Files
javis_bot/tests/test_voice_worker_brain.py
javis-bot 46e4e75d05 feat(voice_worker): a short voice prompt and one judgement before each v2 answer
- compact_prompt: the v2 persona in a tenth of v1's words, answers as short as
  possible, the full tool guidance only on turns with a data tool. A greeting's
  reply prompt went from about 4,500 to 1,400 tokens on Strata.
- intent_judge: the tool router judges whether the turn needs tools; the
  planner runs only when it does. prepare_intent runs the pick ahead on a
  partial transcript (worker message prepare), and the reply reuses it.
- Without stream_browser the screen-share tools are not offered.
- The time context carries the 12-hour form (10:37 was read as pm).
- A <tool_call> block written as the reply is parsed as a call, not spoken.

Evals (Qwen3.6 on Strata, EVAL_COMPACT_PROMPT=1 EVAL_INTENT_JUDGE=1):
221 passed, 4 failed (known failures); new evals/test_voice_reply_length.py
(v2 answers 234 chars against v1's 416) and test_reply_language_default.py.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 12:25:15 +09:00

1108 lines
49 KiB
Python

"""Voice worker history and brain messages (voice_worker.spec.md, "History and brain").
Over a real local WebSocket with the fake STT/VAD/TTS engines of test_voice_worker_server and a
fake brain that records the history context it was given. Behaviours: presence and finals reach
the history store, chat events are kept, a reply answers with the asker's own scope, a voice answer
is stored and can be spoken, and the worker reports when it has no brain.
"""
from __future__ import annotations
import asyncio
import json
import pytest
from jarvis.memory.history import HistoryStore, Scope
from voice_worker.protocol import decode_tts_frame
from voice_worker.server import VoiceWorker
from tests.test_voice_worker_server import TOKEN, AmplitudeVad, FakeTts, LevelTranscriber, hello, recv_until, stream
websockets = pytest.importorskip("websockets")
from websockets.asyncio.client import connect # noqa: E402
class FakeBrain:
def __init__(self, answer=lambda user, text: f"answer for {text}"):
self.answer = answer
self.calls = []
def reply(self, user, text, history, chat=None):
self.calls.append((user, text, history))
self.chat = chat
return self.answer(user, text)
def addressed(self, user, text, names):
self.calls.append(("addressed", user, text, names))
return "jarvis" in text
def rank(self, candidates):
self.calls.append(("rank", [c["id"] for c in candidates]))
return [c["id"] for c in reversed(candidates)]
@pytest.fixture
def store(tmp_path):
s = HistoryStore(str(tmp_path / "history.sqlite"))
yield s
s.close()
async def with_worker(body, store=None, brain=None, tts=None, **kw):
worker = VoiceWorker(TOKEN, LevelTranscriber(), AmplitudeVad, tts, history=store, brain=brain, **kw)
server = await worker.serve("127.0.0.1", 0)
port = server.sockets[0].getsockname()[1]
try:
return await asyncio.wait_for(body(f"ws://127.0.0.1:{port}"), 15)
finally:
server.close()
await server.wait_closed()
await worker.stop()
async def send(ws, **msg):
await ws.send(json.dumps(msg))
async def say(ws, speaker, level):
await stream(ws, speaker, level, 600)
await send(ws, type="speaking_end", speaker=speaker)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "final" and m["speaker"] == speaker)
def texts(records):
return [r.text for r in records]
@pytest.mark.unit
def test_finals_in_a_voice_channel_are_stored_under_the_speakers_name(store):
async def body(url):
async with connect(url) as ws:
await hello(ws, stt={"partials": False})
await say(ws, "100", 0.9) # before Jarvis is in a channel: not stored
await send(ws, type="voice_join", channel="vc1")
await send(ws, type="member_join", user="100", name="Alice")
await say(ws, "100", 0.6)
asyncio.run(with_worker(body, store=store))
recs = store.records_of("100")
assert texts(recs) == ["level 0.6"]
assert recs[0].author_name == "Alice" and recs[0].channel_id == "vc1" and recs[0].kind == "voice"
@pytest.mark.unit
def test_reply_answers_with_the_askers_own_scope_and_stores_the_voice_answer(store):
brain = FakeBrain()
async def body(url):
async with connect(url) as ws:
await hello(ws, stt={"partials": False})
await send(ws, type="voice_join", channel="vc1")
await send(ws, type="member_join", user="100", name="Alice")
await say(ws, "100", 0.9)
await send(ws, type="member_join", user="200", name="Bob")
await say(ws, "200", 0.6)
await send(ws, type="reply", id="r1", user="100", text="what did we say", via="voice", speak=False)
seen = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
await send(ws, type="reply", id="r2", user="200", text="and me", via="voice", speak=False)
seen += await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
return seen
seen = asyncio.run(with_worker(body, store=store, brain=brain))
answers = [m for m in seen if isinstance(m, dict) and m.get("type") == "reply_text"]
assert answers[0] == {"type": "reply_text", "id": "r1", "text": "answer for what did we say", "failed": False}
# Alice was there when Bob spoke, so she sees both; Bob joined after Alice spoke.
alice_ctx, bob_ctx = brain.calls[0][2], brain.calls[1][2]
assert any("Alice: level 0.9" in x for x in alice_ctx.recent)
assert any("Bob: level 0.6" in x for x in alice_ctx.recent)
assert not any("Alice: level 0.9" in x for x in bob_ctx.recent)
# The voice answer is Alice's own record now.
assert "answer for what did we say" in texts(store.records_of("100"))
@pytest.mark.unit
def test_spoken_reply_comes_back_as_tts_under_the_same_id(store):
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="voice_join", channel="vc1")
await send(ws, type="reply", id="r9", user="100", text="hello", via="voice", speak=True)
return await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "tts_done")
seen = asyncio.run(with_worker(body, store=store, brain=FakeBrain(), tts=FakeTts()))
frames = [decode_tts_frame(m) for m in seen if isinstance(m, bytes)]
assert frames and all(rid == "r9" for rid, _, _ in frames)
done = [m for m in seen if isinstance(m, dict) and m.get("type") == "tts_done"][0]
assert done["id"] == "r9" and not done["failed"]
@pytest.mark.unit
def test_chat_answers_are_not_stored_by_the_worker_but_reported_messages_are(store):
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="reply", id="c1", user="100", text="hi", via="chat", speak=False)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
await send(ws, type="chat_message", channel="ch1", message_id="m1", author="100", name="Alice",
text="first", ts=1_800_000_000.0)
await send(ws, type="chat_message", channel="ch1", message_id="m2", author="900", name="Jarvis",
text="answer for hi", ts=1_800_000_001.0, to="100")
await send(ws, type="chat_edit", channel="ch1", message_id="m1", text="first, edited")
await send(ws, type="chat_message", channel="ch1", message_id="m3", author="100", name="Alice",
text="gone soon", ts=1_800_000_002.0)
await send(ws, type="chat_delete", channel="ch1", message_id="m3")
# A round trip makes sure the worker handled everything above.
await send(ws, type="reply", id="c2", user="100", text="x", via="chat", speak=False)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "c2")
asyncio.run(with_worker(body, store=store, brain=FakeBrain()))
assert texts(store.records_of("100")) == ["first, edited", "answer for hi"]
@pytest.mark.unit
def test_leaving_the_voice_channel_stops_storing_and_closes_presence(store):
async def body(url):
async with connect(url) as ws:
await hello(ws, stt={"partials": False})
await send(ws, type="voice_join", channel="vc1")
await send(ws, type="member_join", user="100", name="Alice")
await send(ws, type="voice_leave")
await say(ws, "200", 0.6)
asyncio.run(with_worker(body, store=store))
assert store.records_of("200") == []
assert store.recent("100", now=4_000_000_000, scope=Scope(), minutes=10 ** 9) == []
@pytest.mark.unit
def test_memory_off_gives_the_brain_no_history(store):
brain = FakeBrain()
async def body(url):
async with connect(url) as ws:
await hello(ws, stt={"partials": False}, memory={"off": True})
await send(ws, type="voice_join", channel="vc1")
await say(ws, "100", 0.9)
await send(ws, type="reply", id="r1", user="100", text="q", via="voice", speak=False)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
await send(ws, type="memory_config", off=False)
await send(ws, type="reply", id="r2", user="100", text="q", via="voice", speak=False)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r2")
asyncio.run(with_worker(body, store=store, brain=brain))
off_ctx, on_ctx = brain.calls[0][2], brain.calls[1][2]
assert off_ctx.recent == [] and off_ctx.matches == []
assert any("level 0.9" in x for x in on_ctx.recent)
@pytest.mark.unit
@pytest.mark.parametrize("answer", [lambda u, t: None, lambda u, t: 1 / 0])
def test_a_failed_brain_answer_is_reported(store, answer):
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="reply", id="r1", user="100", text="q", via="voice", speak=True)
return await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
seen = asyncio.run(with_worker(body, store=store, brain=FakeBrain(answer), tts=FakeTts()))
assert seen[-1] == {"type": "reply_text", "id": "r1", "text": "", "failed": True}
@pytest.mark.unit
def test_without_a_brain_reply_is_an_error():
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="voice_join", channel="vc1")
await send(ws, type="reply", id="r1", user="100", text="q", via="voice", speak=False)
return await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "error")
seen = asyncio.run(with_worker(body))
assert seen[-1]["code"] == "brain_unavailable" and seen[-1]["id"] == "r1"
@pytest.mark.unit
def test_malformed_history_messages_are_rejected(store):
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="member_join", user=5)
await send(ws, type="reply", id="r1", user="100", via="radio")
return [await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "error")
for _ in range(2)]
seen = asyncio.run(with_worker(body, store=store, brain=FakeBrain()))
assert [s[-1]["code"] for s in seen] == ["bad_message", "bad_message"]
class SlowTranscriber(LevelTranscriber):
def __init__(self, delay):
super().__init__()
self.delay = delay
def transcribe(self, clips, opts):
import time as _t
_t.sleep(self.delay)
return super().transcribe(clips, opts)
@pytest.mark.unit
def test_a_slow_final_keeps_the_channel_time_and_name_of_when_it_was_said(store):
async def body(url):
async with connect(url) as ws:
await hello(ws, stt={"partials": False})
await send(ws, type="voice_join", channel="vc1")
await send(ws, type="member_join", user="100", name="Alice")
await stream(ws, "100", 0.9, 600)
await send(ws, type="speaking_end", speaker="100")
# While the model is still decoding: a late joiner arrives, then Jarvis moves.
await asyncio.sleep(0.15)
await send(ws, type="member_join", user="300", name="Late")
await send(ws, type="member_join", user="100", name="Renamed")
await send(ws, type="voice_join", channel="vc2")
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "final")
worker = VoiceWorker(TOKEN, SlowTranscriber(0.6), AmplitudeVad, None, history=store)
async def run():
server = await worker.serve("127.0.0.1", 0)
port = server.sockets[0].getsockname()[1]
try:
await asyncio.wait_for(body(f"ws://127.0.0.1:{port}"), 15)
finally:
server.close()
await server.wait_closed()
await worker.stop()
asyncio.run(run())
recs = store.records_of("100")
assert [(r.channel_id, r.author_name, r.text) for r in recs] == [("vc1", "Alice", "level 0.9")]
# The late joiner was not there when it was said.
assert store.recent("300", now=4_000_000_000, scope=Scope(), minutes=10 ** 9) == []
@pytest.mark.unit
def test_presence_left_open_by_a_previous_run_is_closed_at_start(store):
store.presence_join("old", "vc1", 1_000.0) # the worker was killed with "old" still present
async def body(url):
async with connect(url) as ws:
await hello(ws, stt={"partials": False})
await send(ws, type="voice_join", channel="vc1")
await send(ws, type="member_join", user="100", name="Alice")
await say(ws, "100", 0.9)
asyncio.run(with_worker(body, store=store))
assert texts(store.recent("100", now=4_000_000_000, scope=Scope(), minutes=10 ** 9)) == ["level 0.9"]
assert store.recent("old", now=4_000_000_000, scope=Scope(), minutes=10 ** 9) == []
@pytest.mark.unit
def test_the_brain_prints_no_text_from_its_thread(tmp_path, monkeypatch, capfd):
pytest.importorskip("mcp") # the real brain needs the jarvis dependencies (--brain)
import sys
import threading
from jarvis.memory.history import HistoryContext
from voice_worker import brain as brain_mod
cfg = tmp_path / "config.json"
cfg.write_text(json.dumps({"db_path": str(tmp_path / "jarvis.db")}))
monkeypatch.setenv("JARVIS_CONFIG_PATH", str(cfg))
def fake_engine(db, cfg, tts, text, mem, language=None, history=None, legacy_memory=True, **kw):
print(f"answer text {text}")
print(f"debug {text}", file=sys.stderr)
# Tools print from helper threads too (the web fetch pool).
helper = threading.Thread(target=lambda: print(f"helper {text}", file=sys.stderr))
helper.start()
helper.join()
return f"answer to {text}"
import jarvis.reply.engine as engine
monkeypatch.setattr(engine, "run_reply_engine", fake_engine)
b = brain_mod.JarvisBrain()
out = []
t = threading.Thread(target=lambda: out.append(b.reply("100", "secret question", HistoryContext())))
t.start()
t.join()
print("other output stays")
captured = capfd.readouterr()
assert out == ["answer to secret question"]
assert "secret question" not in captured.out + captured.err
assert "other output stays" in captured.out
@pytest.mark.unit
def test_the_brain_turns_off_the_single_user_memory(tmp_path, monkeypatch):
pytest.importorskip("mcp") # the real brain needs the jarvis dependencies (--brain)
from jarvis.memory.history import HistoryContext
from voice_worker import brain as brain_mod
cfg = tmp_path / "config.json"
cfg.write_text(json.dumps({"db_path": str(tmp_path / "jarvis.db")}))
monkeypatch.setenv("JARVIS_CONFIG_PATH", str(cfg))
seen = {}
def fake_engine(db, cfg, tts, text, mem, language=None, history=None, legacy_memory=True, **kw):
seen["legacy_memory"] = legacy_memory
return "ok"
import jarvis.reply.engine as engine
monkeypatch.setattr(engine, "run_reply_engine", fake_engine)
brain_mod.JarvisBrain().reply("100", "q", HistoryContext())
assert seen["legacy_memory"] is False
@pytest.mark.unit
def test_the_brain_answers_in_korean_by_default_even_with_a_v1_english_voice(tmp_path, monkeypatch):
"""v1's ``tts_engine`` (an English-only Piper voice by default) must not lock v2 replies to
English: v2 speaks through the worker's own TTS. Replies come in the default reply language
(Korean) unless the asker asks for another, and tools get its ISO code."""
pytest.importorskip("mcp") # the real brain needs the jarvis dependencies (--brain)
from jarvis.memory.history import HistoryContext
from jarvis.system_prompt import default_language_directive, reply_language_directive
from voice_worker import brain as brain_mod
cfg = tmp_path / "config.json"
cfg.write_text(json.dumps({"db_path": str(tmp_path / "jarvis.db"), "tts_engine": "piper"}))
monkeypatch.setenv("JARVIS_CONFIG_PATH", str(cfg))
monkeypatch.delenv("OUTPUT_LANGUAGE", raising=False)
seen = {}
def fake_engine(db, cfg, tts, text, mem, language=None, history=None, legacy_memory=True,
default_language=None, compact_prompt=False, **kw):
seen["directive"] = reply_language_directive(None, cfg.tts_engine, default_language)
seen["language"] = language
seen["compact"] = compact_prompt
seen["stream_browser"] = cfg.stream_browser
return "ok"
import jarvis.reply.engine as engine
monkeypatch.setattr(engine, "run_reply_engine", fake_engine)
brain_mod.JarvisBrain().reply("100", "what time is it?", HistoryContext())
assert seen["directive"] == default_language_directive("Korean")
assert seen["language"] == "ko"
# The short voice prompt, and no v1 screen-share browser (it needs v1's Node bot).
assert seen["compact"] is True and seen["stream_browser"] is False
# Without a default the brain answers in the asker's language.
brain_mod.JarvisBrain(language=None, reply_language=None).reply("100", "q", HistoryContext())
assert seen["directive"] is None and seen["language"] is None
@pytest.mark.unit
def test_the_brain_tells_tools_who_is_asking(tmp_path, monkeypatch):
pytest.importorskip("mcp") # the real brain needs the jarvis dependencies (--brain)
from jarvis.memory.history import HistoryContext
from jarvis.reply import turn_state
from voice_worker import brain as brain_mod
cfg = tmp_path / "config.json"
cfg.write_text(json.dumps({"db_path": str(tmp_path / "jarvis.db")}))
monkeypatch.setenv("JARVIS_CONFIG_PATH", str(cfg))
seen = []
def fake_engine(db, cfg, tts, text, mem, language=None, history=None, legacy_memory=True, **kw):
seen.append(turn_state.get_user())
return "ok"
import jarvis.reply.engine as engine
monkeypatch.setattr(engine, "run_reply_engine", fake_engine)
b = brain_mod.JarvisBrain()
b.reply("100", "q", HistoryContext())
b.reply("200", "q", HistoryContext())
assert seen == ["100", "200"]
assert turn_state.get_user() is None # nothing left behind on this thread
CHANNELS = {"type": "chat_channels", "bot": "900", "channels": [
{"id": "c1", "name": "일정", "description": "약속", "read": True, "send": True, "edit": True,
"react": True, "delete": False, "pin": False}]}
@pytest.mark.unit
def test_the_brain_gets_the_registered_channels_and_their_messages(store):
seen = {}
def answer(user, text):
chat = brain.chat
seen["channels"] = chat.channels()
seen["read"] = chat.read("c1", 5)
seen["search"] = chat.search(["c1"], "회의", 5)
seen["turn"] = chat.turn
return "ok"
brain = FakeBrain(answer)
async def body(url):
async with connect(url) as ws:
await hello(ws)
await ws.send(json.dumps(CHANNELS))
await send(ws, type="chat_message", channel="c1", message_id="m1", author="200", name="Bob",
text="회의는 3시", ts=1_800_000_000.0)
await send(ws, type="reply", id="r1", user="100", text="q", via="chat", speak=False)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
await send(ws, type="reply_delivered", id="r1", ok=True)
await send(ws, type="reply", id="r2", user="100", text="q", via="chat", speak=False)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r2")
asyncio.run(with_worker(body, store=store, brain=brain))
assert [c["name"] for c in seen["channels"]] == ["일정"] and seen["channels"][0]["send"] is True
assert [(m["message_id"], m["author_name"], m["text"]) for m in seen["read"]] == [("m1", "Bob", "회의는 3시")]
assert [m["text"] for m in seen["search"]] == ["회의는 3시"]
assert seen["turn"] == 2 # the same person's second reply
@pytest.mark.unit
def test_a_chat_action_goes_to_the_app_and_its_answer_comes_back(store):
results = []
brain = FakeBrain(lambda u, t: results.append(brain.chat.act("send", "c1", text="안녕")) or "done")
async def body(url):
async with connect(url) as ws:
await hello(ws)
await ws.send(json.dumps(CHANNELS))
await send(ws, type="reply", id="r1", user="100", text="q", via="chat", speak=False)
seen = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "chat_action")
action = seen[-1]
await send(ws, type="chat_action_result", id=action["id"], ok=True, message_ids=["555"])
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
return action
action = asyncio.run(with_worker(body, store=store, brain=brain))
assert {k: action[k] for k in ("action", "channel", "text")} == {"action": "send", "channel": "c1", "text": "안녕"}
assert results == [{"ok": True, "message_ids": ["555"]}]
@pytest.mark.unit
def test_a_chat_action_the_app_never_answers_fails(store):
results = []
brain = FakeBrain(lambda u, t: results.append(brain.chat.act("send", "c1", text="x")) or "done")
async def body(url):
async with connect(url) as ws:
await hello(ws)
await ws.send(json.dumps(CHANNELS))
await send(ws, type="reply", id="r1", user="100", text="q", via="chat", speak=False)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
worker = VoiceWorker(TOKEN, LevelTranscriber(), AmplitudeVad, None, history=store, brain=brain,
chat_timeout_s=0.3)
async def run():
server = await worker.serve("127.0.0.1", 0)
port = server.sockets[0].getsockname()[1]
try:
await asyncio.wait_for(body(f"ws://127.0.0.1:{port}"), 15)
finally:
server.close()
await server.wait_closed()
await worker.stop()
asyncio.run(run())
assert results and results[0]["ok"] is False and results[0]["error"]
@pytest.mark.unit
def test_a_late_answer_from_an_earlier_connection_is_not_taken_for_a_new_action(store):
results = []
brain = FakeBrain(lambda u, t: results.append(brain.chat.act("send", "c1", text="x")) or "done")
worker = VoiceWorker(TOKEN, LevelTranscriber(), AmplitudeVad, None, history=store, brain=brain,
chat_timeout_s=0.5)
async def ask(url):
async with connect(url) as ws:
await hello(ws)
await ws.send(json.dumps(CHANNELS))
await send(ws, type="reply", id="r1", user="100", text="q", via="chat", speak=False)
seen = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "chat_action")
return ws, seen[-1]["id"]
async def run():
server = await worker.serve("127.0.0.1", 0)
url = f"ws://127.0.0.1:{server.sockets[0].getsockname()[1]}"
try:
async with connect(url) as ws: # first connection: asked, never answered, then gone
await hello(ws)
await ws.send(json.dumps(CHANNELS))
await send(ws, type="reply", id="r1", user="100", text="q", via="chat", speak=False)
old = (await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "chat_action"))[-1]["id"]
async with connect(url) as ws: # second connection: the old answer arrives late
await hello(ws)
await ws.send(json.dumps(CHANNELS))
await send(ws, type="reply", id="r2", user="100", text="q", via="chat", speak=False)
new = (await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "chat_action"))[-1]["id"]
await send(ws, type="chat_action_result", id=old, ok=True, message_ids=["stale"])
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r2")
return old, new
finally:
server.close()
await server.wait_closed()
await worker.stop()
old, new = asyncio.run(asyncio.wait_for(run(), 15))
assert old != new
assert all(r.get("message_ids") != ["stale"] for r in results), results
assert results[-1]["ok"] is False
@pytest.mark.unit
def test_addressed_and_rank_answer_from_the_brain(store):
brain = FakeBrain()
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="addressed", id="q1", user="100", text="jarvis, lights off")
await send(ws, type="addressed", id="q2", user="100", text="did you see that")
await send(ws, type="rank", id="k1", candidates=[{"id": "7", "user": "100", "text": "a"},
{"id": "8", "user": "200", "text": "b"}])
got = {}
while len(got) < 3:
m = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") in
("addressed_result", "rank_result"))
got[m[-1]["id"]] = m[-1]
return got
got = asyncio.run(with_worker(body, store=store, brain=brain))
assert got["q1"] == {"type": "addressed_result", "id": "q1", "addressed": True, "failed": False}
assert got["q2"]["addressed"] is False and got["q2"]["failed"] is False
assert got["k1"] == {"type": "rank_result", "id": "k1", "order": ["8", "7"], "failed": False}
@pytest.mark.unit
def test_addressed_and_rank_do_not_wait_behind_a_running_reply(store):
import threading
release = threading.Event()
def slow(user, text):
release.wait(10)
return "late"
brain = FakeBrain(slow)
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="reply", id="r1", user="100", text="q", via="voice", speak=False)
await send(ws, type="rank", id="k1", candidates=[{"id": "7", "user": "100", "text": "a"},
{"id": "8", "user": "200", "text": "b"}])
seen = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "rank_result")
release.set()
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
return seen
seen = asyncio.run(with_worker(body, store=store, brain=brain))
assert not any(isinstance(m, dict) and m.get("type") == "reply_text" for m in seen)
@pytest.mark.unit
@pytest.mark.parametrize("fails", [lambda *a: None, lambda *a: 1 / 0])
def test_a_failed_judgement_or_ranking_is_reported(store, fails):
brain = FakeBrain()
brain.addressed = fails
brain.rank = fails
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="addressed", id="q1", user="100", text="x")
a = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "addressed_result")
await send(ws, type="rank", id="k1", candidates=[{"id": "7", "user": "100", "text": "a"}])
r = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "rank_result")
return a[-1], r[-1]
a, r = asyncio.run(with_worker(body, store=store, brain=brain))
assert a == {"type": "addressed_result", "id": "q1", "addressed": False, "failed": True}
assert r == {"type": "rank_result", "id": "k1", "order": [], "failed": True}
@pytest.mark.unit
def test_a_reply_cancelled_while_the_brain_thinks_is_not_spoken(store):
import threading
release = threading.Event()
def slow(user, text):
release.wait(10)
return f"answer for {text}"
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="voice_join", channel="vc1")
await send(ws, type="reply", id="r1", user="100", text="q", via="voice", speak=True)
await send(ws, type="tts_cancel", id="r1")
await send(ws, type="reply", id="r2", user="100", text="next", via="voice", speak=True)
await asyncio.sleep(0.1)
release.set()
return await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "tts_done"
and m["id"] == "r2")
seen = asyncio.run(with_worker(body, store=store, brain=FakeBrain(slow), tts=FakeTts()))
frames = {decode_tts_frame(m)[0] for m in seen if isinstance(m, bytes)}
done = {m["id"]: m for m in seen if isinstance(m, dict) and m.get("type") == "tts_done"}
assert frames == {"r2"}
assert done["r1"]["cancelled"] and not done["r2"]["cancelled"]
replies = [m["id"] for m in seen if isinstance(m, dict) and m.get("type") == "reply_text"]
assert replies == ["r1", "r2"], "the text still comes back; only the speech is skipped"
@pytest.mark.unit
def test_the_brains_judgements_use_its_model_and_map_back_to_ids(monkeypatch):
import voice_worker.judges as judges
from voice_worker.brain import JarvisBrain
asked = []
def fake_llm(base_url, model, system, user, **kw):
asked.append((base_url, model, system, user))
return "2, 1" if "<requests>" in user else "Yes."
monkeypatch.setattr(judges, "call_llm_direct", fake_llm)
brain = JarvisBrain.__new__(JarvisBrain)
brain.cfg = type("Cfg", (), {"ollama_base_url": "http://s/v1", "ollama_chat_model": "m"})()
assert brain.addressed("100", "불 꺼 줘", ["자비스"]) is True
assert brain.rank([{"id": "7", "user": "1", "text": "a"}, {"id": "8", "user": "2", "text": "b"}]) == ["8", "7"]
assert asked[0][:2] == ("http://s/v1", "m") and "자비스" in asked[0][2]
assert asked[0][3] == "<utterance>불 꺼 줘</utterance>", "the heard text is fenced as data"
@pytest.mark.unit
@pytest.mark.parametrize("raw,expected", [("no", False), ("NO.", False), ("maybe", None), (None, None)])
def test_an_unclear_judgement_is_none(monkeypatch, raw, expected):
import voice_worker.judges as judges
monkeypatch.setattr(judges, "call_llm_direct", lambda *a, **k: raw)
assert judges.is_addressed("u", "m", [], "x") is expected
@pytest.mark.unit
def test_a_ranking_keeps_left_out_requests_in_order_and_fails_without_numbers(monkeypatch):
import voice_worker.judges as judges
monkeypatch.setattr(judges, "call_llm_direct", lambda *a, **k: "3 then 9")
assert judges.rank_requests("u", "m", ["a", "b", "c"]) == [2, 0, 1]
monkeypatch.setattr(judges, "call_llm_direct", lambda *a, **k: "none of them")
assert judges.rank_requests("u", "m", ["a", "b"]) is None
@pytest.mark.unit
def test_a_voice_answer_does_not_wait_behind_a_chat_answer(store):
import threading
release = threading.Event()
running = []
def answer(user, text):
running.append(text)
if text == "chat question":
release.wait(10)
return f"answer for {text}"
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="reply", id="c1", user="100", text="chat question", via="chat", speak=False)
await asyncio.sleep(0.1)
await send(ws, type="reply", id="v1", user="200", text="voice question", via="voice", speak=False)
seen = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "v1")
release.set()
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "c1")
return seen
seen = asyncio.run(with_worker(body, store=store, brain=FakeBrain(answer)))
assert seen[-1]["text"] == "answer for voice question"
assert not any(isinstance(m, dict) and m.get("id") == "c1" for m in seen), "the chat answer was still running"
@pytest.mark.unit
@pytest.mark.parametrize("via", ["voice", "chat"])
def test_answers_of_one_kind_are_written_one_at_a_time(store, via):
import threading
active, peak = [0], [0]
lock = threading.Lock()
def answer(user, text):
with lock:
active[0] += 1
peak[0] = max(peak[0], active[0])
import time as _t
_t.sleep(0.05)
with lock:
active[0] -= 1
return "ok"
async def body(url):
async with connect(url) as ws:
await hello(ws)
for i in range(3):
await send(ws, type="reply", id=f"r{i}", user="100", text="q", via=via, speak=False)
for i in range(3):
await recv_until(ws, lambda m, i=i: isinstance(m, dict) and m.get("id") == f"r{i}")
asyncio.run(with_worker(body, store=store, brain=FakeBrain(answer)))
assert peak[0] == 1
@pytest.mark.unit
def test_one_persons_voice_and_chat_answers_are_written_one_at_a_time(store):
# Their dialogue memory and delete confirmations ("the very next message") follow the order
# they asked, so a person's second question starts only after their first is answered.
import threading
release = threading.Event()
running = []
def answer(user, text):
running.append(text)
if text == "chat question":
release.wait(10)
return f"answer for {text}"
async def body(url):
async with connect(url) as ws:
await hello(ws)
await send(ws, type="reply", id="c1", user="100", text="chat question", via="chat", speak=False)
await asyncio.sleep(0.1)
await send(ws, type="reply", id="v1", user="100", text="voice question", via="voice", speak=False)
await asyncio.sleep(0.3)
started = list(running)
release.set()
seen = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "v1")
return started, [m["id"] for m in seen if isinstance(m, dict) and m.get("type") == "reply_text"]
started, order = asyncio.run(with_worker(body, store=store, brain=FakeBrain(answer)))
assert started == ["chat question"], "the voice question started before the same person's chat answer"
assert order == ["c1", "v1"]
DELETABLE = {"type": "chat_channels", "bot": "900", "channels": [
{"id": "c1", "name": "일정", "description": "약속", "read": True, "send": True, "edit": True,
"react": True, "delete": True, "pin": False}]}
def delete_tool_brain(release, question="지울까요?"):
"""A brain whose reply runs the real chatChannels tool: "지워" asks to delete message m1 (and waits
for ``release`` before answering), "응, 지워" confirms. The consent check always agrees."""
from unittest.mock import patch
from jarvis.reply import turn_state
from jarvis.tools.builtin.chat_channels import ChatChannelsTool
class Cfg:
ollama_base_url = "http://x"
ollama_chat_model = "m"
results = []
def answer(user, text):
turn_state.reset()
turn_state.set_chat(brain.chat)
args = {"action": "delete", "channel": "c1", "message_id": "m1", "confirm": text == "응, 지워"}
try:
with patch("jarvis.tools.builtin.chat_channels.call_llm_direct", return_value="YES"):
res = ChatChannelsTool().execute(db=None, cfg=Cfg(), tool_args=args, system_prompt="",
original_prompt=text, redacted_text=text, max_retries=0,
user_print=lambda m: None)
finally:
turn_state.reset()
results.append((text, res.success))
if text == "지워":
release.wait(10)
return question if text == "지워" else "done"
brain = FakeBrain(answer)
return brain, results
@pytest.mark.unit
def test_a_yes_sent_before_the_delete_question_was_answered_does_not_delete(store):
import threading
release = threading.Event()
brain, results = delete_tool_brain(release)
async def body(url):
async with connect(url) as ws:
await hello(ws)
await ws.send(json.dumps(DELETABLE))
await send(ws, type="reply", id="r1", user="100", text="지워", via="chat", speak=False)
await asyncio.sleep(0.2)
# Typed before the question reached them: it cannot be their answer to it.
await send(ws, type="reply", id="r2", user="100", text="응, 지워", via="voice", speak=False)
await asyncio.sleep(0.1)
release.set()
return await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r2")
seen = asyncio.run(with_worker(body, store=store, brain=brain))
assert not any(isinstance(m, dict) and m.get("type") == "chat_action" for m in seen)
assert results == [("지워", True), ("응, 지워", False)]
@pytest.mark.unit
def test_a_yes_after_the_delete_question_deletes(store):
import threading
release = threading.Event()
release.set()
brain, results = delete_tool_brain(release)
async def body(url):
async with connect(url) as ws:
await hello(ws)
await ws.send(json.dumps(DELETABLE))
await send(ws, type="reply", id="r1", user="100", text="지워", via="chat", speak=False)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r1")
await send(ws, type="reply_delivered", id="r1", ok=True) # the app posted the question
await send(ws, type="reply", id="r2", user="100", text="응, 지워", via="voice", speak=False)
seen = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "chat_action")
await send(ws, type="chat_action_result", id=seen[-1]["id"], ok=True)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r2")
return seen[-1]
action = asyncio.run(with_worker(body, store=store, brain=brain))
assert (action["action"], action["channel"], action["message_id"]) == ("delete", "c1", "m1")
assert results == [("지워", True), ("응, 지워", True)]
@pytest.mark.unit
def test_a_yes_after_a_delete_question_they_never_got_does_not_delete(store):
import threading
release = threading.Event()
release.set()
brain, results = delete_tool_brain(release, question="") # the answer failed: nothing was asked
async def body(url):
async with connect(url) as ws:
await hello(ws)
await ws.send(json.dumps(DELETABLE))
await send(ws, type="reply", id="r1", user="100", text="지워", via="chat", speak=False)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r1")
await send(ws, type="reply", id="r2", user="100", text="응, 지워", via="voice", speak=False)
return await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r2")
seen = asyncio.run(with_worker(body, store=store, brain=brain))
assert not any(isinstance(m, dict) and m.get("type") == "chat_action" for m in seen)
assert results == [("지워", True), ("응, 지워", False)]
def run_delete_then_yes(store, ask, after, question="지울까요?", tts=None, **kw):
"""The person asks to delete (``ask``: the reply message fields), ``after(ws)`` plays what happened
to the question, then they say "응, 지워" by voice. The chat actions that went out."""
import threading
release = threading.Event()
release.set()
brain, results = delete_tool_brain(release, question=question)
async def body(url):
async with connect(url) as ws:
await hello(ws)
await ws.send(json.dumps(DELETABLE))
await send(ws, type="reply", id="r1", user="100", text="지워", **ask)
await after(ws)
await send(ws, type="reply", id="r2", user="100", text="응, 지워", via="voice", speak=False)
seen = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r2")
return [m for m in seen if isinstance(m, dict) and m.get("type") == "chat_action"]
actions = asyncio.run(with_worker(body, store=store, brain=brain, tts=tts, **kw))
return actions, results
def until(pred):
async def after(ws):
await recv_until(ws, pred)
return after
@pytest.mark.unit
def test_a_yes_after_a_question_that_could_not_be_spoken_does_not_delete(store):
# No TTS engine: the spoken question never reached them.
actions, results = run_delete_then_yes(
store, dict(via="voice", speak=True),
until(lambda m: isinstance(m, dict) and m.get("code") == "tts_unavailable"))
assert actions == [] and results == [("지워", True), ("응, 지워", False)]
@pytest.mark.unit
def test_a_yes_after_a_question_whose_speech_failed_does_not_delete(store):
actions, results = run_delete_then_yes(
store, dict(via="voice", speak=True),
until(lambda m: isinstance(m, dict) and m.get("type") == "tts_done" and m["failed"]),
question="FAIL 지울까요?", tts=FakeTts())
assert actions == [] and results == [("지워", True), ("응, 지워", False)]
@pytest.mark.unit
def test_a_yes_after_a_spoken_question_deletes(store):
import threading
release = threading.Event()
release.set()
brain, results = delete_tool_brain(release)
async def body(url):
async with connect(url) as ws:
await hello(ws)
await ws.send(json.dumps(DELETABLE))
await send(ws, type="reply", id="r1", user="100", text="지워", via="voice", speak=True)
done = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "tts_done")
assert not done[-1]["failed"] and not done[-1]["cancelled"]
await send(ws, type="reply", id="r2", user="100", text="응, 지워", via="voice", speak=False)
seen = await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "chat_action")
await send(ws, type="chat_action_result", id=seen[-1]["id"], ok=True)
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r2")
return seen[-1]
action = asyncio.run(with_worker(body, store=store, brain=brain, tts=FakeTts()))
assert action["action"] == "delete" and results == [("지워", True), ("응, 지워", True)]
@pytest.mark.unit
def test_a_yes_after_a_chat_question_the_app_could_not_post_does_not_delete(store):
async def not_posted(ws):
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r1")
await send(ws, type="reply_delivered", id="r1", ok=False)
actions, results = run_delete_then_yes(store, dict(via="chat", speak=False), not_posted)
assert actions == [] and results == [("지워", True), ("응, 지워", False)]
@pytest.mark.unit
def test_a_chat_question_the_app_never_reports_counts_as_not_posted(store):
async def no_report(ws):
await recv_until(ws, lambda m: isinstance(m, dict) and m.get("id") == "r1")
await asyncio.sleep(0.3)
actions, results = run_delete_then_yes(store, dict(via="chat", speak=False), no_report,
delivery_timeout_s=0.1)
assert actions == [] and results == [("지워", True), ("응, 지워", False)]
class PreparingBrain(FakeBrain):
"""Each prepare waits until the test lets it finish."""
def __init__(self):
super().__init__()
import threading
self.prepared = []
self.release = threading.Event()
def prepare(self, user, text, chat):
self.prepared.append((user, text, chat))
self.release.wait(5)
@pytest.mark.unit
def test_prepare_runs_one_at_a_time_per_user_and_only_the_newest_different_text_waits(store):
brain = PreparingBrain()
async def body(url):
async with connect(url) as ws:
await hello(ws, stt={"partials": False})
await send(ws, type="prepare", user="100", text="자비스 오늘")
await asyncio.sleep(0.2)
await send(ws, type="prepare", user="100", text="자비스 오늘 서울")
await send(ws, type="prepare", user="100", text="자비스 오늘 서울 날씨 어때")
await send(ws, type="prepare", user="200", text="자비스 안녕")
await asyncio.sleep(0.2)
assert sorted(brain.prepared) == [("100", "자비스 오늘", False), ("200", "자비스 안녕", False)]
brain.release.set()
await asyncio.sleep(0.3)
# About the same as the one just prepared: not prepared again.
await send(ws, type="prepare", user="100", text="자비스 오늘 서울 날씨 어때?")
await asyncio.sleep(0.3)
await send(ws, type="reply", id="r1", user="100", text="x", via="voice", speak=False)
return await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
seen = asyncio.run(with_worker(body, store=store, brain=brain))
assert [t for u, t, _ in brain.prepared if u == "100"] == ["자비스 오늘", "자비스 오늘 서울 날씨 어때"]
assert not [m for m in seen if isinstance(m, dict) and m.get("type") == "error"], "nothing is sent back"
@pytest.mark.unit
def test_a_brain_without_prepare_ignores_it(store):
async def body(url):
async with connect(url) as ws:
await hello(ws, stt={"partials": False})
await send(ws, type="prepare", user="100", text="자비스 안녕")
await send(ws, type="reply", id="r1", user="100", text="hi", via="voice", speak=False)
return await recv_until(ws, lambda m: isinstance(m, dict) and m.get("type") == "reply_text")
seen = asyncio.run(with_worker(body, store=store, brain=FakeBrain()))
assert not [m for m in seen if isinstance(m, dict) and m.get("type") == "error"]
def _jarvis_brain(tmp_path, monkeypatch, plans):
"""A real JarvisBrain with the engine's judge and reply replaced: ``plans`` maps text to a pick."""
pytest.importorskip("mcp") # the real brain needs the jarvis dependencies (--brain)
from voice_worker import brain as brain_mod
import jarvis.reply.engine as engine
cfg = tmp_path / "config.json"
cfg.write_text(json.dumps({"db_path": str(tmp_path / "jarvis.db")}))
monkeypatch.setenv("JARVIS_CONFIG_PATH", str(cfg))
seen = {"judged": [], "prepared": []}
def fake_prepare(cfg, text, mem, chat_available):
seen["judged"].append((text, chat_available))
return plans.get(text)
def fake_engine(db, cfg, tts, text, mem, prepared_tools=None, intent_judge=False, **kw):
seen["prepared"].append(prepared_tools)
seen["intent_judge"] = intent_judge
mem.add_message("user", text)
mem.add_message("assistant", "ok")
return "ok"
monkeypatch.setattr(engine, "prepare_intent", fake_prepare)
monkeypatch.setattr(engine, "run_reply_engine", fake_engine)
return brain_mod.JarvisBrain(), seen
@pytest.mark.unit
def test_a_reply_uses_the_pick_prepared_for_about_the_same_text(tmp_path, monkeypatch):
from jarvis.memory.history import HistoryContext
plan = ["getWeather"]
b, seen = _jarvis_brain(tmp_path, monkeypatch, {"자비스 오늘 서울 날씨 어": plan})
b.prepare("100", "자비스 오늘 서울 날씨 어", True)
assert seen["judged"] == [("자비스 오늘 서울 날씨 어", True)]
b.reply("100", "자비스, 오늘 서울 날씨 어때?", HistoryContext())
assert seen["prepared"] == [plan] and seen["intent_judge"] is True
# Used once: the next reply judges for itself.
b.reply("100", "자비스, 오늘 서울 날씨 어때?", HistoryContext())
assert seen["prepared"][-1] is None
@pytest.mark.unit
def test_a_prepared_pick_is_not_used_for_other_words_another_person_or_a_changed_dialogue(tmp_path, monkeypatch):
from jarvis.memory.history import HistoryContext
plan = ["webSearch"]
b, seen = _jarvis_brain(tmp_path, monkeypatch, {"자비스 오늘 서울 날씨 어때": plan})
b.prepare("100", "자비스 오늘 서울 날씨 어때", False)
b.reply("100", "자비스, 내일 회의 몇 시지?", HistoryContext())
b.prepare("100", "자비스 오늘 서울 날씨 어때", False)
b.reply("200", "자비스 오늘 서울 날씨 어때", HistoryContext())
assert seen["prepared"] == [None, None]
b.prepare("100", "자비스 오늘 서울 날씨 어때", False)
b._memory("100").add_message("user", "something said meanwhile")
b.reply("100", "자비스 오늘 서울 날씨 어때", HistoryContext())
assert seen["prepared"][-1] is None
@pytest.mark.unit
def test_a_reply_waits_for_a_pick_still_being_prepared_for_its_text(tmp_path, monkeypatch):
import threading
import time as _time
from jarvis.memory.history import HistoryContext
plan = [] # "no tools" is a pick too
b, seen = _jarvis_brain(tmp_path, monkeypatch, {})
import jarvis.reply.engine as engine
def slow_prepare(cfg, text, mem, chat_available):
_time.sleep(0.4)
return plan
monkeypatch.setattr(engine, "prepare_intent", slow_prepare)
t = threading.Thread(target=b.prepare, args=("100", "자비스 날씨 어때", False))
t.start()
_time.sleep(0.1)
b.reply("100", "자비스 날씨 어때?", HistoryContext())
t.join()
assert seen["prepared"] == [plan]