mirror of
https://github.com/headroomlabs-ai/headroom.git
synced 2026-08-27 14:17:10 -04:00
## Description Two related changes to the compression seams, plus the review fixes for both. Supersedes #2661 and #2662, which are closed in favour of this branch — the fixes are inseparable from the code they fix, so reviewing them together is cheaper than landing two PRs and patching them afterwards. **1. A registered lossless provider now competes on the general path.** The `headroom.transforms.lossless_provider` seam was only ever consulted from `_lossless_compact_excluded`, gated on `DEFAULT_EXCLUDE_TOOLS` (`config.py:216` — `Read/Grep/Glob/Write/Edit/WebSearch/WebFetch`). Gateway traffic carries the caller's own tool names — LiteLLM's `headroom` guardrail (https://docs.litellm.ai/docs/proxy/headroom) posts requests containing tools like `search_docs` / `run_ci` / `fetch_rows` — so a registered provider was structurally unreachable for every gateway/sidecar deployment. The seam existed; nothing could get to it. **2. `POST /v1/compress` is marker-free by default.** A CCR marker is only useful to a caller that also injects the `headroom_retrieve` tool AND can reach `/v1/retrieve`. Neither holds here: tool injection lives in the provider request handlers (`handlers/anthropic.py:1894`), never in `handle_compress`; and every `/v1/retrieve*` route is `Depends(_require_loopback)` (`server.py:4422, 4470, 4749, 4781`) with no remote opt-in — `HEADROOM_COMPRESS_ALLOW_REMOTE` drops the loopback dependency on `/v1/compress` only. So a gateway forwards a `Retrieve more: hash=…` pointer the model cannot follow, and the proxy pays a CCR store write nobody reads. `config.mode="ccr"` opts back in. Closes # ## Type of Change - [ ] Bug fix (non-breaking change that fixes an issue) - [x] New feature (non-breaking change that adds functionality) - [x] Breaking change (fix or feature that would cause existing functionality to change) - [ ] Documentation update - [ ] Performance improvement - [ ] Code refactoring (no functional changes) ## Changes Made ### Seam — `content_router.py`, `lossless_provider.py` - `_lossless_first` (STAGE 0, every block on every path) consults `get_lossless_provider()` and keeps whichever output is smaller. **Strict no-op when no provider is registered**, which is the default; and because it is best-of rather than authoritative, a provider can never do worse than the built-in folds. - Malformed provider output can no longer escape. Every shape check runs inside the `try`: result must be `None`, or a 2-element tuple/list of two `str`. Anything else is ignored at debug level. (Previously the unpack sat outside the `try`, so a 3-tuple raised `ValueError` up through `TransformPipeline.apply`, which re-raises.) - Empty / whitespace-only candidates are rejected rather than silently replacing block content. - **Providers are never offered diff content.** Diff folding is subtractive with no inverse check and a reflowed hunk breaks `git apply` — the same reason the built-in `diff` fold is restricted at `content_router.py:2481`. - The third-party `kind` label is sanitised against `^[a-z0-9_]{1,32}$` before reaching `transforms_applied` and the per-strategy metric dicts, so a caller-controlled string cannot explode Prometheus label cardinality. `fullmatch`, not `match`: `$` also matches before a trailing newline, which would put a newline in a label. - `set_lossless_provider(provider, *, verifier=None)` — in lossless-only mode, where STAGE 0's output is final and there is no marker to recover from, a registered verifier must confirm the fold or the candidate is dropped. No verifier registered = today's behaviour. `provider=None` clears both. - The provider is invoked once per block, not twice (`_has_lossless_fold` probes `_lossless_first` and discards the result, then STAGE 0 recomputes). Bounded memo, wholesale clear on overflow, no lock — a race costs one redundant fold. The memo keys on the provider registration generation, so registering or clearing a provider after a block was already folded takes effect. - The seam docstring now records that the provider runs on the general path and inside the parallel compression pool, so it must be thread-safe as well as deterministic. ### Route — `handlers/openai.py`, `server.py` - `_derived_compress_pipeline(key, **overrides)` replaces the copy-pasted pipeline-derivation block; `_no_ccr_pipeline` (the new default) and `_lossy_inline_pipeline` both use it. - The default pipeline is **built at startup** and included in `_eager_preload_transforms`, so a fresh pod does not pay ContentRouter construction and compressor load on its first request, inside the compression-executor budget. - An unrecognised `config.mode` returns 400 naming the valid values instead of silently falling back to the default. - Claude-family model names resolve their context limit from the Anthropic provider. Real divergence: `bedrock/anthropic.claude-3-5-sonnet` is 200000 there and 128000 on the OpenAI provider. The tokenizer still comes from the OpenAI pipeline's provider — a separate, larger change, noted in a comment. - Documents why the derived router deliberately does **not** share the base router's compression cache: keys do not encode CCR-marker mode, so sharing would leak marker-laden entries into the marker-free path. ### Behavior change A `/v1/compress` caller that relied on default markers now gets none. The only in-tree caller that can resolve them is the TypeScript SDK (`sdk/typescript/src/client.ts:398 retrieve`, `:422 handleToolCall`); it needs `config: {"mode": "ccr"}` to keep today's behaviour, and landing that SDK default in the same release would leave only gateway callers — for whom markers were never resolvable — seeing a difference. No `HEADROOM_COMPRESS_DEFAULT_MODE` compat env deliberately: a flag nobody sets becomes permanent debt, and the wire-level `mode` already covers the one caller that needs it. ## Testing - [x] Unit tests pass (`pytest`) - [x] Linting passes (`ruff check .`) - [x] Type checking passes (`mypy headroom`) - [x] New tests added for new functionality - [x] Manual testing performed ### Test Output ```text $ python -m pytest tests/test_lossless_first_dispatch.py tests/test_lossless_excluded_compaction.py \ tests/test_lossless_mode.py tests/test_lossless_then_lossy.py tests/test_lossless_diff_fold_guard.py \ tests/test_bash_search_lossless_fold.py tests/test_proxy_compress_endpoint.py \ tests/test_ccr_row_drop_store_bridge.py tests/test_gateway_sidecar_ports.py tests/test_compress_api.py \ tests/test_platform_stabilization_functional.py tests/test_proxy_eager_preload_bind.py \ tests/test_proxy_warmup.py tests/test_router_registry_smartcrusher.py -q 189 passed in 32.04s $ ruff check <all 9 changed files> All checks passed! $ ruff format --check <all 9 changed files> 9 files already formatted $ mypy headroom/transforms/content_router.py headroom/transforms/lossless_provider.py \ headroom/proxy/handlers/openai.py headroom/proxy/server.py Success: no issues found in 4 source files ``` New tests cover, one concern each: every malformed provider shape; empty and whitespace-only results; diff content never reaching a provider (call-recording); `kind` sanitisation including the trailing-newline case; the verifier accepting / rejecting / raising; clearing a provider clearing its verifier; single provider invocation per block; memo invalidation on re-registration; unknown and valid `mode` values; the default pipeline existing before any request; and Claude vs OpenAI context-limit resolution with `token_budget` precedence preserved. ## Real Behavior Proof - **Environment:** macOS arm64, Python 3.12.6, proxy built from `_proxy_config_from_env()` with the default `coding` savings profile; Kompress both disabled and offloaded to a remote `kompress-v2-base` endpoint; `HEADROOM_COMPRESSION_TIMEOUT_SECONDS=300`. - **Steps:** `POST /v1/compress` over `TestClient` with OpenAI-shaped payloads under non-excluded tool names (`run_ci`, `list_files`, `code_search`, `fetch_rows`) — a CI log with ANSI escapes and repeated lines, a 160-path listing, a 150-line grep dump, a 150-row JSON array; plus a second payload with a RAG user blob, a 200-row JSON tool result and a 300-line log. Ran with and without a provider registered via `set_lossless_provider`. - **Observed — seam reachability:** | Kompress | no provider registered | provider registered | |---|---|---| | off | 19,284 → 10,265 tokens (46.8%) | 19,284 → **7,879 (59.1%)** | | on (remote) | 19,284 → 9,366 tokens (51.4%) | 19,284 → **7,146 (62.9%)** | Before this change the right-hand column was identical to the left — the registered provider was never called on this payload. - **Observed — marker-free default costs nothing:** | Config | tokens | saved | |---|---|---| | markers on (previous default) | 37,791 → 24,415 | 35.4% | | markers off (new default) | 37,791 → 24,415 | **35.4% — identical** | | `mode="lossy_inline"` | 37,791 → 25,129 | 33.5% | | `--lossless` | 37,791 → 35,100 | 7.1% | - **Observed — memo staleness, before the fix:** registering a provider that folds a grep block to 5 bytes left the block at its 1496-byte built-in fold, and clearing a provider kept serving the provider's output. Both correct after keying on the registration generation. - **Observed — context limit:** `bedrock/anthropic.claude-3-5-sonnet` resolves 200000 via the Anthropic provider, 128000 via the OpenAI provider. - **Not tested:** `tests/test_transforms_content_router.py` was not run — it does not complete on this machine, wedging on its 5th test while that test passes in 5.7s alone. Verified pre-existing before this work: with the diff stashed, the clean tree stalled at the identical test, and it stalled the same way under `HF_HUB_OFFLINE=1 HEADROOM_OFFLINE=1`. The machine was also out of disk at the time, which may be the real cause rather than the suspected native-detector deadlock (#575) — worth a separate issue either way. CI should be the arbiter here. ## Review Readiness - [x] I have performed a self-review - [x] This PR is ready for human review ## Checklist - [x] My code follows the project's style guidelines - [x] I have performed a self-review of my code - [x] I have commented my code, particularly in hard-to-understand areas - [ ] I have made corresponding changes to the documentation - [x] My changes generate no new warnings - [x] I have added tests that prove my fix is effective or that my feature works - [x] New and existing unit tests pass locally with my changes - [x] I did **not** edit `CHANGELOG.md` ## Additional Notes Two pre-existing test files needed adjusting, both direct consequences rather than scope creep: - `test_platform_stabilization_functional.py::test_v1_compress_success_reports_actual_metrics` patches `openai_pipeline.apply`, which the default mode no longer routes through. **This test was already failing on the marker-free-default commit** before any of the fixes — my original test selection missed it. It now patches the pipeline the route actually uses. - `test_proxy_eager_preload_bind.py` substitutes fake pipelines to control exactly what the preload walks; the eager build injected the real derived router's statuses into an exact-equality assertion. Its shared helper now clears the derived cache, preserving each test's intent without weakening an assertion. Docs unchecked — follow-ups worth doing in the same release: document `config.mode` values in `docs/content/docs/litellm.mdx` and `wiki/proxy.md`; the TS SDK `mode:"ccr"` default; and an operational note that `COMPRESSION_TIMEOUT_SECONDS` defaults to 30 (`helpers.py:687`) while a remote ML endpoint makes one sequential call per unit — on a large payload it trips the executor timeout and the handler fails open, returning `compression_skipped: true` with `tokens_before: 0`, which reads as "nothing to save" rather than "we gave up". Those zeroed counters are misleading and worth a separate fix.
491 lines
19 KiB
Python
491 lines
19 KiB
Python
"""Issue #389: SmartCrusher row-drop CCR hash → Python compression_store bridge.
|
|
|
|
The row-drop and opaque-blob paths in the Rust SmartCrusher emit
|
|
``<<ccr:HASH ...>>`` markers and stash the original payload in the
|
|
Rust process-local CCR store. The Python proxy's ``/v1/retrieve``
|
|
endpoint queries the Python ``compression_store`` (not the Rust one),
|
|
so without a bridge every retrieve call for a Rust-emitted marker
|
|
returns 404.
|
|
|
|
These tests pin the bridge:
|
|
|
|
1. Unit: lossy crush of a 200-item array populates the Python
|
|
compression_store keyed by the same hash that's in the marker, with
|
|
the canonical original retrievable via ``store.retrieve(hash)``.
|
|
|
|
2. Integration: ``/v1/compress`` followed by ``/v1/retrieve/{hash}``
|
|
on the same hash returns 200 with the original content. This is
|
|
the failing case from the issue's reproducer.
|
|
|
|
3. Edge: opaque-blob markers (the ``<<ccr:HASH,KIND,SIZE>>`` shape)
|
|
also bridge — the document walker emits these for long strings.
|
|
|
|
If these regress, ``/v1/retrieve`` silently 404s for every
|
|
SmartCrusher-emitted marker even though the data is held in the Rust
|
|
store. The LLM follows the marker, gets nothing, and the proxy's CCR
|
|
contract is broken at the bridge.
|
|
"""
|
|
|
|
from __future__ import annotations
|
|
|
|
import json
|
|
|
|
import pytest
|
|
|
|
|
|
def _build_extension() -> None:
|
|
try:
|
|
from headroom._core import SmartCrusher # noqa: F401
|
|
except ImportError:
|
|
pytest.skip(
|
|
"headroom._core not built — run `bash scripts/build_rust_extension.sh`",
|
|
allow_module_level=True,
|
|
)
|
|
|
|
|
|
_build_extension()
|
|
|
|
|
|
# Skip if fastapi not available — needed for integration tests but not unit.
|
|
try:
|
|
import fastapi # noqa: F401
|
|
|
|
_HAS_FASTAPI = True
|
|
except ImportError:
|
|
_HAS_FASTAPI = False
|
|
|
|
|
|
# ─── Unit tests: shim populates Python store ─────────────────────────────
|
|
|
|
|
|
def test_lossy_crush_populates_python_compression_store() -> None:
|
|
"""The cornerstone unit test: trigger a row-drop via the Python
|
|
SmartCrusher shim, then verify the Python compression_store has
|
|
an entry keyed by the marker's hash with the canonical original.
|
|
"""
|
|
from headroom.cache.compression_store import (
|
|
get_compression_store,
|
|
reset_compression_store,
|
|
)
|
|
from headroom.config import CCRConfig
|
|
from headroom.config import SmartCrusherConfig as PyConfig
|
|
from headroom.transforms.smart_crusher import SmartCrusher
|
|
|
|
reset_compression_store()
|
|
try:
|
|
crusher = SmartCrusher(PyConfig(), ccr_config=CCRConfig(), with_compaction=False)
|
|
|
|
# 60 items is well above adaptive_k → lossy path fires.
|
|
original = [{"id": i, "status": "ok", "tag": "alpha"} for i in range(60)]
|
|
original_json = json.dumps(original)
|
|
|
|
result = crusher.crush_array_json(original_json)
|
|
|
|
# Sanity: lossy path actually fired.
|
|
assert result["ccr_hash"] is not None, (
|
|
f"expected lossy drop, got strategy={result['strategy_info']!r}"
|
|
)
|
|
ccr_hash = result["ccr_hash"]
|
|
# The marker text embeds the same hash.
|
|
assert ccr_hash in result["dropped_summary"], (
|
|
f"marker {result['dropped_summary']!r} should embed hash {ccr_hash}"
|
|
)
|
|
|
|
# The bridge: Python compression_store now holds an entry keyed
|
|
# by the Rust-emitted hash.
|
|
store = get_compression_store()
|
|
entry = store.retrieve(ccr_hash)
|
|
assert entry is not None, (
|
|
f"compression_store has no entry for hash {ccr_hash!r}; "
|
|
f"the bridge dropped the row-drop hash on the floor"
|
|
)
|
|
|
|
# The entry's original content is the canonical-JSON
|
|
# serialization of the original array — same bytes the Rust
|
|
# store has under the same hash.
|
|
rust_canonical = crusher.ccr_get(ccr_hash)
|
|
assert rust_canonical is not None, "Rust store lost the entry"
|
|
assert entry.original_content == rust_canonical, (
|
|
"Python store's canonical bytes diverged from Rust store"
|
|
)
|
|
|
|
# Round-trip: parse the canonical and compare to input.
|
|
retrieved = json.loads(entry.original_content)
|
|
assert retrieved == original
|
|
finally:
|
|
reset_compression_store()
|
|
|
|
|
|
def test_smart_crush_content_populates_python_store() -> None:
|
|
"""Same bridge but driven through the runtime path the proxy
|
|
actually uses: `_smart_crush_content` (not `crush_array_json`).
|
|
This is what `apply()` invokes per-message."""
|
|
from headroom.cache.compression_store import (
|
|
get_compression_store,
|
|
reset_compression_store,
|
|
)
|
|
from headroom.config import CCRConfig
|
|
from headroom.config import SmartCrusherConfig as PyConfig
|
|
from headroom.transforms.smart_crusher import SmartCrusher
|
|
|
|
reset_compression_store()
|
|
try:
|
|
crusher = SmartCrusher(PyConfig(), ccr_config=CCRConfig(), with_compaction=False)
|
|
|
|
# Mix of "ok" + occasional "error" → variance the lossy path
|
|
# can latch onto. A purely-uniform array gets a `skip:unique_
|
|
# entities_no_signal` strategy and never row-drops.
|
|
original = [
|
|
{
|
|
"id": i,
|
|
"level": "error" if i % 30 == 0 else "info",
|
|
"msg": f"line {i}",
|
|
}
|
|
for i in range(80)
|
|
]
|
|
content = json.dumps(original)
|
|
|
|
crushed, was_modified, info = crusher._smart_crush_content(content)
|
|
assert was_modified, f"expected lossy modification, got info={info!r}"
|
|
assert "<<ccr:" in crushed, (
|
|
f"expected CCR marker in output (info={info!r}): {crushed[:200]!r}"
|
|
)
|
|
|
|
# Pull the hash from the rendered output via JSON parse.
|
|
# The output is a JSON array with a `_ccr_dropped` sentinel.
|
|
parsed = json.loads(crushed)
|
|
assert isinstance(parsed, list), f"expected JSON array, got {type(parsed).__name__}"
|
|
sentinel = parsed[-1]
|
|
assert isinstance(sentinel, dict) and "_ccr_dropped" in sentinel, (
|
|
f"expected _ccr_dropped sentinel as last element, got {sentinel!r}"
|
|
)
|
|
marker_text = sentinel["_ccr_dropped"]
|
|
assert marker_text.startswith("<<ccr:") and "rows_offloaded>>" in marker_text
|
|
|
|
# Extract hash by structural slice (no regex).
|
|
ccr_hash = marker_text[len("<<ccr:") :].split(" ", 1)[0]
|
|
assert all(c in "0123456789abcdef" for c in ccr_hash), f"hash should be hex: {ccr_hash!r}"
|
|
|
|
# The Python store now has the entry under this hash.
|
|
store = get_compression_store()
|
|
entry = store.retrieve(ccr_hash)
|
|
assert entry is not None, (
|
|
f"_smart_crush_content didn't bridge hash {ccr_hash!r} to "
|
|
f"Python store; /v1/retrieve would 404"
|
|
)
|
|
|
|
# Original is recoverable byte-for-byte from the bridged entry.
|
|
retrieved = json.loads(entry.original_content)
|
|
assert retrieved == original
|
|
finally:
|
|
reset_compression_store()
|
|
|
|
|
|
def test_passthrough_does_not_populate_store() -> None:
|
|
"""Below adaptive_k → no row drop → no Python store write.
|
|
Pins that we don't accidentally store on every crush call."""
|
|
from headroom.cache.compression_store import (
|
|
get_compression_store,
|
|
reset_compression_store,
|
|
)
|
|
from headroom.config import CCRConfig
|
|
from headroom.config import SmartCrusherConfig as PyConfig
|
|
from headroom.transforms.smart_crusher import SmartCrusher
|
|
|
|
reset_compression_store()
|
|
try:
|
|
crusher = SmartCrusher(PyConfig(), ccr_config=CCRConfig(), with_compaction=False)
|
|
|
|
# 3 items: well below the threshold; no compression happens.
|
|
small = json.dumps([{"id": i} for i in range(3)])
|
|
crusher._smart_crush_content(small)
|
|
|
|
store = get_compression_store()
|
|
stats = store.get_stats()
|
|
assert stats["entry_count"] == 0, (
|
|
f"passthrough crush should not write to compression_store, "
|
|
f"got {stats['entry_count']} entries"
|
|
)
|
|
finally:
|
|
reset_compression_store()
|
|
|
|
|
|
def test_marker_disabled_skips_python_store() -> None:
|
|
"""`ccr_config.enabled=False` flips off marker emission AND store
|
|
writes on the Rust side. The bridge should have nothing to do."""
|
|
from headroom.cache.compression_store import (
|
|
get_compression_store,
|
|
reset_compression_store,
|
|
)
|
|
from headroom.config import CCRConfig
|
|
from headroom.config import SmartCrusherConfig as PyConfig
|
|
from headroom.transforms.smart_crusher import SmartCrusher
|
|
|
|
reset_compression_store()
|
|
try:
|
|
# Markers disabled → Rust skips both marker emission and store write.
|
|
crusher = SmartCrusher(
|
|
PyConfig(),
|
|
ccr_config=CCRConfig(enabled=False),
|
|
with_compaction=False,
|
|
)
|
|
|
|
original = [{"id": i, "status": "ok"} for i in range(60)]
|
|
crushed, was_modified, _info = crusher._smart_crush_content(json.dumps(original))
|
|
|
|
# Compression still happens (rows still drop), but no marker.
|
|
assert was_modified
|
|
assert "<<ccr:" not in crushed, (
|
|
f"expected no marker when ccr_config.enabled=False, got: {crushed[:200]!r}"
|
|
)
|
|
|
|
# And the Python store stays empty — bridge had nothing to mirror.
|
|
store = get_compression_store()
|
|
assert store.get_stats()["entry_count"] == 0
|
|
finally:
|
|
reset_compression_store()
|
|
|
|
|
|
def test_distinct_payloads_get_distinct_python_store_entries() -> None:
|
|
"""Two unrelated payloads → two row drops → two entries under
|
|
distinct hashes in the Python store; both retrievable independently."""
|
|
from headroom.cache.compression_store import (
|
|
get_compression_store,
|
|
reset_compression_store,
|
|
)
|
|
from headroom.config import CCRConfig
|
|
from headroom.config import SmartCrusherConfig as PyConfig
|
|
from headroom.transforms.smart_crusher import SmartCrusher
|
|
|
|
reset_compression_store()
|
|
try:
|
|
crusher = SmartCrusher(PyConfig(), ccr_config=CCRConfig(), with_compaction=False)
|
|
|
|
a = [{"id": i, "tag": "alpha"} for i in range(50)]
|
|
b = [{"id": i, "tag": "beta"} for i in range(50)]
|
|
|
|
ra = crusher.crush_array_json(json.dumps(a))
|
|
rb = crusher.crush_array_json(json.dumps(b))
|
|
|
|
ha, hb = ra["ccr_hash"], rb["ccr_hash"]
|
|
assert ha and hb and ha != hb
|
|
|
|
store = get_compression_store()
|
|
ea = store.retrieve(ha)
|
|
eb = store.retrieve(hb)
|
|
assert ea is not None and eb is not None
|
|
assert json.loads(ea.original_content) == a
|
|
assert json.loads(eb.original_content) == b
|
|
finally:
|
|
reset_compression_store()
|
|
|
|
|
|
# ─── compression_store: explicit_hash parameter ───────────────────────────
|
|
|
|
|
|
def test_compression_store_explicit_hash_round_trips() -> None:
|
|
"""The new `explicit_hash` parameter on `store.store()` keys the
|
|
entry by the caller-supplied hash instead of MD5(original)[:24]."""
|
|
from headroom.cache.compression_store import (
|
|
get_compression_store,
|
|
reset_compression_store,
|
|
)
|
|
|
|
reset_compression_store()
|
|
try:
|
|
store = get_compression_store()
|
|
# SmartCrusher emits 12-char SHA-256 hashes — much shorter than
|
|
# the default MD5[:24].
|
|
explicit = "abc123def456"
|
|
returned = store.store(
|
|
original='[{"id":1}]',
|
|
compressed="<<placeholder>>",
|
|
explicit_hash=explicit,
|
|
)
|
|
assert returned == explicit
|
|
entry = store.retrieve(explicit)
|
|
assert entry is not None
|
|
assert entry.original_content == '[{"id":1}]'
|
|
finally:
|
|
reset_compression_store()
|
|
|
|
|
|
def test_compression_store_explicit_hash_rejects_non_hex() -> None:
|
|
"""Non-hex `explicit_hash` raises ValueError. No silent fallback
|
|
to MD5 — that would re-introduce the marker/store mismatch."""
|
|
from headroom.cache.compression_store import (
|
|
get_compression_store,
|
|
reset_compression_store,
|
|
)
|
|
|
|
reset_compression_store()
|
|
try:
|
|
store = get_compression_store()
|
|
with pytest.raises(ValueError, match="hex string"):
|
|
store.store(
|
|
original="x",
|
|
compressed="y",
|
|
explicit_hash="NOT_HEX!@#",
|
|
)
|
|
with pytest.raises(ValueError, match="hex string"):
|
|
store.store(
|
|
original="x",
|
|
compressed="y",
|
|
explicit_hash="",
|
|
)
|
|
finally:
|
|
reset_compression_store()
|
|
|
|
|
|
# ─── Integration test: /v1/compress → /v1/retrieve via FastAPI ──────────
|
|
|
|
|
|
@pytest.mark.skipif(not _HAS_FASTAPI, reason="fastapi not installed")
|
|
def test_v1_compress_then_v1_retrieve_resolves_marker_hash() -> None:
|
|
"""End-to-end issue #389 reproducer:
|
|
1. POST a 200-item tool message to /v1/compress.
|
|
2. Parse the `<<ccr:HASH N_rows_offloaded>>` marker out of the
|
|
compressed messages.
|
|
3. GET /v1/retrieve/{hash} — must return 200 with the original.
|
|
|
|
Before the fix this returns 404 because the Rust crusher's marker
|
|
points at a hash the Python compression_store never received.
|
|
"""
|
|
from fastapi.testclient import TestClient
|
|
|
|
from headroom.cache.compression_store import reset_compression_store
|
|
from headroom.proxy.server import ProxyConfig, create_app
|
|
|
|
reset_compression_store()
|
|
config = ProxyConfig(
|
|
optimize=True,
|
|
cache_enabled=False,
|
|
rate_limit_enabled=False,
|
|
cost_tracking_enabled=False,
|
|
)
|
|
app = create_app(config)
|
|
|
|
# Build a payload similar to the issue's reproducer — 200 items
|
|
# with enough variation to trigger the lossy path. The Rust
|
|
# crusher's adaptive_k will keep ~15 and drop the rest.
|
|
#
|
|
# The blob is unique-per-item and long relative to the key names so
|
|
# the lossless Table/CSV path (which wins by stripping repeated keys
|
|
# when it saves >= lossless_min_savings_ratio) cannot clear the bar —
|
|
# this test exists to exercise the LOSSY row-drop path and its
|
|
# Rust -> Python CCR store bridge.
|
|
items = [
|
|
{
|
|
"id": i,
|
|
"score": 0.99 if i % 30 == 0 else 0.6,
|
|
"msg": f"Result {i:03d}{' error' if i % 30 == 0 else ' ok'}",
|
|
"blob": f"payload-{i:04d}-" + "".join(chr(97 + (i * 7 + j) % 26) for j in range(240)),
|
|
}
|
|
for i in range(200)
|
|
]
|
|
req = {
|
|
"model": "gpt-4o",
|
|
# /v1/compress is marker-free by default (no gateway caller can resolve a
|
|
# marker); mode="ccr" is the opt-in for callers that run the retrieve loop.
|
|
"config": {"mode": "ccr"},
|
|
"messages": [
|
|
{"role": "user", "content": "Get items"},
|
|
{
|
|
"role": "assistant",
|
|
"content": None,
|
|
"tool_calls": [
|
|
{
|
|
"id": "c1",
|
|
"type": "function",
|
|
"function": {"name": "get", "arguments": "{}"},
|
|
}
|
|
],
|
|
},
|
|
{"role": "tool", "tool_call_id": "c1", "content": json.dumps(items)},
|
|
],
|
|
}
|
|
|
|
try:
|
|
with TestClient(app, base_url="http://127.0.0.1", client=("127.0.0.1", 12345)) as client:
|
|
resp = client.post("/v1/compress", json=req)
|
|
assert resp.status_code == 200, resp.text
|
|
body = resp.json()
|
|
|
|
# The compressed messages should embed at least one CCR marker.
|
|
messages_blob = json.dumps(body["messages"])
|
|
assert "<<ccr:" in messages_blob, (
|
|
f"no CCR marker after /v1/compress on a 200-item array; "
|
|
f"compression didn't fire as expected: "
|
|
f"transforms={body.get('transforms_applied')}"
|
|
)
|
|
|
|
# Pull the hash with a substring scan (no regex).
|
|
start = messages_blob.find("<<ccr:") + len("<<ccr:")
|
|
end = start
|
|
while end < len(messages_blob) and messages_blob[end] in ("0123456789abcdef"):
|
|
end += 1
|
|
ccr_hash = messages_blob[start:end]
|
|
assert ccr_hash, "couldn't extract hash from marker"
|
|
|
|
# The retrieve stats endpoint should now show ≥ 1 entry.
|
|
stats_resp = client.get("/v1/retrieve/stats")
|
|
assert stats_resp.status_code == 200
|
|
stats = stats_resp.json()
|
|
assert stats["store"]["entry_count"] >= 1, (
|
|
f"compression_store empty after /v1/compress; stats={stats!r}"
|
|
)
|
|
|
|
# The actual /v1/retrieve call from the issue:
|
|
retrieve_resp = client.post("/v1/retrieve", json={"hash": ccr_hash})
|
|
assert retrieve_resp.status_code == 200, (
|
|
f"/v1/retrieve for marker hash {ccr_hash!r} returned "
|
|
f"{retrieve_resp.status_code} ({retrieve_resp.text}). "
|
|
f"The Rust→Python store bridge dropped the entry."
|
|
)
|
|
retrieve_body = retrieve_resp.json()
|
|
assert retrieve_body["hash"] == ccr_hash
|
|
# The retrieved content should parse back to a JSON array
|
|
# of the original items.
|
|
retrieved_items = json.loads(retrieve_body["original_content"])
|
|
assert isinstance(retrieved_items, list)
|
|
# The original was 200 items. The Rust hash is over the
|
|
# canonical-JSON form of the parsed input — should round-trip.
|
|
assert len(retrieved_items) == 200, (
|
|
f"expected 200 items in retrieved content, got {len(retrieved_items)}"
|
|
)
|
|
# Spot-check the first item.
|
|
assert retrieved_items[0]["id"] == 0
|
|
|
|
# And the GET shape (used by some clients) returns the same.
|
|
get_resp = client.get(f"/v1/retrieve/{ccr_hash}")
|
|
assert get_resp.status_code == 200
|
|
get_body = get_resp.json()
|
|
assert get_body["hash"] == ccr_hash
|
|
finally:
|
|
reset_compression_store()
|
|
|
|
|
|
@pytest.mark.skipif(not _HAS_FASTAPI, reason="fastapi not installed")
|
|
def test_v1_retrieve_unknown_hash_still_404() -> None:
|
|
"""Sanity: unknown hashes still return 404 (the bridge doesn't
|
|
accidentally make the store too permissive)."""
|
|
from fastapi.testclient import TestClient
|
|
|
|
from headroom.cache.compression_store import reset_compression_store
|
|
from headroom.proxy.server import ProxyConfig, create_app
|
|
|
|
reset_compression_store()
|
|
config = ProxyConfig(
|
|
optimize=False,
|
|
cache_enabled=False,
|
|
rate_limit_enabled=False,
|
|
cost_tracking_enabled=False,
|
|
)
|
|
app = create_app(config)
|
|
|
|
try:
|
|
with TestClient(app, base_url="http://127.0.0.1", client=("127.0.0.1", 12345)) as client:
|
|
resp = client.post("/v1/retrieve", json={"hash": "deadbeef0000"})
|
|
assert resp.status_code == 404
|
|
finally:
|
|
reset_compression_store()
|