Files
FabledScribe/src/fabledassistant/services/knowledge.py
T

359 lines
13 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.
"""Knowledge service — unified query across notes, people, places, and lists."""
import logging
from sqlalchemy import func, select
from fabledassistant.models import async_session
from fabledassistant.models.note import Note
logger = logging.getLogger(__name__)
_SNIPPET_LEN = 200
def _note_to_item(note: Note) -> dict:
meta = note.entity_meta or {}
item: dict = {
"id": note.id,
"note_type": note.entity_type,
"title": note.title,
"snippet": (note.body or "")[:_SNIPPET_LEN],
"tags": note.tags or [],
"project_id": note.project_id,
"metadata": meta,
"created_at": note.created_at.isoformat(),
"updated_at": note.updated_at.isoformat(),
}
# Type-specific convenience fields
if note.entity_type == "person":
item["relationship"] = meta.get("relationship", "")
item["email"] = meta.get("email", "")
item["phone"] = meta.get("phone", "")
item["birthday"] = meta.get("birthday", "")
item["organization"] = meta.get("organization", "")
item["address"] = meta.get("address", "")
elif note.entity_type == "place":
item["address"] = meta.get("address", "")
item["phone"] = meta.get("phone", "")
item["hours"] = meta.get("hours", "")
item["website"] = meta.get("website", "")
item["category"] = meta.get("category", "")
elif note.entity_type == "list":
# Parse markdown task list syntax into structured items
body = note.body or ""
list_items = []
for line in body.split("\n"):
stripped = line.strip()
if stripped.startswith("- [ ] ") or stripped.startswith("- [x] ") or stripped.startswith("- [X] "):
checked_item = not stripped.startswith("- [ ] ")
list_items.append({"text": stripped[6:], "checked": checked_item})
item["list_items"] = list_items
item["item_count"] = len(list_items)
item["checked_count"] = sum(1 for i in list_items if i["checked"])
item["body"] = body
# Task fields — override note_type and add status/priority/due_date
if note.is_task:
item["note_type"] = "task"
item["task_kind"] = note.task_kind
item["status"] = note.status
item["priority"] = note.priority
item["due_date"] = note.due_date.isoformat() if note.due_date else None
return item
def _apply_type_filter(stmt, note_type: str | None):
"""Apply the type facet to a Note select.
'task' = any task (status not null); 'plan' = a task with task_kind='plan';
any other non-empty type = a non-task note of that note_type; None = all.
"""
if note_type == "task":
return stmt.where(Note.status.isnot(None))
if note_type == "plan":
return stmt.where(Note.status.isnot(None)).where(Note.task_kind == "plan")
if note_type:
return stmt.where(Note.note_type == note_type).where(Note.status.is_(None))
return stmt
async def query_knowledge(
user_id: int,
note_type: str | None,
tags: list[str],
sort: str,
q: str | None,
limit: int,
offset: int,
) -> tuple[list[dict], int]:
"""Query knowledge objects (non-task notes) with filters.
Returns (items, total_count).
"""
# Semantic search path — scores take priority over sort
if q:
return await _semantic_knowledge_search(
user_id, q, note_type=note_type, tags=tags, limit=limit, offset=offset
)
async with async_session() as session:
base = select(Note).where(Note.user_id == user_id)
base = _apply_type_filter(base, note_type)
for tag in tags:
base = base.where(Note.tags.contains([tag]))
# Count before pagination
count_stmt = select(func.count()).select_from(base.subquery())
total: int = (await session.execute(count_stmt)).scalar_one()
# Apply sort
if sort == "created":
base = base.order_by(Note.created_at.desc())
elif sort == "alpha":
base = base.order_by(Note.title.asc())
elif sort == "type":
base = base.order_by(Note.note_type.asc(), Note.updated_at.desc())
else: # modified (default)
base = base.order_by(Note.updated_at.desc())
rows = list((await session.execute(base.limit(limit).offset(offset))).scalars().all())
return [_note_to_item(n) for n in rows], total
async def _semantic_knowledge_search(
user_id: int,
q: str,
note_type: str | None,
tags: list[str],
limit: int,
offset: int,
) -> tuple[list[dict], int]:
"""Hybrid search: keyword matches first (title/body ILIKE), then semantic results.
Exact keyword matches always rank above semantic-only matches so that
searching for a name like "Weston" surfaces the note with that title
before conceptually related notes.
"""
# 1. Keyword search — title and body ILIKE
keyword_notes: list[Note] = []
try:
async with async_session() as session:
pattern = f"%{q}%"
base = (
select(Note)
.where(Note.user_id == user_id)
.where(Note.title.ilike(pattern) | Note.body.ilike(pattern))
)
base = _apply_type_filter(base, note_type)
for tag in tags:
base = base.where(Note.tags.contains([tag]))
# Title matches first, then body-only matches, newest first within each
base = base.order_by(
Note.title.ilike(pattern).desc(),
Note.updated_at.desc(),
).limit(limit * 2)
keyword_notes = list((await session.execute(base)).scalars().all())
except Exception:
logger.warning("Keyword search failed", exc_info=True)
# 2. Semantic search — conceptual similarity
semantic_notes: list[Note] = []
try:
from fabledassistant.services.embeddings import semantic_search_notes
is_task_filter = True if note_type in ("task", "plan") else (False if note_type else None)
candidates = await semantic_search_notes(
user_id=user_id,
query=q,
limit=min(200, limit * 4),
threshold=0.3,
is_task=is_task_filter,
)
for _score, note in candidates:
if note_type == "task" and not note.is_task:
continue
elif note_type == "plan" and (not note.is_task or note.task_kind != "plan"):
continue
elif note_type and note_type not in ("task", "plan") and note.entity_type != note_type:
continue
if tags and not all(t in (note.tags or []) for t in tags):
continue
semantic_notes.append(note)
except Exception:
logger.warning("Semantic search unavailable, using keyword results only", exc_info=True)
# 3. Merge — keyword matches first, then semantic (deduplicated)
seen_ids: set[int] = set()
merged: list[Note] = []
for note in keyword_notes:
if note.id not in seen_ids:
seen_ids.add(note.id)
merged.append(note)
for note in semantic_notes:
if note.id not in seen_ids:
seen_ids.add(note.id)
merged.append(note)
total = len(merged)
page_items = merged[offset: offset + limit]
return [_note_to_item(n) for n in page_items], total
async def get_knowledge_tags(user_id: int, note_type: str | None = None) -> list[str]:
"""Return all distinct tags used across knowledge objects for this user."""
async with async_session() as session:
base = (
select(func.unnest(Note.tags).label("tag"))
.where(Note.user_id == user_id)
)
base = _apply_type_filter(base, note_type)
stmt = base.distinct().order_by("tag")
rows = list((await session.execute(stmt)).scalars().all())
return [r for r in rows if r]
async def get_knowledge_counts(user_id: int, tags: list[str] | None = None) -> dict[str, int]:
"""Return per-type count of knowledge objects for the sidebar display."""
async with async_session() as session:
# Count non-task types
stmt = (
select(Note.note_type, func.count(Note.id))
.where(Note.user_id == user_id)
.where(Note.status.is_(None))
.where(Note.note_type.in_(["note", "person", "place", "list"]))
.group_by(Note.note_type)
)
if tags:
for tag in tags:
stmt = stmt.where(Note.tags.contains([tag]))
rows = list((await session.execute(stmt)).all())
counts = {row[0]: row[1] for row in rows}
# Count tasks separately (is_task = status IS NOT NULL)
task_stmt = (
select(func.count(Note.id))
.where(Note.user_id == user_id)
.where(Note.status.isnot(None))
)
if tags:
for tag in tags:
task_stmt = task_stmt.where(Note.tags.contains([tag]))
task_count: int = (await session.execute(task_stmt)).scalar_one()
counts["task"] = task_count
# Plans are a subset of tasks (task_kind='plan'); counted for the facet
# but NOT added to total to avoid double-counting against "task".
plan_stmt = (
select(func.count(Note.id))
.where(Note.user_id == user_id)
.where(Note.status.isnot(None))
.where(Note.task_kind == "plan")
)
if tags:
for tag in tags:
plan_stmt = plan_stmt.where(Note.tags.contains([tag]))
counts["plan"] = (await session.execute(plan_stmt)).scalar_one()
for t in ("note", "person", "place", "list", "task", "plan"):
counts.setdefault(t, 0)
counts["total"] = sum(counts[t] for t in ("note", "person", "place", "list", "task"))
return counts
async def query_knowledge_ids(
user_id: int,
note_type: str | None,
tags: list[str],
sort: str,
q: str | None,
limit: int = 100,
offset: int = 0,
) -> tuple[list[int], int]:
"""Return note IDs only — cheap query for the two-tier pagination feed."""
if q:
# Re-use semantic search, extract IDs in rank order
items, total = await _semantic_knowledge_search(
user_id, q, note_type=note_type, tags=tags,
limit=limit, offset=offset,
)
return [item["id"] for item in items], total
async with async_session() as session:
base = select(Note.id).where(Note.user_id == user_id)
base = _apply_type_filter(base, note_type)
for tag in tags:
base = base.where(Note.tags.contains([tag]))
count_stmt = select(func.count()).select_from(base.subquery())
total: int = (await session.execute(count_stmt)).scalar_one()
if sort == "created":
base = base.order_by(Note.created_at.desc())
elif sort == "alpha":
base = base.order_by(Note.title.asc())
elif sort == "type":
base = base.order_by(Note.note_type.asc(), Note.updated_at.desc())
else:
base = base.order_by(Note.updated_at.desc())
ids = list((await session.execute(base.limit(limit).offset(offset))).scalars().all())
return ids, total
async def get_knowledge_by_ids(user_id: int, ids: list[int]) -> list[dict]:
"""Fetch full items for the given IDs, preserving the requested order."""
if not ids:
return []
async with async_session() as session:
stmt = (
select(Note)
.where(Note.user_id == user_id)
.where(Note.id.in_(ids))
)
rows = list((await session.execute(stmt)).scalars().all())
by_id = {n.id: n for n in rows}
return [_note_to_item(by_id[i]) for i in ids if i in by_id]
async def get_people_and_places_context(user_id: int) -> str:
"""Return a compact summary of known people and places for LLM system prompt injection."""
async with async_session() as session:
stmt = (
select(Note)
.where(Note.user_id == user_id)
.where(Note.note_type.in_(["person", "place"]))
.where(Note.status.is_(None))
.order_by(Note.title.asc())
.limit(50)
)
rows = list((await session.execute(stmt)).scalars().all())
if not rows:
return ""
people = [n for n in rows if n.entity_type == "person"]
places = [n for n in rows if n.entity_type == "place"]
lines = []
if people:
parts = []
for p in people:
meta = p.entity_meta or {}
rel = meta.get("relationship", "")
parts.append(f"{p.title}" + (f" ({rel})" if rel else ""))
lines.append("Known people: " + ", ".join(parts))
if places:
parts = []
for p in places:
meta = p.entity_meta or {}
addr = meta.get("address", "")
parts.append(f"{p.title}" + (f" {addr}" if addr else ""))
lines.append("Known places: " + "; ".join(parts))
return "\n".join(lines)