app.py

import gradio as gr
from transformers import pipeline
from sentence_transformers import SentenceTransformer
import pandas as pd
import numpy as np
import zipfile, os, re, torch

# -----------------------------
# Load Mistral (FP16, GPU if available)
# -----------------------------
llm = pipeline(
    "text-generation",
    model="mistralai/Mistral-7B-Instruct-v0.2",
    torch_dtype=torch.float16,
    device_map="auto"
)

# -----------------------------
# Load embedding model
# -----------------------------
embedding_model = SentenceTransformer("nlpaueb/legal-bert-base-uncased")

# -----------------------------
# Extract ZIP with provincial legal texts
# -----------------------------
zip_path = "/app/provinces.zip"
extract_folder = "/app/provinces_texts"

if os.path.exists(extract_folder):
    import shutil
    shutil.rmtree(extract_folder)

with zipfile.ZipFile(zip_path, "r") as z:
    z.extractall(extract_folder)

date_pattern = re.compile(r"(\d{4}[-_]\d{2}[-_]\d{2})")

# -----------------------------
# Parse documents
# -----------------------------
def parse_metadata_and_content(raw):
    if "CONTENT:" not in raw:
        raise ValueError("Missing CONTENT block")

    header, content = raw.split("CONTENT:", 1)
    metadata = {}
    pdfs = []

    for line in header.split("\n"):
        if ":" in line and not line.startswith("-"):
            k, v = line.split(":", 1)
            metadata[k.strip().upper()] = v.strip()
        elif line.strip().startswith("-"):
            pdfs.append(line.strip())

    if pdfs:
        metadata["PDF_LINKS"] = "\n".join(pdfs)

    return metadata, content.strip()

documents = []
for root, dirs, files in os.walk(extract_folder):
    for filename in files:
        if not filename.endswith(".txt") or filename.startswith("._"):
            continue

        path = os.path.join(root, filename)
        try:
            raw = open(path, "r", encoding="latin-1").read()
            metadata, content = parse_metadata_and_content(raw)

            for p in [x.strip() for x in content.split("\n\n") if x.strip()]:
                documents.append({
                    "source_title": metadata.get("SOURCE_TITLE", "Unknown"),
                    "province": metadata.get("PROVINCE", "Unknown"),
                    "last_updated": metadata.get("LAST_UPDATED", "Unknown"),
                    "url": metadata.get("URL", "N/A"),
                    "pdf_links": metadata.get("PDF_LINKS", ""),
                    "text": p
                })

        except Exception as e:
            print("Skipped:", path, e)

print("Loaded paragraphs:", len(documents))

# -----------------------------
# Build embeddings dataframe
# -----------------------------
df = pd.DataFrame(documents)
texts = df["text"].tolist()
embeddings = embedding_model.encode(texts).astype("float16")

df["Embedding"] = list(embeddings)
print("Embedding index ready:", len(df))

# -----------------------------
# Retrieval
# -----------------------------
def retrieve_with_pandas(query, province=None, top_k=2):
    query_emb = embedding_model.encode([query])[0]

    subset = df if province is None else df[df["province"] == province]

    subset = subset.copy()
    subset["Similarity"] = subset["Embedding"].apply(
        lambda x: np.dot(query_emb, x) /
                  (np.linalg.norm(query_emb) * np.linalg.norm(x))
    )

    return subset.sort_values("Similarity", ascending=False).head(top_k)

# -----------------------------
# Province detection
# -----------------------------
def detect_province(q):
    provinces = {
        "yukon": "Yukon", "alberta": "Alberta", "bc": "British Columbia",
        "british columbia": "British Columbia", "manitoba": "Manitoba",
        "newfoundland": "Newfoundland and Labrador",
        "saskatchewan": "Saskatchewan", "sask": "Saskatchewan",
        "ontario": "Ontario", "pei": "Prince Edward Island",
        "quebec": "Quebec", "new brunswick": "New Brunswick",
        "nova scotia": "Nova Scotia", "nunavut": "Nunavut",
        "northwest territories": "Northwest Territories"
    }
    q = q.lower()
    for key, prov in provinces.items():
        if key in q:
            return prov
    return None

# -----------------------------
# Filters
# -----------------------------
def is_disallowed(q):
    banned = ["kill", "suicide", "bomb", "weapon", "harm yourself"]
    return any(b in q.lower() for b in banned)

def is_off_topic(q):
    keys = ["tenant","landlord","rent","evict","lease","repair","notice","unit"]
    return not any(k in q.lower() for k in keys)

# -----------------------------
# Intro (sent once)
# -----------------------------
INTRO = (
    "Hi! I'm a Canadian rental housing assistant. I help summarize and explain "
    "information from Residential Tenancies Acts across Canada.\n\n"
    "**Note:** I'm not a lawyer — this is not legal advice.\n\n"
)

# -----------------------------
# RAG Generation
# -----------------------------
def generate_with_rag(query):
    if is_disallowed(query):
        return "Sorry — I can’t help with harmful topics."

    if is_off_topic(query):
        return "Sorry — I only answer questions about Canadian tenancy law."

    prov = detect_province(query)
    docs = retrieve_with_pandas(query, province=prov, top_k=2)

    if len(docs) == 0:
        return "I couldn’t find anything relevant in the tenancy database."

    context = " ".join(docs["text"].tolist())

    prompt = f"""
Use only the context below. Do NOT invent laws.

Context:
{context}

Question:
{query}

Answer conversationally:
"""

    out = llm(prompt, max_new_tokens=150)[0]["generated_text"]
    answer = out.split("Answer conversationally:", 1)[-1].strip()

    return answer

# -----------------------------
# Gradio Chat (Intro only once)
# -----------------------------
def start_chat():
    return [(None, INTRO)]

def respond(msg, history):
    answer = generate_with_rag(msg)
    history.append((msg, answer))
    return history

with gr.Blocks() as demo:
    chatbot = gr.Chatbot(value=start_chat())
    inp = gr.Textbox(label="Ask a question:")

    inp.submit(respond, [inp, chatbot], chatbot)

demo.launch(share=True)