Update app.py

app.py

@@ -1,51 +1,170 @@
-# -----------------------------
-# Imports
-# -----------------------------
-from langchain_openai import ChatOpenAI
-from langchain_core.caches import BaseCache  # Must be imported before model_rebuild
-from langchain.chains import LLMChain
-from langchain.prompts import PromptTemplate
+import gradio as gr
+from langchain_community.vectorstores import Chroma
+from langchain_core.documents import Document
+from langchain_huggingface import HuggingFaceEmbeddings
+import json
 import os
 
-# -----------------------------
-# Fix for "class not fully defined" Pydantic error
-# -----------------------------
-ChatOpenAI.model_rebuild()
-
-# -----------------------------
-# API key setup
-# -----------------------------
-# For local use: ensure OPENAI_API_KEY is set in your environment
-# For Hugging Face Space: add it under Settings → Secrets
-if "OPENAI_API_KEY" not in os.environ:
-    raise ValueError("OPENAI_API_KEY not found. Please set it in environment or Space secrets.")
-
-# -----------------------------
-# Initialise model
-# -----------------------------
-llm = ChatOpenAI(
-    model="gpt-4o-mini",  # or "gpt-4o", "gpt-4-turbo" etc.
-    temperature=0,
-)
+# -------------------------------
+# CONFIGURATION
+# -------------------------------
+os.environ["OPENAI_API_KEY"] = os.getenv("OPENAI_API_KEY", "sk-your-key-here")
+
+# -------------------------------
+# LOAD DATA AND BUILD RAG CHAIN
+# -------------------------------
+def clean_metadata(metadata):
+    """Convert list values to comma-separated strings for ChromaDB compatibility"""
+    cleaned = {}
+    for key, value in metadata.items():
+        if isinstance(value, list):
+            cleaned[key] = ", ".join(str(v) for v in value)
+        elif isinstance(value, (str, int, float, bool)) or value is None:
+            cleaned[key] = value
+        else:
+            cleaned[key] = str(value)
+    return cleaned
 
-# -----------------------------
-# Example prompt and chain
-# -----------------------------
-prompt = PromptTemplate(
-    input_variables=["topic"],
-    template="Summarise the main challenges and opportunities of using AI in {topic}."
+print("Loading documents...")
+docs = []
+with open("helpwildlife_rag.jsonl", "r", encoding="utf-8") as f:
+    for line in f:
+        entry = json.loads(line)
+        metadata = entry.get("metadata", {})
+        docs.append(Document(
+            page_content=entry["text"],
+            metadata=clean_metadata(metadata)
+        ))
+print(f"✓ Loaded {len(docs)} documents")
+
+print("Loading embedding model...")
+embeddings = HuggingFaceEmbeddings(
+    model_name="sentence-transformers/all-MiniLM-L6-v2"
 )
 
-chain = LLMChain(
-    llm=llm,
-    prompt=prompt
+print("Building vector store...")
+try:
+    vectorstore = Chroma.from_documents(docs, embedding=embeddings)
+except TypeError:
+    vectorstore = Chroma.from_documents(docs, embedding_function=embeddings)
+
+retriever = vectorstore.as_retriever(search_kwargs={"k": 3})
+print("✓ Vector store created")
+
+# Prompt template as string
+prompt_template = """
+You are a compassionate wildlife advice assistant.
+Your only source of information is the text provided in the CONTEXT section below.
+Do not use outside knowledge, guesses, or general reasoning.
+If the context does not contain enough information to answer fully,
+say "I'm not sure based on the available guidance."
+
+Never suggest killing or harming any animal.
+Always emphasise that taking an injured or distressed animal to a local wildlife rescue
+or 24/7 animal hospital is the safest and most humane course of action.
+Discourage people from trying to handle or treat the animal themselves,
+and note that general veterinary clinics may euthanise wild animals unnecessarily.
+
+Adopt a voice of empathy and respect for all life, consistent with vegan principles:
+there are no animals that are pests, vermin, or unworthy of care.
+Respond clearly and calmly, with brief, practical, step-by-step guidance suitable for the public.
+
+---------------------
+CONTEXT (from HelpWildlife data file):
+{context}
+---------------------
+QUESTION:
+{question}
+
+YOUR ANSWER:
+"""
+
+# Global LLM variable - will be initialized lazily
+_llm = None
+
+def get_llm():
+    """Lazy load the LLM only when first needed"""
+    global _llm
+    if _llm is None:
+        from langchain_openai import ChatOpenAI
+        from langchain_core.prompts import ChatPromptTemplate
+        from langchain_core.output_parsers import StrOutputParser
+        from langchain_core.runnables import RunnablePassthrough
+        
+        # Try to fix pydantic issue
+        try:
+            from langchain_core.caches import BaseCache
+            ChatOpenAI.model_rebuild()
+        except Exception:
+            pass
+        
+        llm = ChatOpenAI(model_name="gpt-4o-mini", temperature=0)
+        prompt = ChatPromptTemplate.from_template(prompt_template)
+        
+        def format_docs(docs):
+            return "\n\n".join(doc.page_content for doc in docs)
+        
+        _llm = (
+            {
+                "context": retriever | format_docs,
+                "question": RunnablePassthrough()
+            }
+            | prompt
+            | llm
+            | StrOutputParser()
+        )
+    return _llm
+
+print("✓ Setup complete - LLM will initialize on first use")
+
+# -------------------------------
+# GRADIO INTERFACE
+# -------------------------------
+def ask_wildlife_question(question):
+    """Process a wildlife question and return an answer"""
+    if not question.strip():
+        return "Please enter a question about wildlife."
+    
+    try:
+        rag_chain = get_llm()
+        answer = rag_chain.invoke(question)
+        return answer
+    except Exception as e:
+        return f"Error: {str(e)}\n\nPlease check your OpenAI API key is set correctly in Space secrets."
+
+# Example questions
+examples = [
+    "I found a baby hedgehog out during the day. What should I do?",
+    "There's a bird that seems injured in my garden. How can I help?",
+    "I found a baby bird on the ground. Should I put it back in the nest?",
+    "A fox is limping in my backyard. What should I do?",
+    "How do I know if a wild animal needs help?"
+]
+
+# Create Gradio interface
+demo = gr.Interface(
+    fn=ask_wildlife_question,
+    inputs=gr.Textbox(
+        label="Ask a Wildlife Question",
+        placeholder="e.g., I found a baby bird on the ground...",
+        lines=3
+    ),
+    outputs=gr.Textbox(
+        label="Compassionate Advice",
+        lines=10
+    ),
+    title="🦔 Wildlife Rescue Assistant",
+    description="""
+    Ask questions about helping wildlife in distress. This assistant provides compassionate, 
+    evidence-based advice prioritizing the wellbeing of all animals.
+    
+    ⚠️ **Important**: This tool provides general guidance only. For urgent situations, 
+    contact your local wildlife rescue or 24/7 animal hospital immediately.
+    """,
+    examples=examples,
+    theme=gr.themes.Soft(),
+    allow_flagging="never"
 )
 
-# -----------------------------
-# Run the chain
-# -----------------------------
 if __name__ == "__main__":
-    topic = "government recordkeeping"
-    result = chain.run(topic=topic)
-    print("=== Model Output ===")
-    print(result)
+    demo.launch()