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(proxy): report deferred Kompress status and promote health from cache (#2564)

AxelRayd50cfabedca2
diff --git a/headroom/proxy/server.py b/headroom/proxy/server.py
index b069335e0..59fe81cfb 100644
--- a/headroom/proxy/server.py
+++ b/headroom/proxy/server.py
@@ -1708,14 +1708,16 @@ class HeadroomProxy(
self.warmup.merge_transform_status(transform_status)
# Update internal status from eager loading results
- if eager_status.get("kompress") == "enabled":
- self._kompress_status = "enabled"
+ if eager_status.get("kompress") in {"enabled", "deferred"}:
+ self._kompress_status = eager_status["kompress"]
if eager_status.get("code_aware") == "enabled":
self._code_aware_status = "enabled"
# Log component status
if self._kompress_status == "enabled":
logger.info("Kompress: ENABLED (ModernBERT token compressor)")
+ elif self._kompress_status == "deferred":
+ logger.info("Kompress: DEFERRED (model loads on first request)")
elif self.config.optimize:
logger.info("Kompress: not installed (pip install headroom-ai[ml] for ML compression)")
@@ -2630,6 +2632,21 @@ def create_app(config: ProxyConfig | None = None) -> FastAPI:
return True
proxy.warmup.kompress.mark_loaded(handle=compressor, backend=backend)
return True
+
+ try:
+ from headroom.transforms.kompress_compressor import HF_MODEL_ID, _kompress_cache
+ except ImportError:
+ return True
+
+ cached = _kompress_cache.get(HF_MODEL_ID)
+ if cached is not None:
+ try:
+ model, _tokenizer, backend = cached
+ except (TypeError, ValueError):
+ pass
+ else:
+ if backend and proxy.warmup.kompress.status != "loaded":
+ proxy.warmup.kompress.mark_loaded(handle=model, backend=backend)
return True
def _health_checks() -> dict[str, dict[str, Any]]:
diff --git a/tests/test_proxy_eager_preload_bind.py b/tests/test_proxy_eager_preload_bind.py
index c168197a4..45b05c549 100644
--- a/tests/test_proxy_eager_preload_bind.py
+++ b/tests/test_proxy_eager_preload_bind.py
@@ -11,6 +11,7 @@ bind still happens and transforms fall back to lazy loading.
from __future__ import annotations
+import logging
import threading
import time
@@ -118,3 +119,25 @@ async def test_startup_merges_warmup_for_normal_transforms(monkeypatch):
assert proxy._kompress_status == "enabled"
finally:
await proxy.shutdown()
+
+
+async def test_startup_reports_deferred_kompress(caplog):
+ proxy = _make_proxy(optimize=True)
+ proxy.anthropic_pipeline = _FakePipeline([_FastTransform({"kompress": "deferred"})])
+ proxy.openai_pipeline = _FakePipeline([])
+
+ try:
+ # Proxy setup disables propagation on the ``headroom`` logger, so
+ # attach caplog's handler directly to the logger that emits this line.
+ server_mod.logger.addHandler(caplog.handler)
+ try:
+ with caplog.at_level(logging.INFO, logger=server_mod.logger.name):
+ await proxy.startup()
+ finally:
+ server_mod.logger.removeHandler(caplog.handler)
+
+ assert proxy._kompress_status == "deferred"
+ assert "Kompress: DEFERRED (model loads on first request)" in caplog.messages
+ assert not any("Kompress: not installed" in message for message in caplog.messages)
+ finally:
+ await proxy.shutdown()
diff --git a/tests/test_proxy_health.py b/tests/test_proxy_health.py
index af459535e..e2a2d89d1 100644
--- a/tests/test_proxy_health.py
+++ b/tests/test_proxy_health.py
@@ -3,6 +3,7 @@ from fastapi.testclient import TestClient
from headroom.proxy.models import ProxyConfig
from headroom.proxy.server import create_app
+from headroom.transforms import kompress_compressor
class _ReadyCompressor:
@@ -88,6 +89,31 @@ def test_readyz_promotes_deferred_kompress_after_runtime_load(monkeypatch):
}
[email protected]("attached", [False, True])
+def test_readyz_promotes_kompress_from_module_cache(monkeypatch, attached):
+ model = object()
+ monkeypatch.setattr(
+ kompress_compressor,
+ "_kompress_cache",
+ {kompress_compressor.HF_MODEL_ID: (model, object(), "onnx")},
+ )
+ compressor = _ReadyCompressor(ready=False) if attached else None
+ app, proxy = _health_app(monkeypatch, compressor)
+ proxy.warmup.kompress.info["source_status"] = "deferred"
+
+ payload = TestClient(app).get("/readyz").json()
+
+ assert payload["checks"]["kompress"] == {
+ "enabled": True,
+ "ready": True,
+ "status": "healthy",
+ "backend": "onnx",
+ }
+ assert proxy.warmup.kompress.handle is model
+ if compressor is not None:
+ assert compressor.calls == ["is_ready"]
+
+
def test_readyz_promotes_remote_kompress_backend(monkeypatch):
compressor = _ReadyCompressor(backend="remote")
app, proxy = _health_app(monkeypatch)