pyr0ball
89a58ec9b0
Bi-encoder embeddings collapse a whole query into one vector, losing multi-part reasoning structure — queries like "the procedure for setting an IP on an AVC-X" or "what is the action economy for a fighter casting a spell while prone" lose nuance. Agent-ModernColBERT is a late-interaction retriever: per-token embeddings, scored via MaxSim at query time, built specifically for agentic/multi-hop queries. Implements Option A from the issue (in-process, via `pylate`) rather than Option B (managed cf-orch service) — cf-orch already has `agent-moderncolbert` registered in model_registry.yaml with a `pagepiper/retrieve` assignment in assignments.yaml pointing at it and referencing this issue directly, someone had already pre-wired that side. - app/services/colbert_index.py: new ColBERTIndex class, mirrors BM25Index's dirty-flag/rebuild-from-SQLite pattern exactly — no separate per-shelve indexing step needed, just mark_dirty() on the same callback that already marks BM25 dirty. - app/services/retriever.py: hybrid_search's semantic half now merges BM25 with ColBERT MaxSim scores (min-max normalized per-batch, since MaxSim is unbounded unlike the old sqlite-vec L2-distance path) instead of Ollama-embed + sqlite-vec cosine. BM25 merge/rank/per-doc-cap/ adjacent-chunk-window logic is unchanged. - app/main.py / app/deps.py: per-user ColBERTIndex registry, same pattern as the existing per-user BM25Index registry. - Existing BYOK tier gate preserved exactly (llm is None check) — this is a retrieval-technology swap, not a tier/licensing change. The ColBERT model runs locally via pylate with no Ollama dependency, but gating still follows product tiering. - 12 new tests. pylate is intentionally NOT installed in the dev/test env — see the cf-sysadmin skill's "Known Gotchas" for why (installing it directly into the shared `cf` conda env broke several other services' torch/transformers pins on 2026-07-10). Tests inject fake pylate modules via sys.modules instead. Known follow-up (not addressed here): shelve scripts still compute and store Ollama embeddings into `page_vecs` at shelve time — that table is no longer read by search/chat now that retrieval uses the ColBERT index. Removing the now-redundant embedding step is separate cleanup. Closes: #8
84 lines
2.5 KiB
Python
84 lines
2.5 KiB
Python
# app/deps.py
|
|
"""FastAPI dependency providers."""
|
|
from __future__ import annotations
|
|
|
|
import sqlite3
|
|
from dataclasses import dataclass
|
|
from pathlib import Path
|
|
from typing import Generator
|
|
|
|
from fastapi import Depends, Request
|
|
|
|
from app.config import DATA_DIR, LOCAL_USER_ID
|
|
from app.services.bm25_index import BM25Index
|
|
from app.services.colbert_index import ColBERTIndex
|
|
|
|
|
|
@dataclass
|
|
class UserCtx:
|
|
"""Per-request context routing DB paths, BM25, and ColBERT to the right user."""
|
|
|
|
user_id: str
|
|
db_path: str
|
|
vec_db_path: str
|
|
data_dir: Path
|
|
watch_dir: Path
|
|
bm25: BM25Index
|
|
colbert: ColBERTIndex
|
|
|
|
|
|
_user_startup_done: set[str] = set()
|
|
|
|
|
|
def _run_user_startup(user_id: str, user_dir: Path) -> None:
|
|
"""Run migrations and vec schema check once per process lifetime per user."""
|
|
if user_id in _user_startup_done:
|
|
return
|
|
_user_startup_done.add(user_id)
|
|
from app.config import VEC_DIMENSIONS
|
|
from app.startup import apply_migrations, check_and_rebuild_vec_schema
|
|
apply_migrations(str(user_dir / "pagepiper.db"))
|
|
check_and_rebuild_vec_schema(
|
|
str(user_dir / "pagepiper_vecs.db"), VEC_DIMENSIONS, str(user_dir / "pagepiper.db")
|
|
)
|
|
|
|
|
|
def get_user_ctx(request: Request) -> UserCtx:
|
|
"""Resolve the per-user data directory, DB paths, and BM25 instance for this request."""
|
|
import app.main as _main
|
|
from app.cloud_session import CLOUD_MODE
|
|
|
|
if CLOUD_MODE:
|
|
from app.cloud_session import resolve_authenticated_user
|
|
from app.config import user_data_dir
|
|
user_id = resolve_authenticated_user(request)
|
|
user_dir = user_data_dir(user_id)
|
|
_run_user_startup(user_id, user_dir)
|
|
watch_dir = user_dir / "books"
|
|
watch_dir.mkdir(parents=True, exist_ok=True)
|
|
else:
|
|
from app.config import WATCH_DIR
|
|
user_id = LOCAL_USER_ID
|
|
user_dir = DATA_DIR
|
|
watch_dir = WATCH_DIR
|
|
|
|
return UserCtx(
|
|
user_id=user_id,
|
|
db_path=str(user_dir / "pagepiper.db"),
|
|
vec_db_path=str(user_dir / "pagepiper_vecs.db"),
|
|
data_dir=user_dir,
|
|
watch_dir=watch_dir,
|
|
bm25=_main._get_bm25_for(user_id),
|
|
colbert=_main._get_colbert_for(user_id, user_dir),
|
|
)
|
|
|
|
|
|
def get_db(ctx: UserCtx = Depends(get_user_ctx)) -> Generator[sqlite3.Connection, None, None]:
|
|
conn = sqlite3.connect(ctx.db_path, check_same_thread=False)
|
|
conn.execute("PRAGMA foreign_keys = ON")
|
|
conn.execute("PRAGMA journal_mode = WAL")
|
|
conn.row_factory = sqlite3.Row
|
|
try:
|
|
yield conn
|
|
finally:
|
|
conn.close()
|