Fix Docker build: create kb_docs at runtime

Dockerfile

@@ -1,22 +1,23 @@
 FROM python:3.10-slim
 
-# Set working directory
 WORKDIR /app
 
-# Install system dependencies
+# System dependencies
 RUN apt-get update && apt-get install -y \
     git \
     && rm -rf /var/lib/apt/lists/*
 
-# Install Python dependencies
+# Python dependencies
 COPY requirements.txt .
 RUN pip install --no-cache-dir -r requirements.txt
 
-# Copy application code
+# Application code
 COPY app.py rag.py ingest.py guardrails.py config.py ./
-COPY kb_docs ./kb_docs
 
-# Hugging Face Spaces expects port 7860
+# Create empty kb_docs directory (PDFs added later via HF UI)
+RUN mkdir -p kb_docs
+
+# Hugging Face Spaces port
 EXPOSE 7860
 
 # Start FastAPI

app.py

@@ -1,6 +1,5 @@
 from fastapi import FastAPI
 from pydantic import BaseModel
-
 from rag import ask_rag_with_status
 
 app = FastAPI(title="HubRAG API")
@@ -10,10 +9,7 @@ class Question(BaseModel):
 
 @app.get("/")
 def root():
-    return {
-        "status": "ok",
-        "message": "RAG API is running"
-    }
+    return {"status": "ok", "message": "RAG API running"}
 
 @app.post("/chat")
 def chat(req: Question):

ingest.py

@@ -1,7 +1,11 @@
 from langchain_community.document_loaders import DirectoryLoader, PyPDFLoader
 from langchain_text_splitters import RecursiveCharacterTextSplitter
+import os
 
 def load_and_split_docs(path="kb_docs"):
+    if not os.path.exists(path):
+        return []
+
     loader = DirectoryLoader(
         path,
         glob="**/*.pdf",
@@ -9,9 +13,11 @@ def load_and_split_docs(path="kb_docs"):
     )
     docs = loader.load()
 
+    if not docs:
+        return []
+
     splitter = RecursiveCharacterTextSplitter(
         chunk_size=800,
         chunk_overlap=100
     )
-
     return splitter.split_documents(docs)

rag.py

@@ -4,46 +4,48 @@ from langchain.schema import SystemMessage, HumanMessage
 
 from ingest import load_and_split_docs
 
-print("⏳ Indexing documents...")
-
-embeddings = HuggingFaceEmbeddings(
-    model_name="sentence-transformers/all-MiniLM-L6-v2"
-)
+print("⏳ Loading documents...")
 
 documents = load_and_split_docs()
 
-vectorstore = Chroma.from_documents(
-    documents,
-    embedding=embeddings
+embeddings = HuggingFaceEmbeddings(
+    model_name="sentence-transformers/all-MiniLM-L6-v2"
 )
 
-retriever = vectorstore.as_retriever(search_kwargs={"k": 3})
+if documents:
+    vectorstore = Chroma.from_documents(
+        documents,
+        embedding=embeddings
+    )
+    retriever = vectorstore.as_retriever(search_kwargs={"k": 3})
+else:
+    retriever = None
 
 llm = ChatHuggingFace(
     repo_id="microsoft/Phi-4-mini-instruct",
     temperature=0.2
 )
 
-print("✅ Indexing complete.")
+print("✅ RAG initialized.")
 
 def ask_rag_with_status(question: str):
-    status = [
-        "🔍 Retrieving relevant documents...",
-        "🧠 Building context...",
-        "💭 Thinking..."
-    ]
+    if not retriever:
+        return {
+            "status": ["⚠️ No documents uploaded yet"],
+            "answer": "Please upload PDF files to the kb_docs folder."
+        }
 
     docs = retriever.get_relevant_documents(question)
     context = "\n\n".join(d.page_content for d in docs)
 
     messages = [
-        SystemMessage(content="You are a helpful assistant. Answer using the context."),
+        SystemMessage(content="Answer using the provided context."),
         HumanMessage(content=f"Context:\n{context}\n\nQuestion: {question}")
     ]
 
     response = llm.invoke(messages)
 
     return {
-        "status": status,
+        "status": ["🔍 Retrieved documents", "🧠 Generating answer"],
         "answer": response.content
     }