DEBUG: FlashRank

functions.py

@@ -198,15 +198,11 @@ def answerQuery(query: str, vectorstore: str, llmModel: str = "llama3-70b-8192")
         vectorstore=vectorstore,
         docstore=store,
         child_splitter=RecursiveCharacterTextSplitter(),
-        search_kwargs={"k": 20}
-    )
-    compressor = FlashrankRerank()
-    retriever = ContextualCompressionRetriever(
-        base_compressor=compressor, base_retriever=retriever
+        search_kwargs={"k": 4, "score_threshold": 0}
     )
     baseChain = (
             {"context": RunnableLambda(lambda x: x["question"]) | retriever | RunnableLambda(format_docs),
-             "question": RunnablePassthrough(), "chatHistory": RunnablePassthrough()}
+             "question": RunnableLambda(lambda x: x["question"]), "chatHistory": RunnableLambda(lambda x: x["chatHistory"])}
             | prompt
             | ChatGroq(model=llmModel, temperature=0.75, max_tokens=512)
             | StrOutputParser()
@@ -289,7 +285,7 @@ def getLinks(url: str, timeout=30):
 def getTextFromImagePDF(pdfBytes):
     def getText(image):
         global reader
-        return "\n".join([text[1] for text in reader.readtext(np.array(image), paragraph=True)])
+        return "\n".join([text[1] for text in reader.readtext(np.array(image), paragraph=True, x_ths = 0)])
     allImages = convert_from_bytes(pdfBytes)
     texts = [getText(image) for image in allImages]
     return "\n\n\n".join(texts)