app.py

import os
import re
import time
import requests
import pandas as pd
import matplotlib
matplotlib.use("Agg")
import matplotlib.pyplot as plt
import gradio as gr

from bs4 import BeautifulSoup
from langchain_core.documents import Document
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain_huggingface import HuggingFaceEmbeddings
from langchain_community.vectorstores import Chroma
from rank_bm25 import BM25Okapi
from sentence_transformers import CrossEncoder

PERSIST_DIR = "k8s_chroma_db"

URLS = {
    "pods": "https://kubernetes.io/docs/concepts/workloads/pods/",
    "deployments": "https://kubernetes.io/docs/concepts/workloads/controllers/deployment/",
    "services": "https://kubernetes.io/docs/concepts/services-networking/service/",
    "namespaces": "https://kubernetes.io/docs/concepts/overview/working-with-objects/namespaces/",
    "nodes": "https://kubernetes.io/docs/concepts/architecture/nodes/",
    "statefulsets": "https://kubernetes.io/docs/concepts/workloads/controllers/statefulset/",
    "rbac": "https://kubernetes.io/docs/reference/access-authn-authz/rbac/",
    "persistent-volumes": "https://kubernetes.io/docs/concepts/storage/persistent-volumes/",
    "ingress": "https://kubernetes.io/docs/concepts/services-networking/ingress/",
    "autoscaling": "https://kubernetes.io/docs/tasks/run-application/horizontal-pod-autoscale/",
}

def scrape_page(name, url):
    try:
        r = requests.get(url, timeout=20)
        soup = BeautifulSoup(r.text, "html.parser")
        content = soup.find("div", class_="td-content")
        if not content:
            return None
        text = content.get_text(separator="\n").strip()
        return Document(page_content=text, metadata={"doc_id": name, "url": url})
    except:
        return None

def build_or_load_kb():
    embedding_model = HuggingFaceEmbeddings(
        model_name="sentence-transformers/all-MiniLM-L6-v2"
    )
    
    if os.path.isdir(PERSIST_DIR):
        vectordb = Chroma(
            embedding_function=embedding_model,
            persist_directory=PERSIST_DIR
        )
        data = vectordb._collection.get(include=["documents", "metadatas"])
        chunks = [
            Document(page_content=doc, metadata=meta)
            for doc, meta in zip(data["documents"], data["metadatas"])
        ]
        return vectordb, chunks

    docs = []
    for name, url in URLS.items():
        d = scrape_page(name, url)
        if d:
            docs.append(d)

    splitter = RecursiveCharacterTextSplitter(chunk_size=900, chunk_overlap=200)
    chunks = splitter.split_documents(docs)

    vectordb = Chroma.from_documents(
        chunks,
        embedding_model,
        persist_directory=PERSIST_DIR
    )
    return vectordb, chunks

vectordb, chunks = build_or_load_kb()

bm25_corpus = [doc.page_content.split() for doc in chunks]
bm25 = BM25Okapi(bm25_corpus)

reranker = CrossEncoder("cross-encoder/ms-marco-MiniLM-L-12-v2")

retriever = vectordb.as_retriever(
    search_type="similarity_score_threshold",
    search_kwargs={"k": 8, "score_threshold": 0.4},
)

def hybrid_search(query):
    vresults = retriever.invoke(query)
    tokens = query.lower().split()
    bm_scores = bm25.get_scores(tokens)
    bm_ranked = sorted(zip(bm_scores, chunks), key=lambda x: x[0], reverse=True)
    bmresults = [d for _, d in bm_ranked[:5]]

    combined = vresults + bmresults
    seen = set()
    unique = []
    for d in combined:
        key = (d.metadata.get("doc_id"), d.page_content[:50])
        if key not in seen:
            seen.add(key)
            unique.append(d)

    if not unique:
        return []

    pairs = [(query, d.page_content) for d in unique]
    scores = reranker.predict(pairs)
    ranked = sorted(zip(scores, unique), key=lambda x: x[0], reverse=True)[:5]

    for s, d in ranked:
        d.metadata["rerank_score"] = float(s)

    return [d for _, d in ranked]

def call_llm(prompt):
    api_key = os.getenv("OPENROUTER_API_KEY")
    if not api_key:
        return "⚠️ Missing OPENROUTER_API_KEY environment variable.\nGroundedness: 0%"

    try:
        res = requests.post(
            "https://openrouter.ai/api/v1/chat/completions",
            headers={
                "Authorization": f"Bearer {api_key}",
                "HTTP-Referer": "https://huggingface.co/",
                "X-Title": "Kubernetes RAG Assistant"
            },
            json={
                "model": "meta-llama/llama-3.1-8b-instruct",
                "messages": [{"role": "user", "content": prompt}],
                "max_tokens": 300,
                "temperature": 0.2
            }
        )
        res.raise_for_status()
        return res.json()["choices"][0]["message"]["content"]
    except Exception as e:
        return f"⚠️ LLM Error: {e}\nGroundedness: 0%"

def build_context(query):
    docs = hybrid_search(query)
    if not docs:
        return "", [], []
    ctx, srcs, scores = "", [], []
    for i, d in enumerate(docs, start=1):
        ctx += f"[{i}] {d.page_content[:900]}\nSource: {d.metadata['url']}\n\n"
        srcs.append(f"[{i}] → {d.metadata['url']}")
        scores.append(d.metadata["rerank_score"])
    return ctx, srcs, scores

def init_metrics():
    return {"q":[], "lat":[], "tok":[], "g":[], "cit":[], "r":[], "type":[]}

def classify(q):
    q=q.lower()
    return "how-to" if "how" in q else ("debug" if "error" in q else "general")

def answer(q, history, metrics):
    if metrics is None: metrics = init_metrics()
    start  = time.time()

    ctx, srcs, scores = build_context(q)
    if not ctx:
        txt = "Not in docs.\nGroundedness: 0%"
        history.append((q, txt))
        return history,"",metrics
        
    prompt = f"""Use context ONLY. Cite every sentence as [n].
User question: {q}

Context:
{ctx}
Groundedness MUST appear as: Groundedness: XX%"""
    txt = call_llm(prompt)

    latency = time.time() - start
    grounded = int(re.search(r"Groundedness:\s*(\d+)%", txt).group(1)) if "Groundedness" in txt else 0
    tokens = len(txt.split())
    cites = len(set(re.findall(r"\[(\d+)\]", txt)))
    avg = sum(scores)/len(scores)

    final = txt+"\n\nSources:\n"+"\n".join(srcs)
    history.append((q, final))

    metrics["q"].append(q)
    metrics["lat"].append(latency)
    metrics["tok"].append(tokens)
    metrics["g"].append(grounded)
    metrics["cit"].append(cites)
    metrics["r"].append(avg)
    metrics["type"].append(classify(q))

    return history,"",metrics

def render(metrics):
    if len(metrics["q"])==0: return [],0,0,0
    rows=[[
        i+1, metrics["q"][i], round(metrics["lat"][i],3),
        metrics["tok"][i], metrics["g"][i],
        round(metrics["r"][i],2), metrics["cit"][i], metrics["type"][i]
    ] for i in range(len(metrics["q"]))]
    avgL=sum(metrics["g"])/len(metrics["g"])
    avgG=sum(metrics["lat"])/len(metrics["lat"])
    avgT=sum(metrics["tok"])/len(metrics["tok"])
    return rows,avgL,avgG,avgT

metrics_state = gr.State(init_metrics())

with gr.Blocks(title="Kubernetes RAG Assistant") as app:
    gr.Markdown("# ☸ Kubernetes RAG Assistant")
    with gr.Tab("Chat"):
        chat = gr.Chatbot()
        inp  = gr.Textbox(label="Ask anything about Kubernetes")
        clear= gr.Button("Reset")
        inp.submit(answer,[inp,chat,metrics_state],[chat,inp,metrics_state])
        clear.click(lambda: ([], "", init_metrics()), None, [chat,inp,metrics_state])

    with gr.Tab("Analytics"):
        table   = gr.DataFrame(headers=["ID","Query","Latency","Tokens","Grounded","Relevance","Citations","Type"])
        avgL    = gr.Number(label="Avg Groundedness")
        avgG    = gr.Number(label="Avg Latency")  
        avgT    = gr.Number(label="Avg Tokens")
        refresh = gr.Button("Update Dashboard")
        refresh.click(render,[metrics_state],[table,avgL,avgG,avgT])

app.launch()