File size: 6,695 Bytes
65de2c3 4d95d6b 65de2c3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 | import chromadb
from chromadb.utils.embedding_functions import SentenceTransformerEmbeddingFunction
import hashlib
import json
import os
import shutil
from errors import get_logger, GenerAIError, ErrorCode, fmt_exc
log = get_logger("knowledge_base")
DB_PATH = "./database"
COUNTS_FILE = f"{DB_PATH}/query_counts.json"
SEEDED_FLAG = f"{DB_PATH}/.seeded_v2"
EMBED_MODEL = "paraphrase-multilingual-MiniLM-L12-v2"
def _load_counts() -> dict:
if os.path.exists(COUNTS_FILE):
try:
with open(COUNTS_FILE) as f:
data = json.load(f)
log.debug("Caricati %d contatori da %s", len(data), COUNTS_FILE)
return data
except Exception as e:
log.warning("[%s] Impossibile leggere %s — %s. Uso contatori vuoti.", ErrorCode.KB_INIT_FAILED.value, COUNTS_FILE, fmt_exc(e))
return {}
return {}
def _save_counts(counts: dict):
os.makedirs(DB_PATH, exist_ok=True)
try:
with open(COUNTS_FILE, "w") as f:
json.dump(counts, f)
except Exception as e:
log.error("[%s] Impossibile salvare contatori: %s", ErrorCode.KB_WRITE_FAILED.value, fmt_exc(e))
def _normalize(text: str) -> str:
return " ".join(text.lower().strip().split()[:12])
class KnowledgeBase:
ARCHIVE_THRESHOLD = 3
def __init__(self):
log.info("Inizializzazione KnowledgeBase in: %s", os.path.abspath(DB_PATH))
try:
os.makedirs(DB_PATH, exist_ok=True)
except Exception as e:
raise GenerAIError(ErrorCode.KB_INIT_FAILED, f"Impossibile creare la directory DB: {fmt_exc(e)}", cause=e)
# Wipe old schema if needed
if not os.path.exists(SEEDED_FLAG):
old_flag = f"{DB_PATH}/.seeded"
if os.path.exists(old_flag):
os.remove(old_flag)
log.info("Rimosso flag vecchio schema.")
chroma_dir = os.path.join(DB_PATH, "chroma.sqlite3")
if os.path.exists(chroma_dir):
os.remove(chroma_dir)
log.info("Rimosso database ChromaDB obsoleto.")
try:
log.debug("Caricamento modello embedding: %s", EMBED_MODEL)
ef = SentenceTransformerEmbeddingFunction(model_name=EMBED_MODEL)
self._client = chromadb.PersistentClient(path=DB_PATH)
self._col = self._client.get_or_create_collection(
name="generai",
embedding_function=ef,
metadata={"hnsw:space": "cosine"},
)
log.info("ChromaDB pronto. Documenti in memoria: %d", self._col.count())
except Exception as e:
raise GenerAIError(ErrorCode.KB_INIT_FAILED, f"ChromaDB non avviato: {fmt_exc(e)}", cause=e)
self._counts = _load_counts()
self._seed_if_needed()
def _seed_if_needed(self):
if os.path.exists(SEEDED_FLAG):
return
open(SEEDED_FLAG, "w").close()
log.info("KB vuota — seeding disabilitato, si parte da zero.")
# ── Public API ─────────────────────────────────────────────────────────────
def search(self, query: str, n_results: int = 3) -> list[dict]:
count = self._col.count()
if count == 0:
log.debug("KB vuota, nessuna ricerca eseguita.")
return []
try:
res = self._col.query(
query_texts=[query],
n_results=min(n_results, count),
include=["metadatas", "distances"],
)
metas = res["metadatas"][0]
dists = res["distances"][0]
log.debug("KB search per %r → %d risultati (best dist=%.3f)", query, len(metas), dists[0] if dists else -1)
return [
{"answer": metas[i].get("answer", ""), "metadata": metas[i], "distance": dists[i]}
for i in range(len(metas))
]
except Exception as e:
err = GenerAIError(ErrorCode.KB_SEARCH_FAILED, f"Ricerca KB fallita: {fmt_exc(e)}", cause=e)
err.log(log)
return []
def increment_and_should_archive(self, query: str) -> bool:
key = _normalize(query)
self._counts[key] = self._counts.get(key, 0) + 1
_save_counts(self._counts)
count = self._counts[key]
log.debug("Query %r vista %d volte (threshold=%d)", key, count, self.ARCHIVE_THRESHOLD)
return count >= self.ARCHIVE_THRESHOLD
def get_query_count(self, query: str) -> int:
return self._counts.get(_normalize(query), 0)
def add(self, query: str, content: str, source_url: str = "") -> str:
doc_id = hashlib.md5(_normalize(query).encode()).hexdigest()
try:
self._col.upsert(
documents=[query],
ids=[doc_id],
metadatas=[{
"query": query[:300],
"answer": content[:4000],
"source": source_url[:500],
"feedback_score": 0,
}],
)
log.info("Documento archiviato (id=%s) per query: %r", doc_id[:8], query[:60])
except Exception as e:
err = GenerAIError(ErrorCode.KB_WRITE_FAILED, f"Impossibile salvare in KB: {fmt_exc(e)}", cause=e)
err.log(log)
return doc_id
def reinforce(self, doc_id: str, positive: bool):
action = "rinforzo positivo" if positive else "eliminazione"
log.info("Feedback (%s) per doc_id=%s", action, doc_id[:8])
try:
result = self._col.get(ids=[doc_id], include=["metadatas"])
if not result["ids"]:
log.warning("[%s] doc_id=%s non trovato in KB.", ErrorCode.KB_REINFORCE_FAILED.value, doc_id[:8])
return
if positive:
meta = result["metadatas"][0]
meta["feedback_score"] = meta.get("feedback_score", 0) + 1
self._col.update(ids=[doc_id], metadatas=[meta])
log.debug("feedback_score aggiornato a %d", meta["feedback_score"])
else:
self._col.delete(ids=[doc_id])
key = _normalize(result["metadatas"][0].get("query", ""))
self._counts.pop(key, None)
_save_counts(self._counts)
log.info("Documento eliminato dalla KB.")
except Exception as e:
err = GenerAIError(ErrorCode.KB_REINFORCE_FAILED, f"Feedback fallito: {fmt_exc(e)}", cause=e)
err.log(log)
def count(self) -> int:
return self._col.count()
|