml

@@ -70,9 +70,10 @@ minijinja = { version = "1.0.16", default-features = false, features = [
 pretty_assertions = "1.4.1"
 rand = "0.8.5"
 regex = "1.10.4"
-reqwest = { version = "0.12.5", features = [
+reqwest = { version = "0.12.12", features = [
   "json",
   "native-tls-vendored",
+  "native-tls-alpn",
   "stream",
 ] }
 scopeguard = "1.2.0"

engine/baml-runtime/Cargo.toml

@@ -103,7 +103,7 @@ colored = { version = "2.1.0", default-features = false, features = [
 ] }
 futures-timer = { version = "3.0.3", features = ["wasm-bindgen"] }
 js-sys = "0.3.69"
-reqwest = { version = "0.12.5", features = ["stream", "json"] }
+reqwest = { version = "0.12.12", features = ["stream", "json", "native-tls-vendored", "native-tls-alpn"] }
 # 
 send_wrapper = { version = "0.6.0", features = ["futures"] }
 serde-wasm-bindgen = "0.6.5"

engine/baml-runtime/src/internal/llm_client/primitive/google/googleai_client.rs

@@ -168,7 +168,11 @@ impl SseResponseTrait for GoogleAIClient {
 
                         if let Some(choice) = event.candidates.get(0) {
                             let part_index = content_part(&model_id);
-                            if let Some(content) = choice.content.as_ref().and_then(|c| c.parts.get(part_index)) {
+                            if let Some(content) = choice
+                                .content
+                                .as_ref()
+                                .and_then(|c| c.parts.get(part_index))
+                            {
                                 inner.content += &content.text;
                             }
                             if let Some(FinishReason::Stop) = choice.finish_reason.as_ref() {
@@ -458,11 +462,11 @@ impl ToProviderMessage for GoogleAIClient {
 /// The Google Gemini 2 model has an experimental feature
 /// called Flash Thinking Mode, which is turned on in a particular
 /// named model: gemini-2.0-flash-thinking-exp-1219
-/// 
+///
 /// When run in this mode, Gemini returns `candidates` with 2 parts each.
 /// Part 0 is the chain of thought, part 1 is the actual output.
 /// Other Gemini models put the output data in part 0.
-/// 
+///
 /// TODO: Explicitly represent Flash Thinking Mode response and
 /// do more thorough checking for the content part.
 /// For examples of how to introspect the response more safely, see:
@@ -473,4 +477,4 @@ fn content_part(model_name: &str) -> usize {
     } else {
         0
     }
-}
+}

engine/baml-runtime/src/internal/llm_client/traits/mod.rs

@@ -150,8 +150,8 @@ where
 {
     #[allow(async_fn_in_trait)]
     async fn single_call(&self, ctx: &RuntimeContext, prompt: &RenderedPrompt) -> LLMResponse {
-        if let RenderedPrompt::Chat(chat) = &prompt {
-            match process_media_urls(
+        match prompt {
+            RenderedPrompt::Chat(chat) => match process_media_urls(
                 self.model_features().resolve_media_urls,
                 true,
                 None,
@@ -160,15 +160,10 @@ where
             )
             .await
             {
-                Ok(messages) => return self.chat(ctx, &messages).await,
-                Err(e) => {
-                    return LLMResponse::InternalFailure(format!("Error occurred:\n\n{:?}", e))
-                }
-            }
-        }
+                Ok(messages) => self.chat(ctx, &messages).await,
+                Err(e) => LLMResponse::InternalFailure(format!("Error occurred:\n\n{:?}", e)),
+            },
 
-        match prompt {
-            RenderedPrompt::Chat(p) => self.chat(ctx, p).await,
             RenderedPrompt::Completion(p) => self.completion(ctx, p).await,
         }
     }