streamlit chat

Modules/rag.py

@@ -4,6 +4,7 @@ from dotenv import load_dotenv
 load_dotenv() # load .env api keys 
 
 mistral_api_key = os.getenv("MISTRAL_API_KEY")
+print("mistral_api_key", mistral_api_key)
 
 from langchain_community.document_loaders import PyPDFLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
@@ -24,6 +25,9 @@ from langchain.embeddings.sentence_transformer import SentenceTransformerEmbeddi
 from langchain_community.tools import DuckDuckGoSearchRun
 from pathlib import Path
 
+from huggingface_hub import login
+login(token=os.getenv("HUGGING_FACE_TOKEN"))
+
 def load_chunk_persist_pdf() -> Chroma:
     
     pdf_folder_path = os.path.join(os.getcwd(),Path("data/pdf/"))
@@ -38,7 +42,7 @@ def load_chunk_persist_pdf() -> Chroma:
     os.makedirs("data/chroma_store/", exist_ok=True)
     vectorstore = Chroma.from_documents(
         documents=chunked_documents,
-        embedding=MistralAIEmbeddings(),
+        embedding=MistralAIEmbeddings(api_key=mistral_api_key),
         persist_directory= os.path.join(os.getcwd(),Path("data/chroma_store/"))
     )
     vectorstore.persist()

app.py

@@ -16,13 +16,6 @@ mistral_api_key = os.getenv("MISTRAL_API_KEY")
 from Modules.PoseEstimation import pose_estimator
 from utils import save_uploaded_file
 
-def format_messages(messages):
-    formatted_messages = ""
-    for message in messages:
-        role = message["role"]
-        content = message["content"]
-        formatted_messages += f"{role}: {content}\n"
-    return formatted_messages
 
 st.set_page_config(layout="wide", initial_sidebar_state="collapsed")
 # Create two columns
@@ -41,41 +34,31 @@ base_chain = prompt | llm
 
 # First column containers
 with col1:
-    st.subheader("Audio Recorder")
-    recorded = False
-    temp_path = 'data/temp_audio/audio_file.wav'
-    wav_audio_data = st_audiorec()
-    if wav_audio_data is not None:
-        with open(temp_path, 'wb') as f:
-            # Write the audio data to the file
-            f.write(wav_audio_data)
-        instruction = transcribe(temp_path)
-        print(instruction)
-        recorded = True
+    
+    st.subheader("LLM answering")
 
+    if "messages" not in st.session_state:
+        st.session_state.messages = []
 
-    st.subheader("LLM answering")
-    if recorded:
-        if "messages" not in st.session_state:
-            st.session_state.messages = []
-        for message in st.session_state.messages:
-            with st.chat_message(message["role"]):
-                st.markdown(message["content"])
+    for message in st.session_state.messages:
+        with st.chat_message(message["role"]):
+            st.markdown(message["content"])
 
-        st.session_state.messages.append({"role": "user", "content": instruction})
+    if prompt := st.chat_input("What is up?"):
+        st.session_state.messages.append({"role": "user", "content": prompt})
         with st.chat_message("user"):
-            st.markdown(instruction)
+            st.markdown(prompt)
 
         with st.chat_message("assistant"):
             # Build answer from LLM
-            direction = router_chain.invoke({"question":instruction})
+            direction = router_chain.invoke({"question":prompt})
             if direction=='fitness_advices':
                 response = rag_chain.invoke(
-                            instruction
+                            prompt
                             )
             elif direction=='smalltalk':
                 response = base_chain.invoke(
-                    {"question":instruction}
+                    {"question":prompt}
                 ).content
             # elif direction =='movement_analysis':
             #     response = agent_executor.invoke(
@@ -100,13 +83,14 @@ with col2:
         ask_video.empty()
         _left, mid, _right = st.columns(3)
         with mid:
-            st.video(video_uploaded)
-            apply_pose = st.button("Apply Pose Estimation")
-
-        if apply_pose:
-            with st.spinner("Processing video"):
-                keypoints = pose_estimator.get_keypoints_from_keypoints(pose_estimator.model, video_uploaded)
+            if os.path.exists('runs'):
+                st.video(os.path.join('runs', 'pose', 'predict', 'squat.mp4'), loop=True)
+            else :
+                st.video(video_uploaded)
             
 
     st.subheader("Graph Displayer")
-        # TO DO
+    if os.path.exists('fig'):
+        file_list = os.listdir('fig')
+        for file in file_list:
+            st.image(os.path.join('fig', file))

requirements.txt

@@ -13,4 +13,7 @@ chromadb
 langgraph
 langchainhub
 pypdf
-duckduckgo-search
+duckduckgo-search
+python-dotenv
+pypdf
+chromadb