oclaw/runtime/tools/experts/productivity/kb_tools.py
oliver d14e9d3596 feat(persistence): PostgreSQL assistant store, chat persist fixes, gateway scripts
- Add SQLAlchemy Core repos, pg adapter/compat, assistant_store factory, Alembic bootstrap and migration/cutover scripts.

- Harden chat_message writes (NUL scrub for PG), turn_uuid on attempt failure, WS turn_runner fallbacks and gateway executed_turn_uuid init.

- start_gateway: log paths, PS7 stderr handling via cmd, background stdout/stderr redirect; runtime assistant_runtime_log_dir export.

- Ops: clear_all_chat_sessions with PG-only --postgresql and env-gated wipe; clear_postgres_chat_sessions.ps1.

- Tests: SA repos, pg compat, persist fallback, smoke env isolation; CI and docs touch-ups.

Co-authored-by: Cursor <cursoragent@cursor.com>
2026-05-14 09:33:50 +08:00

97 lines
4.2 KiB
Python

from __future__ import annotations
import hashlib
from typing import Any
from svc.config.paths import db_path
from svc.embeddings.embedding_client import build_default_embedding_client
from svc.persistence.sqlite_store import SqliteStore
from svc.persistence.assistant_store import get_assistant_store
from runtime.tools.base import ToolSpec
def _chunk_id(source: str, text: str) -> str:
raw = f"{source}\n{text}".encode("utf-8", errors="ignore")
return hashlib.sha1(raw).hexdigest()
def kb_add_tool() -> ToolSpec:
def handler(args: dict[str, Any]) -> dict[str, Any]:
try:
tenant_id = str(args.get("tenant_id") or "").strip()
user_id = str(args.get("user_id") or "").strip()
text = str(args.get("text") or "").strip()
title = str(args.get("title") or "").strip()
if not tenant_id or not user_id or not text:
return {"ok": False, "error": "tenant_id, user_id, text are required"}
source = f"builtin:tenant:{tenant_id}:kb"
if title:
source = f"{source}:{title[:48]}"
cid = _chunk_id(source, text)
store = get_assistant_store()
store.upsert_knowledge_chunk(
chunk_id=cid,
source=source,
content=text,
metadata={"tenant_id": tenant_id, "user_id": user_id, "title": title, "source": source},
)
client = build_default_embedding_client()
emb = client.embed(text[:8000])
store.upsert_knowledge_embedding(chunk_id=cid, model=emb.model, vector=emb.vector)
return {"ok": True, "chunk_id": cid, "source": source, "embedding_model": emb.model}
except Exception as e:
return {"ok": False, "error": f"{type(e).__name__}: {e}"}
return ToolSpec(
name="kb_add",
description="Add a knowledge snippet for a tenant into the vector knowledge base.",
parameters={
"type": "object",
"properties": {"tenant_id": {"type": "string"}, "user_id": {"type": "string"}, "title": {"type": "string", "description": "Optional title/label."}, "text": {"type": "string", "description": "Knowledge content to store."}},
"required": ["tenant_id", "user_id", "text"],
"additionalProperties": False,
},
handler=handler,
tags=frozenset({"productivity", "rag", "write"}),
)
def kb_search_tool() -> ToolSpec:
def handler(args: dict[str, Any]) -> dict[str, Any]:
try:
tenant_id = str(args.get("tenant_id") or "").strip()
query = str(args.get("query") or "").strip()
limit = int(args.get("limit") or 3)
if not tenant_id or not query:
return {"ok": False, "error": "tenant_id and query are required"}
store = get_assistant_store()
from runtime.orchestration.memory import retrieve_context
rows = retrieve_context(store, query, limit=max(1, min(limit, 6)))
filtered = [r for r in rows if str(r.get("source") or "").startswith(f"builtin:tenant:{tenant_id}:")]
hits = filtered[: max(1, min(limit, 6))]
if not hits:
like_rows = store.search_knowledge(query=query, limit=max(1, min(limit, 6)))
hits = [r for r in like_rows if str(r.get("source") or "").startswith(f"builtin:tenant:{tenant_id}:")][: max(1, min(limit, 6))]
refs = []
for h in hits:
refs.append({"source": str(h.get("source") or ""), "snippet": str(h.get("content") or "")[:240]})
return {"ok": True, "hits": refs}
except Exception as e:
return {"ok": False, "error": f"{type(e).__name__}: {e}"}
return ToolSpec(
name="kb_search",
description="Search tenant knowledge base and return citations/snippets.",
parameters={
"type": "object",
"properties": {"tenant_id": {"type": "string"}, "query": {"type": "string"}, "limit": {"type": "integer", "default": 3}},
"required": ["tenant_id", "query"],
"additionalProperties": False,
},
handler=handler,
tags=frozenset({"productivity", "rag"}),
)
__all__ = ["kb_add_tool", "kb_search_tool"]