bedrock: Inference Config updates (#35808)

Fixes #36866

- Updated internal naming for Claude 4 models to be consistent.
- Corrected max output tokens for Anthropic Bedrock models to match docs

Shoutout to @tlehn for noticing the bug, and finding the resolution.

Release Notes:

- bedrock: Fixed inference config errors causing Opus 4 Thinking and
Opus 4.1 Thinking to fail (thanks [@tlehn](https://github.com/tlehn) and
[@5herlocked](https://github.com/5herlocked])
- bedrock: Fixed an issue which prevented Rules / System prompts not
functioning with Bedrock models (thanks
[@tlehn](https://github.com/tlehn) and
[@5herlocked](https://github.com/5herlocked])

crates/bedrock/src/bedrock.rs

@@ -3,6 +3,7 @@ mod models;
 use anyhow::{Context, Error, Result, anyhow};
 use aws_sdk_bedrockruntime as bedrock;
 pub use aws_sdk_bedrockruntime as bedrock_client;
+use aws_sdk_bedrockruntime::types::InferenceConfiguration;
 pub use aws_sdk_bedrockruntime::types::{
     AnyToolChoice as BedrockAnyToolChoice, AutoToolChoice as BedrockAutoToolChoice,
     ContentBlock as BedrockInnerContent, Tool as BedrockTool, ToolChoice as BedrockToolChoice,
@@ -17,7 +18,8 @@ pub use bedrock::types::{
     ConverseOutput as BedrockResponse, ConverseStreamOutput as BedrockStreamingResponse,
     ImageBlock as BedrockImageBlock, Message as BedrockMessage,
     ReasoningContentBlock as BedrockThinkingBlock, ReasoningTextBlock as BedrockThinkingTextBlock,
-    ResponseStream as BedrockResponseStream, ToolResultBlock as BedrockToolResultBlock,
+    ResponseStream as BedrockResponseStream, SystemContentBlock as BedrockSystemContentBlock,
+    ToolResultBlock as BedrockToolResultBlock,
     ToolResultContentBlock as BedrockToolResultContentBlock,
     ToolResultStatus as BedrockToolResultStatus, ToolUseBlock as BedrockToolUseBlock,
 };
@@ -58,6 +60,20 @@ pub async fn stream_completion(
         response = response.set_tool_config(request.tools);
     }
 
+    let inference_config = InferenceConfiguration::builder()
+        .max_tokens(request.max_tokens as i32)
+        .set_temperature(request.temperature)
+        .set_top_p(request.top_p)
+        .build();
+
+    response = response.inference_config(inference_config);
+
+    if let Some(system) = request.system {
+        if !system.is_empty() {
+            response = response.system(BedrockSystemContentBlock::Text(system));
+        }
+    }
+
     let output = response
         .send()
         .await

crates/bedrock/src/models.rs

@@ -151,12 +151,12 @@ impl Model {
 
     pub fn id(&self) -> &str {
         match self {
-            Model::ClaudeSonnet4 => "claude-4-sonnet",
-            Model::ClaudeSonnet4Thinking => "claude-4-sonnet-thinking",
-            Model::ClaudeOpus4 => "claude-4-opus",
-            Model::ClaudeOpus4_1 => "claude-4-opus-1",
-            Model::ClaudeOpus4Thinking => "claude-4-opus-thinking",
-            Model::ClaudeOpus4_1Thinking => "claude-4-opus-1-thinking",
+            Model::ClaudeSonnet4 => "claude-sonnet-4",
+            Model::ClaudeSonnet4Thinking => "claude-sonnet-4-thinking",
+            Model::ClaudeOpus4 => "claude-opus-4",
+            Model::ClaudeOpus4_1 => "claude-opus-4-1",
+            Model::ClaudeOpus4Thinking => "claude-opus-4-thinking",
+            Model::ClaudeOpus4_1Thinking => "claude-opus-4-1-thinking",
             Model::Claude3_5SonnetV2 => "claude-3-5-sonnet-v2",
             Model::Claude3_5Sonnet => "claude-3-5-sonnet",
             Model::Claude3Opus => "claude-3-opus",
@@ -359,14 +359,12 @@ impl Model {
     pub fn max_output_tokens(&self) -> u64 {
         match self {
             Self::Claude3Opus | Self::Claude3Sonnet | Self::Claude3_5Haiku => 4_096,
-            Self::Claude3_7Sonnet
-            | Self::Claude3_7SonnetThinking
-            | Self::ClaudeSonnet4
-            | Self::ClaudeSonnet4Thinking
-            | Self::ClaudeOpus4
-            | Model::ClaudeOpus4Thinking
+            Self::Claude3_7Sonnet | Self::Claude3_7SonnetThinking => 128_000,
+            Self::ClaudeSonnet4 | Self::ClaudeSonnet4Thinking => 64_000,
+            Self::ClaudeOpus4
+            | Self::ClaudeOpus4Thinking
             | Self::ClaudeOpus4_1
-            | Model::ClaudeOpus4_1Thinking => 128_000,
+            | Self::ClaudeOpus4_1Thinking => 32_000,
             Self::Claude3_5SonnetV2 | Self::PalmyraWriterX4 | Self::PalmyraWriterX5 => 8_192,
             Self::Custom {
                 max_output_tokens, ..
@@ -784,10 +782,10 @@ mod tests {
         );
 
         // Test thinking models have different friendly IDs but same request IDs
-        assert_eq!(Model::ClaudeSonnet4.id(), "claude-4-sonnet");
+        assert_eq!(Model::ClaudeSonnet4.id(), "claude-sonnet-4");
         assert_eq!(
             Model::ClaudeSonnet4Thinking.id(),
-            "claude-4-sonnet-thinking"
+            "claude-sonnet-4-thinking"
         );
         assert_eq!(
             Model::ClaudeSonnet4.request_id(),