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(transforms): stop routing timestamped log rows as grep matches (#3748)

Xmon Daid971f7c3983b
diff --git a/headroom/transforms/content_detector.py b/headroom/transforms/content_detector.py
index 55a855013..898e8746b 100644
--- a/headroom/transforms/content_detector.py
+++ b/headroom/transforms/content_detector.py
@@ -21,6 +21,8 @@ import re
from dataclasses import dataclass
from enum import Enum
+from .lossless_compaction import _TIMESTAMP_ROW_RE
+
class ContentType(Enum):
"""Types of content that can be compressed."""
@@ -509,11 +511,20 @@ def _is_search_result_line(line: str) -> bool:
a file path: no angle brackets and no ``=`` (rules out markup tags and
``key=value:12:`` log lines).
+ A timestamped log row is never grep output: since #3419 the lossless
+ fold skips those rows, so a payload classified as search here falls
+ through to the lossy SearchCompressor, which keeps 5 rows per "file"
+ and prints the minute back as an integer (#3736). Reject any line the
+ lossless fold already knows is a timestamp row, reusing the same regex
+ so the two guards cannot drift apart.
+
``grep -A``/``-B``/``-C`` context lines — both the real GNU shape
(``path-NN-content``) and the reported ``path:NN-content`` shape — are
accepted via the context predicates so code in them routes to the search
compressor instead of the prose path (#3580).
"""
+ if _TIMESTAMP_ROW_RE.match(line):
+ return False
if _SEARCH_RESULT_PATTERN.match(line) or _GREP_COLON_DASH_PATTERN.match(line):
return _prefix_looks_like_path(line.split(":", 1)[0])
return _is_grep_context_line(line)
diff --git a/tests/test_transforms/test_content_detector_timestamp_search.py b/tests/test_transforms/test_content_detector_timestamp_search.py
new file mode 100644
index 000000000..23b66a272
--- /dev/null
+++ b/tests/test_transforms/test_content_detector_timestamp_search.py
@@ -0,0 +1,80 @@
+"""Regression for #3736: ISO-8601 timestamped logs must not route as search.
+
+The colon branch of ``_is_search_result_line`` accepts ``^[^\\s:]+:\\d+:``
+whenever the prefix passes ``_prefix_looks_like_path``. A bare ISO date-hour
+prefix (``2026-09-23T10``) passes that guard, so every row of a timestamped
+log counts as a grep match (ratio 1.0). Since #3419 the lossless fold skips
+timestamp rows, so the payload falls through to the lossy SearchCompressor,
+which keeps 5 of 2,000 lines and prints the minute back as an integer
+(``10:0:00Z`` — a time that does not exist in the input).
+"""
+
+from __future__ import annotations
+
+from headroom.transforms.content_detector import (
+ ContentType,
+ _is_search_result_line,
+ _try_detect_search,
+ detect_content_type,
+)
+
+
+def _timestamped_log(n: int = 200) -> str:
+ return "\n".join(
+ f"2026-09-23T10:{i // 60:02d}:{i % 60:02d}Z INFO worker[{i % 8}] "
+ f"processed batch {i} in {(i * 13) % 900}ms"
+ for i in range(n)
+ )
+
+
+def test_timestamp_row_is_not_a_search_result_line() -> None:
+ line = "2026-09-23T10:00:00Z INFO worker[0] processed batch 0 in 0ms"
+ assert not _is_search_result_line(line)
+
+
+def test_timestamped_log_does_not_route_to_search() -> None:
+ payload = _timestamped_log()
+ assert _try_detect_search(payload) is None
+ assert detect_content_type(payload).content_type is not ContentType.SEARCH_RESULTS
+
+
+def test_space_separated_timestamp_log_does_not_route_to_search() -> None:
+ payload = "\n".join(
+ f"2026-09-23 10:{i // 60:02d}:{i % 60:02d} [INFO] worker {i} processed batch {i}"
+ for i in range(50)
+ )
+ assert _try_detect_search(payload) is None
+ assert detect_content_type(payload).content_type is not ContentType.SEARCH_RESULTS
+
+
+def test_genuine_grep_output_still_routes_to_search() -> None:
+ payload = "\n".join(
+ [
+ "src/app.py:10:def main():",
+ "src/app.py:20:print('hello')",
+ "README.md:5:usage docs",
+ ]
+ )
+ result = _try_detect_search(payload)
+ assert result is not None
+ assert result.content_type is ContentType.SEARCH_RESULTS
+
+
+def test_grep_on_timestamped_file_names_still_routes_to_search() -> None:
+ """A path prefix that embeds a date is still a path, not a timestamp row."""
+ payload = "\n".join(
+ [
+ "logs/2026-09-23.log:42:ERROR something failed",
+ "logs/2026-09-23.log:43:INFO recovered",
+ "app/2026-09-24.log:7:WARN retry",
+ ]
+ )
+ assert _is_search_result_line("logs/2026-09-23.log:42:ERROR something failed")
+ result = _try_detect_search(payload)
+ assert result is not None
+ assert result.content_type is ContentType.SEARCH_RESULTS
+
+
+def test_extensionless_paths_still_match() -> None:
+ assert _is_search_result_line("Makefile:12:all:")
+ assert _is_search_result_line("Dockerfile:3:FROM scratch")