Merge pull request #43 from LyzrCore/imp/data-analyzr

DataAnalyzr improvements - big PR

build/lib/lyzr/__init__.py

@@ -5,7 +5,6 @@
 from lyzr.base.vector_store import LyzrVectorStoreIndex
 from lyzr.formula_generator import FormulaGen
 from lyzr.data_analyzr import DataAnalyzr
-from lyzr.data_analyzr import DataConnector
 from lyzr.voicebot import VoiceBot
 from lyzr.qa.search_agent import SearchAgent
 from lyzr.summarizer import Summarizer
@@ -19,7 +18,6 @@
     "ChatBot",
     "FormulaGen",
     "DataAnalyzr",
-    "DataConnector",
     "VoiceBot",
     "SearchAgent",
     "Summarizer",

build/lib/lyzr/base/__init__.py

@@ -1,4 +1,3 @@
-from lyzr.base.file_utils import read_file, describe_dataset
 from lyzr.base.llm import LyzrLLMFactory, LiteLLM
 from lyzr.base.llms import LLM, get_model
 from lyzr.base.service import LyzrService
@@ -13,8 +12,6 @@
     "LyzrVectorStoreIndex",
     "LLM",
     "get_model",
-    "read_file",
-    "describe_dataset",
     "LyzrRetriever",
     "LiteLLM",
     "LyzrPromptFactory",

build/lib/lyzr/base/errors.py

@@ -35,3 +35,15 @@ class ValidationError(Exception):
     """Raise for validations"""
 
     pass
+
+
+class AnalysisFailedError(Exception):
+    """Raise for failed analysis"""
+
+    pass
+
+
+class PromptError(ValueError):
+    """Raise for prompt errors"""
+
+    pass

build/lib/lyzr/base/llm.py

@@ -1,8 +1,14 @@
+"""
+Classes and functions for interacting with LLMs.
+"""
+
 # standard library imports
 import logging
+import traceback
 from typing import Union, Literal, Sequence
 
 # third-party imports
+import litellm
 from llama_index.llms import LiteLLM
 from llama_index.llms.base import LLM
 from llama_index.llms.base import ChatMessage as LlamaChatMessage
@@ -12,18 +18,49 @@
 from lyzr.base.errors import ImproperUsageError
 
 DEFAULT_LLM = "gpt-4-0125-preview"
+litellm.drop_params = True
 
 
 class LyzrLLMFactory:
+    """A factory class for creating instances of LiteLLM."""
+
     @staticmethod
-    def from_defaults(model: str = DEFAULT_LLM, **kwargs) -> LLM:
+    def from_defaults(model: str = DEFAULT_LLM, **kwargs) -> LiteLLM:
         # model_type -> api_type
         # model_name -> model
         # model_prompts -> Sequence[ChatMessage]
         return LiteLLM(model=model, **kwargs)
 
 
 class LiteLLM(LiteLLM):
+    """
+    LiteLLM is a lightweight language model interface that supports chat,
+    text-to-speech (TTS), and speech-to-text (STT) functionalities.
+    Extends the `LiteLLM` class.
+
+    Properties:
+        _tts_kwargs (dict): Returns a dictionary of keyword arguments for TTS, including the voice setting.
+        _model_type (Literal["chat", "tts", "stt"]): Determines the type of model based on the model name.
+
+    Methods:
+        set_model_kwargs(model_kwargs: dict, force: Union[bool, dict] = True) -> dict:
+            Sets the model-specific keyword arguments. If `force` is a boolean, it applies to all arguments; otherwise, it can be a dictionary specifying which arguments to forcefully set.
+
+        set_messages(messages: Sequence[ChatMessage]):
+            Sets the messages for the chat model.
+
+        run(**kwargs):
+            Executes the model based on its type (chat, TTS, or STT) with the provided keyword arguments.
+
+        chat_complete(messages: Sequence[ChatMessage], stream: bool = False, logger: logging.Logger = None, **kwargs):
+            Completes a chat interaction with the provided messages. Supports streaming and logging.
+
+        tts(tts_input, voice: Literal["echo", "alloy", "fable", "onyx", "nova", "shimmer"], **kwargs):
+            Converts text to speech using the specified voice. Requires input text and voice.
+
+        stt(audiofile, **kwargs):
+            Converts speech to text from the provided audio file.
+    """
 
     @property
     def _tts_kwargs(self) -> dict:
@@ -50,7 +87,7 @@ def set_model_kwargs(
             if (not force.get(arg, True)) and (arg in all_kwargs):
                 continue
             if arg in ["temperature", "max_tokens"]:
-                self.__dict__[arg] = model_kwargs[arg]
+                setattr(self, arg, model_kwargs[arg])
             self.additional_kwargs[arg] = model_kwargs[arg]
 
     def set_messages(self, messages: Sequence[ChatMessage]):
@@ -59,7 +96,7 @@ def set_messages(self, messages: Sequence[ChatMessage]):
     def run(self, **kwargs):
         if self._model_type == "chat":
             return self.chat_complete(
-                messages=kwargs.pop("messages", self.__dict__.get("messages", None)),
+                messages=kwargs.pop("messages", getattr(self, "messages", None)),
                 stream=kwargs.pop("stream", False),
                 logger=self.additional_kwargs.pop("logger", kwargs.pop("logger", None)),
                 **kwargs,
@@ -84,22 +121,50 @@ def chat_complete(
         if not messages:
             raise ImproperUsageError("Please provide messages for chat.")
         llama_messages = [
-            LlamaChatMessage(role=msg.role.value, content=msg.content)
+            LlamaChatMessage(role=msg.role.value, content=msg.content.strip())
             for msg in messages
         ]
         if stream:
-            response = self._stream_chat(messages=llama_messages, **kwargs)
-            return response  # TODO: Convert response to ChatMessage
+            return self._stream_chat(messages=llama_messages, **kwargs)
         else:
-            response = self._chat(messages=llama_messages, **kwargs)
-            return ChatResponse(
-                message=ChatMessage(
-                    role=response.message.role, content=response.message.content
-                ),
-                raw=response.raw,
-                delta=response.delta,
-                additional_kwargs=response.additional_kwargs,
-            )
+            try:
+                response = self._chat(messages=llama_messages, **kwargs)
+                logger.info(
+                    f"LLM chat response received: {response.message.content}",
+                    extra={
+                        "function": "chat_complete",
+                        "input_kwargs": {
+                            "messages": messages,
+                            "stream": stream,
+                            **kwargs,
+                        },
+                        "response": response.message.content,
+                    },
+                )
+                return ChatResponse(
+                    message=ChatMessage(
+                        role=response.message.role, content=response.message.content
+                    ),
+                    raw=response.raw,
+                    delta=response.delta,
+                    additional_kwargs=response.additional_kwargs,
+                )
+            except Exception as e:
+                logger.error(
+                    f"Error with getting response from LLM. {e.__class__.__name__}: {e}.",
+                    extra={
+                        "function": "chat_complete",
+                        "traceback": traceback.format_exc().splitlines(),
+                        "input_kwargs": {
+                            "messages": messages,
+                            "stream": stream,
+                            **kwargs,
+                        },
+                    },
+                )
+                return None
+            finally:
+                self.additional_kwargs["logger"] = logger
 
     def tts(
         self,

build/lib/lyzr/base/llms.py

@@ -6,7 +6,7 @@
 from openai import OpenAI
 
 # local imports
-from lyzr.base.prompt_dep import get_prompt_text
+from lyzr.base.prompt_old import get_prompt_text
 from lyzr.base.errors import MissingValueError