feat: Add parallel tool calling support for Meta/Llama models

Add support for the parallel_tool_calls parameter to enable parallel
function calling in Meta/Llama models, improving performance for
multi-tool workflows.

## Changes

- Add parallel_tool_calls class parameter to OCIGenAIBase (default: False)
- Add parallel_tool_calls parameter to bind_tools() method
- Support hybrid approach: class-level default + per-binding override
- Pass is_parallel_tool_calls to OCI API in MetaProvider
- Add validation for Cohere models (raises error if attempted)

## Testing

- 9 comprehensive unit tests (all passing)
- 4 integration tests with live OCI API (all passing)
- No regression in existing tests

## Usage

Class-level default:
  llm = ChatOCIGenAI(
      model_id="meta.llama-3.3-70b-instruct",
      parallel_tool_calls=True
  )

Per-binding override:
  llm_with_tools = llm.bind_tools(
      [tool1, tool2, tool3],
      parallel_tool_calls=True
  )

## Benefits

- Up to N× speedup for N independent tool calls
- Backward compatible (default: False)
- Clear error messages for unsupported models
- Follows existing parameter patterns

Author: fede-kamel

libs/oci/README.md

@@ -62,7 +62,7 @@ embeddings.embed_query("What is the meaning of life?")
 ```
 
 ### 4. Use Structured Output
-`ChatOCIGenAI` supports structured output. 
+`ChatOCIGenAI` supports structured output.
 
 <sub>**Note:** The default method is `function_calling`. If default method returns `None` (e.g. for Gemini models), try `json_schema` or `json_mode`.</sub>
 
@@ -79,6 +79,30 @@ structured_llm = llm.with_structured_output(Joke)
 structured_llm.invoke("Tell me a joke about programming")
 ```
 
+### 5. Use Parallel Tool Calling (Meta/Llama models only)
+Enable parallel tool calling to execute multiple tools simultaneously, improving performance for multi-tool workflows.
+
+```python
+from langchain_oci import ChatOCIGenAI
+
+# Option 1: Set at class level for all tool bindings
+llm = ChatOCIGenAI(
+    model_id="meta.llama-3.3-70b-instruct",
+    service_endpoint="https://inference.generativeai.us-chicago-1.oci.oraclecloud.com",
+    compartment_id="MY_COMPARTMENT_ID",
+    parallel_tool_calls=True  # Enable parallel tool calling
+)
+
+# Option 2: Set per-binding
+llm = ChatOCIGenAI(model_id="meta.llama-3.3-70b-instruct")
+llm_with_tools = llm.bind_tools(
+    [get_weather, calculate_tip, get_population],
+    parallel_tool_calls=True  # Tools can execute simultaneously
+)
+```
+
+<sub>**Note:** Parallel tool calling is only supported for Meta/Llama models. Cohere models will raise an error if this parameter is used.</sub>
+
 
 ## OCI Data Science Model Deployment Examples
 

libs/oci/langchain_oci/chat_models/oci_generative_ai.py

@@ -342,6 +342,13 @@ def messages_to_oci_params(
 
         This includes conversion of chat history and tool call results.
         """
+        # Cohere models don't support parallel tool calls
+        if kwargs.get("is_parallel_tool_calls"):
+            raise ValueError(
+                "Parallel tool calls are not supported for Cohere models. "
+                "This feature is only available for Meta/Llama models using GenericChatRequest."
+            )
+
         is_force_single_step = kwargs.get("is_force_single_step", False)
         oci_chat_history = []
 
@@ -829,6 +836,10 @@ def _should_allow_more_tool_calls(
                 result["tool_choice"] = self.oci_tool_choice_none()
             # else: Allow model to decide (default behavior)
 
+        # Add parallel tool calls support for Meta/Llama models
+        if "is_parallel_tool_calls" in kwargs:
+            result["is_parallel_tool_calls"] = kwargs["is_parallel_tool_calls"]
+
         return result
 
     def _process_message_content(
@@ -1186,6 +1197,7 @@ def bind_tools(
         tool_choice: Optional[
             Union[dict, str, Literal["auto", "none", "required", "any"], bool]
         ] = None,
+        parallel_tool_calls: Optional[bool] = None,
         **kwargs: Any,
     ) -> Runnable[LanguageModelInput, BaseMessage]:
         """Bind tool-like objects to this chat model.
@@ -1206,6 +1218,11 @@ def bind_tools(
                     {"type": "function", "function": {"name": <<tool_name>>}}:
                 calls <<tool_name>> tool.
                 - False or None: no effect, default Meta behavior.
+            parallel_tool_calls: Whether to enable parallel function calling.
+                If True, the model can call multiple tools simultaneously.
+                If False, tools are called sequentially.
+                If None (default), uses the class-level parallel_tool_calls setting.
+                Only supported for Meta/Llama models using GenericChatRequest.
             kwargs: Any additional parameters are passed directly to
                 :meth:`~langchain_oci.chat_models.oci_generative_ai.ChatOCIGenAI.bind`.
         """
@@ -1215,6 +1232,12 @@ def bind_tools(
         if tool_choice is not None:
             kwargs["tool_choice"] = self._provider.process_tool_choice(tool_choice)
 
+        # Add parallel tool calls support
+        # Use bind-time parameter if provided, else fall back to class default
+        use_parallel = parallel_tool_calls if parallel_tool_calls is not None else self.parallel_tool_calls
+        if use_parallel:
+            kwargs["is_parallel_tool_calls"] = True
+
         return super().bind(tools=formatted_tools, **kwargs)
 
     def with_structured_output(

libs/oci/langchain_oci/llms/oci_generative_ai.py

@@ -120,6 +120,12 @@ class OCIGenAIBase(BaseModel, ABC):
     """Maximum tool calls before forcing final answer.
     Prevents infinite loops while allowing multi-step orchestration."""
 
+    parallel_tool_calls: bool = False
+    """Whether to enable parallel function calling during tool use.
+    If True, the model can call multiple tools simultaneously.
+    Only supported for Meta/Llama models using GenericChatRequest.
+    Default: False for backward compatibility."""
+
     model_config = ConfigDict(
         extra="forbid", arbitrary_types_allowed=True, protected_namespaces=()
     )