UPDATE: YT Transcripts

app.py

@@ -8,6 +8,7 @@ from fastapi.middleware.cors import CORSMiddleware
 from langchain_community.document_loaders import UnstructuredURLLoader
 
 
+
 app = FastAPI(title = "ConversAI", root_path = "/api/v1")
 app.add_middleware(
     CORSMiddleware,
@@ -159,4 +160,19 @@ async def getCount(vectorstore: str):
 
 @app.post("/getYoutubeTranscript")
 async def getYTTranscript(url: str):
-    return getTranscript(url = url)
+    return getTranscript(url = url)
+
+
+@app.post("/analyzeData")
+async def analyzeAndAnswer(query: str, file: UploadFile = File(...)):
+    extension = file.name.split(".")[-1]
+    if extension in ["xls", "xlsx", "xlsm", "xlsb"]:
+        df = pd.read_excel(io.BytesIO(file.read()))
+        return analyzeData(query = query, dataframe = df)
+    elif extension == "csv":
+        df = pd.read_csv(io.BytesIO(file.read()))
+        return analyzeData(query = query, dataframe = df)
+    else:
+        return {
+            "output": "INVALID FILE TYPE"
+        }

functions.py

@@ -8,6 +8,7 @@ from langchain_core.output_parsers import StrOutputParser
 from langchain.retrievers import ParentDocumentRetriever
 from langchain_core.runnables.history import RunnableWithMessageHistory
 from langchain.memory import ChatMessageHistory
+from pandasai import SmartDataframe
 from langchain_core.chat_history import BaseChatMessageHistory
 from langchain.storage import InMemoryStore
 from langchain_community.document_loaders import YoutubeLoader
@@ -298,7 +299,7 @@ def getTextFromImagePDF(pdfBytes):
     return text
 
 
-def getTranscript(url: str):
+def getTranscript(urls: dict[str, str]):
     loader = YoutubeLoader.from_youtube_url(
         url, add_video_info=False
     )
@@ -307,3 +308,10 @@ def getTranscript(url: str):
     except:
         doc = "ENGLISH TRANSCRIPT UNAVAILABLE"
     return doc
+
+
+def analyzeData(query, dataframe):
+    llm = ChatGroq("gemma2-9b-it")
+    df = SmartDataframe(dataframe, config = {"llm": llm, "verbose": False})
+    response = df.chat(query)
+    return response

requirements.txt

@@ -16,6 +16,7 @@ numpy
 PyPDF2
 python-dotenv
 pydantic
+pandasas
 pandas
 easyocr
 youtube-transcript-api