Merge pull request #3 from Surajrs812/dev

Dev

config/config.yaml

@@ -1,7 +1,7 @@
 cloud_provider: "groq" #"ollama"
-llm_provider: "groq" #"ollama"
-llm_model: "llama-3.1-70b-versatile" #gemma:2b"
-embedding_model: "mxbai-embed-large"
+llm_provider: "ollama" #"ollama"
+llm_model: "gemma2:2b" #"llama-3.1-70b-versatile" #gemma:2b"
+embedding_model: "nomic-embed-text:latest"
 
 token_limit: 1000
 target_words: 100

src/doc_loaders/doc_loader.py

@@ -1,4 +1,8 @@
 from langchain_community.document_loaders import PyPDFLoader, WebBaseLoader, DirectoryLoader
+# from .multimedia_loader import MultimediaLoader
+import logging
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
 
 class DocumentLoader:
     def __init__(self, source: str, type: str, filter: str = None):
@@ -34,8 +38,11 @@ def load(self) -> None:
             self._load_webpage()
         elif self.type == "directory":
             self._load_directory()
-        else:
-            raise ValueError("Invalid source. Must be a PDF file or a URL.")
+        # else:
+            # loader = MultimediaLoader(self.source)
+            # loader()
+
+
 
     def _load_pdf(self) -> None:
         """

src/doc_loaders/multimedia_loader.py

@@ -1,11 +1,13 @@
 # We will assume that the models are cached already
 
 import os
+import sys 
+sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), '..', '..')))
 import tempfile
 from tqdm import tqdm
 import logging
 from src.chunking.audiochunking import split_audio
-import whisper
+import fast_whisper
 from pytubefix import YouTube
 
 logging.basicConfig(level=logging.INFO)

src/models/models.py

@@ -4,8 +4,10 @@
 from dotenv import load_dotenv
 from langchain_groq import ChatGroq
 from langchain_community.embeddings import OllamaEmbeddings
-import openai 
-import ollama 
+# import openai  
+from langchain_ollama import ChatOllama
+# from langchain_openai import OpenAI
+from langchain_openai import AzureOpenAI
 
 # Set up logging
 logging.basicConfig(level=logging.INFO)
@@ -18,6 +20,7 @@ def __init__(self, config_path):
         :param config_path: Path to the YAML configuration file.
         """
         load_dotenv()
+
         try:
             with open(config_path, 'r') as file:
                 self.config = yaml.safe_load(file)
@@ -26,81 +29,90 @@ def __init__(self, config_path):
             logger.error("Failed to load configuration: %s", e)
             raise
 
-        self.llm_groq = None
+        self.llm = None
         self.embedding_model = None
 
-        if self.config['cloud_provider'] == 'groq':
+        if self.config['llm_provider'] == 'groq':
             self.load_llm_groq()
 
-        if self.config['cloud_provider'] == 'ollama':
+        if self.config['llm_provider'] == 'ollama':
             self.load_llm_ollama()
 
-        if self.config['cloud_provider'] == 'openai':
+        if self.config['llm_provider'] == 'openai':
             self.load_llm_openai()
 
 
-    def get_llm_response(self, prompt):
-        if self.config['cloud_provider'] == 'groq':
-            return self.llm_groq.invoke(prompt).content
-        if self.config['cloud_provider'] == 'ollama':
-            return self.llm_ollama(prompt)
-        if self.config['cloud_provider'] == 'openai':
-            return self.llm_openai(prompt)
+    # def get_llm_response(self):
+    #     if self.config['llm_provider'] == 'groq':
+    #         return self.llm_groq.invoke(prompt).content
+    #     if self.config['llm_provider'] == 'ollama':
+    #         return self.llm_ollama(prompt)
+    #     if self.config['llm_provider'] == 'openai':
+    #         return self.llm_openai(prompt)
 
     def load_llm_groq(self):
         """
         Lazily loads the LLM Groq model based on the configuration if it hasn't been loaded yet.
         :return: The loaded LLM Groq model.
         """
-        if not self.llm_groq:
+        if not self.llm:
             try:
                 logger.info("Loading Groq LLM model...")
-                self.llm_groq = ChatGroq(
+                self.llm = ChatGroq(
                     model_name=self.config['llm_model'],
                     api_key=os.getenv("GROQ_API_KEY")
                 )
                 logger.info("Groq LLM model loaded successfully.")
+
             except KeyError as e:
                 logger.error("Missing required config key for LLM: %s", e)
                 raise
             except Exception as e:
                 logger.error("Error loading Groq LLM model: %s", e)
                 raise
 
-    def load_llm_openai(self,prompt):
+    def load_llm_openai(self):
         """
         Loads the OpenAI model.
         :return: The loaded OpenAI model.
         """
         if not self.llm:
             try:
                 logger.info("Loading OpenAI model...")
-                openai.api_key = os.getenv("OPENAI_API_KEY")
-
-                completion = openai.ChatCompletion.create(
-                model=self.config['llm_model'],  # Make sure the model is set in config
-                messages=prompt
-            )
+                # openai.api_key = os.getenv("OPENAI_API_KEY")
+                # os.environ["OPENAI_API_VERSION"] = "2023-12-01-preview"
+                # os.environ["AZURE_OPENAI_ENDPOINT"] = "..."
+                # os.environ["AZURE_OPENAI_API_KEY"] = "..."
+                print("AZURE_OPENAI_DEPLOYMENT:", os.getenv("AZURE_OPENAI_DEPLOYMENT"))
+                load_dotenv()
+                # print("AZURE_OPENAI_DEPLOYMENT:", os.getenv("AZURE_OPENAI_DEPLOYMENT"))
+                self.llm = AzureOpenAI(
+                    deployment_name=os.getenv("AZURE_OPENAI_DEPLOYMENT"),  # The deployment name of the model
+                    )
                 logger.info("OpenAI model loaded successfully.")
-                return completion.choices[0].message['content']
+
             except KeyError as e:
                 logger.error("Missing required config key for OpenAI LLM: %s", e)
                 raise
             except Exception as e:
                 logger.error("Error loading OpenAI LLM model: %s", e)
                 raise
 
-    def load_llm_ollama(self,prompt):
+    def load_llm_ollama(self):
         """
         Loads the Ollama LLM model.
         :return: The loaded Ollama LLM model.
         """
         if not self.llm:
             try:
                 logger.info("Loading Ollama LLM model...")
-                response = ollama.chat(model=self.config['llm_model'], messages=prompt)
+
+                self.llm = ChatOllama(
+                                model=self.config['llm_model'],
+                                temperature=0,
+                                )
                 logger.info("Ollama LLM model loaded successfully.")
-                return response['message']['content']             
+
             except KeyError as e:
                 logger.error("Missing required config key for Ollama LLM: %s", e)
                 raise
@@ -132,12 +144,12 @@ def count_tokens(self, text):
         :param text: The text to count tokens for.
         :return: Number of tokens in the text.
         """
-        if not self.llm_groq:
+        if not self.llm:
             logger.warning("Groq LLM model is not loaded. Loading the model first...")
             self.load_llm_groq()
 
         try:
-            num_tokens = self.llm_groq.get_num_tokens(text)
+            num_tokens = self.llm.get_num_tokens(text)
             logger.info("Successfully counted %d tokens for the given text.", num_tokens)
             return num_tokens
         except Exception as e:
@@ -151,9 +163,9 @@ def count_tokens(self, text):
         model_manager = ModelManager('config/config.yaml')
         model_manager.load_embedding_model()
 
-        print("Token count:", model_manager.count_tokens("Hello world."))
-        print("Groq LLM Response:", model_manager.llm_groq.invoke("Hello world."))
+        # print("Token count:", model_manager.count_tokens("Hello world."))
+        print("Groq LLM Response:", model_manager.llm.invoke("Hello world."))
         print("Embedding Result:", model_manager.embedding_model.embed_documents(["Hello world."]))
-        print("LLM content:", model_manager.get_llm_response("Hello world."))
+        # print("LLM content:", model_manager.get_llm_response("Hello world."))
     except Exception as e:
         logger.error("An error occurred: %s", e)

src/models/test.py

@@ -0,0 +1,55 @@
+# from langchain_text_splitters import TokenTextSplitter
+
+# text_splitter = TokenTextSplitter(chunk_size=10, chunk_overlap=0)
+
+# texts = text_splitter.split_text("heloo welcone otd jvbk =works")
+# print(texts)
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_ollama import ChatOllama
+
+template = """Question: {question}
+
+Answer: Let's think step by step."""
+
+prompt = ChatPromptTemplate.from_template(template)
+llm = ChatOllama(
+    model="gemma2:2b",
+    temperature=0,
+)
+
+
+
+print(llm.invoke([
+    (
+        "system",
+        "You are a helpful assistant that translates English to French. Translate the user sentence.",
+    ),
+    ("human", "I love programming."),
+]))

src/summarize.py

@@ -88,7 +88,7 @@ def __call__(self, source: str, type: str) -> dict:
         self.combined_content = " ".join(cluster_content.values())
         prompt = self.prompts['create_summary_prompt'].format(combined_content=self.combined_content)
 
-        final_summary = self.model_manager.get_llm_response(prompt)
+        final_summary = self.model_manager.llm.invoke(prompt).content
 
         # Step 7: Perform analysis on the document
         chunk_words, total_chunks, total_words, total_tokens, tokens_sent_tokens = self.get_analysis()
@@ -133,7 +133,7 @@ def find_suitable_theme(self, chunk_text):
         """
         prompt = self.prompts['find_suitable_theme_prompt'].format(chunk_text=chunk_text)
         logger.info("Finding suitable theme for chunk: %s", chunk_text)
-        return self.model_manager.get_llm_response(prompt)
+        return self.model_manager.llm.invoke(prompt).content
 
     def find_themes_for_clusters_slow(self, chunks, representatives):
         """
@@ -192,7 +192,7 @@ def find_themes_for_clusters(self, chunks, representatives):
         prompt = self.prompts['find_suitable_theme_prompt_multiple'].format(first_representative_chunk=first_representative_chunk)
 
         # Step 3: Call the LLM once for all clusters
-        response = self.model_manager.get_llm_response(prompt)
+        response = self.model_manager.llm.invoke(prompt).content
 
 
         print(response)
@@ -232,9 +232,9 @@ def main():
     summarizer = Summarizer(config_path)
     print(summarizer.find_suitable_theme("Who is John Galt!"))
 
-    #data = summarizer('https://mitrarobot.com',"web")
-    data = summarizer('https://abc7.com/read-harris-trump-presidential-debate-transcript/15289001/','web')
-    #data = summarizer('https://www.whitehouse.gov/state-of-the-union-2024/',"web")
+    data = summarizer('https://medium.com/@balajivis/whats-so-challenging-about-building-chatbots-drawing-lessons-from-the-trenches-1ca7343c6e3d',"web")
+    # data = summarizer('https://abc7.com/read-harris-trump-presidential-debate-transcript/15289001/','web')
+    # data = summarizer('https://www.whitehouse.gov/state-of-the-union-2024/',"web")
     #data = summarizer('https://d18rn0p25nwr6d.cloudfront.net/CIK-0001921963/77018dae-bae9-4c33-8eaf-fa6685991719.pdf',"pdf")
 
     create_final_report(data,report_path='reports/final_report.pdf')

temp.txt

@@ -1 +1,5 @@
 set PYTHONPATH=E:\Invento Projects\BrahmaSumm
+clip
+kotaemon
+detectron
+unstructured

todo.md

@@ -4,8 +4,8 @@
 - [x] Provide a nice output option such as PDF
 - [ ] In Config provide option to summarize as small/medium/large
 - [ ] Video https://pypi.org/project/audio-extract/ 
-- [X] Youtube, audio
-- [ ] Handle pandas dataframe for excel and optionally explore pandas ai
+- [x] Youtube, audio
+- [x] Handle pandas dataframe for excel and optionally explore pandas ai
 - [ ] Use unstructured document loader https://python.langchain.com/v0.2/docs/integrations/providers/unstructured/
 - [ ] explore summarization with LIDA https://microsoft.github.io/lida/
 - [ ] Table extraction
@@ -14,7 +14,7 @@
 - [ ] Handle different loaders -- csv, arxiv
 - [ ] Improve clustering with more flexible organization as well as remove outliers
 - [ ] Give proper error for various things including not finding GROQ key or not finding the Ollama model
-- [ ] Support other models for LLM and embeddings [Openai, Ollama, Anthropic]
+- [x] Support other models for LLM and embeddings [Openai, Ollama, Anthropic]
 - [ ] Provide smaller summmarizer model
 - [ ] Multi-document summaries
 - [ ] Fix the bug when the target words is large