wx44wx/AI-VideoAssistant
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Enhance session management and logging configuration

Browse Source 
- Updated .env.example to clarify audio frame size validation and default codec settings.
- Refactored logging setup in main.py to support JSON serialization based on log format configuration.
- Improved session.py to dynamically compute audio frame bytes and include protocol version in session events.
- Added tests to validate session start events and audio frame handling based on chunk size settings.
This commit is contained in:
Xin Wang
2026-03-05 21:44:23 +08:00
parent 1cecbaa172
commit 6b589a1b7c
4 changed files with 105 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
engine/core/session.py
View File

@@ -54,7 +54,7 @@ class Session:
TRACK_AUDIO_IN = "audio_in"
TRACK_AUDIO_OUT = "audio_out"
TRACK_CONTROL = "control"
AUDIO_FRAME_BYTES = 640 # 16k mono pcm_s16le, 20ms
AUDIO_FRAME_BYTES = 640 # Legacy fallback: 16k mono pcm_s16le, 20ms
_METADATA_ALLOWED_TOP_LEVEL_KEYS = {
"overrides",
"dynamicVariables",
@@ -111,6 +111,7 @@ class Session:
self.id = session_id
self.transport = transport
self.use_duplex = use_duplex if use_duplex is not None else settings.duplex_enabled
self.audio_frame_bytes = self._compute_audio_frame_bytes()
self._assistant_id = str(assistant_id or "").strip() or None
self._backend_gateway = backend_gateway or build_backend_adapter_from_settings()
self._history_bridge = SessionHistoryBridge(
@@ -210,11 +211,14 @@ class Session:
)
return
frame_bytes = self.AUDIO_FRAME_BYTES
frame_bytes = getattr(self, "audio_frame_bytes", self._compute_audio_frame_bytes())
if len(audio_bytes) % frame_bytes != 0:
await self._send_error(
"client",
f"Audio frame size must be a multiple of {frame_bytes} bytes (20ms PCM)",
(
f"Audio frame size must be a multiple of {frame_bytes} bytes "
f"({settings.chunk_size_ms}ms PCM @ {settings.sample_rate}Hz)"
),
"audio.frame_size_mismatch",
stage="audio",
retryable=False,
@@ -384,6 +388,7 @@ class Session:
ev(
"session.started",
trackId=self.current_track_id,
protocolVersion=self._public_ws_protocol_version(),
tracks={
"audio_in": self.TRACK_AUDIO_IN,
"audio_out": self.TRACK_AUDIO_OUT,
@@ -1137,6 +1142,7 @@ class Session:
output_mode = str(runtime_output.get("mode") or "").strip().lower() if isinstance(runtime_output, dict) else ""
if output_mode not in {"audio", "text"}:
output_mode = "audio"
output_codec = str(runtime_output.get("codec") or settings.default_codec or "pcm").strip().lower() or "pcm"
tools_allowlist: List[str] = []
runtime_tools = runtime.get("tools", {}) if isinstance(runtime, dict) else {}
@@ -1146,7 +1152,11 @@ class Session:
tools_allowlist = [str(item) for item in allowlist if item is not None and str(item).strip()]
resolved: Dict[str, Any] = {
"output": {"mode": output_mode},
"protocolVersion": self._public_ws_protocol_version(),
"output": {
"mode": output_mode,
"codec": output_codec,
},
"tools": {
"enabled": bool(tools_allowlist),
"count": len(tools_allowlist),
@@ -1162,6 +1172,24 @@ class Session:
return resolved
@staticmethod
def _compute_audio_frame_bytes() -> int:
"""Compute expected PCM frame bytes from SAMPLE_RATE and CHUNK_SIZE_MS."""
sample_rate = max(1, int(getattr(settings, "sample_rate", 16000)))
chunk_ms = max(1, int(getattr(settings, "chunk_size_ms", 20)))
bytes_per_frame = int(round(sample_rate * 2 * (chunk_ms / 1000.0)))
if bytes_per_frame < 2:
bytes_per_frame = 2
if bytes_per_frame % 2 != 0:
bytes_per_frame += 1
return bytes_per_frame
@staticmethod
def _public_ws_protocol_version() -> str:
"""Return public protocol version label announced to clients."""
version = str(getattr(settings, "ws_protocol_version", "v1") or "v1").strip()
return version or "v1"
def _extract_json_obj(self, text: str) -> Optional[Dict[str, Any]]:
"""Best-effort extraction of a JSON object from freeform text."""
try: