Use decoupled way for backend client

tests/test_backend_adapters.py

@@ -0,0 +1,150 @@
+import aiohttp
+import pytest
+
+from app.backend_adapters import (
+    HistoryDisabledBackendAdapter,
+    HttpBackendAdapter,
+    NullBackendAdapter,
+    build_backend_adapter,
+)
+
+
+@pytest.mark.asyncio
+async def test_build_backend_adapter_without_url_returns_null_adapter():
+    adapter = build_backend_adapter(
+        backend_url=None,
+        backend_mode="auto",
+        history_enabled=True,
+        timeout_sec=3,
+    )
+    assert isinstance(adapter, NullBackendAdapter)
+
+    assert await adapter.fetch_assistant_config("assistant_1") is None
+    assert (
+        await adapter.create_call_record(
+            user_id=1,
+            assistant_id="assistant_1",
+            source="debug",
+        )
+        is None
+    )
+    assert (
+        await adapter.add_transcript(
+            call_id="call_1",
+            turn_index=0,
+            speaker="human",
+            content="hi",
+            start_ms=0,
+            end_ms=100,
+            confidence=0.9,
+            duration_ms=100,
+        )
+        is False
+    )
+    assert (
+        await adapter.finalize_call_record(
+            call_id="call_1",
+            status="connected",
+            duration_seconds=2,
+        )
+        is False
+    )
+    assert await adapter.search_knowledge_context(kb_id="kb_1", query="hello", n_results=3) == []
+    assert await adapter.fetch_tool_resource("tool_1") is None
+
+
+@pytest.mark.asyncio
+async def test_http_backend_adapter_create_call_record_posts_expected_payload(monkeypatch):
+    captured = {}
+
+    class _FakeResponse:
+        def __init__(self, status=200, payload=None):
+            self.status = status
+            self._payload = payload if payload is not None else {}
+
+        async def __aenter__(self):
+            return self
+
+        async def __aexit__(self, exc_type, exc, tb):
+            return None
+
+        async def json(self):
+            return self._payload
+
+        def raise_for_status(self):
+            if self.status >= 400:
+                raise RuntimeError("http_error")
+
+    class _FakeClientSession:
+        def __init__(self, timeout=None):
+            captured["timeout"] = timeout
+
+        async def __aenter__(self):
+            return self
+
+        async def __aexit__(self, exc_type, exc, tb):
+            return None
+
+        def post(self, url, json=None):
+            captured["url"] = url
+            captured["json"] = json
+            return _FakeResponse(status=200, payload={"id": "call_123"})
+
+    monkeypatch.setattr("app.backend_adapters.aiohttp.ClientSession", _FakeClientSession)
+
+    adapter = build_backend_adapter(
+        backend_url="http://localhost:8100",
+        backend_mode="auto",
+        history_enabled=True,
+        timeout_sec=7,
+    )
+    assert isinstance(adapter, HttpBackendAdapter)
+
+    call_id = await adapter.create_call_record(
+        user_id=99,
+        assistant_id="assistant_9",
+        source="debug",
+    )
+
+    assert call_id == "call_123"
+    assert captured["url"] == "http://localhost:8100/api/history"
+    assert captured["json"] == {
+        "user_id": 99,
+        "assistant_id": "assistant_9",
+        "source": "debug",
+        "status": "connected",
+    }
+    assert isinstance(captured["timeout"], aiohttp.ClientTimeout)
+    assert captured["timeout"].total == 7
+
+
+@pytest.mark.asyncio
+async def test_backend_mode_disabled_forces_null_even_with_url():
+    adapter = build_backend_adapter(
+        backend_url="http://localhost:8100",
+        backend_mode="disabled",
+        history_enabled=True,
+        timeout_sec=7,
+    )
+    assert isinstance(adapter, NullBackendAdapter)
+
+
+@pytest.mark.asyncio
+async def test_history_disabled_wraps_backend_adapter():
+    adapter = build_backend_adapter(
+        backend_url="http://localhost:8100",
+        backend_mode="auto",
+        history_enabled=False,
+        timeout_sec=7,
+    )
+    assert isinstance(adapter, HistoryDisabledBackendAdapter)
+    assert await adapter.create_call_record(user_id=1, assistant_id="a1", source="debug") is None
+    assert await adapter.add_transcript(
+        call_id="c1",
+        turn_index=0,
+        speaker="human",
+        content="hi",
+        start_ms=0,
+        end_ms=10,
+        duration_ms=10,
+    ) is False

tests/test_history_bridge.py

@@ -0,0 +1,147 @@
+import asyncio
+import time
+
+import pytest
+
+from core.history_bridge import SessionHistoryBridge
+
+
+class _FakeHistoryWriter:
+    def __init__(self, *, add_delay_s: float = 0.0, add_result: bool = True):
+        self.add_delay_s = add_delay_s
+        self.add_result = add_result
+        self.created_call_ids = []
+        self.transcripts = []
+        self.finalize_calls = 0
+        self.finalize_statuses = []
+        self.finalize_at = None
+        self.last_transcript_at = None
+
+    async def create_call_record(self, *, user_id: int, assistant_id: str | None, source: str = "debug"):
+        _ = (user_id, assistant_id, source)
+        call_id = "call_test_1"
+        self.created_call_ids.append(call_id)
+        return call_id
+
+    async def add_transcript(
+        self,
+        *,
+        call_id: str,
+        turn_index: int,
+        speaker: str,
+        content: str,
+        start_ms: int,
+        end_ms: int,
+        confidence: float | None = None,
+        duration_ms: int | None = None,
+    ) -> bool:
+        _ = confidence
+        if self.add_delay_s > 0:
+            await asyncio.sleep(self.add_delay_s)
+        self.transcripts.append(
+            {
+                "call_id": call_id,
+                "turn_index": turn_index,
+                "speaker": speaker,
+                "content": content,
+                "start_ms": start_ms,
+                "end_ms": end_ms,
+                "duration_ms": duration_ms,
+            }
+        )
+        self.last_transcript_at = time.monotonic()
+        return self.add_result
+
+    async def finalize_call_record(self, *, call_id: str, status: str, duration_seconds: int) -> bool:
+        _ = (call_id, duration_seconds)
+        self.finalize_calls += 1
+        self.finalize_statuses.append(status)
+        self.finalize_at = time.monotonic()
+        return True
+
+@pytest.mark.asyncio
+async def test_slow_backend_does_not_block_enqueue():
+    writer = _FakeHistoryWriter(add_delay_s=0.15, add_result=True)
+    bridge = SessionHistoryBridge(
+        history_writer=writer,
+        enabled=True,
+        queue_max_size=32,
+        retry_max_attempts=0,
+        retry_backoff_sec=0.01,
+        finalize_drain_timeout_sec=1.0,
+    )
+
+    try:
+        call_id = await bridge.start_call(user_id=1, assistant_id="assistant_1", source="debug")
+        assert call_id == "call_test_1"
+
+        t0 = time.perf_counter()
+        queued = bridge.enqueue_turn(role="user", text="hello world")
+        elapsed_s = time.perf_counter() - t0
+
+        assert queued is True
+        assert elapsed_s < 0.02
+
+        await bridge.finalize(status="connected")
+        assert len(writer.transcripts) == 1
+        assert writer.finalize_calls == 1
+    finally:
+        await bridge.shutdown()
+
+
+@pytest.mark.asyncio
+async def test_failing_backend_retries_but_enqueue_remains_non_blocking():
+    writer = _FakeHistoryWriter(add_delay_s=0.01, add_result=False)
+    bridge = SessionHistoryBridge(
+        history_writer=writer,
+        enabled=True,
+        queue_max_size=32,
+        retry_max_attempts=2,
+        retry_backoff_sec=0.01,
+        finalize_drain_timeout_sec=0.5,
+    )
+
+    try:
+        await bridge.start_call(user_id=1, assistant_id="assistant_1", source="debug")
+        t0 = time.perf_counter()
+        assert bridge.enqueue_turn(role="assistant", text="retry me")
+        elapsed_s = time.perf_counter() - t0
+        assert elapsed_s < 0.02
+
+        await bridge.finalize(status="connected")
+
+        # Initial try + 2 retries
+        assert len(writer.transcripts) == 3
+        assert writer.finalize_calls == 1
+    finally:
+        await bridge.shutdown()
+
+
+@pytest.mark.asyncio
+async def test_finalize_is_idempotent_and_waits_for_queue_drain():
+    writer = _FakeHistoryWriter(add_delay_s=0.05, add_result=True)
+    bridge = SessionHistoryBridge(
+        history_writer=writer,
+        enabled=True,
+        queue_max_size=32,
+        retry_max_attempts=0,
+        retry_backoff_sec=0.01,
+        finalize_drain_timeout_sec=1.0,
+    )
+
+    try:
+        await bridge.start_call(user_id=1, assistant_id="assistant_1", source="debug")
+        assert bridge.enqueue_turn(role="user", text="first")
+
+        ok_1 = await bridge.finalize(status="connected")
+        ok_2 = await bridge.finalize(status="connected")
+
+        assert ok_1 is True
+        assert ok_2 is True
+        assert len(writer.transcripts) == 1
+        assert writer.finalize_calls == 1
+        assert writer.last_transcript_at is not None
+        assert writer.finalize_at is not None
+        assert writer.finalize_at >= writer.last_transcript_at
+    finally:
+        await bridge.shutdown()