Skip to content

Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.

fix(deps): enforce transformers security floor (#2201)

JD Daviscbfa267c5f4f
diff --git a/headroom/cli/wrap.py b/headroom/cli/wrap.py
index b4079ad8a..5ebcdf78a 100644
--- a/headroom/cli/wrap.py
+++ b/headroom/cli/wrap.py
@@ -5615,7 +5615,7 @@ def zcode(
upstream = _detect_zcode_upstream()
anthropic_url, openai_url = _zcode_upstream_to_urls(upstream)
- def _print_zcode_setup(actual_port: int) -> None:
+ def _print_zcode_setup(actual_port: int = port) -> None:
click.echo(f" Detected provider: {upstream.provider_name}")
click.echo(f" Upstream: {upstream.base_url}")
click.echo()
diff --git a/headroom/proxy/handlers/anthropic.py b/headroom/proxy/handlers/anthropic.py
index 897318b5c..027051a7b 100644
--- a/headroom/proxy/handlers/anthropic.py
+++ b/headroom/proxy/handlers/anthropic.py
@@ -1572,9 +1572,7 @@ class AnthropicHandlerMixin:
# Split timeout from other errors: a timeout means the
# compression budget was too tight, not a code bug.
reason = (
- "timeout"
- if isinstance(e, (asyncio.TimeoutError, TimeoutError))
- else "error"
+ "timeout" if isinstance(e, asyncio.TimeoutError | TimeoutError) else "error"
)
self.metrics.record_compression_failed(reason)
@@ -3661,7 +3659,7 @@ class AnthropicHandlerMixin:
# Same fail-open accounting as the single-message path: a
# timeout means the budget was too tight, not a code bug.
reason = (
- "timeout" if isinstance(e, (asyncio.TimeoutError, TimeoutError)) else "error"
+ "timeout" if isinstance(e, asyncio.TimeoutError | TimeoutError) else "error"
)
self.metrics.record_compression_failed(reason)
# Pass through unchanged on failure
diff --git a/pyproject.toml b/pyproject.toml
index 3bf14239a..920a39519 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -76,7 +76,7 @@ proxy = [
"zstandard>=0.20.0", # Decompress zstd request bodies (Codex, etc.)
"websockets>=13.0", # WebSocket proxy for /v1/responses (Codex gpt-5.4+)
"onnxruntime>=1.16.0", # Kompress ONNX INT8 text compression (no torch needed)
- "transformers>=4.30.0,<6.0", # Tokenizer only (for Kompress)
+ "transformers>=5.5.0,<6.0", # Tokenizer only (for Kompress)
"watchdog>=4.0.0", # File watcher for live code graph reindexing (--code-graph)
"sqlite-vec>=0.1.6", # Vector index for memory (--memory). Lightweight, no torch.
]
@@ -105,7 +105,7 @@ ml = [
# PyTorch does not publish wheels for macOS 15 x86_64 at this floor, which
# makes `headroom-ai[all]` unsatisfiable on Intel Macs (#1931).
"torch>=2.12.1; sys_platform != 'darwin' or platform_machine != 'x86_64'",
- "transformers>=4.30.0,<6.0",
+ "transformers>=5.5.0,<6.0",
# transformers >= 5.x requires huggingface-hub >= 1.5.0,<2.0; pinning
# the floor here prevents Kompress from silently falling back to
# "unavailable" when a sibling install (e.g. `pip install
@@ -218,7 +218,7 @@ mcp = [
# Voice filler detection
voice = [
"onnxruntime>=1.16.0",
- "transformers>=4.30.0,<6.0",
+ "transformers>=5.5.0,<6.0",
"torch>=2.12.1; sys_platform != 'darwin' or platform_machine != 'x86_64'",
]
# Voice training (includes voice deps + training extras)
diff --git a/uv.lock b/uv.lock
index f74a040dd..28b84d856 100644
--- a/uv.lock
+++ b/uv.lock
@@ -1501,6 +1501,7 @@ all = [
{ name = "sentence-transformers", marker = "platform_machine != 'x86_64' or sys_platform != 'darwin'" },
{ name = "sentencepiece" },
{ name = "sqlite-vec" },
+ { name = "starlette" },
{ name = "torch", marker = "platform_machine != 'x86_64' or sys_platform != 'darwin'" },
{ name = "trafilatura" },
{ name = "transformers" },
@@ -1660,6 +1661,7 @@ sandbox = [
{ name = "opentelemetry-sdk" },
{ name = "orjson", marker = "platform_python_implementation != 'PyPy'" },
{ name = "sqlite-vec" },
+ { name = "starlette" },
{ name = "trafilatura" },
{ name = "transformers" },
{ name = "tree-sitter" },
@@ -1781,9 +1783,9 @@ requires-dist = [
{ name = "torch", marker = "(platform_machine != 'x86_64' and extra == 'ml') or (sys_platform != 'darwin' and extra == 'ml')", specifier = ">=2.12.1" },
{ name = "torch", marker = "(platform_machine != 'x86_64' and extra == 'voice') or (sys_platform != 'darwin' and extra == 'voice')", specifier = ">=2.12.1" },
{ name = "trafilatura", marker = "extra == 'html'", specifier = ">=1.6.0" },
- { name = "transformers", marker = "extra == 'ml'", specifier = ">=4.30.0,<6.0" },
- { name = "transformers", marker = "extra == 'proxy'", specifier = ">=4.30.0,<6.0" },
- { name = "transformers", marker = "extra == 'voice'", specifier = ">=4.30.0,<6.0" },
+ { name = "transformers", marker = "extra == 'ml'", specifier = ">=5.5.0,<6.0" },
+ { name = "transformers", marker = "extra == 'proxy'", specifier = ">=5.5.0,<6.0" },
+ { name = "transformers", marker = "extra == 'voice'", specifier = ">=5.5.0,<6.0" },
{ name = "tree-sitter", marker = "extra == 'code'", specifier = ">=0.25.2,<0.27" },
{ name = "tree-sitter-language-pack", marker = "extra == 'code'", specifier = ">=0.10.0,<1.0" },
{ name = "uvicorn", marker = "extra == 'dev'", specifier = ">=0.23.0,<1.0" },