Files
aramesh-backend/chatbot/vectorstore.py
T
MMD ce0735d775 feat(stack): full overhaul — hardening, ops tooling, monitoring, analytics, chatbot
Security: deep /api/health (DB+MinIO probes, 503 on failure) + /api/live;
rate-limited public forms with honeypots; real client IP behind Traefik
(--proxy-headers); TipTap URL sanitization on write (nh3); non-root
containers + slim .dockerignore; prod MinIO console bound to 127.0.0.1;
HSTS/nosniff on TLS routers; JWT 8h; no default credentials in code.

Reliability: Alembic migrations (create_all baseline); graceful shutdown
(exec uvicorn); hardened backups (checksums, sanity check, freshness gauge,
optional rclone off-site) + restore.sh + hourly db-backup sidecar (ops
profile); stack.sh ops tooling (deploy/logs/health/backup/restore/monitoring);
.gitattributes LF for shell scripts.

Orchestration: json-file log rotation, pinned images, per-service memory
limits, edge/internal networks, Traefik ping/accesslog/metrics + flood
limiter, conditional minio-setup, pgvector db image, one-shot db-init
(umami + aramesh_chatbot databases).

Chatbot: unified from site-chatbot branch (RAG + pgvector, SSE, ChatWidget).

Monitoring: Prometheus/Alertmanager/Loki/Promtail/Grafana + cAdvisor,
node/postgres/blackbox exporters, provisioned dashboards and 10 alert
rules; alerts delivered to Bale (tapi.bale.ai — Telegram-compatible).

Analytics: Umami on shared Postgres, build-time tracker injection, custom
events (contact/newsletter/phone/chat/lang) via frontend/src/analytics.ts.

Hygiene: CSV export for contact messages and subscribers; upload image
optimization (Pillow); CI script + GitHub-compatible workflow; audit.sh;
gen-env.sh for strong credentials.
2026-09-19 00:24:57 +03:30

212 lines
8.2 KiB
Python
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
"""ذخیره‌ساز برداری pgvector — جایگزین FAISS در مهاجرت به pgvector.
بردارها در یک دیتابیس جدا (aramesh_chatbot) روی همان نمونه Postgres سایت
نگهداری می‌شوند تا دیتابیس سایت فقط‌خواندنی بماند. همگام‌سازی افزایشی است:
فقط سطرهای جدید/تغییریافته امبد می‌شوند (کلید: doc_id پایدار + hash محتوا).
"""
import hashlib
import json
import logging
from typing import Any
from urllib.parse import urlsplit, urlunsplit
from .config import settings
log = logging.getLogger(__name__)
def _strip_driver(dsn: str) -> str:
return dsn.replace("+psycopg", "").replace("+asyncpg", "")
def chatbot_dsn() -> str:
"""DSN دیتابیس اختصاصی چت‌بات — از DATABASE_URL با تعویض نام دیتابیس ساخته می‌شود."""
parts = urlsplit(_strip_driver(settings.database_url))
return urlunsplit(parts._replace(path=f"/{settings.chatbot_db_name}"))
def _admin_dsn() -> str:
"""DSN دیتابیس postgres برای ساخت دیتابیس چت‌بات (فقط یک‌بار)."""
parts = urlsplit(_strip_driver(settings.database_url))
return urlunsplit(parts._replace(path="/postgres"))
def ensure_schema() -> None:
"""ساخت دیتابیس و جدول بردارها — idempotent و در هر شروع اجرا می‌شود."""
import psycopg
from psycopg import errors as pg_errors
try:
with psycopg.connect(_admin_dsn(), autocommit=True, connect_timeout=10) as conn:
try:
conn.execute(f'CREATE DATABASE "{settings.chatbot_db_name}"')
log.info("Created chatbot database '%s'", settings.chatbot_db_name)
except pg_errors.DuplicateDatabase:
pass
except Exception as e:
raise RuntimeError(f"Cannot reach Postgres to ensure chatbot database: {e}") from e
with psycopg.connect(chatbot_dsn(), autocommit=True, connect_timeout=10) as conn:
conn.execute("CREATE EXTENSION IF NOT EXISTS vector")
conn.execute(
f"""
CREATE TABLE IF NOT EXISTS chatbot_chunks (
doc_id TEXT PRIMARY KEY,
content_hash TEXT NOT NULL,
model TEXT NOT NULL,
content JSONB NOT NULL,
embedding vector({settings.embedding_dim}) NOT NULL
)
"""
)
conn.execute(
"""
CREATE INDEX IF NOT EXISTS chatbot_chunks_embedding_hnsw
ON chatbot_chunks USING hnsw (embedding vector_cosine_ops)
"""
)
def sync_chunks(chunks: list[dict], force: bool = False) -> dict:
"""همگام‌سازی افزایشی قطعات با جدول pgvector.
- سطرهایی که دیگر وجود ندارند حذف می‌شوند
- سطرهای جدید یا تغییر یافته (hash یا مدل امبدینگ عوض شده) امبد و upsert می‌شوند
- force=True همه سطرها را از نو امبد می‌کند
خروجی: آمار عملیات برای لاگ.
"""
import psycopg
from psycopg.types.json import Jsonb
from .embeddings import get_embeddings
model = settings.embedding_model
wanted = {c["doc_id"]: c for c in chunks}
with psycopg.connect(chatbot_dsn(), autocommit=True, connect_timeout=10) as conn:
existing: dict[str, str] = {}
if not force:
for row in conn.execute("SELECT doc_id, content_hash, model FROM chatbot_chunks").fetchall():
if row["model"] == model:
existing[row["doc_id"]] = row["content_hash"]
else:
conn.execute("DELETE FROM chatbot_chunks")
to_delete = [doc_id for doc_id in existing if doc_id not in wanted]
if to_delete:
conn.execute(
"DELETE FROM chatbot_chunks WHERE doc_id = ANY(%s)",
[to_delete],
)
to_upsert = [
c for c in chunks
if existing.get(c["doc_id"]) != c["content_hash"]
]
inserted = updated = 0
if to_upsert:
vectors = get_embeddings().embed_documents([c["text"] for c in to_upsert])
for chunk, vec in zip(to_upsert, vectors):
literal = "[" + ",".join(f"{x:.7f}" for x in vec) + "]"
conn.execute(
"""
INSERT INTO chatbot_chunks (doc_id, content_hash, model, content, embedding)
VALUES (%s, %s, %s, %s, %s::vector)
ON CONFLICT (doc_id) DO UPDATE SET
content_hash = EXCLUDED.content_hash,
model = EXCLUDED.model,
content = EXCLUDED.content,
embedding = EXCLUDED.embedding
""",
(chunk["doc_id"], chunk["content_hash"], model, Jsonb(chunk), literal),
)
if chunk["doc_id"] in existing:
updated += 1
else:
inserted += 1
stats = {"inserted": inserted, "updated": updated, "deleted": len(to_delete),
"unchanged": len(wanted) - len(to_upsert)}
log.info("Vector sync: %s", stats)
return stats
def fetch_all_chunks() -> list[dict]:
"""همه قطعات ذخیره‌شده (برای ساخت BM25 در حافظه)."""
import psycopg
with psycopg.connect(chatbot_dsn(), autocommit=True, connect_timeout=10) as conn:
rows = conn.execute(
"SELECT doc_id, content_hash, content FROM chatbot_chunks ORDER BY doc_id"
).fetchall()
return [
{
"doc_id": r["doc_id"],
"content_hash": r["content_hash"],
**(r["content"] if isinstance(r["content"], dict) else json.loads(r["content"])),
}
for r in rows
]
def vector_count() -> int:
"""تعداد بردارهای ذخیره‌شده — برای healthz (خطا → ۰)."""
try:
import psycopg
with psycopg.connect(chatbot_dsn(), autocommit=True, connect_timeout=5) as conn:
return int(conn.execute("SELECT count(*) FROM chatbot_chunks").fetchone()[0])
except Exception:
return 0
def dense_search(query: str, top_k: int) -> list[tuple[dict, float]]:
"""جست‌وجوی معنایی SQL با فاصله کسینوسی — خروجی: (سند، شباهت).
خطا (مثلاً دیتابیس هنوز آماده نیست) → فهرست خالی تا بازیابی فقط با BM25 ادامه یابد.
"""
try:
import psycopg
from psycopg.rows import dict_row as _dict_row
from .embeddings import get_embeddings
vec = get_embeddings().embed_query(query)
literal = "[" + ",".join(f"{x:.7f}" for x in vec) + "]"
with psycopg.connect(chatbot_dsn(), row_factory=_dict_row, connect_timeout=10) as conn:
rows = conn.execute(
f"""
SELECT doc_id, content, 1 - (embedding <=> %s::vector) AS sim
FROM chatbot_chunks
ORDER BY embedding <=> %s::vector
LIMIT %s
""",
(literal, literal, top_k),
).fetchall()
return [(_row_to_doc(r), float(r["sim"])) for r in rows]
except Exception as e:
log.warning("Dense search failed (%s) — falling back to keyword only", e)
return []
def _row_to_doc(row: Any) -> dict:
content = row["content"] if isinstance(row["content"], dict) else json.loads(row["content"])
return {
"doc_id": row["doc_id"],
"text": content.get("text", ""),
"type": content.get("type", ""),
"title": content.get("title", ""),
"url": content.get("url", ""),
}
def chunk_content_hash(chunk: dict) -> str:
"""هش محتوای قطعه — تغییر متن یا متادیتا یعنی سطر باید به‌روز شود."""
payload = json.dumps(
{k: chunk[k] for k in ("text", "type", "title", "url")},
ensure_ascii=False, sort_keys=True,
)
return hashlib.sha256(payload.encode("utf-8")).hexdigest()[:16]