Merge remote-tracking branch 'origin/main'

CHANGELOG.md

@@ -1,5 +1,44 @@
 # ChangeLog
 
+## [2024-12-18]
+
+### `llama-index-core` [0.12.7]
+
+- fix: add a timeout to langchain callback handler (#17296)
+- fix: make Document serialization event more backward compatible (#17312)
+
+### `llama-index-embeddings-voyageai` [0.3.4]
+
+- Exposing additional keyword arguments for VoyageAI's embedding model (#17315)
+
+### `llama-index-llms-keywordsai` [0.1.0]
+
+- Added KeywordsAI LLM (#16860)
+
+### `llama-index-llms-oci-genai` [0.4.0]
+
+- Add OCI Generative AI tool calling support (#16888)
+
+### `llama-index-llms-openai` [0.3.11]
+
+- support new o1 models (#17307)
+
+### `llama-index-postprocessor-voyageai-rerank` [0.3.1]
+
+- VoyageAI Reranker optional API Key (#17310)
+
+### `llama-index-vector-stores-azureaisearch` [0.3.1]
+
+- improve async search client handling (#17319)
+
+### `llama-index-vector-stores-azurecosmosmongo` [0.4.0]
+
+- CosmosDB insertion timestamp bugfix (#17290)
+
+### `llama-index-vector-stores-azurecosmosnosql` [1.3.0]
+
+- CosmosDB insertion timestamp bugfix (#17290)
+
 ## [2024-12-17]
 
 ### `llama-index-core` [0.12.6]

docs/docs/CHANGELOG.md

@@ -1,10 +1,49 @@
 # ChangeLog
 
+## [2024-12-18]
+
+### `llama-index-core` [0.12.7]
+
+- fix: add a timeout to langchain callback handler (#17296)
+- fix: make Document serialization event more backward compatible (#17312)
+
+### `llama-index-embeddings-voyageai` [0.3.4]
+
+- Exposing additional keyword arguments for VoyageAI's embedding model (#17315)
+
+### `llama-index-llms-keywordsai` [0.1.0]
+
+- Added KeywordsAI LLM (#16860)
+
+### `llama-index-llms-oci-genai` [0.4.0]
+
+- Add OCI Generative AI tool calling support (#16888)
+
+### `llama-index-llms-openai` [0.3.11]
+
+- support new o1 models (#17307)
+
+### `llama-index-postprocessor-voyageai-rerank` [0.3.1]
+
+- VoyageAI Reranker optional API Key (#17310)
+
+### `llama-index-vector-stores-azureaisearch` [0.3.1]
+
+- improve async search client handling (#17319)
+
+### `llama-index-vector-stores-azurecosmosmongo` [0.4.0]
+
+- CosmosDB insertion timestamp bugfix (#17290)
+
+### `llama-index-vector-stores-azurecosmosnosql` [1.3.0]
+
+- CosmosDB insertion timestamp bugfix (#17290)
+
 ## [2024-12-17]
 
 ### `llama-index-core` [0.12.6]
 
-- [bug fix] Ensure that StopEvent gets cleared from Context._in_progress["_done"] after a Workflow run (#17300)
+- [bug fix] Ensure that StopEvent gets cleared from Context.\_in_progress["_done"] after a Workflow run (#17300)
 - fix: add a timeout to langchain callback handler (#17296)
 - tweak User vs tool in react prompts (#17273)
 - refact: Refactor Document to be natively multimodal (#17204)

llama-index-core/llama_index/core/__init__.py

@@ -1,6 +1,6 @@
 """Init file of LlamaIndex."""
 
-__version__ = "0.12.6"
+__version__ = "0.12.7"
 
 import logging
 from logging import NullHandler

llama-index-core/pyproject.toml

@@ -46,7 +46,7 @@ name = "llama-index-core"
 packages = [{include = "llama_index"}]
 readme = "README.md"
 repository = "https://github.com/run-llama/llama_index"
-version = "0.12.6"
+version = "0.12.7"
 
 [tool.poetry.dependencies]
 SQLAlchemy = {extras = ["asyncio"], version = ">=1.4.49"}

llama-index-integrations/agent/llama-index-agent-openai/llama_index/agent/openai/step.py

@@ -661,7 +661,6 @@ async def _arun_step(
                 step, task.extra_state["new_memory"], verbose=self._verbose
             )
 
-        # TODO: see if we want to do step-based inputs
         tools = self.get_tools(task.input)
         openai_tools = [tool.metadata.to_openai_tool() for tool in tools]
 
@@ -670,40 +669,46 @@ async def _arun_step(
             task, mode=mode, **llm_chat_kwargs
         )
 
-        # TODO: implement _should_continue
         latest_tool_calls = self.get_latest_tool_calls(task) or []
         latest_tool_outputs: List[ToolOutput] = []
 
         if not self._should_continue(
             latest_tool_calls, task.extra_state["n_function_calls"]
         ):
             is_done = True
-
         else:
             is_done = False
+
+            # Validate all tool calls first
             for tool_call in latest_tool_calls:
-                # Some validation
                 if not isinstance(tool_call, get_args(OpenAIToolCall)):
                     raise ValueError("Invalid tool_call object")
-
                 if tool_call.type != "function":
                     raise ValueError("Invalid tool type. Unsupported by OpenAI")
 
-                # TODO: maybe execute this with multi-threading
-                return_direct = await self._acall_function(
-                    tools,
-                    tool_call,
-                    task.extra_state["new_memory"],
-                    latest_tool_outputs,
-                )
+            # Execute all tool calls in parallel using asyncio.gather
+            tool_results = await asyncio.gather(
+                *[
+                    self._acall_function(
+                        tools,
+                        tool_call,
+                        task.extra_state["new_memory"],
+                        latest_tool_outputs,
+                    )
+                    for tool_call in latest_tool_calls
+                ]
+            )
+
+            # Process results
+            for return_direct in tool_results:
                 task.extra_state["sources"].append(latest_tool_outputs[-1])
 
-                # change function call to the default value, if a custom function was given
-                # as an argument (none and auto are predefined by OpenAI)
+                # change function call to the default value if a custom function was given
                 if tool_choice not in ("auto", "none"):
                     tool_choice = "auto"
                 task.extra_state["n_function_calls"] += 1
 
+                # If any tool call requests direct return and it's the only call
                 if return_direct and len(latest_tool_calls) == 1:
                     is_done = True
                     response_str = latest_tool_outputs[-1].content
@@ -723,7 +728,6 @@ async def _arun_step(
             [
                 step.get_next_step(
                     step_id=str(uuid.uuid4()),
-                    # NOTE: input is unused
                     input=None,
                 )
             ]

llama-index-integrations/agent/llama-index-agent-openai/pyproject.toml

@@ -28,7 +28,7 @@ exclude = ["**/BUILD"]
 license = "MIT"
 name = "llama-index-agent-openai"
 readme = "README.md"
-version = "0.4.0"
+version = "0.4.1"
 
 [tool.poetry.dependencies]
 python = ">=3.9,<4.0"