app.py

import os
import streamlit as st
import pandas as pd
import openai
import sqlite3
import json
import numpy as np
import datetime
import re
from langchain.chains import RetrievalQA
from langchain.schema import Document
from langchain_core.retrievers import BaseRetriever
from pydantic import Field
from langchain_openai import ChatOpenAI
from langchain.prompts import ChatPromptTemplate

DB_PATH = "json_vector.db"

# Read API keys
OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
OPENROUTER_API_KEY = os.getenv("OPENROUTER_API_KEY")  # NEW

EMBEDDING_MODEL = "text-embedding-ada-002"

if "ingested_batches" not in st.session_state:
    st.session_state.ingested_batches = 0
if "messages" not in st.session_state:
    st.session_state.messages = []
if "json_links" not in st.session_state:
    st.session_state.json_links = []
if "json_link_details" not in st.session_state:
    st.session_state.json_link_details = {}

st.set_page_config(page_title="Chat with Your JSON Vectors (Hybrid, Clean)", layout="wide")
st.title("Chat with Your Vectorized JSON Files")

uploaded_files = st.file_uploader(
    "Upload JSON files in batches (any structure)", type="json", accept_multiple_files=True
)

def flatten_json_obj(obj, parent_key="", sep="."):
    items = {}
    if isinstance(obj, dict):
        for k, v in obj.items():
            new_key = f"{parent_key}{sep}{k}" if parent_key else k
            if (
                k.lower() in {"customer", "user", "email", "username"} and
                isinstance(v, str) and "@" in v
            ):
                local = v.split("@")[0]
                local_clean = re.sub(r'[^a-zA-Z0-9]', ' ', local)
                parts = [part for part in local_clean.split() if part]
                if parts:
                    items[new_key + "_name"] = parts[0].lower()
                    items[new_key + "_all_names"] = " ".join(parts).lower()
            items.update(flatten_json_obj(v, new_key, sep=sep))
    elif isinstance(obj, list):
        for i, v in enumerate(obj):
            new_key = f"{parent_key}{sep}{i}" if parent_key else str(i)
            items.update(flatten_json_obj(v, new_key, sep=sep))
    else:
        items[parent_key] = obj
    return items

def get_embedding(text):
    client = openai.OpenAI(api_key=OPENAI_API_KEY)
    response = client.embeddings.create(input=[text], model=EMBEDDING_MODEL)
    return response.data[0].embedding

def ensure_table():
    conn = sqlite3.connect(DB_PATH)
    cursor = conn.cursor()
    cursor.execute("""
    CREATE TABLE IF NOT EXISTS json_records (
        id INTEGER PRIMARY KEY AUTOINCREMENT,
        batch_time TEXT,
        source_file TEXT,
        raw_json TEXT,
        flat_text TEXT,
        embedding BLOB
    )
    """)
    conn.commit()
    conn.close()

def ingest_json_files(files):
    ensure_table()
    rows = []
    batch_time = datetime.datetime.utcnow().isoformat()
    for file in files:
        file.seek(0)
        raw = json.load(file)
        source_name = file.name
        records = raw if isinstance(raw, list) else [raw]
        for rec in records:
            flat = flatten_json_obj(rec)
            flat_text = "; ".join([f"{k}: {v}" for k, v in flat.items()])
            rows.append((batch_time, source_name, json.dumps(rec), flat_text))
    if not rows:
        st.warning("No records found in uploaded files!")
        return
    df = pd.DataFrame(rows, columns=["batch_time", "source_file", "raw_json", "flat_text"])
    st.write(f"Flattened {len(df)} records. Generating embeddings (this may take time, please wait)...")
    df["embedding"] = df["flat_text"].apply(get_embedding)
    conn = sqlite3.connect(DB_PATH)
    cursor = conn.cursor()
    for _, row in df.iterrows():
        emb_bytes = np.array(row.embedding, dtype=np.float32).tobytes()
        cursor.execute("""
            INSERT INTO json_records (batch_time, source_file, raw_json, flat_text, embedding)
            VALUES (?, ?, ?, ?, ?)
        """, (row.batch_time, row.source_file, row.raw_json, row.flat_text, emb_bytes))
    conn.commit()
    conn.close()
    st.success(f"Ingested and indexed {len(df)} new records!")
    st.session_state.ingested_batches += 1

if uploaded_files and st.button("Ingest batch to database"):
    ingest_json_files(uploaded_files)

def query_vector_db(user_query, top_k=5):
    query_emb = get_embedding(user_query)
    conn = sqlite3.connect(DB_PATH)
    cursor = conn.cursor()
    cursor.execute("SELECT id, batch_time, source_file, raw_json, flat_text, embedding FROM json_records")
    results = []
    for row in cursor.fetchall():
        db_emb = np.frombuffer(row[5], dtype=np.float32)
        if len(db_emb) != len(query_emb): continue
        sim = np.dot(query_emb, db_emb) / (np.linalg.norm(query_emb) * np.linalg.norm(db_emb))
        results.append((sim, row))
    conn.close()
    results = sorted(results, reverse=True)[:top_k]
    docs = []
    for sim, row in results:
        meta = {
            "id": row[0],
            "batch_time": str(row[1]),
            "source_file": row[2],
            "similarity": f"{sim:.4f} (embedding)",
            "raw_json": row[3],
        }
        docs.append(Document(page_content=row[4], metadata=meta))
    return docs

def python_fuzzy_match(user_query, top_k=5):
    query_terms = set(user_query.lower().replace("@", " ").replace(".", " ").split())
    conn = sqlite3.connect(DB_PATH)
    cursor = conn.cursor()
    cursor.execute("SELECT id, batch_time, source_file, raw_json, flat_text FROM json_records")
    results = []
    for row in cursor.fetchall():
        flat_text = row[4].lower()
        score = sum(any(term in flat_text for term in query_terms) for term in query_terms)
        if score > 0:
            results.append((score, row))
    conn.close()
    results = sorted(results, reverse=True)[:top_k]
    docs = []
    for score, row in results:
        meta = {
            "id": row[0],
            "batch_time": str(row[1]),
            "source_file": row[2],
            "similarity": f"{score} (fuzzy)",
            "raw_json": row[3],
        }
        docs.append(Document(page_content=row[4], metadata=meta))
    return docs

def extract_main_entity(question):
    import re
    quoted = re.findall(r"['\"]([^'\"]+)['\"]", question)
    if quoted:
        return quoted[0].lower()
    email = re.findall(r"\b[A-Za-z0-9._%+-]+@[A-Za-z0-9.-]+\.[A-Z|a-z]{2,}\b", question)
    if email:
        return email[0].lower().split('@')[0]
    tokens = re.findall(r"\b([A-Za-z0-9]+)\b", question)
    stopwords = {"how", "much", "did", "spend", "was", "the", "is", "in", "on", "for", "a", "an", "of", "to", "with"}
    keywords = [t.lower() for t in tokens if t.lower() not in stopwords]
    if not keywords:
        return ""
    return max(keywords, key=len)

def filter_records_by_entity(records, entity):
    if not entity:
        return records
    matches = []
    for doc in records:
        if entity in doc.page_content.lower():
            matches.append(doc)
        elif any(entity in v.lower() for v in doc.page_content.split(';')):
            matches.append(doc)
    return matches if matches else records

def hybrid_query(user_query, top_k=5):
    vector_docs = query_vector_db(user_query, top_k=top_k)
    fuzzy_docs = python_fuzzy_match(user_query, top_k=top_k)
    all_docs = []
    seen_ids = set()
    for doc in (vector_docs + fuzzy_docs):
        doc_id = doc.metadata.get("id")
        if doc_id not in seen_ids:
            all_docs.append(doc)
            seen_ids.add(doc_id)
    entity = extract_main_entity(user_query)
    entity_docs = filter_records_by_entity(all_docs, entity) if entity else all_docs
    if entity_docs:
        doc = entity_docs[0]
        return [doc]
    else:
        return all_docs[:1]

class HybridRetriever(BaseRetriever):
    top_k: int = Field(default=5)
    def _get_relevant_documents(self, query, run_manager=None, **kwargs):
        return hybrid_query(query, self.top_k)

system_prompt = (
    "You are a JSON data assistant. "
    "If the question mentions a name or email (e.g. Johnny), match it to any field value (even as part of an email) "
    "and answer directly using the record's fields. "
    "For example, if 'customer: johnny.appleseed@gmail.com' and the question is about Johnny, you should use that record."
    "If you can't find the answer, reply: 'I don’t have that information.'"
    "Never make up data. Never ask for clarification."
)
prompt = ChatPromptTemplate.from_messages([
    ("system", system_prompt),
    ("human", "Here are the most relevant records:\n{context}\n\nQuestion: {question}")
])

# --- LLM PROVIDER SELECTION --- # NEW/MODIFIED FOR LLM SELECTION
llm_provider = st.selectbox(
    "Select LLM Provider",
    options=["OpenAI GPT-4", "Mistral (OpenRouter)"],
    index=0,
    help="Choose which LLM to use for answering your questions."
)

def get_llm(llm_provider):
    if llm_provider == "OpenAI GPT-4":
        return ChatOpenAI(
            model="gpt-4.1",
            openai_api_key=OPENAI_API_KEY,
            temperature=0,
        )
    else:  # "Mistral (OpenRouter)"
        return ChatOpenAI(
            model="mistralai/mistral-small-3.1-24b-instruct:free",  # Or another Mistral model if desired
            openai_api_key=OPENROUTER_API_KEY,
            openai_api_base="https://openrouter.ai/api/v1",
            temperature=0,
        )

llm = get_llm(llm_provider)

retriever = HybridRetriever(top_k=5)
qa_chain = RetrievalQA.from_chain_type(
    llm=llm,
    retriever=retriever,
    chain_type_kwargs={"prompt": prompt},
    return_source_documents=True,
)

st.markdown("### Ask any question about your data, just like ChatGPT.")

def show_tiny_json_links():
    # Only show for the last assistant answer if there are matching JSONs
    if not st.session_state.json_links:
        return
    for idx, link_key in enumerate(st.session_state.json_links):
        label = st.session_state.json_link_details[link_key]['label']
        rec = st.session_state.json_link_details[link_key]['record']
        expander_label = f"<span style='font-size:11px; color:#444; text-decoration:underline;'>[view JSON]</span> <span style='font-size:10px; color:#aaa'>{label}</span>"
        with st.expander(label="", expanded=False):
            st.markdown(expander_label, unsafe_allow_html=True)
            st.code(json.dumps(rec, indent=2), language="json")
    st.session_state.json_links = []
    st.session_state.json_link_details = {}

for msg in st.session_state.messages:
    if msg["role"] == "user":
        st.markdown(f"<div style='color: #4F8BF9;'><b>User:</b> {msg['content']}</div>", unsafe_allow_html=True)
    elif msg["role"] == "assistant":
        st.markdown(f"<div style='color: #1C6E4C;'><b>Agent:</b> {msg['content']}</div>", unsafe_allow_html=True)
        show_tiny_json_links()

def send_message():
    user_input = st.session_state.temp_input.strip()
    if not user_input:
        return
    st.session_state.messages.append({"role": "user", "content": user_input})
    with st.spinner("Thinking..."):
        result = qa_chain({"query": user_input})
        answer = result['result']
        st.session_state.messages.append({"role": "assistant", "content": answer})
        docs = result['source_documents']
        link_keys = []
        link_details = {}
        for idx, doc in enumerate(docs):
            link_key = f"json_{doc.metadata['id']}_{idx}"
            rec = json.loads(doc.metadata["raw_json"])
            label = f"{doc.metadata['source_file']} | Similarity: {doc.metadata['similarity']}"
            link_details[link_key] = {"label": label, "record": rec}
            link_keys.append(link_key)
        st.session_state.json_links = link_keys
        st.session_state.json_link_details = link_details
    st.session_state.temp_input = ""

st.text_input("Your message:", key="temp_input", on_change=send_message)

if st.button("Clear chat"):
    st.session_state.messages = []
    st.session_state.json_links = []
    st.session_state.json_link_details = {}

st.info(f"Batches ingested so far (this session): {st.session_state.ingested_batches}")