feat(transcribe,texts): voice notes and round videos to the bot come with a transcript, beavergram finds them by sha256 and telegram premium transcribes

tests/test_transcribe.py

@@ -0,0 +1,179 @@
+import hashlib
+import tempfile
+from pathlib import Path
+
+import pytest
+from aiohttp import web
+from aiohttp.test_utils import TestServer
+from beaver_gateway.transcribe import Audio
+
+from beaver_agent.transcribe import BeavergramTranscriber
+
+OWNER = 777
+VOICE = b"OggS voice"
+DIGEST = hashlib.sha256(VOICE).hexdigest()
+
+
+class Beavergram:
+    """Поддельный beavergram: аккаунты, поиск по sha256 и расшифровка."""
+
+    def __init__(self, *, empty_lookups=0, kind="voice", text=None, fail=None):
+        self.empty_lookups = empty_lookups
+        self.kind = kind
+        self.text = text
+        self.fail = fail
+        self.lookups = []
+        self.transcribes = []
+        self.auth = set()
+
+    def app(self):
+        app = web.Application(middlewares=[self.gate])
+        app.router.add_get("/api/accounts", self.accounts)
+        app.router.add_get("/api/media/lookup", self.lookup)
+        app.router.add_post("/api/media/transcribe", self.transcribe)
+        return app
+
+    @web.middleware
+    async def gate(self, request, handler):
+        self.auth.add(request.headers.get("Authorization"))
+        if self.fail:
+            return web.json_response({"detail": "nope"}, status=self.fail)
+        return await handler(request)
+
+    async def accounts(self, request):
+        return web.json_response(
+            [
+                {"account_id": 1, "tg_user_id": 5, "is_bot": True},
+                {"account_id": 2, "tg_user_id": OWNER, "is_bot": False},
+            ]
+        )
+
+    async def lookup(self, request):
+        self.lookups.append(dict(request.query))
+        if len(self.lookups) <= self.empty_lookups:
+            return web.json_response([])
+        return web.json_response(
+            [
+                {
+                    "account_id": 2,
+                    "chat_id": 42,
+                    "message_id": 9,
+                    "kind": self.kind,
+                    "extracted_text": self.text,
+                }
+            ]
+        )
+
+    async def transcribe(self, request):
+        self.transcribes.append(await request.json())
+        return web.json_response({"status": "done", "text": "привет из премиума"})
+
+
+@pytest.fixture
+def voice():
+    path = Path(tempfile.mkdtemp(prefix="beaver-bg-")) / "1-vu1.ogg"
+    path.write_bytes(VOICE)
+    return path
+
+
+async def serve(fake):
+    server = TestServer(fake.app())
+    await server.start_server()
+    return server
+
+
+def transcriber(server):
+    return BeavergramTranscriber(
+        origin=str(server.make_url("")).rstrip("/"),
+        token="t0k",
+        tg_user_id=OWNER,
+        find_for=2.0,
+        find_every=0.01,
+    )
+
+
+async def test_finds_the_voice_after_a_retry_and_asks_for_a_transcript(voice):
+    fake = Beavergram(empty_lookups=2)
+    server = await serve(fake)
+    try:
+        stt = transcriber(server)
+        assert await stt(Audio(voice, "audio/ogg", 3)) == "привет из премиума"
+    finally:
+        await server.close()
+    assert len(fake.lookups) == 3
+    assert fake.lookups[-1] == {"sha256": DIGEST, "account_id": "2"}
+    assert fake.transcribes == [
+        {"account_id": 2, "chat_id": 42, "message_id": 9, "wait": 60}
+    ]
+    assert fake.auth == {"Bearer t0k"}
+
+
+async def test_an_existing_transcript_comes_back_without_a_transcribe(voice):
+    fake = Beavergram(kind="video_note", text="уже расшифровано")
+    server = await serve(fake)
+    try:
+        assert (
+            await transcriber(server)(Audio(voice, "video/mp4")) == "уже расшифровано"
+        )
+    finally:
+        await server.close()
+    assert fake.transcribes == []
+
+
+async def test_music_and_other_files_are_not_touched(voice):
+    fake = Beavergram()
+    server = await serve(fake)
+    try:
+        stt = transcriber(server)
+        for media_type in ("audio/mpeg", "audio/mp4", None):
+            assert await stt(Audio(voice, media_type)) is None
+        assert fake.lookups == []
+        fake.kind = "audio"
+        assert await stt(Audio(voice, "audio/ogg")) is None
+    finally:
+        await server.close()
+    assert len(fake.lookups) == 1 and fake.transcribes == []
+
+
+async def test_nothing_found_in_time_is_none(voice):
+    fake = Beavergram(empty_lookups=1000)
+    server = await serve(fake)
+    try:
+        stt = transcriber(server)
+        stt.find_for = 0.05
+        assert await stt(Audio(voice, "audio/ogg")) is None
+    finally:
+        await server.close()
+    assert len(fake.lookups) >= 2 and fake.transcribes == []
+
+
+@pytest.mark.parametrize("status", [401, 500])
+async def test_http_errors_are_none_not_raised(voice, status):
+    fake = Beavergram(fail=status)
+    server = await serve(fake)
+    try:
+        assert await transcriber(server)(Audio(voice, "audio/ogg")) is None
+    finally:
+        await server.close()
+    assert fake.auth == {"Bearer t0k"}
+
+
+def test_from_env_needs_the_token_and_the_owner():
+    url = "https://tg.example.com/mcp?token=" + "a" * 64
+    stt = BeavergramTranscriber.from_env(
+        {"BEAVERGRAM_MCP": url, "TELEGRAM_USER_ID": "777"}
+    )
+    assert stt is not None
+    assert (stt.origin, stt.token, stt.tg_user_id) == (
+        "https://tg.example.com",
+        "a" * 64,
+        777,
+    )
+    assert BeavergramTranscriber.from_env({"TELEGRAM_USER_ID": "777"}) is None
+    assert (
+        BeavergramTranscriber.from_env(
+            {"BEAVERGRAM_MCP": "https://tg.example.com/mcp", "TELEGRAM_USER_ID": "777"}
+        )
+        is None
+    )
+    assert BeavergramTranscriber.from_env({"BEAVERGRAM_MCP": url}) is None