Add bedrock model ID fallback

lib/ruby_llm/protocols/converse/chat.rb

@@ -81,7 +81,7 @@ def parse_completion_body(data, raw:)
             cache_write_tokens: usage['cacheWriteInputTokens'],
             thinking_tokens: reasoning_tokens(usage),
             finish_reason: data['stopReason'],
-            model: data['modelId'],
+            model: data['modelId'] || @model&.id,
             raw: raw
           )
         end

spec/ruby_llm/protocols/converse/chat_spec.rb

@@ -73,6 +73,25 @@
       expect(message.finish_reason).to eq('guardrail_intervened')
     end
 
+    it 'falls back to the request model when Bedrock omits modelId' do
+      response_body = {
+        'output' => {
+          'message' => {
+            'content' => [{ 'text' => 'Hi!' }]
+          }
+        },
+        'usage' => {}
+      }
+
+      provider = double(config: RubyLLM.config, connection: nil)
+      model = instance_double(RubyLLM::Model, id: 'openai.gpt-oss-120b-1:0')
+      protocol = RubyLLM::Protocols::Converse.new(provider, model)
+      response = instance_double(Faraday::Response, body: response_body)
+      message = protocol.send(:parse_completion_body, response_body, raw: response)
+
+      expect(message.model).to eq('openai.gpt-oss-120b-1:0')
+    end
+
     it 'extracts thinking tokens from top-level reasoningTokens' do
       response_body = {
         'output' => {