Skip to content

"Vibe-Trading: Your Personal Trading Agent"

fix(session): apply search limits to distinct conversations

Rudy Celeklifa016870b5ec
diff --git a/agent/src/session/search.py b/agent/src/session/search.py
index b1ca0e457..98bb74986 100644
--- a/agent/src/session/search.py
+++ b/agent/src/session/search.py
@@ -19,7 +19,6 @@ import time
from dataclasses import dataclass
from datetime import datetime
from pathlib import Path
-from typing import List, Optional
from src.config.paths import get_runtime_root
@@ -41,7 +40,11 @@ def _prepare_search_text(text: str) -> str:
def _search_snippet(content: str, query: str) -> str:
"""Highlight query terms in the original text, never in expanded tokens."""
terms = sorted(set(_QUERY_TERM.findall(query)), key=len, reverse=True)
- pattern = re.compile("|".join(re.escape(term) for term in terms), re.IGNORECASE) if terms else None
+ pattern = (
+ re.compile("|".join(re.escape(term) for term in terms), re.IGNORECASE)
+ if terms
+ else None
+ )
first = pattern.search(content) if pattern else None
start = max(0, first.start() - 100) if first else 0
end = min(len(content), start + 400)
@@ -99,7 +102,7 @@ class SessionSearchIndex:
"""
self.db_path = db_path
self.db_path.parent.mkdir(parents=True, exist_ok=True)
- self._conn: Optional[sqlite3.Connection] = None
+ self._conn: sqlite3.Connection | None = None
self._init_db()
def _get_conn(self) -> sqlite3.Connection:
@@ -135,7 +138,9 @@ class SessionSearchIndex:
""")
# Serialize upgrade across processes. Keep original messages for readable
# snippets; rebuild the derived index once, including historical rows.
- conn.create_function("prepare_session_text", 1, _prepare_search_text, deterministic=True)
+ conn.create_function(
+ "prepare_session_text", 1, _prepare_search_text, deterministic=True
+ )
with conn:
conn.execute("BEGIN IMMEDIATE")
columns = {row[1] for row in conn.execute("PRAGMA table_info(messages)")}
@@ -143,28 +148,38 @@ class SessionSearchIndex:
if migrating:
for name in ("messages_ai", "messages_ad", "messages_au"):
conn.execute(f"DROP TRIGGER IF EXISTS {name}")
- conn.execute("ALTER TABLE messages ADD COLUMN search_content TEXT NOT NULL DEFAULT ''")
- conn.execute("UPDATE messages SET search_content = prepare_session_text(content)")
+ conn.execute(
+ "ALTER TABLE messages ADD COLUMN search_content TEXT NOT NULL DEFAULT ''"
+ )
+ conn.execute(
+ "UPDATE messages SET search_content = prepare_session_text(content)"
+ )
conn.execute("DROP TABLE IF EXISTS messages_fts")
conn.execute("""
CREATE VIRTUAL TABLE IF NOT EXISTS messages_fts
USING fts5(search_content, content=messages, content_rowid=id)
""")
- conn.execute("""CREATE TRIGGER IF NOT EXISTS messages_ai AFTER INSERT ON messages BEGIN
+ conn.execute(
+ """CREATE TRIGGER IF NOT EXISTS messages_ai AFTER INSERT ON messages BEGIN
INSERT INTO messages_fts(rowid, search_content) VALUES (new.id, new.search_content);
- END""")
- conn.execute("""CREATE TRIGGER IF NOT EXISTS messages_ad AFTER DELETE ON messages BEGIN
+ END"""
+ )
+ conn.execute(
+ """CREATE TRIGGER IF NOT EXISTS messages_ad AFTER DELETE ON messages BEGIN
INSERT INTO messages_fts(messages_fts, rowid, search_content)
VALUES ('delete', old.id, old.search_content);
- END""")
+ END"""
+ )
if migrating:
- conn.execute("INSERT INTO messages_fts(messages_fts) VALUES ('rebuild')")
+ conn.execute(
+ "INSERT INTO messages_fts(messages_fts) VALUES ('rebuild')"
+ )
def index_session(
self,
session_id: str,
title: str = "",
- ts: Optional[float] = None,
+ ts: float | None = None,
) -> None:
"""Upsert a session record.
@@ -193,8 +208,9 @@ class SessionSearchIndex:
)
conn.commit()
- def index_message(self, session_id: str, role: str, content: str,
- tool_name: Optional[str] = None) -> None:
+ def index_message(
+ self, session_id: str, role: str, content: str, tool_name: str | None = None
+ ) -> None:
"""Index a single message.
Args:
@@ -234,10 +250,12 @@ class SessionSearchIndex:
Returns:
FTS5-safe MATCH expression.
"""
- tokens = [_prepare_search_text(term).strip() for term in _QUERY_TERM.findall(query)]
+ tokens = [
+ _prepare_search_text(term).strip() for term in _QUERY_TERM.findall(query)
+ ]
return " OR ".join(f'"{token}"' for token in tokens) if tokens else '""'
- def search(self, query: str, max_sessions: int = 3) -> List[SearchMatch]:
+ def search(self, query: str, max_sessions: int = 3) -> list[SearchMatch]:
"""Full-text search across all sessions.
Args:
@@ -264,16 +282,15 @@ class SessionSearchIndex:
JOIN sessions s ON s.id = m.session_id
WHERE messages_fts MATCH ?
ORDER BY rank
- LIMIT ?
""",
- (fts_query, max_sessions * 5),
+ (fts_query,),
)
except sqlite3.OperationalError as exc:
logger.warning("FTS5 search failed: %s", exc)
return []
seen: dict[str, SearchMatch] = {}
- for row in cursor.fetchall():
+ for row in cursor:
sid = row[0]
if sid in seen:
continue
@@ -328,14 +345,18 @@ class SessionSearchIndex:
title = session_data.get("title", "")
try:
- ts = datetime.fromisoformat(session_data.get("created_at", "")).timestamp()
+ ts = datetime.fromisoformat(
+ session_data.get("created_at", "")
+ ).timestamp()
except (ValueError, TypeError):
ts = time.time()
self.index_session(sid, title, ts=ts)
if messages_file.exists():
- for line in messages_file.read_text(encoding="utf-8").strip().splitlines():
+ for line in (
+ messages_file.read_text(encoding="utf-8").strip().splitlines()
+ ):
if not line.strip():
continue
try:
@@ -347,7 +368,8 @@ class SessionSearchIndex:
count += 1
except json.JSONDecodeError:
continue
- except Exception as exc:
+ # One corrupt session must not abort the remaining reindex.
+ except Exception as exc: # noqa: BLE001
logger.warning("Failed to index session %s: %s", session_dir.name, exc)
return count
@@ -367,7 +389,7 @@ class SessionSearchIndex:
self._conn = None
-_shared_index: Optional[SessionSearchIndex] = None
+_shared_index: SessionSearchIndex | None = None
_shared_lock = _threading.Lock()
diff --git a/agent/tests/test_session_search_distinct_limit.py b/agent/tests/test_session_search_distinct_limit.py
new file mode 100644
index 000000000..ea0a2c643
--- /dev/null
+++ b/agent/tests/test_session_search_distinct_limit.py
@@ -0,0 +1,75 @@
+"""Session search limits sessions rather than an approximate message sample."""
+
+import json
+from pathlib import Path
+
+from src.session.search import SessionSearchIndex
+from src.tools.session_search_tool import SessionSearchTool
+
+
+def _populate(index: SessionSearchIndex) -> None:
+ for session_id, count in [("long", 25), ("second", 1), ("third", 1)]:
+ index.index_session(session_id, session_id, ts=1_700_000_000)
+ for _ in range(count):
+ index.index_message(session_id, "user", "portfolio volatility")
+
+
+def test_actual_tool_returns_distinct_sessions_after_long_conversation(
+ tmp_path: Path, monkeypatch
+) -> None:
+ index = SessionSearchIndex(tmp_path / "sessions.db")
+ try:
+ _populate(index)
+ monkeypatch.setattr("src.session.search.get_shared_index", lambda: index)
+ result = json.loads(
+ SessionSearchTool().execute(query="volatility", max_results=3)
+ )
+ assert result["status"] == "ok"
+ assert {row["session_id"] for row in result["results"]} == {
+ "long",
+ "second",
+ "third",
+ }
+ assert len(result["results"]) == 3
+ assert all(">>>volatility<<<" in row["snippet"] for row in result["results"])
+ assert len(index.search("volatility", max_sessions=1)) == 1
+ finally:
+ index.close()
+
+
+def test_ranked_cursor_stops_after_requested_distinct_sessions(tmp_path: Path) -> None:
+ index = SessionSearchIndex(tmp_path / "sessions.db")
+ try:
+ for i in range(30):
+ index.index_session(str(i), str(i))
+ index.index_message(str(i), "user", "volatility")
+ real = index._get_conn()
+ consumed = []
+
+ class Cursor:
+ def __iter__(self):
+ for row in self.cursor:
+ consumed.append(row)
+ yield row
+
+ def fetchall(self):
+ raise AssertionError("Search must not materialize every message")
+
+ class Connection:
+ def execute(self, sql, params):
+ wrapper = Cursor()
+ wrapper.cursor = real.execute(sql, params)
+ return wrapper
+
+ index._conn = Connection()
+ matches = index.search("volatility", max_sessions=3)
+ index._conn = real
+ assert len(matches) == 3
+ assert len(consumed) == 3
+ assert [match.rank for match in matches] == sorted(
+ match.rank for match in matches
+ )
+ finally:
+ if index._conn is not real:
+ index._conn = real
+ index.close()