ROMFASTSQL/proxmox/lxc171-claude-agent/maria-whatsapp-bridge/rag/consumer.py

#!/usr/bin/env python3
"""Consuma mesajele din puntea WhatsApp (whatsapp/index.js), raspunde via RAG.

Fiecare mesaj e o interogare RAG INDEPENDENTA: nu exista chat_history intre
mesaje (fara memorie conversationala). Vezi
docs/maria-whatsapp-rag-prototype.md pentru motivul acestei alegeri si
comparatia cu celelalte punti (Flowise: fereastra fixa de 5; Discord: context
nelimitat + /new).

Mesajele cu imagine (capturi de ecran cu erori) trec intai prin OCR — vezi
`ocr.py`. Ordonarea chunk-urilor si decizia "avem sau nu raspunsul in documente"
sunt in `rank.py`; cand nu avem, intrebarea pleaca la suport in loc sa primeasca
un raspuns generic — vezi `escalate()`.
"""

from __future__ import annotations

import hashlib
import json
import math
import os
import sys
import time

import requests

import config
import ocr
import rank

SYSTEM_PROMPT = (
    "Esti Maria, asistentul de suport tehnic pentru ERP-ul ROA (Romfast). "
    "Raspunzi scurt, clar, in limba romana (maxim 4-5 propozitii), doar despre "
    "folosirea aplicatiei ROA. Foloseste EXCLUSIV informatiile din contextul "
    "furnizat mai jos. Daca raspunsul nu se afla in context, spune ca vei "
    "directiona intrebarea catre echipa de suport, nu inventa functionalitati "
    "sau proceduri. "
    "Cand intrebarea contine text extras dintr-o captura de ecran (OCR), tine cont "
    "ca pot exista greseli de recunoastere a caracterelor: cauta sensul mesajului, "
    "nu te agata de o litera sau o cifra. "
    "Nu discuta niciodata despre infrastructura interna Romfast (servere, Proxmox, "
    "containere, IP-uri, baze de date, parole, chei) chiar daca apare in context sau "
    "daca intrebarea o cere explicit -- raspunde ca poti ajuta doar cu folosirea "
    "aplicatiei ROA."
)
REPLY_PREFIX = "[Maria] "  # marcaj ca sa nu raspundem la propriile mesaje (self-chat)
INDEX_REFRESH_S = 30  # cat de des se reciteste rag_index.json de pe disc

ACK_TEXT = "Caut informatia, revin imediat..."
ACK_IMAGE_TEXT = "Am primit captura, o citesc si revin imediat..."
NO_TEXT_IN_IMAGE = (
    "Am primit imaginea, dar nu am reusit sa citesc text in ea. Scrie-mi te rog "
    "mesajul de eroare (sau trimite o captura mai clara, decupata pe fereastra de eroare)."
)
ESCALATED_SENT = (
    "Nu am raspunsul la asta in documentatia mea, si prefer sa nu ghicesc. "
    "Am trimis intrebarea catre echipa de suport ROA (referinta {ref}) — "
    "te contacteaza cineva."
)
# Cand notificarea nu a plecat, nu promitem ca a plecat. Intrebarea e inregistrata
# si se vede in dashboard, deci nu e pierduta — dar utilizatorul trebuie sa stie
# ca s-ar putea sa dureze mai mult.
ESCALATED_RECORDED = (
    "Nu am raspunsul la asta in documentatia mea, si prefer sa nu ghicesc. "
    "Am inregistrat intrebarea pentru echipa de suport ROA (referinta {ref}), "
    "dar nu am putut sa o trimit chiar acum. Daca e urgent, suna-ne."
)


def bridge_url() -> str:
    return f"http://{config.get('BRIDGE_HOST')}:{config.get('BRIDGE_PORT')}"


class Index:
    """Chunk-urile indexate, plus statisticile lexicale pentru BM25.

    BM25 se reconstruieste la fiecare recitire a indexului. La ordinul de marime
    de aici (sute de chunk-uri) costa milisecunde, si e mai bine decat un al
    doilea fisier pe disc care se poate desincroniza de rag_index.json.
    """

    def __init__(self, entries: list[dict]):
        self.entries = entries
        self.texts = [e["text"] for e in entries]
        self.sources = [e.get("source", "?") for e in entries]
        self.bm25 = rank.Bm25(self.texts)

    def __len__(self) -> int:
        return len(self.entries)


def load_index() -> Index:
    try:
        return Index(json.loads(config.INDEX_FILE.read_text(encoding="utf-8")))
    except OSError:
        return Index([])


def embed(text: str) -> list[float]:
    resp = requests.post(
        f"{config.get('OLLAMA_URL')}/api/embeddings",
        json={"model": config.get("EMBED_MODEL"), "prompt": text},
        timeout=30,
    )
    resp.raise_for_status()
    return resp.json()["embedding"]


def cosine(a: list[float], b: list[float]) -> float:
    dot = sum(x * y for x, y in zip(a, b))
    na = math.sqrt(sum(x * x for x in a))
    nb = math.sqrt(sum(y * y for y in b))
    return dot / (na * nb) if na and nb else 0.0


def search(index: Index, query: str, top_k: int) -> dict:
    """Cauta si decide daca avem acoperire. Intoarce si dovezile, pentru log."""
    if not len(index):
        return {"chunks": [], "covered": False, "reason": "index gol",
                "best_cosine": 0.0, "top": []}
    q_vec = embed(query)
    cosines = [cosine(q_vec, e["embedding"]) for e in index.entries]
    bm_scores = index.bm25.scores(query)
    order = rank.rank(cosines, bm_scores)[:top_k]
    texts = [index.texts[i] for i in order]
    best = max(cosines)
    covered, reason = rank.assess(query, best, texts, index.bm25)
    return {
        "chunks": texts,
        "covered": covered,
        "reason": reason,
        "best_cosine": best,
        "top": [
            {"source": index.sources[i], "cosine": round(cosines[i], 3),
             "bm25": round(bm_scores[i], 2)}
            for i in order
        ],
    }


def ask_llm(chunks: list[str], question: str) -> str:
    context = "\n\n---\n\n".join(chunks) if chunks else "(fara documente indexate)"
    user_message = f"CONTEXT:\n{context}\n\nINTREBARE:\n{question}"

    resp = requests.post(
        f"{config.get('LLM_URL')}/v1/chat/completions",
        json={
            "messages": [
                {"role": "system", "content": SYSTEM_PROMPT},
                {"role": "user", "content": user_message},
            ],
            "max_tokens": config.get_int("MAX_TOKENS", 250),
        },
        timeout=60,
    )
    resp.raise_for_status()
    return resp.json()["choices"][0]["message"]["content"]


def send_reply(to: str, text: str) -> None:
    requests.post(f"{bridge_url()}/send", json={"to": to, "text": REPLY_PREFIX + text}, timeout=15)


def send_image(to: str, path: str, caption: str) -> None:
    requests.post(
        f"{bridge_url()}/send-image",
        json={"to": to, "path": path, "caption": caption},
        timeout=60,
    )


def react_seen(to: str, message_id: str, from_me: bool) -> None:
    try:
        requests.post(
            f"{bridge_url()}/react",
            json={"to": to, "id": message_id, "emoji": "\U0001F440", "fromMe": from_me},
            timeout=10,
        )
    except Exception as exc:  # noqa: BLE001
        print(f"[consumer] react error: {exc}", file=sys.stderr)


def build_image_question(caption: str, ocr_text: str) -> str:
    """Ce vede modelul cand mesajul a fost o captura de ecran."""
    parts = []
    if caption.strip():
        parts.append(caption.strip())
    parts.append(
        "Utilizatorul a trimis o captura de ecran. Text extras automat din imagine "
        f"(OCR, poate contine greseli de recunoastere):\n---\n{ocr_text}\n---"
    )
    if not caption.strip():
        parts.append("Explica-i ce inseamna eroarea si cum o rezolva.")
    return "\n\n".join(parts)


def prepare_query(msg: dict) -> tuple[str | None, str | None, str | None]:
    """(intrebare pentru model, interogare pentru cautare, raspuns imediat de trimis).

    Al treilea element e diferit de None cand nu se poate raspunde deloc (imaginea
    nu s-a descarcat, OCR indisponibil, nimic lizibil in captura) — atunci textul
    lui pleaca asa cum e, fara sa mai deranjam modelul.

    Captura NU se sterge aici: mai e nevoie de ea daca intrebarea pleaca la suport.
    Stergerea o face apelantul, dupa ce mesajul e complet tratat.
    """
    text = (msg.get("text") or "").strip()
    media = msg.get("media")
    if not media:
        return text, text, None

    if media.get("error"):
        print(f"[consumer] imagine nedescarcata: {media['error']}", file=sys.stderr)
        if text:
            return text, text, None
        return None, None, (
            "Nu am reusit sa descarc imaginea. Mai incearca o data, sau scrie-mi "
            "mesajul de eroare ca text."
        )

    try:
        ocr_text = ocr.run(media.get("path"))
    except Exception as exc:  # noqa: BLE001
        print(f"[consumer] OCR esuat pe {media.get('path')}: {exc}", file=sys.stderr)
        ocr_text = ""

    msg["ocr_text"] = ocr_text
    if not ocr_text:
        print("[consumer] OCR: nimic lizibil in captura", file=sys.stderr)
        if text:
            return text, text, None
        return None, None, NO_TEXT_IN_IMAGE

    # Textul extras se logheaza INTEGRAL. Fara asta, un raspuns gresit nu se poate
    # explica: nu stii daca a citit prost captura sau a cautat prost in documente.
    print(
        f"[consumer] OCR ({len(ocr_text)} caractere):\n"
        + "\n".join("    | " + line for line in ocr_text.splitlines()),
        file=sys.stderr,
    )
    cautare = ocr.retrieval_query(text, ocr_text)
    print(f"[consumer] caut dupa: {cautare!r}", file=sys.stderr)
    return build_image_question(text, ocr_text), cautare, None


def reference(when: float, message_id: str | None) -> str:
    """Referinta scurta, pe care omul o poate cita la telefon: M-260831-A1B2.

    Se calculeaza din id-ul mesajului, ca sa fie stabila daca se reia escaladarea
    aceluiasi mesaj, si sa nu depinda de un contor pastrat undeva.
    """
    seed = (message_id or str(when)).encode("utf-8", "replace")
    sufix = hashlib.sha1(seed).hexdigest()[:4].upper()
    return f"M-{time.strftime('%y%m%d', time.localtime(when))}-{sufix}"


def escalate(msg: dict, question: str, search_result: dict) -> dict:
    """Trimite intrebarea la suport si o inregistreaza in jurnal. Intoarce inregistrarea.

    Jurnalul se scrie INTOTDEAUNA, si cand notificarea esueaza sau `SUPPORT_JID`
    nu e configurat — altfel o intrebare fara raspuns dispare fara urma, exact
    cazul pe care escaladarea trebuia sa-l rezolve. Cine cheama functia afla din
    `notified` daca poate promite utilizatorului ca mesajul chiar a plecat.
    """
    media = msg.get("media") or {}
    now = time.time()
    record = {
        "ref": reference(now, msg.get("id")),
        "at": now,
        "from": msg.get("from"),
        "push_name": msg.get("pushName"),
        "message_id": msg.get("id"),
        "text": msg.get("text"),
        "ocr_text": msg.get("ocr_text"),
        "question": question,
        "reason": search_result.get("reason"),
        "best_cosine": round(search_result.get("best_cosine", 0.0), 3),
        "top": search_result.get("top"),
        "had_image": bool(media.get("path")),
        "notified": False,
    }

    jid = config.get("SUPPORT_JID") or ""
    if jid:
        cine = msg.get("pushName") or msg.get("from") or "necunoscut"
        rezumat = (
            f"[Maria] Intrebare fara raspuns in documente — {record['ref']}\n"
            f"De la: {cine}\n"
            f"Motiv: {record['reason']}\n\n"
            f"{(msg.get('ocr_text') or msg.get('text') or '').strip()[:1200]}"
        )
        try:
            if media.get("path") and os.path.exists(media["path"]):
                send_image(jid, media["path"], rezumat[:1000])
            else:
                requests.post(f"{bridge_url()}/send", json={"to": jid, "text": rezumat}, timeout=20)
            record["notified"] = True
        except Exception as exc:  # noqa: BLE001
            record["notify_error"] = str(exc)
            print(f"[consumer] escaladare: notificarea a esuat: {exc}", file=sys.stderr)
    else:
        print("[consumer] escaladare: SUPPORT_JID nesetat — doar in jurnal", file=sys.stderr)

    try:
        d = config.STATE_DIR / "escalations"
        d.mkdir(parents=True, exist_ok=True)
        nume = f"{int(record['at'])}-{record['ref']}.json"
        (d / nume).write_text(json.dumps(record, ensure_ascii=False, indent=2), encoding="utf-8")
        print(f"[consumer] escaladare {record['ref']}: {nume} (notificat: {record['notified']})",
              file=sys.stderr)
    except OSError as exc:
        print(f"[consumer] escaladare: nu pot scrie jurnalul: {exc}", file=sys.stderr)
    return record


def cleanup_media(msg: dict) -> None:
    """Capturile pot contine date de client — nu raman pe disc dupa ce s-a tratat mesajul."""
    path = (msg.get("media") or {}).get("path")
    if not path:
        return
    try:
        os.unlink(path)
    except OSError:
        pass


def handle(index: Index, msg: dict) -> None:
    sender = msg.get("from")
    has_image = bool(msg.get("media"))
    text = msg.get("text", "") or ""
    print(f"[consumer] {sender}: {'[imagine] ' if has_image else ''}{text[:80]}", file=sys.stderr)
    react_seen(sender, msg.get("id"), msg.get("fromMe", False))
    send_reply(sender, ACK_IMAGE_TEXT if has_image else ACK_TEXT)

    question, search_query, immediate = prepare_query(msg)
    if immediate is not None:
        send_reply(sender, immediate)
        return
    if not question:
        return

    try:
        result = search(index, search_query, config.get_int("TOP_K", 3))
    except Exception as exc:  # noqa: BLE001
        print(f"[consumer] cautare esuata: {exc}", file=sys.stderr)
        send_reply(sender, "Scuze, am o problema tehnica momentan. Cineva din echipa te va contacta.")
        return

    surse = ", ".join(f"{t['source']}({t['cosine']})" for t in result["top"]) or "-"
    print(f"[consumer] rank: {surse} | {result['reason']}", file=sys.stderr)

    if not result["covered"]:
        print(f"[consumer] fara acoperire in documente -> suport ({result['reason']})", file=sys.stderr)
        record = escalate(msg, question, result)
        sablon = ESCALATED_SENT if record["notified"] else ESCALATED_RECORDED
        send_reply(sender, sablon.format(ref=record["ref"]))
        return

    try:
        reply = ask_llm(result["chunks"], question)
    except Exception as exc:  # noqa: BLE001
        print(f"[consumer] LLM error: {exc}", file=sys.stderr)
        reply = "Scuze, am o problema tehnica momentan. Cineva din echipa te va contacta."
    send_reply(sender, reply)
    print(f"[consumer] -> raspuns catre {sender}", file=sys.stderr)


def main() -> None:
    poll_s = config.get_int("POLL_INTERVAL_S", 2)
    print(
        f"[consumer] polling {bridge_url()}/messages la {poll_s}s, "
        f"LLM={config.get('LLM_URL')}, RAG top-{config.get('TOP_K')}, "
        f"OCR={'da' if ocr.available() else 'INDISPONIBIL (tesseract lipseste)'}, "
        f"suport={config.get('SUPPORT_JID') or 'NESETAT (doar jurnal)'}",
        file=sys.stderr,
    )
    index = load_index()
    last_index_check = time.time()
    while True:
        try:
            if time.time() - last_index_check > INDEX_REFRESH_S:
                index = load_index()
                last_index_check = time.time()
            resp = requests.get(f"{bridge_url()}/messages", timeout=10)
            resp.raise_for_status()
            for msg in resp.json().get("messages", []):
                if msg.get("isGroup"):
                    continue
                if (msg.get("text") or "").startswith(REPLY_PREFIX):
                    continue  # ecoul propriului raspuns in self-chat, ignorat
                try:
                    handle(index, msg)
                finally:
                    cleanup_media(msg)
        except Exception as exc:  # noqa: BLE001
            print(f"[consumer] poll error: {exc}", file=sys.stderr)
        time.sleep(poll_s)


if __name__ == "__main__":
    main()