dohertj2/chat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

merge: T58 scene compression + thread emission on close

Browse Source
This commit is contained in:
Joseph Doherty
2026-04-26 20:21:01 -04:00
parent 8aeadfd0e4 343f305587
commit 8efbcdf6c3
2 changed files with 357 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test_per_pov_summary.py
+255 -5
View File
@@ -504,13 +504,15 @@ async def test_close_with_no_guest_matches_phase1(tmp_path):
"relationship_summary": "BotA leaned in supportively.",
}
)
no_threads = json.dumps({"candidates": []})
with open_db(db) as conn:
_seed_single_bot_scene(conn)
project(conn)
# canned has 2 entries to detect any over-call; the assertion below
# confirms only one was consumed.
client = MockLLMClient(canned=[canned, canned])
# 1 host-POV entry + 1 thread-detection entry (T58.2) + 1 spare
# to detect any over-call. Assertion below confirms exactly two
# were consumed.
client = MockLLMClient(canned=[canned, no_threads, canned])
await apply_scene_close_summary(
conn,
client,
@@ -520,8 +522,8 @@ async def test_close_with_no_guest_matches_phase1(tmp_path):
host_bot_id="bot_a",
)
# Exactly one classifier call -> exactly one canned entry consumed,
# leaving the second untouched.
# Host POV + thread detection -> exactly two canned entries
# consumed, leaving the spare untouched.
assert len(client._canned) == 1
# Host memory rewritten with the per-POV summary content.
@@ -845,3 +847,251 @@ async def test_group_summary_skipped_when_no_guest(tmp_path):
"SELECT 1 FROM event_log WHERE kind = 'group_node_updated'"
).fetchall()
assert rows == []
# ---------------------------------------------------------------------------
# T58: significance-driven quote retention + thread detection on close.
# ---------------------------------------------------------------------------
def _seed_single_bot_scene_no_memory(conn) -> None:
"""Like ``_seed_single_bot_scene`` but skips the memory_written event so
callers can seed memories with custom significance / text themselves."""
append_event(conn, kind="bot_authored", payload=_bot_payload("bot_a", "BotA"))
append_event(
conn,
kind="you_authored",
payload={"name": "Me", "pronouns": "they/them", "persona": "engineer"},
)
append_event(
conn,
kind="chat_created",
payload={
"id": "chat_bot_a",
"host_bot_id": "bot_a",
"initial_time": "2026-04-26T20:00:00+00:00",
"narrative_anchor": "Day 1",
"weather": "",
},
)
append_event(
conn,
kind="container_created",
payload={
"chat_id": "chat_bot_a",
"name": "office",
"type": "workplace",
"properties": {},
},
)
append_event(
conn,
kind="scene_opened",
payload={
"chat_id": "chat_bot_a",
"container_id": 1,
"started_at": "2026-04-26T20:00:00+00:00",
"participants": ["you", "bot_a"],
},
)
append_event(
conn,
kind="edge_update",
payload={
"source_id": "bot_a",
"target_id": "you",
"chat_id": "chat_bot_a",
},
)
append_event(
conn,
kind="user_turn",
payload={
"chat_id": "chat_bot_a",
"prose": "Quick chat about the deadline",
"segments": [],
},
)
append_event(
conn,
kind="assistant_turn",
payload={
"chat_id": "chat_bot_a",
"speaker_id": "bot_a",
"text": "It's going to be okay.",
"truncated": False,
"user_turn_id": 1,
},
)
def _seed_memory(conn, *, pov_summary: str, significance: int) -> None:
append_event(
conn,
kind="memory_written",
payload={
"owner_id": "bot_a",
"chat_id": "chat_bot_a",
"scene_id": 1,
"pov_summary": pov_summary,
"witness_you": 1,
"witness_host": 1,
"witness_guest": 0,
"significance": significance,
},
)
@pytest.mark.asyncio
async def test_low_significance_scene_omits_quotes(tmp_path):
"""When the scene's max-turn-significance is < 2, the per-POV summary
rewrite collapses fully — no "Key quotes:" suffix is appended."""
db = tmp_path / "t.db"
apply_migrations(db)
canned = json.dumps(
{
"summary": "BotA had a low-key chat with you.",
"knowledge_facts": [],
"relationship_summary": "Nothing major shifted.",
}
)
no_threads = json.dumps({"candidates": []})
with open_db(db) as conn:
_seed_single_bot_scene_no_memory(conn)
_seed_memory(conn, pov_summary="Maya rambled about coffee", significance=1)
_seed_memory(conn, pov_summary="Maya glanced at the clock", significance=0)
project(conn)
client = MockLLMClient(canned=[canned, no_threads])
await apply_scene_close_summary(
conn,
client,
classifier_model="x",
chat_id="chat_bot_a",
scene_id=1,
host_bot_id="bot_a",
)
rows = conn.execute(
"SELECT pov_summary FROM memories WHERE scene_id = 1"
).fetchall()
assert rows
for (pov,) in rows:
assert "Key quotes:" not in pov
assert "BotA had a low-key chat" in pov
@pytest.mark.asyncio
async def test_high_significance_scene_includes_top_3_quotes(tmp_path):
"""When max-turn-significance is >= 2, each per-POV summary text gains
a "Key quotes:" suffix listing the top-3 highest-significance memory
rows verbatim, ordered by (significance DESC, id ASC)."""
db = tmp_path / "t.db"
apply_migrations(db)
canned = json.dumps(
{
"summary": "BotA had a heavy talk with you.",
"knowledge_facts": [],
"relationship_summary": "Things shifted.",
}
)
no_threads = json.dumps({"candidates": []})
with open_db(db) as conn:
_seed_single_bot_scene_no_memory(conn)
# Insertion order matches id ASC. Top-3 by (sig DESC, id ASC):
# quote 1 (sig 3) -> quote 2 (sig 2, lower id) -> quote 4 (sig 2,
# higher id). quote 3 (sig 1) is dropped.
_seed_memory(conn, pov_summary="Maya quote one", significance=3)
_seed_memory(conn, pov_summary="Maya quote two", significance=2)
_seed_memory(conn, pov_summary="Maya quote three", significance=1)
_seed_memory(conn, pov_summary="Maya quote four", significance=2)
project(conn)
client = MockLLMClient(canned=[canned, no_threads])
await apply_scene_close_summary(
conn,
client,
classifier_model="x",
chat_id="chat_bot_a",
scene_id=1,
host_bot_id="bot_a",
)
rows = conn.execute(
"SELECT pov_summary FROM memories WHERE scene_id = 1"
).fetchall()
assert rows
for (pov,) in rows:
assert "Key quotes:" in pov
assert '"Maya quote one"' in pov
assert '"Maya quote two"' in pov
assert '"Maya quote four"' in pov
# The sig-1 quote falls outside the top-3 cap.
assert '"Maya quote three"' not in pov
# Ordering: sig 3 first, then the two sig-2s by id ASC.
i_one = pov.index('"Maya quote one"')
i_two = pov.index('"Maya quote two"')
i_four = pov.index('"Maya quote four"')
assert i_one < i_two < i_four
@pytest.mark.asyncio
async def test_thread_detection_emits_events(tmp_path, monkeypatch):
"""On scene close, ``detect_threads`` is invoked and each "open"
candidate yields a ``thread_opened`` event with a fresh thread_id."""
from chat.services import thread_detection as td_mod
canned = json.dumps(
{
"summary": "BotA noticed something unresolved.",
"knowledge_facts": [],
"relationship_summary": "Tension lingered.",
}
)
async def fake_detect_threads(client, **kwargs):
return td_mod.ThreadDetectionResult(
candidates=[
td_mod.ThreadCandidate(
action="open",
title="Test thread",
summary="A test",
existing_thread_id=None,
),
]
)
monkeypatch.setattr(td_mod, "detect_threads", fake_detect_threads)
db = tmp_path / "t.db"
apply_migrations(db)
with open_db(db) as conn:
_seed_single_bot_scene(conn)
project(conn)
client = MockLLMClient(canned=[canned])
await apply_scene_close_summary(
conn,
client,
classifier_model="x",
chat_id="chat_bot_a",
scene_id=1,
host_bot_id="bot_a",
)
rows = conn.execute(
"SELECT payload_json FROM event_log WHERE kind = 'thread_opened'"
).fetchall()
assert len(rows) == 1
payload = json.loads(rows[0][0])
assert payload["title"] == "Test thread"
assert payload["summary"] == "A test"
assert payload["chat_id"] == "chat_bot_a"
assert payload["thread_id"].startswith("thr_")
# The threads-table projection ran via append_and_apply.
from chat.state.threads import list_open_threads
open_threads = list_open_threads(conn, "chat_bot_a")
assert len(open_threads) == 1
assert open_threads[0]["title"] == "Test thread"