update chat vector

Dockerfile

@@ -20,6 +20,9 @@ RUN apt-get update && \
     build-essential \
     && rm -rf /var/lib/apt/lists/*
 
+# Install supervisord
+RUN apt-get update && apt-get install -y supervisor && rm -rf /var/lib/apt/lists/*
+
 # Create non-root user to avoid git permission issues
 RUN useradd -m appuser && chown -R appuser /app
 
@@ -47,6 +50,9 @@ RUN pip install --no-cache-dir --upgrade pip && \
 # Copy the entire app source code
 COPY . .
 
+# Copy supervisord config
+COPY supervisord.conf /etc/supervisor/conf.d/supervisord.conf
+
 # Create necessary directories
 RUN mkdir -p vector_store logs
 
@@ -60,5 +66,5 @@ RUN pip install torch torchvision torchaudio --index-url https://download.pytorc
 HEALTHCHECK --interval=30s --timeout=30s --start-period=5s --retries=3 \
     CMD curl -f http://localhost:7860/docs || exit 1
 
-# Run the FastAPI app via Uvicorn
-CMD ["uvicorn", "app.main:app", "--host", "0.0.0.0", "--port", "7860", "--workers", "1"]
+# Run supervisord
+CMD ["/usr/bin/supervisord", "-c", "/etc/supervisor/conf.d/supervisord.conf"]

app/agent/custom_chatbot.py

@@ -7,7 +7,8 @@ from dotenv import load_dotenv
 from langchain_groq import ChatGroq
 
 from langchain_openai import OpenAIEmbeddings
-from langchain_community.vectorstores import FAISS
+from langchain_community.vectorstores import SupabaseVectorStore
+from supabase.client import create_client
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain.chains.combine_documents import create_stuff_documents_chain
 from langchain_core.prompts import ChatPromptTemplate
@@ -27,9 +28,24 @@ if not groq_api_key:
     logger.error("❌ GROQ_API_KEY is not set in the environment.")
     raise RuntimeError("GROQ_API_KEY must be set in .env or environment variables.")
 
+# Supabase Initialization
+supabase_url = os.getenv("SUPABASE_URL")
+supabase_key = os.getenv("SUPABASE_KEY")
+
+if not supabase_url or not supabase_key:
+    logger.error("❌ SUPABASE_URL or SUPABASE_KEY is not set in the environment.")
+    raise RuntimeError("SUPABASE_URL and SUPABASE_KEY must be set in .env or environment variables.")
+
+try:
+    supabase_client = create_client(supabase_url, supabase_key)
+    logger.info("✅ Supabase client initialized successfully.")
+except Exception as e:
+    logger.exception("❌ Failed to initialize Supabase client.")
+    raise
+
 # LLM Initialization
 try:
-    llm = ChatGroq(groq_api_key=groq_api_key, model_name="Llama3-8b-8192")
+    llm = ChatGroq(groq_api_key=groq_api_key, model_name="llama-3.3-70b-versatile")
     logger.info("✅ ChatGroq LLM initialized successfully.")
 except Exception as e:
     logger.exception("❌ Failed to initialize ChatGroq LLM.")
@@ -45,31 +61,20 @@ Answer the question based only on the provided context.
 Question: {input}
 """)
 
+generic_prompt_template = ChatPromptTemplate.from_template("""
+You are a helpful AI assistant. Answer the following question:
+<context>
+{context}
+</context>
+
+Question: {input}
+""")
+
 # Input schema
 class ChatRequest(BaseModel):
     query: str
     user_id: int
 
-# Load vector store for a given user
-def load_user_vector_store(user_id: int):
-    user_path = f"vector_store/user_{user_id}"
-    index_file = os.path.join(user_path, "index.faiss")
-
-    logger.info(f"🔍 Looking for vector store at {index_file}")
-    if not os.path.exists(index_file):
-        msg = f"No vector store found for user {user_id}"
-        logger.warning(f"📭 {msg}")
-        raise FileNotFoundError(msg)
-
-    try:
-        embeddings = OpenAIEmbeddings()
-        vector_store = FAISS.load_local(user_path, embeddings,allow_dangerous_deserialization=True)
-        logger.info(f"📦 Vector store loaded for user {user_id}")
-        return vector_store
-    except Exception as e:
-        logger.exception(f"❌ Failed to load vector store for user {user_id}")
-        raise
-
 # Main chatbot endpoint
 @router.post("/custom-chatbot")
 async def custom_chatbot(request: ChatRequest):
@@ -78,25 +83,28 @@ async def custom_chatbot(request: ChatRequest):
     logger.info(f"🤖 Received query from user {user_id}: {query}")
 
     try:
-        vector_store = load_user_vector_store(user_id)
-        retriever = vector_store.as_retriever()
+        embeddings = OpenAIEmbeddings()
+        # Changed table_name and query_name to fixed values, and added filter to query
+        vector_store = SupabaseVectorStore(client=supabase_client, embedding=embeddings, table_name="documents", query_name="match_documents")
 
+        retriever = vector_store.as_retriever(search_kwargs={"filter": {"user_id": user_id}})
         doc_chain = create_stuff_documents_chain(llm, prompt_template)
         rag_chain = create_retrieval_chain(retriever, doc_chain)
 
         response = rag_chain.invoke({"input": query})
-        logger.info(f"✅ Response generated for user {user_id}")
+        logger.info(f"✅ Response generated for user {user_id} using RAG.")
 
         return {
             "answer": response["answer"],
             "sources": [doc.page_content for doc in response.get("context", [])],
         }
 
-    except FileNotFoundError as e:
-        logger.warning(f"🚫 {e}")
-        raise HTTPException(status_code=404, detail=str(e))
-
     except Exception as e:
-        logger.exception("❌ Unexpected error in custom chatbot endpoint.")
-        raise HTTPException(status_code=500, detail="Internal server error")
+        logger.warning(f"🤷 Falling back to generic response due to: {e}")
+        generic_chain = create_stuff_documents_chain(llm, generic_prompt_template)
+        generic_response = generic_chain.invoke({"input": query, "context": []}) # Pass empty context for generic response
+        return {
+            "answer": generic_response,
+            "sources": []
+        }
  

app/main.py

@@ -52,11 +52,6 @@ async def root():
         "status": "running"
     }
 
-@app.get("/health")
-async def health_check():
-    """Health check endpoint for Hugging Face Spaces"""
-    return {"status": "healthy", "service": "dubsway-video-ai"}
-
 @app.on_event("startup")
 async def startup_event():
     logger.info("✅ FastAPI app started")

app/utils/agentic_integration.py

@@ -221,7 +221,7 @@ class MCPToolManager:
         if self.groq_api_key:
             try:
                 from langchain_groq import ChatGroq
-                llm = ChatGroq(groq_api_key=self.groq_api_key, model_name="llama3-8b-8192")
+                llm = ChatGroq(groq_api_key=self.groq_api_key, model_name="llama-3.3-70b-versatile")
                 # This would use Groq for sentiment analysis
                 return {"positive": 0.6, "negative": 0.2, "neutral": 0.2}
             except:
@@ -234,7 +234,7 @@ class MCPToolManager:
         if self.groq_api_key:
             try:
                 from langchain_groq import ChatGroq
-                llm = ChatGroq(groq_api_key=self.groq_api_key, model_name="llama3-8b-8192")
+                llm = ChatGroq(groq_api_key=self.groq_api_key, model_name="llama-3.3-70b-versatile")
                 # This would use Groq for topic extraction
                 return ["technology", "innovation", "business"]
             except:
@@ -247,7 +247,7 @@ class MCPToolManager:
         if self.groq_api_key:
             try:
                 from langchain_groq import ChatGroq
-                llm = ChatGroq(groq_api_key=self.groq_api_key, model_name="llama3-8b-8192")
+                llm = ChatGroq(groq_api_key=self.groq_api_key, model_name="llama-3.3-70b-versatile")
                 # This would use Groq to add context
                 return f"Enhanced context for: {content}"
             except:
@@ -299,7 +299,7 @@ def integrate_agentic_analysis():
     
     Benefits:
     - Multi-modal analysis (audio + visual)
-    - Context-aware summarization using Groq Llama3-8b-8192
+    - Context-aware summarization using Groq llama-3.3-70b-versatile
     - Beautiful, comprehensive reports
     - Enhanced vector embeddings for better RAG
     - Web search integration for context

app/utils/enhanced_analysis.py

@@ -82,7 +82,7 @@ class MultiModalAnalyzer:
             
         self.llm = ChatGroq(
             groq_api_key=groq_api_key,
-            model_name="llama3-8b-8192",
+            model_name="llama-3.3-70b-versatile",
             temperature=0.1,
             max_tokens=2000
         )
@@ -325,7 +325,7 @@ Based on the analysis, consider:
 - Target audience insights
 
 ---
-*Report generated on {datetime.now().strftime('%Y-%m-%d %H:%M:%S')} using Groq Llama3-8b-8192*
+*Report generated on {datetime.now().strftime('%Y-%m-%d %H:%M:%S')} using Groq llama-3.3-70b-versatile*
         """
         
         return report_template

app/utils/lightweight_agentic.py

@@ -74,7 +74,7 @@ class LightweightAgenticProcessor:
             # Initialize Groq
             llm = ChatGroq(
                 groq_api_key=self.groq_api_key,
-                model_name="llama3-8b-8192",
+                model_name="llama-3.3-70b-versatile",
                 temperature=0.1,
                 max_tokens=1000
             )
@@ -128,7 +128,7 @@ class LightweightAgenticProcessor:
 ## 📝 Basic Summary
 {summary}
 
-## 🤖 Enhanced Analysis (Groq Llama3-8b-8192)
+## 🤖 Enhanced Analysis (Groq llama-3.3-70b-versatile)
 {enhanced_analysis.get('enhanced_analysis', 'Analysis not available')}
 
 ## 🎯 Key Insights
@@ -136,13 +136,13 @@ class LightweightAgenticProcessor:
 
 ## 📊 Analysis Details
 - **Processing Method**: Lightweight Agentic Analysis
-- **LLM Provider**: Groq Llama3-8b-8192
+- **LLM Provider**: Groq llama-3.3-70b-versatile
 - **Enhanced Features**: Text-based analysis and reasoning
 - **Topics**: {', '.join(enhanced_analysis.get('topics', ['General']))}
 - **Sentiment**: {enhanced_analysis.get('sentiment', {})}
 
 ---
-*Report generated using Groq Llama3-8b-8192*
+*Report generated using Groq llama-3.3-70b-versatile*
             """
         else:
             return f"""

app/utils/whisper_llm.py

@@ -2,6 +2,7 @@ import os
 import logging
 import requests
 import tempfile
+import uuid
 import torch
 from transformers import pipeline
 from faster_whisper import WhisperModel
@@ -9,7 +10,11 @@ from faster_whisper import WhisperModel
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain_openai import OpenAIEmbeddings
 from langchain_core.documents import Document
-from langchain_community.vectorstores import FAISS
+from supabase.client import create_client
+try:
+    import av  # Optional: used to pre-check audio streams for robustness
+except Exception:  # pragma: no cover
+    av = None
 
 from sqlalchemy import select
 from sqlalchemy.ext.asyncio import AsyncSession
@@ -26,14 +31,26 @@ if not logger.handlers:
 
 # Whisper Model Initialization
 def get_whisper_model():
-    if torch.cuda.is_available():
-        device = "cuda"
-        compute_type = "float32"
-        logger.info("GPU detected: Using CUDA with float32 compute")
+    # Allow overrides via env vars
+    env_device = os.getenv("FASTER_WHISPER_DEVICE")
+    env_compute = os.getenv("FASTER_WHISPER_COMPUTE")
+
+    if env_device:
+        device = env_device
+        logger.info(f"Using device from env FASTER_WHISPER_DEVICE={env_device}")
+    else:
+        if torch.cuda.is_available():
+            device = "cuda"
+            logger.info("GPU detected: Using CUDA")
+        else:
+            device = "cpu"
+            logger.warning("GPU not available: Falling back to CPU")
+
+    if env_compute:
+        compute_type = env_compute
+        logger.info(f"Using compute_type from env FASTER_WHISPER_COMPUTE={env_compute}")
     else:
-        device = "cpu"
-        compute_type = "int8"
-        logger.warning("GPU not available: Falling back to CPU with int8 compute")
+        compute_type = "float32" if device == "cuda" else "int8"
 
     try:
         model = WhisperModel("base", device=device, compute_type=compute_type)
@@ -45,6 +62,21 @@ def get_whisper_model():
 
 whisper_model = get_whisper_model()
 
+# Supabase Initialization
+supabase_url = os.getenv("SUPABASE_URL")
+supabase_key = os.getenv("SUPABASE_KEY")
+
+if not supabase_url or not supabase_key:
+    logger.error("❌ SUPABASE_URL or SUPABASE_KEY is not set in the environment.")
+    raise RuntimeError("SUPABASE_URL and SUPABASE_KEY must be set in .env or environment variables.")
+
+try:
+    supabase_client = create_client(supabase_url, supabase_key)
+    logger.info("✅ Supabase client initialized successfully.")
+except Exception as e:
+    logger.exception("❌ Failed to initialize Supabase client.")
+    raise
+
 # Summarizer
 try:
     summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
@@ -152,50 +184,65 @@ async def analyze(video_url: str, user_id: int, db: AsyncSession):
 
     # Step 2: Transcribe
     try:
+        # Optional pre-check: ensure the file has an audio stream
+        if av is not None:
+            try:
+                with av.open(tmp_path) as container:
+                    has_audio = any(s.type == "audio" for s in container.streams)
+                if not has_audio:
+                    logger.error("No valid audio stream in file; skipping transcription")
+                    raise IndexError("No audio stream")
+            except IndexError:
+                raise
+            except Exception:
+                # If PyAV check fails, continue and let transcribe attempt
+                pass
+
         logger.info("Transcribing audio with Faster-Whisper...")
-        
+
         # Get transcription result
         result = whisper_model.transcribe(tmp_path)
-        
+
         # Handle different return formats from faster-whisper
         if isinstance(result, tuple):
             segments, info = result
         else:
-            # If it's not a tuple, it might be just segments
             segments = result
             info = None
-            
+
         # Extract text from segments
         if segments:
             text = " ".join(segment.text for segment in segments if hasattr(segment, 'text') and segment.text)
         else:
             text = ""
-            
+
         logger.info(f"Transcription completed. Length: {len(text)} characters.")
-        
+
         # Log additional info if available
         if info:
             logger.info(f"Transcription info: language={getattr(info, 'language', 'unknown')}, language_probability={getattr(info, 'language_probability', 'unknown')}")
-            
+
         # Handle empty transcription
         if not text or len(text.strip()) == 0:
             logger.warning("Transcription resulted in empty text, using fallback")
             text = "No speech detected in video"
-            
+
+    except IndexError:
+        logger.error("No valid audio stream in file; skipping transcription")
+        text = "Transcription failed - video may be corrupted or have no audio"
     except Exception as e:
         logger.error(f"Transcription failed: {e}")
         logger.error(f"Error type: {type(e)}")
         import traceback
         logger.error(f"Traceback: {traceback.format_exc()}")
-        
         # Provide fallback text instead of failing completely
         logger.warning("Using fallback text due to transcription failure")
         text = "Transcription failed - video may be corrupted or have no audio"
-        
-        # Clean up temp file
+    finally:
+        # Always attempt to clean up temp file
         try:
             os.unlink(tmp_path)
-        except:
+        except Exception:
             pass
 
     # Step 3: Summarize
@@ -217,33 +264,28 @@ async def analyze(video_url: str, user_id: int, db: AsyncSession):
         except:
             pass
 
-    # Step 4: Save to FAISS store
+    # Step 4: Save to Supabase vector store (explicit user_id)
     try:
-        logger.info("Creating/updating FAISS vector store for user...")
-        documents = [Document(page_content=summary)]
-        embeddings = OpenAIEmbeddings()
-
-        user_vector_path = f"vector_store/user_{user_id}"
-        os.makedirs(user_vector_path, exist_ok=True)
-
-        if os.path.exists(os.path.join(user_vector_path, "index.faiss")):
-            # Load existing vector store - safe to use allow_dangerous_deserialization
-            # since we're loading our own created files
-            vector_store = FAISS.load_local(user_vector_path, embeddings, allow_dangerous_deserialization=True)
-            vector_store.add_documents(documents)
+        logger.info("Saving summary to Supabase vector store for user...")
+        if not summary or not summary.strip():
+            logger.warning("Empty summary; skipping Supabase insert")
         else:
-            # Create new vector store
-            vector_store = FAISS.from_documents(documents, embeddings)
+            embeddings = OpenAIEmbeddings()
+            embedding_vector = embeddings.embed_query(summary)
+
+            document_id = str(uuid.uuid4())
+            payload = {
+                "id": document_id,
+                "user_id": user_id,
+                "content": summary,
+                "embedding": embedding_vector,
+                "metadata": {"user_id": user_id, "video_url": video_url},
+            }
+            supabase_client.table("documents").insert(payload).execute()
+            logger.info(f"Summary saved to Supabase for user: {user_id}")
 
-        vector_store.save_local(user_vector_path)
-        logger.info(f"Vector store saved at: {user_vector_path}")
     except Exception as e:
-        logger.error(f"Failed to create vector store: {e}")
-        # Clean up temp file
-        try:
-            os.unlink(tmp_path)
-        except:
-            pass
+        logger.error(f"Failed to save to Supabase vector store: {e}")
         raise
 
     # Clean up temp file

requirements-hf.txt

@@ -3,7 +3,7 @@ fastapi==0.104.1
 uvicorn[standard]==0.24.0
 python-dotenv==1.0.0
 python-multipart==0.0.6
-pydantic[email]==2.5.0
+pydantic[email]==2.11.7
 
 # Database
 asyncpg==0.29.0
@@ -23,7 +23,7 @@ langchain-groq==0.0.1
 
 # Embedding & vector DB
 sentence-transformers==2.2.2
-faiss-cpu==1.7.4
+supabase==2.18.1
 
 # Transcription
 faster-whisper==1.0.1

requirements-old.txt

@@ -0,0 +1,33 @@
+
+
+# Tools & Transcription
+transformers
+whisper
+faster-whisper==1.0.1
+ctranslate2>=4.6.0
+PyPDF2
+pypdf
+reportlab
+bs4
+beautifulsoup4
+
+# Enhanced Analysis & MCP/ACP
+opencv-python
+pillow
+duckduckgo-search
+wikipedia-api
+easyocr
+langchain-groq
+timm
+
+# Optional
+sse-starlette
+wikipedia
+arxiv
+cassio
+streamlit
+supabase 
+# CUDA-enabled Torch (installed separately)
+# torch==2.2.2+cu121 and torchvision==0.17.2+cu121
+# Must be installed via pip with specific index:
+# pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

requirements-windows.txt

@@ -0,0 +1,73 @@
+# Windows-compatible requirements file
+# This avoids compilation issues by using pre-compiled packages
+
+# Core FastAPI dependencies
+fastapi==0.104.1
+uvicorn[standard]==0.24.0
+python-dotenv==1.0.0
+python-multipart==0.0.6
+pydantic[email]==2.11.7
+requests==2.31.0
+boto3==1.34.0
+
+# Database
+asyncpg==0.29.0
+sqlalchemy==2.0.23
+aiosqlite==0.19.0
+databases
+psycopg2-binary
+
+# Auth
+passlib[bcrypt]==1.7.4
+python-jose[cryptography]==3.3.0
+
+# LLM & RAG - using compatible versions
+langchain==0.1.13
+langchain-openai==0.1.7
+langchain-community==0.0.38
+langchain-core==0.1.53
+langchain-groq==0.0.1
+langchainhub
+langserve
+langchain-objectbox
+
+# Embedding & vector DB
+sentence-transformers==2.2.2
+supabase==2.18.1
+chromadb
+
+# Transcription - using latest compatible versions
+faster-whisper==1.0.1
+ctranslate2>=4.6.0
+transformers==4.35.2
+whisper
+
+# PDF & Reports
+PyPDF2==3.0.1
+reportlab==4.0.7
+beautifulsoup4==4.12.2
+bs4
+pypdf
+
+# Enhanced Analysis
+duckduckgo-search==4.1.1
+wikipedia-api==0.6.0
+timm==0.9.12
+opencv-python
+pillow
+easyocr
+
+# Optional
+sse-starlette
+wikipedia
+arxiv
+cassio
+streamlit
+
+# Pre-compiled numpy and scipy to avoid compilation
+numpy>=1.24.0,<2.0
+scipy>=1.10.0
+
+# CUDA-enabled Torch (install separately if needed)
+# pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
+

requirements.txt

@@ -1,66 +1,68 @@
-# Core
-fastapi
-uvicorn
-boto3
-requests
-python-dotenv
-python-multipart
-pydantic[email]>=1.10,<2.0  # V1 for compatibility with many frameworks
+# Merged requirements from requirements.txt and requirements-hf.txt
 
-# DB
+# Core FastAPI dependencies
+fastapi==0.104.1
+uvicorn[standard]==0.24.0
+python-dotenv==1.0.0
+python-multipart==0.0.6
+pydantic[email]==2.11.7
+requests==2.31.0
+boto3==1.34.0
+
+# Database
 asyncpg
-sqlalchemy>=2.0
+aiosqlite
 databases
 psycopg2-binary
-aiosqlite
 
 # Auth
-passlib[bcrypt]
-python-jose[cryptography]
+passlib[bcrypt]==1.7.4
+python-jose[cryptography]==3.3.0
 
 # LLM & RAG
 langchain==0.1.13
 langchain-openai==0.1.7
 langchain-community==0.0.38
 langchain-core==0.1.53
-langchain-groq
+langchain-groq==0.0.1
 langchainhub
 langserve
 langchain-objectbox
 
 # Embedding & vector DB
 sentence-transformers==2.2.2
-faiss-cpu
+supabase==2.18.1
 chromadb
 
-# Tools & Transcription
-transformers
-whisper
+# Transcription
 faster-whisper==1.0.1
-ctranslate2>=4.0,<5
-PyPDF2
-pypdf
-reportlab
-bs4
-beautifulsoup4
+ctranslate2
+transformers==4.35.2
+whisper # from requirements.txt, not explicitly in hf
 
-# Enhanced Analysis & MCP/ACP
-opencv-python
-pillow
-duckduckgo-search
-wikipedia-api
-easyocr
-langchain-groq
-timm
+# PDF & Reports
+PyPDF2==3.0.1
+reportlab==4.0.7
+beautifulsoup4==4.12.2
+bs4 # from requirements.txt
+pypdf # from requirements.txt
+
+# Enhanced Analysis
+duckduckgo-search==4.1.1
+wikipedia-api==0.6.0
+timm==0.9.12
+opencv-python # from requirements.txt
+pillow # from requirements.txt
+easyocr # from requirements.txt
 
 # Optional
+faiss-gpu
 sse-starlette
 wikipedia
 arxiv
 cassio
 streamlit
-
 # CUDA-enabled Torch (installed separately)
 # torch==2.2.2+cu121 and torchvision==0.17.2+cu121
 # Must be installed via pip with specific index:
-# pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
+# pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121

supervisord.conf

@@ -0,0 +1,22 @@
+[supervisord]
+nodaemon=true
+logfile=/dev/null
+loglevel=info
+
+[program:web]
+command=uvicorn app.main:app --host 0.0.0.0 --port 7860 --workers 1
+directory=/app
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+stderr_logfile=/dev/stderr
+stderr_logfile_maxbytes=0
+environment=PYTHONUNBUFFERED=1,PYTHONDONTWRITEBYTECODE=1
+
+[program:worker]
+command=python worker/daemon.py
+directory=/app
+stdout_logfile=/dev/stdout
+stdout_logfile_maxbytes=0
+stderr_logfile=/dev/stderr
+stderr_logfile_maxbytes=0
+environment=PYTHONUNBUFFERED=1,PYTHONDONTWRITEBYTECODE=1