blaxel 0.2.37__py3-none-any.whl → 0.2.38rc122__py3-none-any.whl

blaxel/__init__.py

@@ -4,8 +4,8 @@ from .core.common.autoload import autoload
 from .core.common.env import env
 from .core.common.settings import settings
 
-__version__ = "0.2.37"
-__commit__ = "a09d7ea6c786074ceaa4639ce6415958524a7a5f"
+__version__ = "0.2.38.pre-122"
+__commit__ = "e7e19d2c82b2b0c2a44914245734597adcc00f49"
 __sentry_dsn__ = "https://9711de13cd02b285ca4378c01de8dc30@o4508714045276160.ingest.us.sentry.io/4510461121462272"
 __all__ = ["autoload", "settings", "env"]
 

blaxel/llamaindex/model.py

@@ -2,7 +2,7 @@ from __future__ import annotations
 
 import os
 from logging import getLogger
-from typing import TYPE_CHECKING, Any, Sequence
+from typing import TYPE_CHECKING, Any, Dict, List, Sequence, Union
 
 from blaxel.core import bl_model as bl_model_core
 from blaxel.core import settings
@@ -20,20 +20,61 @@ if TYPE_CHECKING:
         CompletionResponseAsyncGen,
         CompletionResponseGen,
     )
+    from llama_index.core.llms.llm import ToolSelection
+    from llama_index.core.tools.types import BaseTool
+
+# Runtime imports needed for class inheritance and construction
+from llama_index.core.base.llms.types import LLMMetadata  # noqa: E402
+from llama_index.core.llms.function_calling import FunctionCallingLLM  # noqa: E402
+from pydantic import PrivateAttr  # noqa: E402
 
 logger = getLogger(__name__)
 
+DEFAULT_CONTEXT_WINDOW = 128000
+DEFAULT_NUM_OUTPUT = 4096
+
+
+class TokenRefreshingLLM(FunctionCallingLLM):
+    """Wrapper for LlamaIndex LLMs that refreshes token before each call.
 
-class TokenRefreshingWrapper:
-    """Base wrapper class that refreshes token before each call."""
+    Inherits from FunctionCallingLLM to maintain type compatibility with
+    LlamaIndex's agents and components that validate isinstance(model, LLM).
+    """
+
+    _model_config_data: dict = PrivateAttr(default_factory=dict)
+    _wrapped: Any = PrivateAttr(default=None)
 
     def __init__(self, model_config: dict):
-        self.model_config = model_config
-        self.wrapped_model = self._create_model()
+        super().__init__()
+        self._model_config_data = model_config
+        self._wrapped = self._create_model()
+
+    @classmethod
+    def class_name(cls) -> str:
+        return "TokenRefreshingLLM"
+
+    @property
+    def wrapped_model(self) -> Any:
+        """Access the underlying wrapped LLM model."""
+        return self._wrapped
+
+    @property
+    def metadata(self) -> LLMMetadata:
+        """Get LLM metadata, with fallback for unknown model names."""
+        try:
+            return self._wrapped.metadata
+        except (ValueError, KeyError) as e:
+            logger.warning(f"Could not get metadata from wrapped model: {e}. Using defaults.")
+            return LLMMetadata(
+                context_window=DEFAULT_CONTEXT_WINDOW,
+                num_output=DEFAULT_NUM_OUTPUT,
+                is_chat_model=True,
+                model_name=self._model_config_data.get("model", "unknown"),
+            )
 
     def _create_model(self):
         """Create the model instance with current token."""
-        config = self.model_config
+        config = self._model_config_data
         model_type = config["type"]
         model = config["model"]
         url = config["url"]
@@ -115,102 +156,106 @@ class TokenRefreshingWrapper:
 
     def _refresh_token(self):
         """Refresh the token and recreate the model if needed."""
-        # Only refresh if using ClientCredentials (which has get_token method)
         current_token = settings.auth.token
 
-        if hasattr(settings.auth, "get_token"):
-            # This will trigger token refresh if needed
-            settings.auth.get_token()
-
         new_token = settings.auth.token
 
-        # If token changed, recreate the model
         if current_token != new_token:
-            self.wrapped_model = self._create_model()
+            self._wrapped = self._create_model()
 
-    def __getattr__(self, name):
-        """Delegate attribute access to wrapped model."""
-        return getattr(self.wrapped_model, name)
+    # --- Core LLM methods with token refresh ---
 
+    def chat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
+        self._refresh_token()
+        return self._wrapped.chat(messages, **kwargs)
 
-class TokenRefreshingLLM(TokenRefreshingWrapper):
-    """Wrapper for LlamaIndex LLMs that refreshes token before each call."""
-
-    async def achat(
-        self,
-        messages: Sequence[ChatMessage],
-        **kwargs: Any,
-    ) -> ChatResponse:
-        """Async chat with token refresh."""
+    async def achat(self, messages: Sequence[ChatMessage], **kwargs: Any) -> ChatResponse:
         self._refresh_token()
-        return await self.wrapped_model.achat(messages, **kwargs)
+        return await self._wrapped.achat(messages, **kwargs)
 
-    def chat(
-        self,
-        messages: Sequence[ChatMessage],
-        **kwargs: Any,
-    ) -> ChatResponse:
-        """Sync chat with token refresh."""
+    def complete(self, prompt: str, formatted: bool = False, **kwargs: Any) -> CompletionResponse:
         self._refresh_token()
-        return self.wrapped_model.chat(messages, **kwargs)
+        return self._wrapped.complete(prompt, formatted=formatted, **kwargs)
 
-    async def astream_chat(
-        self,
-        messages: Sequence[ChatMessage],
-        **kwargs: Any,
-    ) -> ChatResponseAsyncGen:
-        """Async stream chat with token refresh."""
+    async def acomplete(
+        self, prompt: str, formatted: bool = False, **kwargs: Any
+    ) -> CompletionResponse:
         self._refresh_token()
-        async for chunk in self.wrapped_model.astream_chat(messages, **kwargs):
-            yield chunk
+        return await self._wrapped.acomplete(prompt, formatted=formatted, **kwargs)
 
     def stream_chat(
-        self,
-        messages: Sequence[ChatMessage],
-        **kwargs: Any,
+        self, messages: Sequence[ChatMessage], **kwargs: Any
     ) -> ChatResponseGen:
-        """Sync stream chat with token refresh."""
         self._refresh_token()
-        for chunk in self.wrapped_model.stream_chat(messages, **kwargs):
-            yield chunk
+        return self._wrapped.stream_chat(messages, **kwargs)
 
-    async def acomplete(
-        self,
-        prompt: str,
-        **kwargs: Any,
-    ) -> CompletionResponse:
-        """Async complete with token refresh."""
+    async def astream_chat(
+        self, messages: Sequence[ChatMessage], **kwargs: Any
+    ) -> ChatResponseAsyncGen:
         self._refresh_token()
-        return await self.wrapped_model.acomplete(prompt, **kwargs)
+        result = self._wrapped.astream_chat(messages, **kwargs)
+        # Handle both coroutine and async generator patterns
+        if hasattr(result, "__aiter__"):
+            return result
+        return await result
 
-    def complete(
-        self,
-        prompt: str,
-        **kwargs: Any,
-    ) -> CompletionResponse:
-        """Sync complete with token refresh."""
+    def stream_complete(
+        self, prompt: str, formatted: bool = False, **kwargs: Any
+    ) -> CompletionResponseGen:
         self._refresh_token()
-        return self.wrapped_model.complete(prompt, **kwargs)
+        return self._wrapped.stream_complete(prompt, formatted=formatted, **kwargs)
 
     async def astream_complete(
-        self,
-        prompt: str,
-        **kwargs: Any,
+        self, prompt: str, formatted: bool = False, **kwargs: Any
     ) -> CompletionResponseAsyncGen:
-        """Async stream complete with token refresh."""
         self._refresh_token()
-        async for chunk in self.wrapped_model.astream_complete(prompt, **kwargs):
-            yield chunk
+        result = self._wrapped.astream_complete(prompt, formatted=formatted, **kwargs)
+        # Handle both coroutine and async generator patterns
+        if hasattr(result, "__aiter__"):
+            return result
+        return await result
 
-    def stream_complete(
+    # --- FunctionCallingLLM methods (delegate to wrapped model) ---
+
+    def _prepare_chat_with_tools(
         self,
-        prompt: str,
+        tools: Sequence[BaseTool],
+        user_msg: Union[str, ChatMessage, None] = None,
+        chat_history: List[ChatMessage] | None = None,
+        verbose: bool = False,
+        allow_parallel_tool_calls: bool = False,
+        tool_required: Any = None,
         **kwargs: Any,
-    ) -> CompletionResponseGen:
-        """Sync stream complete with token refresh."""
-        self._refresh_token()
-        for chunk in self.wrapped_model.stream_complete(prompt, **kwargs):
-            yield chunk
+    ) -> Dict[str, Any]:
+        if hasattr(self._wrapped, "_prepare_chat_with_tools"):
+            return self._wrapped._prepare_chat_with_tools(
+                tools,
+                user_msg=user_msg,
+                chat_history=chat_history,
+                verbose=verbose,
+                allow_parallel_tool_calls=allow_parallel_tool_calls,
+                tool_required=tool_required,
+                **kwargs,
+            )
+        raise NotImplementedError(
+            f"The wrapped model ({type(self._wrapped).__name__}) does not support function calling"
+        )
+
+    def get_tool_calls_from_response(
+        self,
+        response: ChatResponse,
+        error_on_no_tool_call: bool = True,
+        **kwargs: Any,
+    ) -> List[ToolSelection]:
+        if hasattr(self._wrapped, "get_tool_calls_from_response"):
+            return self._wrapped.get_tool_calls_from_response(
+                response,
+                error_on_no_tool_call=error_on_no_tool_call,
+                **kwargs,
+            )
+        raise NotImplementedError(
+            f"The wrapped model ({type(self._wrapped).__name__}) does not support function calling"
+        )
 
 
 async def bl_model(name, **kwargs):

{blaxel-0.2.37.dist-info → blaxel-0.2.38rc122.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: blaxel
-Version: 0.2.37
+Version: 0.2.38rc122
 Summary: Blaxel - AI development platform SDK
 Project-URL: Homepage, https://blaxel.ai
 Project-URL: Documentation, https://docs.blaxel.ai

{blaxel-0.2.37.dist-info → blaxel-0.2.38rc122.dist-info}/RECORD

@@ -1,4 +1,4 @@
-blaxel/__init__.py,sha256=UGO2L0cKhLfwXPewJ7MF6x92gMzo_-lJRzE-2ApXDF0,413
+blaxel/__init__.py,sha256=CJtHkLCfIuSmNEqG5afzbxuceMG5h0ve2VADXBGizWs,421
 blaxel/core/__init__.py,sha256=CU0gXpVRbuQZNWoCJuuhZS0ZhXPEu0cg-3XzoYMrBm4,1756
 blaxel/core/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 blaxel/core/agents/__init__.py,sha256=MJZga99lU8JWUUPHd4rmUfdo7ALwWgF7CQq95SfT2OI,4456
@@ -446,7 +446,7 @@ blaxel/livekit/model.py,sha256=Qhl3EVa4Uum1IsT3nyMxJwXrZn2eDvCyj6V8RHOaJsQ,1687
 blaxel/livekit/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 blaxel/livekit/tools.py,sha256=QipxGDnKqma_ktzTUpKgzdp5x4bvpQnxEVoqSTBesu8,1053
 blaxel/llamaindex/__init__.py,sha256=iZ3QbZhlwKvP91ChcqSXVkpRrzurMxJoQfKdZFzE2AA,127
-blaxel/llamaindex/model.py,sha256=_wsCiGScqXeMHbxlRDtl6ucwQsrGGM4AGHNFBdIWu1o,7003
+blaxel/llamaindex/model.py,sha256=wDWEXn2oLCjdsJ9aBN6po6-LuiSuYa3-V6SRxBMTKG0,9405
 blaxel/llamaindex/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 blaxel/llamaindex/tools.py,sha256=5wnPX45sgLduk9fPkhQ0rHRGK3Ekrye0MJNtMdw3GII,1024
 blaxel/llamaindex/custom/cohere.py,sha256=8Kfo5BjrND6EFDZXosHNrgrg98ktvxhYG48ztaWptps,18707
@@ -472,7 +472,7 @@ blaxel/telemetry/instrumentation/map.py,sha256=PCzZJj39yiYVYJrxLBNP-NW-tjjYyTijw
 blaxel/telemetry/instrumentation/utils.py,sha256=FGyMY5ZE4f-0JdZpm_R_BCoKLJ18hftz8vsh7ftDwMk,1889
 blaxel/telemetry/log/log.py,sha256=vtzUIFIIj4MTTKUigILDYXN8NHHPOo44OaKukpyIjQg,2407
 blaxel/telemetry/log/logger.py,sha256=IcFWCd1yyWWGAjAd2i0pDYqpZHQ61pmcaQ7Kf4bC8lg,4150
-blaxel-0.2.37.dist-info/METADATA,sha256=Donv60dGoQx0kX2i5GoNSEtgifJIPLgVkhKAe5WZybY,18684
-blaxel-0.2.37.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
-blaxel-0.2.37.dist-info/licenses/LICENSE,sha256=p5PNQvpvyDT_0aYBDgmV1fFI_vAD2aSV0wWG7VTgRis,1069
-blaxel-0.2.37.dist-info/RECORD,,
+blaxel-0.2.38rc122.dist-info/METADATA,sha256=--Hscbw_W5MLB8BsePm4W1NIyMklqQuO-jdV0EBoI1E,18689
+blaxel-0.2.38rc122.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+blaxel-0.2.38rc122.dist-info/licenses/LICENSE,sha256=p5PNQvpvyDT_0aYBDgmV1fFI_vAD2aSV0wWG7VTgRis,1069
+blaxel-0.2.38rc122.dist-info/RECORD,,