update

Dockerfile

@@ -1,7 +1,5 @@
 FROM python:3.10-slim
-
 WORKDIR /app
-
 RUN apt-get update && apt-get install -y git && rm -rf /var/lib/apt/lists/*
 
 COPY requirements.txt .
@@ -10,6 +8,8 @@ RUN pip install --no-cache-dir -r requirements.txt
 COPY app.py rag.py ingest.py config.py ./ 
 COPY frontend ./frontend
 
-EXPOSE 7860
+# This will now succeed because requirements.txt has langchain-chroma
+RUN python ingest.py
 
+EXPOSE 7860
 CMD ["uvicorn", "app:app", "--host", "0.0.0.0", "--port", "7860"]

rag.py

@@ -2,68 +2,34 @@
 import os
 from transformers import pipeline
 from langchain_huggingface import HuggingFaceEmbeddings
-from langchain_chroma import Chroma
+from langchain_chroma import Chroma  # This requires langchain-chroma package
 from config import EMBEDDING_MODEL, LLM_MODEL, CHROMA_DIR
 
-# -----------------------------
-# 1. Initialize Embeddings (LangChain-HuggingFace)
-# -----------------------------
-embeddings = HuggingFaceEmbeddings(
-    model_name=EMBEDDING_MODEL
-)
+# 1. Initialize Embeddings
+embeddings = HuggingFaceEmbeddings(model_name=EMBEDDING_MODEL)
 
-# -----------------------------
-# 2. Load Vector DB (Safe Loading)
-# -----------------------------
-# We expect the DB to be pre-built by ingest.py during Docker build
+# 2. Load Vector DB
 if os.path.exists(CHROMA_DIR) and os.listdir(CHROMA_DIR):
     vectordb = Chroma(
         persist_directory=CHROMA_DIR,
         embedding_function=embeddings
     )
-    print(f"✅ Vector DB loaded from {CHROMA_DIR}")
+    print(f"✅ Vector DB loaded")
 else:
-    print(f"⚠️ Vector DB not found at {CHROMA_DIR}. Please check ingestion.")
     vectordb = None
+    print(f"⚠️ Vector DB not found")
 
-# -----------------------------
-# 3. LLM Pipeline (CPU safe)
-# -----------------------------
-qa_pipeline = pipeline(
-    task="text-generation",
-    model=LLM_MODEL,
-    max_new_tokens=256
-)
+# 3. LLM Pipeline
+qa_pipeline = pipeline(task="text-generation", model=LLM_MODEL, max_new_tokens=256)
 
-# -----------------------------
-# 4. RAG Query Function
-# -----------------------------
 def ask_rag_with_status(question: str):
     if vectordb is None:
-        return "Knowledge base is empty. Technical error during ingestion.", "NO_KB"
+        return "Knowledge base is empty.", "NO_KB"
 
-    # Search for relevant context
     docs = vectordb.similarity_search(question, k=3)
-
-    if not docs:
-        return "No relevant documents found in the knowledge base.", "NO_MATCH"
-
     context = "\n\n".join(d.page_content for d in docs)
 
-    prompt = f"""Use the context below to answer the question accurately.
-
-Context:
-{context}
-
-Question:
-{question}
-
-Answer:"""
-
+    prompt = f"Use the context to answer.\nContext:\n{context}\nQuestion:\n{question}\nAnswer:"
     result = qa_pipeline(prompt)
     
-    # Extract only the generated answer
-    full_text = result[0]["generated_text"]
-    answer = full_text.split("Answer:")[-1].strip()
-
-    return answer, ["Context retrieved", "LLM processed"]
+    return result[0]["generated_text"].split("Answer:")[-1].strip(), "OK"

requirements.txt

@@ -4,12 +4,13 @@ pydantic
 python-dotenv
 langchain>=0.2.17
 langchain-community>=0.2.17
-langchain-huggingface  # Resolves the LangChainDeprecationWarning
+langchain-huggingface
+langchain-chroma
 langchain-text-splitters==0.2.4
 chromadb==0.5.5
 sentence-transformers
 pypdf
-pdfplumber             # This fixes the ImportError
+pdfplumber
 transformers>=4.39.0
 huggingface_hub
 datasets