fix(openai): split embedded tool_calls into standard chunk format

- Standard OpenAI clients expect tool_calls to be streamed as two parts:
  1. Initialization chunk containing 'id', 'type', and 'name', with empty 'arguments'.
  2. Payload chunk(s) containing 'arguments', with 'id', 'type', and 'name' omitted.
- Previously, the proxy was yielding all fields in a single chunk when parsing the custom 'tool_uses' JSON from gpt-5.4, causing strict clients like opencode to fail silently when delegating parallel tasks.
- The proxy now splits the extracted JSON into the correct two-chunk sequence, restoring subagent compatibility.

src/providers/openai.rs

@@ -819,15 +819,37 @@ impl super::Provider for OpenAIProvider {
                                                 };
                                             }
                                             
-                                            // Yield the tool calls
+                                            // Yield the tool calls in two chunks to mimic standard streaming behavior
-                                            // ... (rest of tool call yielding unchanged)
+                                            // Chunk 1: Initialization (id, name)
-                                            let deltas: Vec<crate::models::ToolCallDelta> = embedded_calls.into_iter().enumerate().map(|(idx, tc)| {
+                                            let init_deltas: Vec<crate::models::ToolCallDelta> = embedded_calls.iter().enumerate().map(|(idx, tc)| {
                                                 crate::models::ToolCallDelta {
                                                     index: idx as u32,
-                                                    id: Some(tc.id),
+                                                    id: Some(tc.id.clone()),
                                                     call_type: Some("function".to_string()),
                                                     function: Some(crate::models::FunctionCallDelta {
-                                                        name: Some(tc.function.name),
+                                                        name: Some(tc.function.name.clone()),
+                                                        arguments: Some("".to_string()),
+                                                    }),
+                                                }
+                                            }).collect();
+                                            
+                                            yield ProviderStreamChunk {
+                                                content: String::new(),
+                                                reasoning_content: None,
+                                                finish_reason: None,
+                                                tool_calls: Some(init_deltas),
+                                                model: model.clone(),
+                                                usage: None,
+                                            };
+
+                                            // Chunk 2: Payload (arguments)
+                                            let arg_deltas: Vec<crate::models::ToolCallDelta> = embedded_calls.into_iter().enumerate().map(|(idx, tc)| {
+                                                crate::models::ToolCallDelta {
+                                                    index: idx as u32,
+                                                    id: None,
+                                                    call_type: None,
+                                                    function: Some(crate::models::FunctionCallDelta {
+                                                        name: None,
                                                         arguments: Some(tc.function.arguments),
                                                     }),
                                                 }
@@ -837,7 +859,7 @@ impl super::Provider for OpenAIProvider {
                                                 content: String::new(),
                                                 reasoning_content: None,
                                                 finish_reason: None,
-                                                tool_calls: Some(deltas),
+                                                tool_calls: Some(arg_deltas),
                                                 model: model.clone(),
                                                 usage: None,
                                             };