diff --git a/circuitforge_core/audio/__init__.py b/circuitforge_core/audio/__init__.py
new file mode 100644
index 0000000..b8ceebb
--- /dev/null
+++ b/circuitforge_core/audio/__init__.py
@@ -0,0 +1,29 @@
+"""
+circuitforge_core.audio — shared PCM and audio signal utilities.
+
+MIT licensed. No model weights. No HuggingFace. Dependency: numpy only
+(scipy optional for high-quality resampling).
+
+Consumers:
+  cf-voice   — replaces hand-rolled PCM conversion in stt.py / context.py
+  Sparrow    — torchaudio stitching, export, acoustic analysis
+  Avocet     — audio preprocessing for classifier training corpus
+  Linnet     — chunk accumulation for real-time tone annotation
+"""
+from circuitforge_core.audio.convert import (
+    bytes_to_float32,
+    float32_to_pcm,
+    pcm_to_float32,
+)
+from circuitforge_core.audio.gate import is_silent
+from circuitforge_core.audio.resample import resample
+from circuitforge_core.audio.buffer import ChunkAccumulator
+
+__all__ = [
+    "bytes_to_float32",
+    "float32_to_pcm",
+    "pcm_to_float32",
+    "is_silent",
+    "resample",
+    "ChunkAccumulator",
+]
diff --git a/circuitforge_core/audio/buffer.py b/circuitforge_core/audio/buffer.py
new file mode 100644
index 0000000..aaba3c2
--- /dev/null
+++ b/circuitforge_core/audio/buffer.py
@@ -0,0 +1,67 @@
+"""
+ChunkAccumulator — collect fixed-size audio chunks into a classify window.
+
+Used by cf-voice and Linnet to gather N × 100ms frames before firing
+a classification pass. The window size trades latency against context:
+a 2-second window (20 × 100ms) gives the classifier enough signal to
+detect tone/affect reliably without lagging the conversation.
+"""
+from __future__ import annotations
+
+from collections import deque
+
+import numpy as np
+
+
+class ChunkAccumulator:
+    """Accumulate audio chunks and flush when the window is full.
+
+    Args:
+        window_chunks: Number of chunks to collect before is_ready() is True.
+        dtype:         numpy dtype of the accumulated array. Default float32.
+    """
+
+    def __init__(self, window_chunks: int, *, dtype: np.dtype = np.float32) -> None:
+        if window_chunks < 1:
+            raise ValueError(f"window_chunks must be >= 1, got {window_chunks}")
+        self._window = window_chunks
+        self._dtype = dtype
+        self._buf: deque[np.ndarray] = deque()
+
+    def accumulate(self, chunk: np.ndarray) -> None:
+        """Add a chunk to the buffer. Oldest chunks are dropped once the
+        buffer exceeds window_chunks to bound memory."""
+        self._buf.append(chunk.astype(self._dtype))
+        while len(self._buf) > self._window:
+            self._buf.popleft()
+
+    def is_ready(self) -> bool:
+        """True when window_chunks have been accumulated."""
+        return len(self._buf) >= self._window
+
+    def flush(self) -> np.ndarray:
+        """Concatenate accumulated chunks and reset the buffer.
+
+        Returns:
+            float32 ndarray of concatenated audio.
+
+        Raises:
+            RuntimeError: if fewer than window_chunks have been accumulated.
+        """
+        if not self.is_ready():
+            raise RuntimeError(
+                f"Not enough chunks accumulated: have {len(self._buf)}, "
+                f"need {self._window}. Check is_ready() before calling flush()."
+            )
+        result = np.concatenate(list(self._buf), axis=-1).astype(self._dtype)
+        self._buf.clear()
+        return result
+
+    def reset(self) -> None:
+        """Discard all buffered audio without returning it."""
+        self._buf.clear()
+
+    @property
+    def chunk_count(self) -> int:
+        """Current number of buffered chunks."""
+        return len(self._buf)
diff --git a/circuitforge_core/audio/convert.py b/circuitforge_core/audio/convert.py
new file mode 100644
index 0000000..ed0e23e
--- /dev/null
+++ b/circuitforge_core/audio/convert.py
@@ -0,0 +1,50 @@
+"""
+PCM / float32 conversion utilities.
+
+All functions operate on raw audio bytes or numpy arrays. No torch dependency.
+
+Standard pipeline:
+    bytes (int16 PCM) -> float32 ndarray -> signal processing -> bytes (int16 PCM)
+"""
+from __future__ import annotations
+
+import numpy as np
+
+
+def pcm_to_float32(pcm_bytes: bytes, *, dtype: np.dtype = np.int16) -> np.ndarray:
+    """Convert raw PCM bytes to a float32 numpy array in [-1.0, 1.0].
+
+    Args:
+        pcm_bytes: Raw PCM audio bytes.
+        dtype:     Sample dtype of the input. Default: int16 (standard mic input).
+
+    Returns:
+        float32 ndarray, values in [-1.0, 1.0].
+    """
+    scale = np.iinfo(dtype).max
+    return np.frombuffer(pcm_bytes, dtype=dtype).astype(np.float32) / scale
+
+
+def bytes_to_float32(pcm_bytes: bytes) -> np.ndarray:
+    """Alias for pcm_to_float32 with default int16 dtype.
+
+    Matches the naming used in cf-voice context.py for easier migration.
+    """
+    return pcm_to_float32(pcm_bytes)
+
+
+def float32_to_pcm(audio: np.ndarray, *, dtype: np.dtype = np.int16) -> bytes:
+    """Convert a float32 ndarray in [-1.0, 1.0] to raw PCM bytes.
+
+    Clips to [-1.0, 1.0] before scaling to prevent wraparound distortion.
+
+    Args:
+        audio: float32 ndarray, values nominally in [-1.0, 1.0].
+        dtype: Target PCM sample dtype. Default: int16.
+
+    Returns:
+        Raw PCM bytes.
+    """
+    scale = np.iinfo(dtype).max
+    clipped = np.clip(audio, -1.0, 1.0)
+    return (clipped * scale).astype(dtype).tobytes()
diff --git a/circuitforge_core/audio/gate.py b/circuitforge_core/audio/gate.py
new file mode 100644
index 0000000..7843bd3
--- /dev/null
+++ b/circuitforge_core/audio/gate.py
@@ -0,0 +1,44 @@
+"""
+Energy gate — silence detection via RMS amplitude.
+"""
+from __future__ import annotations
+
+import numpy as np
+
+# Default threshold extracted from cf-voice stt.py.
+# Signals below this RMS level are considered silent.
+_DEFAULT_RMS_THRESHOLD = 0.005
+
+
+def is_silent(
+    audio: np.ndarray,
+    *,
+    rms_threshold: float = _DEFAULT_RMS_THRESHOLD,
+) -> bool:
+    """Return True when the audio clip is effectively silent.
+
+    Uses root-mean-square amplitude as the energy estimate. This is a fast
+    frame-level gate — not a VAD model. Use it to skip inference on empty
+    audio frames before they hit a more expensive transcription or
+    classification pipeline.
+
+    Args:
+        audio:         float32 ndarray, values in [-1.0, 1.0].
+        rms_threshold: Clips with RMS below this value are silent.
+                       Default 0.005 is conservative — genuine speech at
+                       normal mic levels sits well above this.
+
+    Returns:
+        True if silent, False if the clip contains meaningful signal.
+    """
+    if audio.size == 0:
+        return True
+    rms = float(np.sqrt(np.mean(audio.astype(np.float32) ** 2)))
+    return rms < rms_threshold
+
+
+def rms(audio: np.ndarray) -> float:
+    """Return the RMS amplitude of an audio array."""
+    if audio.size == 0:
+        return 0.0
+    return float(np.sqrt(np.mean(audio.astype(np.float32) ** 2)))
diff --git a/circuitforge_core/audio/resample.py b/circuitforge_core/audio/resample.py
new file mode 100644
index 0000000..2a36446
--- /dev/null
+++ b/circuitforge_core/audio/resample.py
@@ -0,0 +1,39 @@
+"""
+Audio resampling — change sample rate of a float32 audio array.
+
+Uses scipy.signal.resample_poly when available (high-quality, anti-aliased).
+Falls back to linear interpolation via numpy when scipy is absent — acceptable
+for 16kHz speech but not for music.
+"""
+from __future__ import annotations
+
+import numpy as np
+
+
+def resample(audio: np.ndarray, from_hz: int, to_hz: int) -> np.ndarray:
+    """Resample audio from one sample rate to another.
+
+    Args:
+        audio:   float32 ndarray, shape (samples,) or (channels, samples).
+        from_hz: Source sample rate in Hz.
+        to_hz:   Target sample rate in Hz.
+
+    Returns:
+        Resampled float32 ndarray at to_hz.
+    """
+    if from_hz == to_hz:
+        return audio.astype(np.float32)
+
+    try:
+        from scipy.signal import resample_poly  # type: ignore[import]
+        from math import gcd
+        g = gcd(from_hz, to_hz)
+        up, down = to_hz // g, from_hz // g
+        return resample_poly(audio.astype(np.float32), up, down, axis=-1)
+    except ImportError:
+        # Numpy linear interpolation fallback — lower quality but no extra deps.
+        # Adequate for 16kHz ↔ 8kHz conversion on speech; avoid for music.
+        n_out = int(len(audio) * to_hz / from_hz)
+        x_old = np.linspace(0, 1, len(audio), endpoint=False)
+        x_new = np.linspace(0, 1, n_out, endpoint=False)
+        return np.interp(x_new, x_old, audio.astype(np.float32)).astype(np.float32)
diff --git a/circuitforge_core/db/base.py b/circuitforge_core/db/base.py
index e3bbf2e..9de3212 100644
--- a/circuitforge_core/db/base.py
+++ b/circuitforge_core/db/base.py
@@ -23,7 +23,7 @@ def get_connection(db_path: Path, key: str = "") -> sqlite3.Connection:
     if cloud_mode and key:
         from pysqlcipher3 import dbapi2 as _sqlcipher  # type: ignore
         conn = _sqlcipher.connect(str(db_path), timeout=30)
-        conn.execute(f"PRAGMA key='{key}'")
+        conn.execute("PRAGMA key=?", (key,))
         return conn
     # timeout=30: retry for up to 30s when another writer holds the lock (WAL mode
     # allows concurrent readers but only one writer at a time).
diff --git a/tests/test_audio/__init__.py b/tests/test_audio/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/tests/test_audio/test_buffer.py b/tests/test_audio/test_buffer.py
new file mode 100644
index 0000000..2b99ab0
--- /dev/null
+++ b/tests/test_audio/test_buffer.py
@@ -0,0 +1,72 @@
+import numpy as np
+import pytest
+from circuitforge_core.audio.buffer import ChunkAccumulator
+
+
+def _chunk(value: float = 0.0, size: int = 1600) -> np.ndarray:
+    return np.full(size, value, dtype=np.float32)
+
+
+def test_not_ready_initially():
+    acc = ChunkAccumulator(window_chunks=3)
+    assert acc.is_ready() is False
+    assert acc.chunk_count == 0
+
+
+def test_ready_after_window_filled():
+    acc = ChunkAccumulator(window_chunks=3)
+    for _ in range(3):
+        acc.accumulate(_chunk())
+    assert acc.is_ready() is True
+    assert acc.chunk_count == 3
+
+
+def test_flush_returns_concatenated():
+    acc = ChunkAccumulator(window_chunks=2)
+    acc.accumulate(_chunk(0.1, size=100))
+    acc.accumulate(_chunk(0.2, size=100))
+    result = acc.flush()
+    assert result.shape == (200,)
+    assert np.allclose(result[:100], 0.1)
+    assert np.allclose(result[100:], 0.2)
+
+
+def test_flush_clears_buffer():
+    acc = ChunkAccumulator(window_chunks=2)
+    acc.accumulate(_chunk())
+    acc.accumulate(_chunk())
+    acc.flush()
+    assert acc.chunk_count == 0
+    assert acc.is_ready() is False
+
+
+def test_flush_raises_when_not_ready():
+    acc = ChunkAccumulator(window_chunks=3)
+    acc.accumulate(_chunk())
+    with pytest.raises(RuntimeError, match="Not enough chunks"):
+        acc.flush()
+
+
+def test_reset_clears_buffer():
+    acc = ChunkAccumulator(window_chunks=2)
+    acc.accumulate(_chunk())
+    acc.accumulate(_chunk())
+    acc.reset()
+    assert acc.chunk_count == 0
+
+
+def test_oldest_dropped_when_overfilled():
+    # Accumulate more than window_chunks — oldest should be evicted
+    acc = ChunkAccumulator(window_chunks=2)
+    acc.accumulate(_chunk(1.0, size=10))  # will be evicted
+    acc.accumulate(_chunk(2.0, size=10))
+    acc.accumulate(_chunk(3.0, size=10))
+    assert acc.chunk_count == 2
+    result = acc.flush()
+    assert np.allclose(result[:10], 2.0)
+    assert np.allclose(result[10:], 3.0)
+
+
+def test_invalid_window_raises():
+    with pytest.raises(ValueError):
+        ChunkAccumulator(window_chunks=0)
diff --git a/tests/test_audio/test_convert.py b/tests/test_audio/test_convert.py
new file mode 100644
index 0000000..6052d52
--- /dev/null
+++ b/tests/test_audio/test_convert.py
@@ -0,0 +1,57 @@
+import numpy as np
+import pytest
+from circuitforge_core.audio.convert import pcm_to_float32, float32_to_pcm, bytes_to_float32
+
+
+def _silence_pcm(n_samples: int = 1024) -> bytes:
+    return (np.zeros(n_samples, dtype=np.int16)).tobytes()
+
+
+def _sine_pcm(freq_hz: float = 440.0, sample_rate: int = 16000, duration_s: float = 0.1) -> bytes:
+    t = np.linspace(0, duration_s, int(sample_rate * duration_s), endpoint=False)
+    samples = (np.sin(2 * np.pi * freq_hz * t) * 16000).astype(np.int16)
+    return samples.tobytes()
+
+
+def test_pcm_to_float32_silence():
+    result = pcm_to_float32(_silence_pcm())
+    assert result.dtype == np.float32
+    assert np.allclose(result, 0.0)
+
+
+def test_pcm_to_float32_range():
+    # Full-scale positive int16 -> ~1.0
+    pcm = np.array([32767], dtype=np.int16).tobytes()
+    result = pcm_to_float32(pcm)
+    assert abs(result[0] - 1.0) < 0.001
+
+    # Full-scale negative int16 -> ~-1.0
+    pcm = np.array([-32768], dtype=np.int16).tobytes()
+    result = pcm_to_float32(pcm)
+    assert abs(result[0] - (-32768 / 32767)) < 0.001
+
+
+def test_pcm_roundtrip():
+    original = _sine_pcm()
+    float_audio = pcm_to_float32(original)
+    recovered = float32_to_pcm(float_audio)
+    # Roundtrip through float32 introduces tiny quantisation error — within 1 LSB
+    orig_arr = np.frombuffer(original, dtype=np.int16)
+    recv_arr = np.frombuffer(recovered, dtype=np.int16)
+    assert np.max(np.abs(orig_arr.astype(np.int32) - recv_arr.astype(np.int32))) <= 1
+
+
+def test_float32_to_pcm_clips():
+    # Values outside [-1.0, 1.0] must be clipped, not wrap.
+    # int16 is asymmetric: max=32767, min=-32768. Scaling by 32767 means
+    # -1.0 → -32767, not -32768 — that's expected and correct.
+    audio = np.array([2.0, -2.0], dtype=np.float32)
+    result = float32_to_pcm(audio)
+    samples = np.frombuffer(result, dtype=np.int16)
+    assert samples[0] == 32767
+    assert samples[1] == -32767
+
+
+def test_bytes_to_float32_alias():
+    pcm = _sine_pcm()
+    assert np.allclose(bytes_to_float32(pcm), pcm_to_float32(pcm))
diff --git a/tests/test_audio/test_gate.py b/tests/test_audio/test_gate.py
new file mode 100644
index 0000000..16c0ebb
--- /dev/null
+++ b/tests/test_audio/test_gate.py
@@ -0,0 +1,38 @@
+import numpy as np
+from circuitforge_core.audio.gate import is_silent, rms
+
+
+def test_silence_detected():
+    audio = np.zeros(1600, dtype=np.float32)
+    assert is_silent(audio) is True
+
+
+def test_speech_level_not_silent():
+    # Sine at 0.1 amplitude — well above 0.005 RMS
+    t = np.linspace(0, 0.1, 1600, endpoint=False)
+    audio = (np.sin(2 * np.pi * 440 * t) * 0.1).astype(np.float32)
+    assert is_silent(audio) is False
+
+
+def test_just_below_threshold():
+    # RMS exactly at 0.004 — should be silent
+    audio = np.full(1600, 0.004, dtype=np.float32)
+    assert is_silent(audio, rms_threshold=0.005) is True
+
+
+def test_just_above_threshold():
+    audio = np.full(1600, 0.006, dtype=np.float32)
+    assert is_silent(audio, rms_threshold=0.005) is False
+
+
+def test_empty_array_is_silent():
+    assert is_silent(np.array([], dtype=np.float32)) is True
+
+
+def test_rms_zero_for_silence():
+    assert rms(np.zeros(100, dtype=np.float32)) == 0.0
+
+
+def test_rms_nonzero_for_signal():
+    audio = np.ones(100, dtype=np.float32) * 0.5
+    assert abs(rms(audio) - 0.5) < 1e-6
diff --git a/tests/test_audio/test_resample.py b/tests/test_audio/test_resample.py
new file mode 100644
index 0000000..bbab4bd
--- /dev/null
+++ b/tests/test_audio/test_resample.py
@@ -0,0 +1,41 @@
+import numpy as np
+import pytest
+from circuitforge_core.audio.resample import resample
+
+
+def _sine(freq_hz: float, sample_rate: int, duration_s: float = 0.5) -> np.ndarray:
+    t = np.linspace(0, duration_s, int(sample_rate * duration_s), endpoint=False)
+    return np.sin(2 * np.pi * freq_hz * t).astype(np.float32)
+
+
+def test_same_rate_is_noop():
+    audio = _sine(440.0, 16000)
+    result = resample(audio, 16000, 16000)
+    assert np.allclose(result, audio, atol=1e-5)
+
+
+def test_output_length_correct():
+    audio = _sine(440.0, 16000, duration_s=1.0)
+    result = resample(audio, 16000, 8000)
+    assert len(result) == 8000
+
+
+def test_upsample_output_length():
+    audio = _sine(440.0, 8000, duration_s=1.0)
+    result = resample(audio, 8000, 16000)
+    assert len(result) == 16000
+
+
+def test_output_dtype_float32():
+    audio = _sine(440.0, 16000)
+    result = resample(audio, 16000, 8000)
+    assert result.dtype == np.float32
+
+
+def test_energy_preserved_approximately():
+    # RMS should be approximately the same after resampling a sine wave
+    audio = _sine(440.0, 16000, duration_s=1.0)
+    result = resample(audio, 16000, 8000)
+    rms_in = float(np.sqrt(np.mean(audio ** 2)))
+    rms_out = float(np.sqrt(np.mean(result ** 2)))
+    assert abs(rms_in - rms_out) < 0.05  # within 5% — resampling is not power-preserving exactly
diff --git a/tests/test_db.py b/tests/test_db.py
index 1b2a018..aa0362e 100644
--- a/tests/test_db.py
+++ b/tests/test_db.py
@@ -1,9 +1,15 @@
+import os
 import sqlite3
 import tempfile
 from pathlib import Path
 import pytest
 from circuitforge_core.db import get_connection, run_migrations
 
+sqlcipher_available = pytest.mark.skipif(
+    __import__("importlib").util.find_spec("pysqlcipher3") is None,
+    reason="pysqlcipher3 not installed",
+)
+
 
 def test_get_connection_returns_sqlite_connection(tmp_path):
     db = tmp_path / "test.db"
@@ -61,3 +67,38 @@ def test_run_migrations_applies_in_order(tmp_path):
     run_migrations(conn, migrations_dir)
     conn.execute("INSERT INTO foo (name) VALUES ('bar')")
     conn.close()
+
+
+# ── SQLCipher PRAGMA key tests (skipped when pysqlcipher3 not installed) ──────
+
+
+@sqlcipher_available
+def test_sqlcipher_key_with_special_chars_does_not_inject(tmp_path, monkeypatch):
+    """Key containing a single quote must not cause a SQL syntax error.
+
+    Regression for: conn.execute(f"PRAGMA key='{key}'") — if key = "x'--"
+    the f-string form produced a broken PRAGMA statement. Parameterized
+    form (PRAGMA key=?) must handle this safely.
+    """
+    monkeypatch.setenv("CLOUD_MODE", "1")
+    db = tmp_path / "enc.db"
+    tricky_key = "pass'word\"--inject"
+    # Must not raise; if the f-string form were used, this would produce
+    # a syntax error or silently set an incorrect key.
+    conn = get_connection(db, key=tricky_key)
+    conn.execute("CREATE TABLE t (x INTEGER)")
+    conn.close()
+
+
+@sqlcipher_available
+def test_sqlcipher_wrong_key_raises(tmp_path, monkeypatch):
+    """Opening an encrypted DB with the wrong key should raise, not silently corrupt."""
+    monkeypatch.setenv("CLOUD_MODE", "1")
+    db = tmp_path / "enc.db"
+    conn = get_connection(db, key="correct-key")
+    conn.execute("CREATE TABLE t (x INTEGER)")
+    conn.close()
+
+    with pytest.raises(Exception):
+        bad = get_connection(db, key="wrong-key")
+        bad.execute("SELECT * FROM t")  # should raise on bad key
diff --git a/tests/test_musicgen/__init__.py b/tests/test_musicgen/__init__.py
new file mode 100644
index 0000000..e69de29
diff --git a/tests/test_musicgen/test_app.py b/tests/test_musicgen/test_app.py
new file mode 100644
index 0000000..84dc2da
--- /dev/null
+++ b/tests/test_musicgen/test_app.py
@@ -0,0 +1,124 @@
+"""
+Tests for the cf-musicgen FastAPI app using mock backend.
+"""
+import io
+import os
+import wave
+
+import pytest
+from fastapi.testclient import TestClient
+
+import circuitforge_core.musicgen.app as musicgen_app
+from circuitforge_core.musicgen.backends.mock import MockMusicGenBackend
+
+
+@pytest.fixture(autouse=True)
+def inject_mock_backend():
+    """Inject mock backend before each test; restore None after."""
+    original = musicgen_app._backend
+    musicgen_app._backend = MockMusicGenBackend()
+    yield
+    musicgen_app._backend = original
+
+
+@pytest.fixture()
+def client():
+    return TestClient(musicgen_app.app)
+
+
+def _silent_wav(duration_s: float = 1.0) -> bytes:
+    n = int(duration_s * 16000)
+    buf = io.BytesIO()
+    with wave.open(buf, "wb") as wf:
+        wf.setnchannels(1)
+        wf.setsampwidth(2)
+        wf.setframerate(16000)
+        wf.writeframes(b"\x00\x00" * n)
+    return buf.getvalue()
+
+
+# ── /health ───────────────────────────────────────────────────────────────────
+
+
+def test_health_returns_ok(client):
+    resp = client.get("/health")
+    assert resp.status_code == 200
+    data = resp.json()
+    assert data["status"] == "ok"
+    assert data["model"] == "mock"
+    assert data["vram_mb"] == 0
+
+
+def test_health_503_when_no_backend(client):
+    musicgen_app._backend = None
+    resp = client.get("/health")
+    assert resp.status_code == 503
+
+
+# ── /continue ─────────────────────────────────────────────────────────────────
+
+
+def test_continue_returns_audio(client):
+    resp = client.post(
+        "/continue",
+        data={"duration_s": "5.0"},
+        files={"audio": ("test.wav", _silent_wav(), "audio/wav")},
+    )
+    assert resp.status_code == 200
+    assert resp.headers["content-type"] == "audio/wav"
+
+
+def test_continue_duration_header(client):
+    resp = client.post(
+        "/continue",
+        data={"duration_s": "7.0"},
+        files={"audio": ("test.wav", _silent_wav(), "audio/wav")},
+    )
+    assert resp.status_code == 200
+    assert float(resp.headers["x-duration-s"]) == pytest.approx(7.0)
+
+
+def test_continue_model_header(client):
+    resp = client.post(
+        "/continue",
+        data={"duration_s": "5.0"},
+        files={"audio": ("test.wav", _silent_wav(), "audio/wav")},
+    )
+    assert resp.headers["x-model"] == "mock"
+
+
+def test_continue_rejects_zero_duration(client):
+    resp = client.post(
+        "/continue",
+        data={"duration_s": "0"},
+        files={"audio": ("test.wav", _silent_wav(), "audio/wav")},
+    )
+    assert resp.status_code == 422
+
+
+def test_continue_rejects_too_long_duration(client):
+    resp = client.post(
+        "/continue",
+        data={"duration_s": "61"},
+        files={"audio": ("test.wav", _silent_wav(), "audio/wav")},
+    )
+    assert resp.status_code == 422
+
+
+def test_continue_rejects_empty_audio(client):
+    resp = client.post(
+        "/continue",
+        data={"duration_s": "5.0"},
+        files={"audio": ("empty.wav", b"", "audio/wav")},
+    )
+    assert resp.status_code == 400
+
+
+def test_continue_503_when_no_backend(client):
+    musicgen_app._backend = None
+    resp = client.post(
+        "/continue",
+        data={"duration_s": "5.0"},
+        files={"audio": ("test.wav", _silent_wav(), "audio/wav")},
+    )
+    assert resp.status_code == 503
diff --git a/tests/test_musicgen/test_mock_backend.py b/tests/test_musicgen/test_mock_backend.py
new file mode 100644
index 0000000..c4fad49
--- /dev/null
+++ b/tests/test_musicgen/test_mock_backend.py
@@ -0,0 +1,97 @@
+"""
+Tests for the mock MusicGen backend and shared audio encode/decode utilities.
+
+All tests run without a GPU or AudioCraft install.
+"""
+import io
+import wave
+
+import pytest
+from circuitforge_core.musicgen.backends.base import (
+    MODEL_MELODY,
+    MODEL_SMALL,
+    MusicGenBackend,
+    MusicContinueResult,
+    make_musicgen_backend,
+)
+from circuitforge_core.musicgen.backends.mock import MockMusicGenBackend
+
+
+# ── Mock backend ──────────────────────────────────────────────────────────────
+
+
+def test_mock_satisfies_protocol():
+    backend = MockMusicGenBackend()
+    assert isinstance(backend, MusicGenBackend)
+
+
+def test_mock_model_name():
+    assert MockMusicGenBackend().model_name == "mock"
+
+
+def test_mock_vram_mb():
+    assert MockMusicGenBackend().vram_mb == 0
+
+
+def _silent_wav(duration_s: float = 1.0, sample_rate: int = 16000) -> bytes:
+    n = int(duration_s * sample_rate)
+    buf = io.BytesIO()
+    with wave.open(buf, "wb") as wf:
+        wf.setnchannels(1)
+        wf.setsampwidth(2)
+        wf.setframerate(sample_rate)
+        wf.writeframes(b"\x00\x00" * n)
+    return buf.getvalue()
+
+
+def test_mock_returns_result():
+    backend = MockMusicGenBackend()
+    result = backend.continue_audio(_silent_wav(), duration_s=5.0)
+    assert isinstance(result, MusicContinueResult)
+
+
+def test_mock_duration_matches_request():
+    backend = MockMusicGenBackend()
+    result = backend.continue_audio(_silent_wav(), duration_s=7.5)
+    assert result.duration_s == 7.5
+
+
+def test_mock_returns_valid_wav():
+    backend = MockMusicGenBackend()
+    result = backend.continue_audio(_silent_wav(), duration_s=2.0)
+    assert result.format == "wav"
+    buf = io.BytesIO(result.audio_bytes)
+    with wave.open(buf, "rb") as wf:
+        assert wf.getnframes() > 0
+
+
+def test_mock_sample_rate():
+    backend = MockMusicGenBackend()
+    result = backend.continue_audio(_silent_wav())
+    assert result.sample_rate == 32000
+
+
+def test_mock_prompt_duration_passthrough():
+    backend = MockMusicGenBackend()
+    result = backend.continue_audio(_silent_wav(), prompt_duration_s=8.0)
+    assert result.prompt_duration_s == 8.0
+
+
+def test_mock_description_ignored():
+    backend = MockMusicGenBackend()
+    # Should not raise regardless of description
+    result = backend.continue_audio(_silent_wav(), description="upbeat jazz")
+    assert result is not None
+
+
+# ── make_musicgen_backend factory ─────────────────────────────────────────────
+
+
+def test_factory_returns_mock_when_flag_set():
+    backend = make_musicgen_backend(mock=True)
+    assert isinstance(backend, MockMusicGenBackend)
+
+
+def test_factory_mock_for_mock_model_name():
+    backend = make_musicgen_backend(model_name="mock", mock=True)
+    assert isinstance(backend, MockMusicGenBackend)