From 04ba2614179b97256abde3ec343b238f7be641a6 Mon Sep 17 00:00:00 2001
From: tjc66666666 <3428979959@qq.com>
Date: Sat, 6 Jun 2026 13:39:30 +0800
Subject: [PATCH 1/4] Implement data URI handling for audio references

Add support for data URI audio references by decoding base64 audio data and saving it to a temporary file.
---
 .../core/provider/sources/openai_source.py    | 22 +++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/astrbot/core/provider/sources/openai_source.py b/astrbot/core/provider/sources/openai_source.py
index 8aa2778f1b..c8af63e7b5 100644
--- a/astrbot/core/provider/sources/openai_source.py
+++ b/astrbot/core/provider/sources/openai_source.py
@@ -367,6 +367,28 @@ async def _audio_ref_to_local_path(self, audio_ref: str) -> tuple[str, list[Path
             return str(target_path), cleanup_paths
         if audio_ref.startswith("file://"):
             return self._file_uri_to_path(audio_ref), cleanup_paths
+        if audio_ref.startswith("data:"):
+            # data URI 格式: data:audio/wav;base64,<base64_data>
+            # 或 data:audio/mp3;base64,<base64_data>
+            try:
+                header, base64_data = audio_ref.split(",", 1)
+                # 从 data URI header 中提取 MIME 类型和格式
+                # 例如 "data:audio/wav;base64" -> "wav"
+                mime_parts = header.removeprefix("data:").split(";")
+                mime_type = mime_parts[0] if mime_parts else "audio/wav"
+                suffix = "." + mime_type.split("/")[-1] if "/" in mime_type else ".wav"
+                if suffix not in (".wav", ".mp3", ".ogg", ".m4a", ".aac", ".flac"):
+                    suffix = ".wav"
+                audio_bytes = base64.b64decode(base64_data)
+                temp_dir = Path(get_astrbot_temp_path())
+                temp_dir.mkdir(parents=True, exist_ok=True)
+                target_path = temp_dir / f"provider_audio_{uuid.uuid4().hex}{suffix}"
+                target_path.write_bytes(audio_bytes)
+                cleanup_paths.append(target_path)
+                return str(target_path), cleanup_paths
+            except Exception as exc:
+                logger.warning("解析 data URI 音频失败: %s，错误: %s", audio_ref[:100], exc)
+                return audio_ref, cleanup_paths
         return audio_ref, cleanup_paths
 
     async def _resolve_audio_part(self, audio_ref: str) -> dict | None:

From 01a9e557ab9de0eee700b8512750f6892a376188 Mon Sep 17 00:00:00 2001
From: tjc66666666 <3428979959@qq.com>
Date: Sat, 6 Jun 2026 14:17:27 +0800
Subject: [PATCH 2/4] Enhance data URI handling for audio formats

---
 .../core/provider/sources/openai_source.py    | 39 +++++++++++++++----
 1 file changed, 31 insertions(+), 8 deletions(-)

diff --git a/astrbot/core/provider/sources/openai_source.py b/astrbot/core/provider/sources/openai_source.py
index c8af63e7b5..34fbf398b5 100644
--- a/astrbot/core/provider/sources/openai_source.py
+++ b/astrbot/core/provider/sources/openai_source.py
@@ -369,16 +369,37 @@ async def _audio_ref_to_local_path(self, audio_ref: str) -> tuple[str, list[Path
             return self._file_uri_to_path(audio_ref), cleanup_paths
         if audio_ref.startswith("data:"):
             # data URI 格式: data:audio/wav;base64,<base64_data>
-            # 或 data:audio/mp3;base64,<base64_data>
+            # 或 data:audio/mpeg;base64,<base64_data>
             try:
+                # 防止过大的 base64 payload 导致内存耗尽
+                if len(audio_ref) > 10 * 1024 * 1024:  # 10 MB
+                    logger.warning(
+                        "data URI 音频过大 (%.1f MB)，将忽略",
+                        len(audio_ref) / (1024 * 1024),
+                    )
+                    return audio_ref, cleanup_paths
+
                 header, base64_data = audio_ref.split(",", 1)
-                # 从 data URI header 中提取 MIME 类型和格式
-                # 例如 "data:audio/wav;base64" -> "wav"
+
+                # 从 data URI header 中提取 MIME 类型
                 mime_parts = header.removeprefix("data:").split(";")
                 mime_type = mime_parts[0] if mime_parts else "audio/wav"
-                suffix = "." + mime_type.split("/")[-1] if "/" in mime_type else ".wav"
-                if suffix not in (".wav", ".mp3", ".ogg", ".m4a", ".aac", ".flac"):
-                    suffix = ".wav"
+
+                # 使用显式映射避免 audio/mpeg 被错误转换为 .mpeg 而非 .mp3
+                mime_to_suffix = {
+                    "audio/wav": ".wav",
+                    "audio/x-wav": ".wav",
+                    "audio/mpeg": ".mp3",
+                    "audio/mp3": ".mp3",
+                    "audio/ogg": ".ogg",
+                    "audio/m4a": ".m4a",
+                    "audio/x-m4a": ".m4a",
+                    "audio/aac": ".aac",
+                    "audio/flac": ".flac",
+                    "audio/x-flac": ".flac",
+                }
+                suffix = mime_to_suffix.get(mime_type, ".wav")
+
                 audio_bytes = base64.b64decode(base64_data)
                 temp_dir = Path(get_astrbot_temp_path())
                 temp_dir.mkdir(parents=True, exist_ok=True)
@@ -386,8 +407,10 @@ async def _audio_ref_to_local_path(self, audio_ref: str) -> tuple[str, list[Path
                 target_path.write_bytes(audio_bytes)
                 cleanup_paths.append(target_path)
                 return str(target_path), cleanup_paths
-            except Exception as exc:
-                logger.warning("解析 data URI 音频失败: %s，错误: %s", audio_ref[:100], exc)
+            except (ValueError, binascii.Error, OSError) as exc:
+                logger.warning(
+                    "解析 data URI 音频失败: %s，错误: %s", audio_ref[:100], exc
+                )
                 return audio_ref, cleanup_paths
         return audio_ref, cleanup_paths
 

From 54fb50b32c34f25cd0abbcc4a77c1d7f786fb000 Mon Sep 17 00:00:00 2001
From: tjc66666666 <3428979959@qq.com>
Date: Sat, 6 Jun 2026 14:21:55 +0800
Subject: [PATCH 3/4] Implement audio data URI tests for WAV, MP3, OGG

Add tests for handling audio data URIs in various formats.
---
 tests/test_openai_source.py | 161 ++++++++++++++++++++++++++++++++++++
 1 file changed, 161 insertions(+)

diff --git a/tests/test_openai_source.py b/tests/test_openai_source.py
index b5587ffb14..0991a8cdfb 100644
--- a/tests/test_openai_source.py
+++ b/tests/test_openai_source.py
@@ -1809,3 +1809,164 @@ async def fake_create(**kwargs):
         assert messages[1] == {"role": "user", "content": "again"}
     finally:
         await provider.terminate()
+
+
+# ---------------------------------------------------------------------------
+# _audio_ref_to_local_path - data URI 处理测试
+# ---------------------------------------------------------------------------
+
+
+def _minimal_wav_bytes() -> bytes:
+    """生成一个最小的有效 WAV 头（44 字节空数据）。"""
+    import struct
+
+    data_size = 0
+    file_size = 36 + data_size  # 4 + 24 + 8 + data_size
+    header = bytearray(44)
+    header[0:4] = b"RIFF"
+    struct.pack_into("<I", header, 4, file_size)
+    header[8:12] = b"WAVE"
+    header[12:16] = b"fmt "
+    struct.pack_into("<I", header, 16, 16)  # PCM chunk size
+    struct.pack_into("<H", header, 20, 1)  # PCM format
+    struct.pack_into("<H", header, 22, 1)  # mono
+    struct.pack_into("<I", header, 24, 44100)  # sample rate
+    struct.pack_into("<I", header, 28, 88200)  # byte rate
+    struct.pack_into("<H", header, 32, 2)  # block align
+    struct.pack_into("<H", header, 34, 16)  # bits per sample
+    header[36:40] = b"data"
+    struct.pack_into("<I", header, 40, data_size)
+    return bytes(header)
+
+
+@pytest.mark.asyncio
+async def test_audio_ref_data_uri_wav(monkeypatch, tmp_path):
+    """data:audio/wav;base64 URI 应被解码为临时 .wav 文件。"""
+    wav_data = _minimal_wav_bytes()
+    b64_data = base64.b64encode(wav_data).decode("utf-8")
+    data_uri = f"data:audio/wav;base64,{b64_data}"
+
+    provider = _make_provider()
+    try:
+        from astrbot.core.utils.astrbot_path import get_astrbot_temp_path
+
+        monkeypatch.setattr(
+            "astrbot.core.provider.sources.openai_source.get_astrbot_temp_path",
+            lambda: str(tmp_path),
+        )
+
+        audio_path, cleanup = await provider._audio_ref_to_local_path(data_uri)
+        assert cleanup, "应该返回需要清理的临时文件路径"
+        assert audio_path.endswith(".wav"), f"后缀应为 .wav，实际为 {audio_path}"
+        assert Path(audio_path).parent == tmp_path
+        assert Path(audio_path).read_bytes() == wav_data, "解码后的内容应与原始 WAV 数据一致"
+        # 清理
+        for p in cleanup:
+            p.unlink(missing_ok=True)
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.asyncio
+async def test_audio_ref_data_uri_mpeg_uses_mp3_suffix(monkeypatch, tmp_path):
+    """data:audio/mpeg;base64 应使用 .mp3 后缀，而非 .mpeg。"""
+    wav_data = _minimal_wav_bytes()
+    b64_data = base64.b64encode(wav_data).decode("utf-8")
+    data_uri = f"data:audio/mpeg;base64,{b64_data}"
+
+    provider = _make_provider()
+    try:
+        monkeypatch.setattr(
+            "astrbot.core.provider.sources.openai_source.get_astrbot_temp_path",
+            lambda: str(tmp_path),
+        )
+
+        audio_path, cleanup = await provider._audio_ref_to_local_path(data_uri)
+        assert cleanup
+        assert audio_path.endswith(".mp3"), (
+            f"audio/mpeg 应映射为 .mp3，实际后缀为 {Path(audio_path).suffix}"
+        )
+        for p in cleanup:
+            p.unlink(missing_ok=True)
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.asyncio
+async def test_audio_ref_data_uri_unknown_mime_falls_back_to_wav(monkeypatch, tmp_path):
+    """未知 MIME 类型应回退为 .wav 后缀。"""
+    wav_data = _minimal_wav_bytes()
+    b64_data = base64.b64encode(wav_data).decode("utf-8")
+    data_uri = f"data:audio/x-unknown;base64,{b64_data}"
+
+    provider = _make_provider()
+    try:
+        monkeypatch.setattr(
+            "astrbot.core.provider.sources.openai_source.get_astrbot_temp_path",
+            lambda: str(tmp_path),
+        )
+
+        audio_path, cleanup = await provider._audio_ref_to_local_path(data_uri)
+        assert cleanup
+        assert audio_path.endswith(".wav"), (
+            f"未知 MIME 类型应回退为 .wav，实际为 {Path(audio_path).suffix}"
+        )
+        for p in cleanup:
+            p.unlink(missing_ok=True)
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.asyncio
+async def test_audio_ref_data_uri_invalid_base64_returns_as_is():
+    """无效 base64 的 data URI 应原样返回且不崩溃。"""
+    data_uri = "data:audio/wav;base64,!!!not_valid_base64!!!"
+
+    provider = _make_provider()
+    try:
+        audio_path, cleanup = await provider._audio_ref_to_local_path(data_uri)
+        assert not cleanup, "无效数据不应产生清理文件"
+        assert audio_path == data_uri, "无效 data URI 应原样返回"
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.asyncio
+async def test_audio_ref_data_uri_oversized_rejected():
+    """超过 10 MB 的 data URI 应被拒绝。"""
+    # 创建一个 ~11 MB 的 data URI
+    large_b64 = "A" * (11 * 1024 * 1024)  # base64 字符串约 11 MB
+    data_uri = f"data:audio/wav;base64,{large_b64}"
+
+    provider = _make_provider()
+    try:
+        audio_path, cleanup = await provider._audio_ref_to_local_path(data_uri)
+        assert not cleanup, "超大 payload 不应产生临时文件"
+        assert audio_path == data_uri, "超大 data URI 应原样返回"
+    finally:
+        await provider.terminate()
+
+
+@pytest.mark.asyncio
+async def test_audio_ref_data_uri_ogg_suffix(monkeypatch, tmp_path):
+    """data:audio/ogg;base64 应使用 .ogg 后缀。"""
+    wav_data = _minimal_wav_bytes()
+    b64_data = base64.b64encode(wav_data).decode("utf-8")
+    data_uri = f"data:audio/ogg;base64,{b64_data}"
+
+    provider = _make_provider()
+    try:
+        monkeypatch.setattr(
+            "astrbot.core.provider.sources.openai_source.get_astrbot_temp_path",
+            lambda: str(tmp_path),
+        )
+
+        audio_path, cleanup = await provider._audio_ref_to_local_path(data_uri)
+        assert cleanup
+        assert audio_path.endswith(".ogg"), (
+            f"audio/ogg 应使用 .ogg 后缀，实际为 {Path(audio_path).suffix}"
+        )
+        for p in cleanup:
+            p.unlink(missing_ok=True)
+    finally:
+        await provider.terminate()

From 293f21c8672e338e2115624fe569fe41fbb18999 Mon Sep 17 00:00:00 2001
From: tjc66666666 <3428979959@qq.com>
Date: Sat, 6 Jun 2026 14:26:32 +0800
Subject: [PATCH 4/4] Refactor assertion message for audio path test

---
 tests/test_openai_source.py | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/tests/test_openai_source.py b/tests/test_openai_source.py
index 0991a8cdfb..5fb7a60482 100644
--- a/tests/test_openai_source.py
+++ b/tests/test_openai_source.py
@@ -1,6 +1,7 @@
 import base64
 import builtins
 from io import BytesIO
+from pathlib import Path
 from types import SimpleNamespace
 
 import pytest
@@ -1859,7 +1860,9 @@ async def test_audio_ref_data_uri_wav(monkeypatch, tmp_path):
         assert cleanup, "应该返回需要清理的临时文件路径"
         assert audio_path.endswith(".wav"), f"后缀应为 .wav，实际为 {audio_path}"
         assert Path(audio_path).parent == tmp_path
-        assert Path(audio_path).read_bytes() == wav_data, "解码后的内容应与原始 WAV 数据一致"
+        assert Path(audio_path).read_bytes() == wav_data, (
+            "解码后的内容应与原始 WAV 数据一致"
+        )
         # 清理
         for p in cleanup:
             p.unlink(missing_ok=True)