Skip to content

Give your AI agent eyes to see the entire internet. Read & search Twitter, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu — one CLI, zero API fees.

security(transcribe): bound media processing

Pnant77d1abe482de
diff --git a/agent_reach/transcribe.py b/agent_reach/transcribe.py
index 0344a05e..a5a28c8c 100644
--- a/agent_reach/transcribe.py
+++ b/agent_reach/transcribe.py
@@ -28,6 +28,9 @@ from agent_reach.config import Config
# Whisper API limit is 25MB; leave headroom for multipart overhead.
SIZE_LIMIT_BYTES = 24 * 1024 * 1024
CHUNK_SECONDS = 600 # 10 min — small enough that boundary cuts rarely lose meaning
+MAX_SOURCE_BYTES = 512 * 1024 * 1024
+MAX_CHUNKS = 24 # 4 hours at the standard 10-minute segment size
+MAX_TOTAL_CHUNK_BYTES = 96 * 1024 * 1024
PROVIDERS = {
"groq": {
@@ -66,6 +69,15 @@ def _require(binary: str) -> None:
raise MissingDependency(f"{binary} not found in PATH")
+def _require_size_at_most(path: Path, limit: int, label: str) -> int:
+ """Return file size or fail before expensive downstream processing."""
+ size = path.stat().st_size
+ if size > limit:
+ limit_mib = limit / (1024 * 1024)
+ raise TranscribeError(f"{label} exceeds safety limit of {limit_mib:g} MiB")
+ return size
+
+
def _run(cmd: List[str], timeout: int = 600) -> None:
"""Run a subprocess, raising TranscribeError on nonzero exit or timeout.
@@ -135,6 +147,9 @@ def download_audio(url: str, out_dir: Path) -> Path:
"m4a",
"--audio-quality",
"0",
+ "--no-playlist",
+ "--max-filesize",
+ str(MAX_SOURCE_BYTES),
"-o",
str(template),
"--",
@@ -144,8 +159,13 @@ def download_audio(url: str, out_dir: Path) -> Path:
)
files = sorted(out_dir.glob("source.*"))
if not files:
- raise TranscribeError("yt-dlp produced no output file")
- return files[0]
+ limit_mib = MAX_SOURCE_BYTES // (1024 * 1024)
+ raise TranscribeError(
+ f"yt-dlp produced no output file (source may exceed {limit_mib} MiB limit)"
+ )
+ audio = files[0]
+ _require_size_at_most(audio, MAX_SOURCE_BYTES, "downloaded source")
+ return audio
def compress_audio(src: Path, out_dir: Path) -> Path:
@@ -290,12 +310,31 @@ def _transcribe_in_dir(source: str, order: List[str], cfg: Config, work_dir: Pat
else:
audio = download_audio(source, work_dir)
+ _require_size_at_most(audio, MAX_SOURCE_BYTES, "source")
compressed = compress_audio(audio, work_dir)
if compressed.stat().st_size <= SIZE_LIMIT_BYTES:
chunks = [compressed]
else:
chunks = chunk_audio(compressed, work_dir)
+ if len(chunks) > MAX_CHUNKS:
+ max_minutes = MAX_CHUNKS * CHUNK_SECONDS // 60
+ raise TranscribeError(
+ f"audio produced {len(chunks)} chunks; safety limit is "
+ f"{MAX_CHUNKS} (~{max_minutes} minutes)"
+ )
+ chunk_sizes = [
+ _require_size_at_most(chunk, SIZE_LIMIT_BYTES, f"chunk {chunk.name}")
+ for chunk in chunks
+ ]
+ total_chunk_bytes = sum(chunk_sizes)
+ if total_chunk_bytes > MAX_TOTAL_CHUNK_BYTES:
+ limit_mib = MAX_TOTAL_CHUNK_BYTES / (1024 * 1024)
+ raise TranscribeError(
+ f"audio chunks total {total_chunk_bytes} bytes; "
+ f"safety limit is {limit_mib:g} MiB"
+ )
+
pieces: List[str] = []
for chunk in chunks:
text = _transcribe_with_fallback(chunk, order, cfg)
diff --git a/tests/test_transcribe.py b/tests/test_transcribe.py
index 748a8011..d4b43ff3 100644
--- a/tests/test_transcribe.py
+++ b/tests/test_transcribe.py
@@ -161,6 +161,22 @@ class TestFallback:
class TestOrchestrator:
+ def test_rejects_oversized_source_before_compression(
+ self, monkeypatch, fake_config, chunk_file
+ ):
+ fake_config.set("groq_api_key", "gsk_test")
+ monkeypatch.setattr(tr, "MAX_SOURCE_BYTES", 4)
+ monkeypatch.setattr(
+ tr,
+ "compress_audio",
+ lambda *_args, **_kwargs: (_ for _ in ()).throw(
+ AssertionError("oversized source must fail before ffmpeg")
+ ),
+ )
+
+ with pytest.raises(tr.TranscribeError, match="source.*limit"):
+ tr.transcribe(str(chunk_file), config=fake_config)
+
def test_local_file_skips_yt_dlp(self, monkeypatch, fake_config, tmp_path, chunk_file):
fake_config.set("groq_api_key", "gsk_test")
@@ -217,6 +233,62 @@ class TestOrchestrator:
)
assert text == "part one\npart two"
+ def test_rejects_too_many_chunks_before_any_provider_call(
+ self, monkeypatch, fake_config, tmp_path, chunk_file
+ ):
+ fake_config.set("groq_api_key", "gsk_test")
+ monkeypatch.setattr(tr, "SIZE_LIMIT_BYTES", 1)
+ compressed = tmp_path / "compressed.m4a"
+ compressed.write_bytes(b"xx")
+ monkeypatch.setattr(tr, "compress_audio", lambda *_args: compressed)
+
+ chunks = []
+ for index in range(tr.MAX_CHUNKS + 1):
+ chunk = tmp_path / f"chunk_{index:03d}.m4a"
+ chunk.write_bytes(b"x")
+ chunks.append(chunk)
+ monkeypatch.setattr(tr, "chunk_audio", lambda *_args: chunks)
+
+ provider_calls = []
+ monkeypatch.setattr(
+ tr,
+ "_transcribe_with_fallback",
+ lambda *_args: provider_calls.append("called") or "text",
+ )
+
+ with pytest.raises(tr.TranscribeError, match="chunks.*limit"):
+ tr.transcribe(str(chunk_file), out_dir=tmp_path / "work", config=fake_config)
+
+ assert provider_calls == []
+
+ def test_rejects_excessive_total_chunk_bytes_before_provider_calls(
+ self, monkeypatch, fake_config, tmp_path, chunk_file
+ ):
+ fake_config.set("groq_api_key", "gsk_test")
+ monkeypatch.setattr(tr, "SIZE_LIMIT_BYTES", 10)
+ monkeypatch.setattr(tr, "MAX_TOTAL_CHUNK_BYTES", 5)
+ compressed = tmp_path / "compressed.m4a"
+ compressed.write_bytes(b"x" * 11)
+ monkeypatch.setattr(tr, "compress_audio", lambda *_args: compressed)
+
+ first = tmp_path / "chunk_000.m4a"
+ second = tmp_path / "chunk_001.m4a"
+ first.write_bytes(b"aaa")
+ second.write_bytes(b"bbb")
+ monkeypatch.setattr(tr, "chunk_audio", lambda *_args: [first, second])
+
+ provider_calls = []
+ monkeypatch.setattr(
+ tr,
+ "_transcribe_with_fallback",
+ lambda *_args: provider_calls.append("called") or "text",
+ )
+
+ with pytest.raises(tr.TranscribeError, match="total.*limit"):
+ tr.transcribe(str(chunk_file), out_dir=tmp_path / "work", config=fake_config)
+
+ assert provider_calls == []
+
def test_no_provider_configured_fails_fast(self, fake_config, chunk_file):
with pytest.raises(tr.NoProviderConfigured):
tr.transcribe(str(chunk_file), config=fake_config)
@@ -298,6 +370,20 @@ class TestOrchestrator:
class TestDownloadAudioSafety:
+ def test_rejects_download_that_exceeds_limit(
+ self, monkeypatch, tmp_path
+ ):
+ monkeypatch.setattr(tr, "_require", lambda _binary: None)
+ monkeypatch.setattr(tr, "MAX_SOURCE_BYTES", 4)
+
+ def fake_run(_cmd, timeout=600):
+ (tmp_path / "source.m4a").write_bytes(b"audio")
+
+ monkeypatch.setattr(tr, "_run", fake_run)
+
+ with pytest.raises(tr.TranscribeError, match="downloaded source.*limit"):
+ tr.download_audio("https://example.com/watch?v=123", tmp_path)
+
def test_rejects_private_network_url_before_yt_dlp(self, monkeypatch, tmp_path):
monkeypatch.setattr(tr, "_require", lambda binary: None)
@@ -323,8 +409,11 @@ class TestDownloadAudioSafety:
assert audio == tmp_path / "source.m4a"
assert "--" in captured["cmd"]
+ assert "--no-playlist" in captured["cmd"]
marker_index = captured["cmd"].index("--")
assert captured["cmd"][marker_index + 1] == "https://example.com/watch?v=123"
+ max_size_index = captured["cmd"].index("--max-filesize")
+ assert captured["cmd"][max_size_index + 1] == str(tr.MAX_SOURCE_BYTES)
def test_preserves_bare_public_urls_supported_by_yt_dlp(self, monkeypatch, tmp_path):
monkeypatch.setattr(tr, "_require", lambda binary: None)