Files
headroom/tests/test_transforms/test_read_lifecycle.py
c21b0813bd fix(compress): stop adding markers the caller did not ask for (#4678)
`/v1/compress` returns a body that someone else's client forwards
unchanged, so anything Headroom adds there reaches the provider. Two
things got added that the caller never asked for.

## Stale-Read retrieval hashes in marker-free modes

Read lifecycle replaces a Read that a later edit made stale with a short
notice. That notice always ended in `Retrieve original: hash=…` and
wrote the original to the CCR store, even in the default and
`lossy_inline` modes. The docs call those modes marker-free, and their
callers have no retrieval tool, so the hash pointed nowhere.

- Marker-free modes keep the notice ("re-read the file for current
content") but drop the hash and skip the store write.
- Lossless routers (proxy `--lossless`,
`ContentRouterConfig(lossless=True)`) skip read lifecycle entirely,
because the notice drops the file's bytes.
- Read maturation now recognises the hashless notice. Otherwise it would
replace the notice with a marker whose hash points at the notice itself.

## Cache breakpoints the client didn't send

On the session path, `normalize_message_cache_control` fell back to
placing `{"type": "ephemeral"}` on the last block message when the
client marked nothing. A caller that chose not to cache got billed cache
writes. `/v1/compress` now passes `place_unmarked=False`: if the client
marked no message, markers are stripped and none are placed. Proxy
handlers keep today's behaviour, and client markers in a misaligned list
still go through the old path.

## Tests

Each new test fails on main.

- `test_marker_free_modes_mark_a_stale_read_without_a_retrieval_hash`,
for the default mode and `lossy_inline`
- `test_lossless_router_keeps_a_stale_read`
- `test_session_turns_add_no_marker_the_caller_did_not_send`
- `test_respects_lifecycle_markers`, extended to the hashless notice

The related suites (gateway, transforms, read maturation, prefix
tracker, cache_control, compress endpoint and session) pass: 1371 passed
and 9 skipped before the review fixes, then 222 of 222 on the touched
files after them. Lint is clean on CI's ruff 0.17.0, and mypy is clean
on the touched modules.

## Not in this PR

- Read maturation (`handlers/anthropic.py`) is not gated on `--lossless`
or `--no-ccr` either. It's the same kind of bug on the proxy path, so it
belongs in a separate fix.
- The "lossless densify only" pipeline in `compress.py` sets
`ccr_enabled=False` without `lossless=True`, so read lifecycle still
runs there.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

<!-- devin-review-badge-begin -->

---

<a href="https://app.devin.ai/review/headroomlabs-ai/headroom/pull/4678"
target="_blank"><picture><source media="(prefers-color-scheme: dark)"
srcset="https://static.devin.ai/assets/gh-devin-review-dark.svg?v=4"><img
src="https://static.devin.ai/assets/gh-devin-review-light.svg?v=4"
alt="Devin Review"></picture></a>
<!-- devin-review-badge-end -->

## Real Behavior Proof

Maintainer qualification on Windows, Python 3.12.13, using an actual
Uvicorn server bound to an owned loopback socket and HTTPX over real
HTTP. The normal `/v1/compress` endpoint and compression pipeline ran
with offline mode and Kompress disabled; no handler, compression stage,
HTTP response or store was mocked. Workspace, configuration, savings and
TOIN paths were isolated. All 243 native inputs and the reused artifact
SHA-256
`5be63a4f31d3e38e7427d28380ecb977616e3f6a3dd6a1c59af5ff6361a05978` were
verified; this was not a fresh native build.

- Exact command / steps: the committed probe is `python -m pytest -q -s
tests/test_compress_runtime_http.py`. The qualifying run used the
maintainer loopback-only launcher with an explicit verified native path
and SHA-256, followed by the four focused
session/API/maturation/lifecycle suites. All 138 tests passed. The probe
starts the actual server, GETs `/health`, and sends nine POSTs to
`/v1/compress`: a Read followed by Edit in default, lossy-inline and
lossless modes, then three session turns without caller cache markers
and three turns with a caller marker. It stops and joins the server in
`finally`.
- Observed result: health and all nine compression requests returned
HTTP 200. Default and lossy-inline emitted stale notices with no
retrieval hash or CCR hashes, saving 3,576 tokens per request. Lossless
preserved the original Read bytes. All three unmarked session turns
returned zero cache markers; all three marked controls retained exactly
one. Session compression saved 7,350 / 15,901 / 25,958 tokens. Server
shutdown joined successfully.
- Before-fix control: the same real-server requests with all six
affected production modules replaced by current main
`200c838c7393d20ae56a0c338f832ed27133c398` failed five assertions:
default and lossy-inline included retrieval hashes, and each unmarked
session turn added a cache marker. Lossless and the three marked
controls passed on main and are not claimed as reproduced regressions.
Candidate source was restored byte-for-byte.
- Not tested: provider forwarding or billing, a live model, deployment
behavior, or the separate proxy read-maturation scope already listed
above. External startup connection attempts were blocked by the
qualifying launcher; this is not a zero-attempt network claim. No author
readiness or self-review declaration is added by this maintainer
evidence.

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: JerrettDavis <mxjerrett@gmail.com>
2026-11-13 09:39:10 -06:00

807 lines
28 KiB
Python
Raw Permalink Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
"""Tests for ReadLifecycleManager - event-driven Read lifecycle management.
Tests covering:
- Disabled by default (backward compatibility)
- Stale detection (file edited after Read)
- Superseded detection (file re-Read)
- Fresh Reads untouched
- Multiple files and complex chains
- OpenAI and Anthropic message formats
- CCR store integration
- Size gating
"""
import json
from headroom.config import ReadLifecycleConfig
from headroom.transforms.read_lifecycle import (
ReadLifecycleManager,
)
# =============================================================================
# Helpers
# =============================================================================
def make_openai_read(tool_call_id: str, file_path: str) -> dict:
"""Create an OpenAI-format assistant message with a Read tool call."""
return {
"role": "assistant",
"content": None,
"tool_calls": [
{
"id": tool_call_id,
"type": "function",
"function": {
"name": "Read",
"arguments": json.dumps({"file_path": file_path}),
},
}
],
}
def make_openai_edit(tool_call_id: str, file_path: str) -> dict:
"""Create an OpenAI-format assistant message with an Edit tool call."""
return {
"role": "assistant",
"content": None,
"tool_calls": [
{
"id": tool_call_id,
"type": "function",
"function": {
"name": "Edit",
"arguments": json.dumps(
{
"file_path": file_path,
"old_string": "old",
"new_string": "new",
}
),
},
}
],
}
def make_openai_write(tool_call_id: str, file_path: str) -> dict:
"""Create an OpenAI-format assistant message with a Write tool call."""
return {
"role": "assistant",
"content": None,
"tool_calls": [
{
"id": tool_call_id,
"type": "function",
"function": {
"name": "Write",
"arguments": json.dumps({"file_path": file_path, "content": "new content"}),
},
}
],
}
def make_openai_tool_result(tool_call_id: str, content: str) -> dict:
"""Create an OpenAI-format tool result message."""
return {
"role": "tool",
"tool_call_id": tool_call_id,
"content": content,
}
def make_anthropic_read(tool_call_id: str, file_path: str) -> dict:
"""Create an Anthropic-format assistant message with a Read tool call."""
return {
"role": "assistant",
"content": [
{
"type": "tool_use",
"id": tool_call_id,
"name": "Read",
"input": {"file_path": file_path},
}
],
}
def make_anthropic_edit(tool_call_id: str, file_path: str) -> dict:
"""Create an Anthropic-format assistant message with an Edit tool call."""
return {
"role": "assistant",
"content": [
{
"type": "tool_use",
"id": tool_call_id,
"name": "Edit",
"input": {
"file_path": file_path,
"old_string": "old",
"new_string": "new",
},
}
],
}
def make_anthropic_tool_result(tool_call_id: str, content: str) -> dict:
"""Create an Anthropic-format user message with a tool_result block."""
return {
"role": "user",
"content": [
{
"type": "tool_result",
"tool_use_id": tool_call_id,
"content": content,
}
],
}
LARGE_CONTENT = "x" * 2000 # Well above min_size_bytes
SMALL_CONTENT = "tiny" # Below min_size_bytes
REJECTED_EDIT = "<tool_use_error>String to replace not found in file.</tool_use_error>"
# =============================================================================
# Tests
# =============================================================================
class TestReadLifecycleDisabled:
"""Verify backward compatibility when disabled."""
def test_disabled_when_explicitly_off(self):
"""Explicitly disabled config: no changes to messages."""
config = ReadLifecycleConfig(enabled=False)
assert config.enabled is False
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
]
result = mgr.apply(messages)
assert result.messages is messages # Same object, not copied
assert result.reads_total == 0
assert result.transforms_applied == []
def test_enabled_by_default(self):
"""Default config has lifecycle enabled."""
config = ReadLifecycleConfig()
assert config.enabled is True
class TestStaleDetection:
"""Read outputs become stale when the file is subsequently edited."""
def test_read_then_edit_makes_stale(self):
"""Read(A) → Edit(A): Read becomes stale."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "edit success"),
]
result = mgr.apply(messages)
assert result.reads_stale == 1
assert result.reads_fresh == 0
# Read content should be replaced with marker
tool_result = result.messages[1]
assert "stale" in tool_result["content"].lower()
assert "/src/app.py" in tool_result["content"]
assert "hash=" in tool_result["content"]
def test_write_makes_read_stale(self):
"""Read(A) → Write(A): Read becomes stale."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_write("w1", "/src/app.py"),
make_openai_tool_result("w1", "write success"),
]
result = mgr.apply(messages)
assert result.reads_stale == 1
assert "stale" in result.messages[1]["content"].lower()
def test_edit_different_file_not_stale(self):
"""Read(A) → Edit(B): Read(A) stays fresh."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_edit("e1", "/src/other.py"),
make_openai_tool_result("e1", "edit success"),
]
result = mgr.apply(messages)
assert result.reads_stale == 0
assert result.reads_fresh == 1
assert result.messages[1]["content"] == LARGE_CONTENT
def test_multiple_reads_all_stale(self):
"""Read(A) × 3 → Edit(A): all 3 Reads become stale."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_read("r2", "/src/app.py"),
make_openai_tool_result("r2", LARGE_CONTENT + "_v2"),
make_openai_read("r3", "/src/app.py"),
make_openai_tool_result("r3", LARGE_CONTENT + "_v3"),
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "edit success"),
]
result = mgr.apply(messages)
# All 3 reads are stale (edit happened after all of them)
assert result.reads_stale == 3
assert result.reads_fresh == 0
def test_compress_stale_disabled(self):
"""compress_stale=False: stale Reads are not replaced."""
config = ReadLifecycleConfig(enabled=True, compress_stale=False)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "edit success"),
]
result = mgr.apply(messages)
# With compress_stale=False but compress_superseded=True,
# Read is superseded by nothing (only one read), and not stale → fresh
assert result.reads_fresh == 1
assert result.messages[1]["content"] == LARGE_CONTENT
class TestSupersededDetection:
"""Read outputs become superseded when the same file is re-Read."""
def test_reread_makes_superseded(self):
"""Read(A) → Read(A): first Read becomes superseded."""
config = ReadLifecycleConfig(enabled=True, compress_superseded=True)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_read("r2", "/src/app.py"),
make_openai_tool_result("r2", LARGE_CONTENT + "_updated"),
]
result = mgr.apply(messages)
assert result.reads_superseded == 1
assert result.reads_fresh == 1
# First read replaced, second read untouched
assert "superseded" in result.messages[1]["content"].lower()
assert result.messages[3]["content"] == LARGE_CONTENT + "_updated"
def test_compress_superseded_disabled(self):
"""compress_superseded=False: superseded Reads not replaced."""
config = ReadLifecycleConfig(enabled=True, compress_superseded=False)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_read("r2", "/src/app.py"),
make_openai_tool_result("r2", LARGE_CONTENT + "_updated"),
]
result = mgr.apply(messages)
# Both reads are fresh (superseded detection disabled)
assert result.reads_fresh == 2
assert result.messages[1]["content"] == LARGE_CONTENT
class TestFreshReads:
"""Fresh Reads must never be modified."""
def test_single_read_stays_fresh(self):
"""One Read, no Edit: stays fresh."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
]
result = mgr.apply(messages)
assert result.reads_fresh == 1
assert result.reads_stale == 0
assert result.reads_superseded == 0
assert result.messages[1]["content"] == LARGE_CONTENT
def test_read_edit_read_chain(self):
"""Read(A) → Edit(A) → Read(A): first stale, second fresh."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "edit success"),
make_openai_read("r2", "/src/app.py"),
make_openai_tool_result("r2", LARGE_CONTENT + "_v2"),
]
result = mgr.apply(messages)
# First read: stale (edit happened after) AND superseded (re-read after)
# → classified as stale (stale takes priority)
assert result.reads_stale == 1
# Second read: fresh (latest, no edit after)
assert result.reads_fresh == 1
assert "stale" in result.messages[1]["content"].lower()
assert result.messages[5]["content"] == LARGE_CONTENT + "_v2"
class TestMultipleFiles:
"""Lifecycle management across multiple files."""
def test_independent_files(self):
"""Read(A) → Edit(A) → Read(B): A stale, B fresh."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "edit success"),
make_openai_read("r2", "/src/utils.py"),
make_openai_tool_result("r2", LARGE_CONTENT + "_utils"),
]
result = mgr.apply(messages)
assert result.reads_stale == 1
assert result.reads_fresh == 1
assert "stale" in result.messages[1]["content"].lower()
assert result.messages[5]["content"] == LARGE_CONTENT + "_utils"
class TestSizeGating:
"""Small Read outputs should be skipped."""
def test_small_read_not_replaced(self):
"""Read output below min_size_bytes: not replaced even if stale."""
config = ReadLifecycleConfig(enabled=True, min_size_bytes=512)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", SMALL_CONTENT), # 4 bytes
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "edit success"),
]
result = mgr.apply(messages)
# Stale but too small to replace
assert result.messages[1]["content"] == SMALL_CONTENT
class TestAnthropicFormat:
"""Lifecycle works with Anthropic message format."""
def test_anthropic_stale_read(self):
"""Anthropic format: Read(A) → Edit(A): Read becomes stale."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_anthropic_read("r1", "/src/app.py"),
make_anthropic_tool_result("r1", LARGE_CONTENT),
make_anthropic_edit("e1", "/src/app.py"),
make_anthropic_tool_result("e1", "edit success"),
]
result = mgr.apply(messages)
assert result.reads_stale == 1
# Check the tool_result block inside the user message was replaced
user_msg = result.messages[1]
tool_result_block = user_msg["content"][0]
assert "stale" in tool_result_block["content"].lower()
assert "hash=" in tool_result_block["content"]
def test_anthropic_fresh_read(self):
"""Anthropic format: single Read stays fresh."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_anthropic_read("r1", "/src/app.py"),
make_anthropic_tool_result("r1", LARGE_CONTENT),
]
result = mgr.apply(messages)
assert result.reads_fresh == 1
user_msg = result.messages[1]
assert user_msg["content"][0]["content"] == LARGE_CONTENT
def test_failed_edit_does_not_make_read_stale(self):
"""A tool_result with is_error means the edit did not happen."""
mgr = ReadLifecycleManager(ReadLifecycleConfig(enabled=True))
failed = make_anthropic_tool_result("e1", REJECTED_EDIT)
failed["content"][0]["is_error"] = True
messages = [
make_anthropic_read("r1", "/src/app.py"),
make_anthropic_tool_result("r1", LARGE_CONTENT),
make_anthropic_edit("e1", "/src/app.py"),
failed,
]
result = mgr.apply(messages)
assert result.reads_stale == 0
assert result.reads_fresh == 1
assert result.messages[1]["content"][0]["content"] == LARGE_CONTENT
def test_successful_edit_after_failed_edit_still_makes_read_stale(self):
mgr = ReadLifecycleManager(ReadLifecycleConfig(enabled=True))
failed = make_anthropic_tool_result("e1", REJECTED_EDIT)
failed["content"][0]["is_error"] = True
messages = [
make_anthropic_read("r1", "/src/app.py"),
make_anthropic_tool_result("r1", LARGE_CONTENT),
make_anthropic_edit("e1", "/src/app.py"),
failed,
make_anthropic_edit("e2", "/src/app.py"),
make_anthropic_tool_result("e2", "edit success"),
]
result = mgr.apply(messages)
assert result.reads_stale == 1
def test_interrupted_edit_still_makes_read_stale(self):
"""An error that is not a pre-run rejection may have changed the file."""
mgr = ReadLifecycleManager(ReadLifecycleConfig(enabled=True))
failed = make_anthropic_tool_result(
"e1",
"Interrupted: the response that made this tool call was stopped while "
"the call was running; it may have partially or fully completed.",
)
failed["content"][0]["is_error"] = True
messages = [
make_anthropic_read("r1", "/src/app.py"),
make_anthropic_tool_result("r1", LARGE_CONTENT),
make_anthropic_edit("e1", "/src/app.py"),
failed,
]
result = mgr.apply(messages)
assert result.reads_stale == 1
def test_rejected_edit_with_text_block_content_keeps_read_fresh(self):
mgr = ReadLifecycleManager(ReadLifecycleConfig(enabled=True))
failed = make_anthropic_tool_result("e1", "")
failed["content"][0]["content"] = [{"type": "text", "text": REJECTED_EDIT}]
failed["content"][0]["is_error"] = True
messages = [
make_anthropic_read("r1", "/src/app.py"),
make_anthropic_tool_result("r1", LARGE_CONTENT),
make_anthropic_edit("e1", "/src/app.py"),
failed,
]
result = mgr.apply(messages)
assert result.reads_stale == 0
assert result.messages[1]["content"][0]["content"] == LARGE_CONTENT
def test_error_result_with_non_string_tool_use_id_is_ignored(self):
mgr = ReadLifecycleManager(ReadLifecycleConfig(enabled=True))
bad = make_anthropic_tool_result("e1", REJECTED_EDIT)
bad["content"][0]["tool_use_id"] = ["e1"]
bad["content"][0]["is_error"] = True
messages = [
make_anthropic_read("r1", "/src/app.py"),
make_anthropic_tool_result("r1", LARGE_CONTENT),
bad,
]
result = mgr.apply(messages)
assert result.reads_fresh == 1
assert result.messages[1]["content"][0]["content"] == LARGE_CONTENT
class TestCCRStoreIntegration:
"""Lifecycle manager stores originals in CCR."""
def test_original_stored_in_ccr(self):
"""When a Read is replaced, original content is stored in CCR."""
class MockStore:
def __init__(self):
self.stored = []
def store(self, **kwargs):
self.stored.append(kwargs)
return "mock_hash_1234567890ab"
mock_store = MockStore()
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config, compression_store=mock_store)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "edit success"),
]
result = mgr.apply(messages)
assert len(mock_store.stored) == 1
assert mock_store.stored[0]["original"] == LARGE_CONTENT
assert mock_store.stored[0]["tool_name"] == "Read"
assert "mock_hash_1234567890ab" in result.messages[1]["content"]
assert result.ccr_hashes == ["mock_hash_1234567890ab"]
def test_no_store_uses_content_hash(self):
"""Without CCR store, marker uses content-derived hash."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config, compression_store=None)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "edit success"),
]
result = mgr.apply(messages)
assert "hash=" in result.messages[1]["content"]
class TestTransformTracking:
"""Lifecycle transforms are tracked correctly."""
def test_transforms_recorded(self):
"""Each replacement generates a transform entry."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_read("r2", "/src/app.py"),
make_openai_tool_result("r2", LARGE_CONTENT),
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "done"),
]
result = mgr.apply(messages)
stale_transforms = [t for t in result.transforms_applied if "stale" in t]
assert len(stale_transforms) == 2 # Both reads are stale
def test_transform_tag_includes_file_path_openai(self):
"""OpenAI-format tag shape is ``read_lifecycle:<state>:<file_path>``."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
make_openai_read("r1", "/src/app.py"),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "done"),
]
result = mgr.apply(messages)
assert "read_lifecycle:stale:/src/app.py" in result.transforms_applied
def test_transform_tag_includes_file_path_anthropic(self):
"""Anthropic-format tag shape matches OpenAI tag shape."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
{
"role": "assistant",
"content": [
{
"type": "tool_use",
"id": "r1",
"name": "Read",
"input": {"file_path": "/src/notes.md"},
}
],
},
{
"role": "user",
"content": [{"type": "tool_result", "tool_use_id": "r1", "content": LARGE_CONTENT}],
},
{
"role": "assistant",
"content": [
{
"type": "tool_use",
"id": "e1",
"name": "Edit",
"input": {
"file_path": "/src/notes.md",
"old_string": "old",
"new_string": "new",
},
}
],
},
{
"role": "user",
"content": [{"type": "tool_result", "tool_use_id": "e1", "content": "done"}],
},
]
result = mgr.apply(messages)
assert "read_lifecycle:stale:/src/notes.md" in result.transforms_applied
def test_transform_tag_preserves_colons_in_path(self):
"""Paths containing ``:`` survive — consumers must bound their split."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
weird_path = "/tmp/has:colon/file.py"
messages = [
make_openai_read("r1", weird_path),
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_edit("e1", weird_path),
make_openai_tool_result("e1", "done"),
]
result = mgr.apply(messages)
tag = next(t for t in result.transforms_applied if t.startswith("read_lifecycle:stale"))
assert tag.split(":", 2) == ["read_lifecycle", "stale", weird_path]
class TestNoFilePathHandling:
"""Reads without parseable file_path should be left alone."""
def test_read_without_file_path(self):
"""Read with no file_path in arguments: treated as unknown, not matched."""
config = ReadLifecycleConfig(enabled=True)
mgr = ReadLifecycleManager(config)
messages = [
{
"role": "assistant",
"content": None,
"tool_calls": [
{
"id": "r1",
"type": "function",
"function": {"name": "Read", "arguments": "{}"},
}
],
},
make_openai_tool_result("r1", LARGE_CONTENT),
make_openai_edit("e1", "/src/app.py"),
make_openai_tool_result("e1", "done"),
]
result = mgr.apply(messages)
# Can't match file_path, so Read is not classified at all
assert result.reads_total == 0
assert result.messages[1]["content"] == LARGE_CONTENT
def _read_then_edit(read_content: str) -> list[dict]:
return [
{
"role": "assistant",
"content": [
{
"type": "tool_use",
"id": "t1",
"name": "Read",
"input": {"file_path": "/tmp/foo.txt"},
}
],
},
{
"role": "user",
"content": [
{
"type": "tool_result",
"tool_use_id": "t1",
"content": read_content,
}
],
},
# Edit the same file -> the Read above becomes STALE.
{
"role": "assistant",
"content": [
{
"type": "tool_use",
"id": "t2",
"name": "Edit",
"input": {"file_path": "/tmp/foo.txt"},
}
],
},
{
"role": "user",
"content": [
{
"type": "tool_result",
"tool_use_id": "t2",
"content": "edited",
}
],
},
]
class TestContentRouterIntegration:
"""Regression: ContentRouter.transform must wire a real CCR store into
ReadLifecycleManager so STALE Read markers resolve via headroom_retrieve."""
def test_stale_read_marker_retrievable_via_compress(self, monkeypatch):
import re
# Force an in-memory backend so the test is hermetic.
monkeypatch.setenv("HEADROOM_CCR_BACKEND", "memory")
from headroom import compress
from headroom.cache.compression_store import (
get_compression_store,
reset_compression_store,
)
reset_compression_store()
try:
large_content = "source line\n" * 500 # above read_lifecycle min_size_bytes
messages = _read_then_edit(large_content)
result = compress(messages, model="claude-sonnet-4-5-20250929")
hashes: list[str] = []
for m in result.messages:
content = m.get("content")
if isinstance(content, list):
for b in content:
if isinstance(b, dict) and b.get("type") == "tool_result":
s = b.get("content", "")
if isinstance(s, str):
hashes.extend(re.findall(r"hash=([a-f0-9]+)", s))
assert hashes, "Expected a STALE Read marker with a hash"
store = get_compression_store()
entry = store.retrieve(hashes[0])
assert entry is not None, "STALE Read marker hash not in CCR store"
assert entry.tool_name == "Read"
assert entry.compression_strategy == "read_lifecycle:stale"
finally:
# Drop the memory-backend singleton so later tests in the suite
# see the env-driven default again.
reset_compression_store()
def test_lossless_router_keeps_a_stale_read(self):
from headroom.providers import OpenAIProvider
from headroom.tokenizer import Tokenizer
from headroom.transforms.content_router import ContentRouter, ContentRouterConfig
tokenizer = Tokenizer(OpenAIProvider().get_token_counter("gpt-4o"), "gpt-4o")
large_content = "source line\n" * 500
router = ContentRouter(ContentRouterConfig(lossless=True))
result = router.apply(_read_then_edit(large_content), tokenizer=tokenizer)
assert result.messages[1]["content"][0]["content"] == large_content