update

Dockerfile

@@ -9,13 +9,12 @@ RUN apt-get update && apt-get install -y git && rm -rf /var/lib/apt/lists/*
 COPY requirements.txt .
 RUN pip install --no-cache-dir -r requirements.txt
 
-# Copy all project files
+# Copy everything (including your config and scripts)
 COPY . .
 
 # ---------------------------------------------------------
 # PRE-BUILD PHASE
-# This downloads models and processes PDFs during the build.
-# This prevents httpx.ReadTimeout errors at runtime.
+# We run these in the container so they are "baked into" the image.
 # ---------------------------------------------------------
 RUN python download_models.py
 RUN python ingest.py

ingest.py

@@ -8,15 +8,20 @@ from langchain_chroma import Chroma
 from config import KB_DIR, HF_DATASET_REPO, EMBEDDING_MODEL, CHROMA_DIR
 
 def run_ingestion():
+    # Clean setup for the container
+    if os.path.exists(KB_DIR): shutil.rmtree(KB_DIR)
+    if os.path.exists(CHROMA_DIR): shutil.rmtree(CHROMA_DIR)
     os.makedirs(KB_DIR, exist_ok=True)
-    # Load dataset - this will use the cached version from build phase
+
+    print(f"⬇️ Loading dataset from {HF_DATASET_REPO}...")
     dataset = load_dataset(HF_DATASET_REPO, split="train")
     
     pdf_paths = []
     for i, row in enumerate(dataset):
         pdf_feature = row.get("pdf")
-        # Access local path directly from HF cache
         src_path = None
+        
+        # Access the cached file path from Hugging Face's internal download
         if isinstance(pdf_feature, dict): src_path = pdf_feature.get("path")
         elif hasattr(pdf_feature, 'filename'): src_path = pdf_feature.filename
 
@@ -24,20 +29,34 @@ def run_ingestion():
             dest_path = os.path.join(KB_DIR, f"doc_{i}.pdf")
             shutil.copy(src_path, dest_path)
             pdf_paths.append(dest_path)
+            print(f"✅ Cached: doc_{i}.pdf")
 
+    print(f"📄 Processing {len(pdf_paths)} documents...")
     docs = []
     for p in pdf_paths:
         try:
             loader = PyPDFLoader(p)
             docs.extend(loader.load())
-        except Exception as e: print(f"❌ Error: {e}")
+        except Exception as e:
+            print(f"❌ Error loading {p}: {e}")
+
+    if not docs:
+        print("❌ Error: No documents successfully loaded.")
+        return
 
-    if not docs: return
+    splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100)
+    splits = splitter.split_documents(docs)
 
-    splits = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=100).split_documents(docs)
+    print(f"🧠 Building Vector DB at {CHROMA_DIR}...")
     embeddings = HuggingFaceEmbeddings(model_name=EMBEDDING_MODEL)
-    Chroma.from_documents(documents=splits, embedding=embeddings, persist_directory=CHROMA_DIR)
-    print("✅ KB Initialized")
+    
+    # This creates the physical folder and files
+    Chroma.from_documents(
+        documents=splits, 
+        embedding=embeddings, 
+        persist_directory=CHROMA_DIR
+    )
+    print("✅ Ingestion complete.")
 
 if __name__ == "__main__":
     run_ingestion()

rag.py

@@ -7,32 +7,30 @@ from config import EMBEDDING_MODEL, LLM_MODEL, CHROMA_DIR
 # 1. Initialize Embeddings
 embeddings = HuggingFaceEmbeddings(model_name=EMBEDDING_MODEL)
 
-# 2. Load Vector DB (Load only)
-if os.path.exists(CHROMA_DIR) and os.listdir(CHROMA_DIR):
+# 2. Load Vector DB
+# Ensure CHROMA_DIR is exactly the same as in ingest.py
+if os.path.exists(CHROMA_DIR) and len(os.listdir(CHROMA_DIR)) > 0:
     vectordb = Chroma(
         persist_directory=CHROMA_DIR,
         embedding_function=embeddings
     )
     print(f"✅ Vector DB loaded from {CHROMA_DIR}")
 else:
+    print(f"⚠️ Vector DB not found at {CHROMA_DIR}")
     vectordb = None
-    print("❌ Vector DB directory is missing or empty")
 
 # 3. LLM Pipeline
 qa_pipeline = pipeline("text-generation", model=LLM_MODEL, max_new_tokens=256)
 
 def ask_rag_with_status(question: str):
     if vectordb is None:
-        return "The knowledge base is not initialized.", "ERROR"
+        return "The knowledge base is not initialized. Please check build logs.", "ERROR"
 
     docs = vectordb.similarity_search(question, k=3)
-    if not docs:
-        return "No relevant information found.", "NO_MATCH"
-
     context = "\n\n".join(d.page_content for d in docs)
-    prompt = f"Context:\n{context}\n\nQuestion: {question}\n\nAnswer:"
+    prompt = f"Use the context to answer Gandhi related questions.\nContext:\n{context}\n\nQuestion: {question}\n\nAnswer:"
     
     result = qa_pipeline(prompt)
     answer = result[0]["generated_text"].split("Answer:")[-1].strip()
 
-    return answer, ["Context retrieved", "OK"]
+    return answer, ["Context retrieved", "Answer generated"]