v0.11.x (#15398)

.github/workflows/build_package.yml

@@ -28,8 +28,6 @@ jobs:
         uses: actions/setup-python@v5
         with:
           python-version: ${{ matrix.python-version }}
-          cache: "poetry"
-          cache-dependency-path: "**/poetry.lock"
       - name: Install deps
         shell: bash
         run: poetry install

.gitignore

@@ -15,3 +15,4 @@ credentials.json
 token.json
 .python-version
 .DS_Store
+/storage/

CHANGELOG.md

 # ChangeLog
 
+## [2024-08-22]
+
+### `llama-index-core` [0.11.0]
+
+- removed deprecated `ServiceContext` -- using this now will print an error with a link to the migration guide
+- removed deprecated `LLMPredictor` -- using this now will print an error, any existing LLM is a drop-in replacement
+- made `pandas` an optional dependency
+
+### `Everything Else`
+
+- bumped the minor version of every package to account for the new version of `llama-index-core`
+
 ## [2024-08-21]
 
 ### `llama-index-core` [0.10.68]

docs/BUILD

+python_sources()
+
+poetry_requirements(
+    name="poetry",
+)
+
+python_requirements(
+    name="reqs",
+)

llama-index-cli/llama_index/cli/rag/base.py

@@ -7,6 +7,7 @@ from pathlib import Path
 from typing import Any, Callable, Dict, Optional, Union, cast
 
 from llama_index.core import (
+    Settings,
     SimpleDirectoryReader,
     VectorStoreIndex,
 )
@@ -16,9 +17,8 @@ from llama_index.core.base.response.schema import (
     StreamingResponse,
     Response,
 )
-from llama_index.core.bridge.pydantic import BaseModel, Field, validator
+from llama_index.core.bridge.pydantic import BaseModel, Field, field_validator
 from llama_index.core.chat_engine import CondenseQuestionChatEngine
-from llama_index.core.indices.service_context import ServiceContext
 from llama_index.core.ingestion import IngestionPipeline
 from llama_index.core.llms import LLM
 from llama_index.core.query_engine import CustomQueryEngine
@@ -100,7 +100,7 @@ class RagCLI(BaseModel):
     class Config:
         arbitrary_types_allowed = True
 
-    @validator("query_pipeline", always=True)
+    @field_validator("query_pipeline", mode="before")
     def query_pipeline_from_ingestion_pipeline(
         cls, query_pipeline: Any, values: Dict[str, Any]
     ) -> Optional[QueryPipeline]:
@@ -127,15 +127,13 @@ class RagCLI(BaseModel):
                     embed_model = transformation
                     break
 
-        service_context = ServiceContext.from_defaults(
-            llm=llm, embed_model=embed_model or "default"
-        )
+        Settings.llm = llm
+        Settings.embed_model = embed_model
+
         retriever = VectorStoreIndex.from_vector_store(
-            ingestion_pipeline.vector_store, service_context=service_context
+            ingestion_pipeline.vector_store,
         ).as_retriever(similarity_top_k=8)
-        response_synthesizer = CompactAndRefine(
-            service_context=service_context, streaming=True, verbose=verbose
-        )
+        response_synthesizer = CompactAndRefine(streaming=True, verbose=verbose)
 
         # define query pipeline
         query_pipeline = QueryPipeline(verbose=verbose)
@@ -151,7 +149,7 @@ class RagCLI(BaseModel):
         query_pipeline.add_link("query", "summarizer", dest_key="query_str")
         return query_pipeline
 
-    @validator("chat_engine", always=True)
+    @field_validator("chat_engine", mode="before")
     def chat_engine_from_query_pipeline(
         cls, chat_engine: Any, values: Dict[str, Any]
     ) -> Optional[CondenseQuestionChatEngine]:

llama-index-cli/llama_index/cli/upgrade/mappings.json

 {
   "StorageContext": "llama_index.core",
-  "ServiceContext": "llama_index.core",
   "ComposableGraph": "llama_index.core",
   "# indicesSummaryIndex": "llama_index.core",
   "VectorStoreIndex": "llama_index.core",

llama-index-cli/pyproject.toml

@@ -32,13 +32,13 @@ maintainers = [
 name = "llama-index-cli"
 packages = [{include = "llama_index/"}]
 readme = "README.md"
-version = "0.1.13"
+version = "0.3.0"
 
 [tool.poetry.dependencies]
 python = ">=3.8.1,<4.0"
-llama-index-core = "^0.10.11.post1"
-llama-index-embeddings-openai = "^0.1.1"
-llama-index-llms-openai = "^0.1.1"
+llama-index-core = "^0.11.0"
+llama-index-embeddings-openai = "^0.2.0"
+llama-index-llms-openai = "^0.2.0"
 
 [tool.poetry.group.dev.dependencies]
 black = {extras = ["jupyter"], version = "<=23.9.1,>=23.7.0"}

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

 """Init file of LlamaIndex."""
 
-__version__ = "0.10.68.post1"
+__version__ = "0.11.0"
 
 import logging
 from logging import NullHandler
@@ -152,8 +152,5 @@ global_handler: Optional[BaseCallbackHandler] = None
 # NOTE: keep for backwards compatibility
 SQLContextBuilder = SQLDocumentContextBuilder
 
-# global service context for ServiceContext.from_defaults()
-global_service_context: Optional[ServiceContext] = None
-
 # global tokenizer
 global_tokenizer: Optional[Callable[[str], list]] = None

llama-index-core/llama_index/core/agent/custom/pipeline_worker.py

@@ -15,7 +15,7 @@ from llama_index.core.agent.types import (
     TaskStepOutput,
 )
 from llama_index.core.base.query_pipeline.query import QueryComponent
-from llama_index.core.bridge.pydantic import BaseModel, Field
+from llama_index.core.bridge.pydantic import BaseModel, Field, ConfigDict
 from llama_index.core.callbacks import (
     CallbackManager,
     trace_method,
@@ -72,14 +72,12 @@ class QueryPipelineAgentWorker(BaseModel, BaseAgentWorker):
 
     """
 
+    model_config = ConfigDict(arbitrary_types_allowed=True)
     pipeline: QueryPipeline = Field(..., description="Query pipeline")
     callback_manager: CallbackManager = Field(..., exclude=True)
     task_key: str = Field("task", description="Key to store task in state")
     step_state_key: str = Field("step_state", description="Key to store step in state")
 
-    class Config:
-        arbitrary_types_allowed = True
-
     def __init__(
         self,
         pipeline: QueryPipeline,

llama-index-core/llama_index/core/agent/custom/simple.py

@@ -19,7 +19,7 @@ from llama_index.core.agent.types import (
     TaskStep,
     TaskStepOutput,
 )
-from llama_index.core.bridge.pydantic import BaseModel, Field, PrivateAttr
+from llama_index.core.bridge.pydantic import BaseModel, Field, PrivateAttr, ConfigDict
 from llama_index.core.callbacks import (
     CallbackManager,
     trace_method,
@@ -55,6 +55,7 @@ class CustomSimpleAgentWorker(BaseModel, BaseAgentWorker):
 
     """
 
+    model_config = ConfigDict(arbitrary_types_allowed=True)
     tools: Sequence[BaseTool] = Field(..., description="Tools to use for reasoning")
     llm: LLM = Field(..., description="LLM to use")
     callback_manager: CallbackManager = Field(
@@ -67,9 +68,6 @@ class CustomSimpleAgentWorker(BaseModel, BaseAgentWorker):
 
     _get_tools: Callable[[str], Sequence[BaseTool]] = PrivateAttr()
 
-    class Config:
-        arbitrary_types_allowed = True
-
     def __init__(
         self,
         tools: Sequence[BaseTool],
@@ -79,18 +77,7 @@ class CustomSimpleAgentWorker(BaseModel, BaseAgentWorker):
         tool_retriever: Optional[ObjectRetriever[BaseTool]] = None,
         **kwargs: Any,
     ) -> None:
-        if len(tools) > 0 and tool_retriever is not None:
-            raise ValueError("Cannot specify both tools and tool_retriever")
-        elif len(tools) > 0:
-            self._get_tools = lambda _: tools
-        elif tool_retriever is not None:
-            tool_retriever_c = cast(ObjectRetriever[BaseTool], tool_retriever)
-            self._get_tools = lambda message: tool_retriever_c.retrieve(message)
-        else:
-            self._get_tools = lambda _: []
-
         callback_manager = callback_manager or CallbackManager([])
-
         super().__init__(
             tools=tools,
             llm=llm,
@@ -100,6 +87,16 @@ class CustomSimpleAgentWorker(BaseModel, BaseAgentWorker):
             **kwargs,
         )
 
+        if len(tools) > 0 and tool_retriever is not None:
+            raise ValueError("Cannot specify both tools and tool_retriever")
+        elif len(tools) > 0:
+            self._get_tools = lambda _: tools
+        elif tool_retriever is not None:
+            tool_retriever_c = cast(ObjectRetriever[BaseTool], tool_retriever)
+            self._get_tools = lambda message: tool_retriever_c.retrieve(message)
+        else:
+            self._get_tools = lambda _: []
+
     @classmethod
     def from_tools(
         cls,

llama-index-core/llama_index/core/agent/custom/simple_function.py

@@ -16,7 +16,7 @@ from llama_index.core.agent.types import (
     TaskStep,
     TaskStepOutput,
 )
-from llama_index.core.bridge.pydantic import BaseModel, Field
+from llama_index.core.bridge.pydantic import BaseModel, Field, ConfigDict
 from llama_index.core.callbacks import (
     CallbackManager,
     trace_method,
@@ -44,6 +44,7 @@ class FnAgentWorker(BaseModel, BaseAgentWorker):
 
     """
 
+    model_config = ConfigDict(arbitrary_types_allowed=True)
     fn: Callable = Field(..., description="Function to run.")
     async_fn: Optional[Callable] = Field(
         None, description="Async function to run. If not provided, will run `fn`."
@@ -56,9 +57,6 @@ class FnAgentWorker(BaseModel, BaseAgentWorker):
 
     verbose: bool = Field(False, description="Verbose mode.")
 
-    class Config:
-        arbitrary_types_allowed = True
-
     def __init__(
         self,
         fn: Callable,

llama-index-core/llama_index/core/agent/react/formatter.py

@@ -13,7 +13,7 @@ from llama_index.core.agent.react.types import (
     ObservationReasoningStep,
 )
 from llama_index.core.base.llms.types import ChatMessage, MessageRole
-from llama_index.core.bridge.pydantic import BaseModel
+from llama_index.core.bridge.pydantic import BaseModel, ConfigDict
 from llama_index.core.tools import BaseTool
 
 logger = logging.getLogger(__name__)
@@ -36,8 +36,7 @@ def get_react_tool_descriptions(tools: Sequence[BaseTool]) -> List[str]:
 class BaseAgentChatFormatter(BaseModel):
     """Base chat formatter."""
 
-    class Config:
-        arbitrary_types_allowed = True
+    model_config = ConfigDict(arbitrary_types_allowed=True)
 
     @abstractmethod
     def format(

llama-index-core/llama_index/core/base/agent/types.py

@@ -7,7 +7,12 @@ from typing import Any, Dict, List, Optional, TYPE_CHECKING
 from llama_index.core.base.base_query_engine import BaseQueryEngine
 from llama_index.core.base.llms.types import ChatMessage
 from llama_index.core.base.response.schema import RESPONSE_TYPE, Response
-from llama_index.core.bridge.pydantic import BaseModel, Field
+from llama_index.core.bridge.pydantic import (
+    BaseModel,
+    Field,
+    SerializeAsAny,
+    ConfigDict,
+)
 from llama_index.core.callbacks import CallbackManager, trace_method
 from llama_index.core.chat_engine.types import (
     BaseChatEngine,
@@ -80,11 +85,11 @@ class TaskStep(BaseModel):
 
     """
 
-    task_id: str = Field(..., diescription="Task ID")
+    task_id: str = Field(..., description="Task ID")
     step_id: str = Field(..., description="Step ID")
     input: Optional[str] = Field(default=None, description="User input")
     # memory: BaseMemory = Field(
-    #     ..., type=BaseMemory, description="Conversational Memory"
+    #     ..., description="Conversational Memory"
     # )
     step_state: Dict[str, Any] = Field(
         default_factory=dict, description="Additional state for a given step."
@@ -155,25 +160,22 @@ class Task(BaseModel):
 
     """
 
-    class Config:
-        arbitrary_types_allowed = True
-
+    model_config = ConfigDict(arbitrary_types_allowed=True)
     task_id: str = Field(
-        default_factory=lambda: str(uuid.uuid4()), type=str, description="Task ID"
+        default_factory=lambda: str(uuid.uuid4()), description="Task ID"
     )
-    input: str = Field(..., type=str, description="User input")
+    input: str = Field(..., description="User input")
 
     # NOTE: this is state that may be modified throughout the course of execution of the task
-    memory: BaseMemory = Field(
+    memory: SerializeAsAny[BaseMemory] = Field(
         ...,
-        type=BaseMemory,
         description=(
             "Conversational Memory. Maintains state before execution of this task."
         ),
     )
 
     callback_manager: CallbackManager = Field(
-        default_factory=CallbackManager,
+        default_factory=lambda: CallbackManager([]),
         exclude=True,
         description="Callback manager for the task.",
     )
@@ -190,8 +192,7 @@ class Task(BaseModel):
 class BaseAgentWorker(PromptMixin, DispatcherSpanMixin):
     """Base agent worker."""
 
-    class Config:
-        arbitrary_types_allowed = True
+    model_config = ConfigDict(arbitrary_types_allowed=True)
 
     def _get_prompts(self) -> PromptDictType:
         """Get prompts."""

llama-index-core/llama_index/core/base/base_query_engine.py

@@ -12,7 +12,7 @@ from llama_index.core.base.query_pipeline.query import (
     validate_and_convert_stringable,
 )
 from llama_index.core.base.response.schema import RESPONSE_TYPE
-from llama_index.core.bridge.pydantic import Field
+from llama_index.core.bridge.pydantic import Field, ConfigDict, SerializeAsAny
 from llama_index.core.callbacks.base import CallbackManager
 from llama_index.core.prompts.mixin import PromptDictType, PromptMixin
 from llama_index.core.schema import NodeWithScore, QueryBundle, QueryType
@@ -108,10 +108,10 @@ class BaseQueryEngine(ChainableMixin, PromptMixin, DispatcherSpanMixin):
 class QueryEngineComponent(QueryComponent):
     """Query engine component."""
 
-    query_engine: BaseQueryEngine = Field(..., description="Query engine")
-
-    class Config:
-        arbitrary_types_allowed = True
+    model_config = ConfigDict(arbitrary_types_allowed=True)
+    query_engine: SerializeAsAny[BaseQueryEngine] = Field(
+        ..., description="Query engine"
+    )
 
     def set_callback_manager(self, callback_manager: CallbackManager) -> None:
         """Set callback manager."""

llama-index-core/llama_index/core/base/base_retriever.py

@@ -11,7 +11,7 @@ from llama_index.core.base.query_pipeline.query import (
     QueryComponent,
     validate_and_convert_stringable,
 )
-from llama_index.core.bridge.pydantic import Field
+from llama_index.core.bridge.pydantic import Field, ConfigDict
 from llama_index.core.callbacks.base import CallbackManager
 from llama_index.core.callbacks.schema import CBEventType, EventPayload
 from llama_index.core.prompts.mixin import (
@@ -27,7 +27,6 @@ from llama_index.core.schema import (
     QueryType,
     TextNode,
 )
-from llama_index.core.service_context import ServiceContext
 from llama_index.core.settings import Settings
 from llama_index.core.utils import print_text
 from llama_index.core.instrumentation import DispatcherSpanMixin
@@ -213,7 +212,10 @@ class BaseRetriever(ChainableMixin, PromptMixin, DispatcherSpanMixin):
         return [
             n
             for n in retrieved_nodes
-            if not ((n.node.hash, n.node.ref_doc_id) in seen or seen.add((n.node.hash, n.node.ref_doc_id)))  # type: ignore[func-returns-value]
+            if not (
+                (n.node.hash, n.node.ref_doc_id) in seen
+                or seen.add((n.node.hash, n.node.ref_doc_id))
+            )  # type: ignore[func-returns-value]
         ]
 
     @dispatcher.span
@@ -304,19 +306,6 @@ class BaseRetriever(ChainableMixin, PromptMixin, DispatcherSpanMixin):
         """
         return self._retrieve(query_bundle)
 
-    def get_service_context(self) -> Optional[ServiceContext]:
-        """Attempts to resolve a service context.
-        Short-circuits at self.service_context, self._service_context,
-        or self._index.service_context.
-        """
-        if hasattr(self, "service_context"):
-            return self.service_context
-        if hasattr(self, "_service_context"):
-            return self._service_context
-        elif hasattr(self, "_index") and hasattr(self._index, "service_context"):
-            return self._index.service_context
-        return None
-
     def _as_query_component(self, **kwargs: Any) -> QueryComponent:
         """Return a query component."""
         return RetrieverComponent(retriever=self)
@@ -325,11 +314,9 @@ class BaseRetriever(ChainableMixin, PromptMixin, DispatcherSpanMixin):
 class RetrieverComponent(QueryComponent):
     """Retriever component."""
 
+    model_config = ConfigDict(arbitrary_types_allowed=True)
     retriever: BaseRetriever = Field(..., description="Retriever")
 
-    class Config:
-        arbitrary_types_allowed = True
-
     def set_callback_manager(self, callback_manager: CallbackManager) -> None:
         """Set callback manager."""
         self.retriever.callback_manager = callback_manager

llama-index-core/llama_index/core/base/embeddings/base.py

@@ -6,7 +6,11 @@ from enum import Enum
 from typing import Any, Callable, Coroutine, List, Optional, Tuple
 
 import numpy as np
-from llama_index.core.bridge.pydantic import Field, validator
+from llama_index.core.bridge.pydantic import (
+    Field,
+    ConfigDict,
+    field_validator,
+)
 from llama_index.core.callbacks.base import CallbackManager
 from llama_index.core.callbacks.schema import CBEventType, EventPayload
 from llama_index.core.constants import (
@@ -63,6 +67,9 @@ def similarity(
 class BaseEmbedding(TransformComponent, DispatcherSpanMixin):
     """Base class for embeddings."""
 
+    model_config = ConfigDict(
+        protected_namespaces=("pydantic_model_",), arbitrary_types_allowed=True
+    )
     model_name: str = Field(
         default="unknown", description="The name of the embedding model."
     )
@@ -70,7 +77,7 @@ class BaseEmbedding(TransformComponent, DispatcherSpanMixin):
         default=DEFAULT_EMBED_BATCH_SIZE,
         description="The batch size for embedding calls.",
         gt=0,
-        lte=2048,
+        le=2048,
     )
     callback_manager: CallbackManager = Field(
         default_factory=lambda: CallbackManager([]), exclude=True
@@ -80,13 +87,9 @@ class BaseEmbedding(TransformComponent, DispatcherSpanMixin):
         description="The number of workers to use for async embedding calls.",
     )
 
-    class Config:
-        arbitrary_types_allowed = True
-
-    @validator("callback_manager", pre=True)
-    def _validate_callback_manager(
-        cls, v: Optional[CallbackManager]
-    ) -> CallbackManager:
+    @field_validator("callback_manager")
+    @classmethod
+    def check_callback_manager(cls, v: CallbackManager) -> CallbackManager:
         if v is None:
             return CallbackManager([])
         return v

llama-index-core/llama_index/core/base/llms/base.py

@@ -17,7 +17,7 @@ from llama_index.core.base.llms.types import (
 from llama_index.core.base.query_pipeline.query import (
     ChainableMixin,
 )
-from llama_index.core.bridge.pydantic import Field, validator
+from llama_index.core.bridge.pydantic import Field, model_validator, ConfigDict
 from llama_index.core.callbacks import CallbackManager
 from llama_index.core.instrumentation import DispatcherSpanMixin
 from llama_index.core.schema import BaseComponent
@@ -26,18 +26,16 @@ from llama_index.core.schema import BaseComponent
 class BaseLLM(ChainableMixin, BaseComponent, DispatcherSpanMixin):
     """BaseLLM interface."""
 
+    model_config = ConfigDict(arbitrary_types_allowed=True)
     callback_manager: CallbackManager = Field(
-        default_factory=CallbackManager, exclude=True
+        default_factory=lambda: CallbackManager([]), exclude=True
     )
 
-    class Config:
-        arbitrary_types_allowed = True
-
-    @validator("callback_manager", pre=True)
-    def _validate_callback_manager(cls, v: CallbackManager) -> CallbackManager:
-        if v is None:
-            return CallbackManager([])
-        return v
+    @model_validator(mode="after")
+    def check_callback_manager(self) -> "BaseLLM":
+        if self.callback_manager is None:
+            self.callback_manager = CallbackManager([])
+        return self
 
     @property
     @abstractmethod

llama-index-core/llama_index/core/base/llms/types.py

 from enum import Enum
 from typing import Any, AsyncGenerator, Generator, Optional, Union, List, Any
 
-from llama_index.core.bridge.pydantic import BaseModel, Field
+from llama_index.core.bridge.pydantic import BaseModel, Field, ConfigDict
 from llama_index.core.constants import DEFAULT_CONTEXT_WINDOW, DEFAULT_NUM_OUTPUTS
 
 try:
@@ -49,7 +49,7 @@ class ChatMessage(BaseModel):
 
     def _recursive_serialization(self, value: Any) -> Any:
         if isinstance(value, (V1BaseModel, V2BaseModel)):
-            return value.dict()
+            return value.model_dump()
         if isinstance(value, dict):
             return {
                 key: self._recursive_serialization(value)
@@ -60,8 +60,11 @@ class ChatMessage(BaseModel):
         return value
 
     def dict(self, **kwargs: Any) -> dict:
+        return self.model_dump(**kwargs)
+
+    def model_dump(self, **kwargs: Any) -> dict:
         # ensure all additional_kwargs are serializable
-        msg = super().dict(**kwargs)
+        msg = super().model_dump(**kwargs)
 
         for key, value in msg.get("additional_kwargs", {}).items():
             value = self._recursive_serialization(value)
@@ -129,6 +132,9 @@ CompletionResponseAsyncGen = AsyncGenerator[CompletionResponse, None]
 
 
 class LLMMetadata(BaseModel):
+    model_config = ConfigDict(
+        protected_namespaces=("pydantic_model_",), arbitrary_types_allowed=True
+    )
     context_window: int = Field(
         default=DEFAULT_CONTEXT_WINDOW,
         description=(

llama-index-core/llama_index/core/base/query_pipeline/query.py

@@ -20,7 +20,7 @@ from llama_index.core.base.llms.types import (
     CompletionResponse,
 )
 from llama_index.core.base.response.schema import Response
-from llama_index.core.bridge.pydantic import BaseModel, Field
+from llama_index.core.bridge.pydantic import BaseModel, Field, ConfigDict
 from llama_index.core.callbacks.base import CallbackManager
 from llama_index.core.schema import NodeWithScore, QueryBundle, TextNode
 
@@ -243,13 +243,11 @@ class QueryComponent(BaseModel):
 class CustomQueryComponent(QueryComponent):
     """Custom query component."""
 
+    model_config = ConfigDict(arbitrary_types_allowed=True)
     callback_manager: CallbackManager = Field(
         default_factory=CallbackManager, description="Callback manager"
     )
 
-    class Config:
-        arbitrary_types_allowed = True
-
     def set_callback_manager(self, callback_manager: CallbackManager) -> None:
         """Set callback manager."""
         self.callback_manager = callback_manager

llama-index-core/llama_index/core/base/response/schema.py

@@ -58,11 +58,11 @@ class PydanticResponse:
 
     def __str__(self) -> str:
         """Convert to string representation."""
-        return self.response.json() if self.response else "None"
+        return self.response.model_dump_json() if self.response else "None"
 
     def __getattr__(self, name: str) -> Any:
         """Get attribute, but prioritize the pydantic  response object."""
-        if self.response is not None and name in self.response.dict():
+        if self.response is not None and name in self.response.model_dump():
             return getattr(self.response, name)
         else:
             return None
@@ -97,7 +97,7 @@ class PydanticResponse:
 
     def get_response(self) -> Response:
         """Get a standard response object."""
-        response_txt = self.response.json() if self.response else "None"
+        response_txt = self.response.model_dump_json() if self.response else "None"
         return Response(response_txt, self.source_nodes, self.metadata)