Files
headroom/tests/test_memory_handler_project_isolation.py
fda75de505 fix(memory): run only the memory tools the proxy injected on the request (#4870)
## Description

With memory enabled (`--memory`, `--learn`, `wrap/init --memory`), the
proxy ran any upstream tool call named
`memory_list`/`memory_save`/`memory_search`/`memory_update`/`memory_delete`,
and hid it from the client. That held even when the proxy had not
offered those tools on the request (`--no-memory-tools`), or when the
client had declared a tool with that name itself.

The proxy now runs only the memory tools it injected on the current
request. Every other call reaches the client unchanged. The proxy does
not run it, does not withhold it, and sends no continuation upstream.
Anthropic streaming already followed this rule
(`_server_memory_tool_names`). That helper is now shared, and every
executor uses it.

**Behaviour change**
- With `--no-memory-tools`, memory tool calls from the model now reach
the client instead of running on the proxy.
- A memory-named tool that the client declares is now the client's. The
proxy no longer runs it on any path, including the streaming path that
used to run it as a side effect.
- Injected memory tools behave as before: they run on the proxy, are
hidden from the client, and the turn continues upstream.

## Type of Change

- [x] Bug fix (non-breaking change that fixes an issue)
- [ ] New feature (non-breaking change that adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to change)
- [ ] Documentation update
- [ ] Performance improvement
- [ ] Code refactoring (no functional changes)

## Changes Made

- `headroom/proxy/memory_handler.py`: `server_memory_tool_names(tools,
client_tools)` moved here from `AnthropicHandlerMixin`. It now reads
Anthropic, Chat and Responses tool shapes through `extract_tool_name`.
`handle_memory_tool_calls` takes a required `tool_names` and skips every
other call.
- `handlers/openai.py`, Chat Completions (non-streaming): records the
injected names and passes them to the executor.
- `handlers/openai.py`, Responses (HTTP): runs a `function_call` only if
its name was injected. Before, any name in `MEMORY_TOOL_NAMES` ran.
- `handlers/openai.py`, Responses over WebSocket: the injected names are
kept per `response.create`, like the user id. The relay holds back and
runs only those calls.
- `handlers/anthropic.py`, non-streaming: passes the injected names it
already computed for streaming.
- `handlers/streaming.py`: removes the branch that ran memory calls
after the stream had already reached the client. It ran only for tools
the proxy had not injected. Memory handling in the stream (SSE
buffering, the Claude Code credential warning) now turns on only when
the proxy injected memory tools.
- Tests: new `tests/test_memory_tools_only_when_injected.py`. Existing
fixtures that relied on the old behaviour now inject the tool they call:
`test_codex_project_context.py`, `test_memory_handler_native_ops.py`,
`test_memory_handler_project_isolation.py`,
`test_memory_tool_stream.py`.

## 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

The new tests use `create_app` with a fake memory backend and a
`MockTransport` upstream. The WebSocket test uses a scripted upstream
socket. For each path (chat, Responses, Responses over WebSocket,
Anthropic, Anthropic streaming) they check two cases. First, injection
is off and the model calls `memory_delete`. Second, injection is on and
the client declares its own `memory_delete`. Each case expects zero
backend calls, exactly one upstream request, and the call in the
client's response. The control cases inject `memory_list` and expect it
to run as before.

On `origin/main` (0368a872):
```text
PASSED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_did_not_inject_reaches_the_client[client_declared-anthropic_stream]
PASSED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_injected_still_runs[chat]
PASSED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_injected_still_runs[responses]
PASSED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_injected_still_runs[anthropic]
PASSED tests/test_memory_tools_only_when_injected.py::test_ws_memory_call_the_proxy_injected_still_runs
FAILED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_did_not_inject_reaches_the_client[not_injected-chat]
FAILED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_did_not_inject_reaches_the_client[not_injected-responses]
FAILED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_did_not_inject_reaches_the_client[not_injected-anthropic]
FAILED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_did_not_inject_reaches_the_client[not_injected-anthropic_stream]
FAILED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_did_not_inject_reaches_the_client[client_declared-chat]
FAILED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_did_not_inject_reaches_the_client[client_declared-responses]
FAILED tests/test_memory_tools_only_when_injected.py::test_memory_call_the_proxy_did_not_inject_reaches_the_client[client_declared-anthropic]
FAILED tests/test_memory_tools_only_when_injected.py::test_ws_memory_call_the_proxy_did_not_inject_reaches_the_client[not_injected]
FAILED tests/test_memory_tools_only_when_injected.py::test_ws_memory_call_the_proxy_did_not_inject_reaches_the_client[client_declared]
9 failed, 5 passed in 3.05s
```
On this branch:
```text
14 passed in 3.23s
```
The existing tests for memory, anthropic, openai and streaming (165
files under `tests/`, not `test_live`) pass on both branches:
```text
origin/main: 3183 passed, 44 skipped in 230.96s
this branch: 3197 passed, 44 skipped in 229.83s
```
`uvx ruff@0.17.3 check` and `format --check` pass on the changed files.
`mypy` on the four changed modules reports no issues.

## Real Behavior Proof

- Environment: macOS, Python 3.12,
`create_app(ProxyConfig(memory_enabled=True,
memory_storage_mode="global", ...))` in-process via
`httpx.ASGITransport`. The upstream is `httpx.MockTransport`, the memory
backend is a recording fake, and `LITELLM_LOCAL_MODEL_COST_MAP=True`.
- Exact command / steps: a harness sends one request per path. The
mocked upstream answers with a tool call to `memory_delete`. The harness
records backend calls, upstream request count, and whether the client
response contains the call.
- Observed result (`origin/main` → this branch):
```text
chat       inject=False memory_delete  upstream_calls=2 backend_calls=[delete] client_saw_tool_call=False  →  upstream_calls=1 backend_calls=[] client_saw_tool_call=True
responses  inject=False memory_delete  upstream_calls=2 backend_calls=[delete] client_saw_tool_call=False  →  upstream_calls=1 backend_calls=[] client_saw_tool_call=True
anthropic  inject=False memory_delete  upstream_calls=2 backend_calls=[delete] client_saw_tool_call=False  →  upstream_calls=1 backend_calls=[] client_saw_tool_call=True
anthropic stream inject=False memory_delete  backend_calls=[delete] client_saw_tool_call=True  →  backend_calls=[] client_saw_tool_call=True
chat       inject=True, client-declared memory_delete  upstream_calls=2 backend_calls=[delete]  →  upstream_calls=1 backend_calls=[]
anthropic  inject=True, client-declared memory_delete  upstream_calls=2 backend_calls=[delete]  →  upstream_calls=1 backend_calls=[]
```
- Not tested: a live provider and a live Codex WebSocket session. The
WebSocket path is covered by the in-process test above.

## Runtime Rollout Safety

- Rollout-managed feature(s): none.
- Minimum rollout channel: n/a.
- Stable/default behavior changed: only for memory-named tool calls that
the proxy did not inject. See "Behaviour change" above. The default
`--memory` setup, where the proxy injects the tools, is unchanged.
- Kill switch / disable path: none needed.
- Unsafe override required: no.
- Qualification impact: none.
- Rollback path: revert this PR.

## 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` — it is generated by
release-please from my Conventional Commit PR title (a CI guard enforces
this)

## Additional Notes

Follow-up, not in this PR: on the Chat, Responses and WebSocket paths, a
single turn can call an injected memory tool and a client tool together.
The proxy then continues upstream without an output for the client's
call. That was already true for a client tool such as `shell`. It now
also applies to a memory-named tool that the client declared, which the
proxy used to run. The Anthropic non-streaming path already handles this
by returning the turn to the client. Bringing the OpenAI paths in line
is a separate change.

The docs need no change: `--no-memory-tools` is already documented as
"Disable automatic memory tool injection", which is now what it does.

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

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

---

<a href="https://app.devin.ai/review/headroomlabs-ai/headroom/pull/4870"
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 -->

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: JerrettDavis <mxjerrett@gmail.com>
2026-11-14 10:47:05 +05:30

413 lines
14 KiB
Python

"""End-to-end isolation test for the per-project memory router (GH #462).
Verifies that two sessions running in different working directories
never see each other's memories — neither at search-time, nor in the
injected ``## Relevant Memories`` block.
The test stubs out the real backend so we don't have to load embedders
or open SQLite files. The interesting invariant is that the router
hands a *different* backend instance to each cwd, and that the handler
calls ``search_memories`` on the backend it received (not on the legacy
``self._backend``).
"""
from __future__ import annotations
import asyncio
from pathlib import Path
from types import SimpleNamespace
from typing import Any
import pytest
from headroom.memory import storage_router as sr_mod
from headroom.proxy.memory_handler import (
MemoryConfig,
MemoryHandler,
MemoryMode,
)
class _FakeBackend:
"""Minimal stand-in for ``LocalBackend`` used by the router cache.
Each instance tracks its own write log so we can prove that
Project A's saves and Project B's saves landed on different
backends.
"""
instances: list[_FakeBackend] = []
def __init__(self, cfg: Any) -> None:
self.cfg = cfg
self.saved_contents: list[str] = []
self.search_results: list[Any] = []
# Tag the backend with its db_path so tests can assert on it.
self.db_path = getattr(cfg, "db_path", "<unknown>")
_FakeBackend.instances.append(self)
async def _ensure_initialized(self) -> None:
return None
async def search_memories(self, **kwargs: Any) -> list[Any]:
return list(self.search_results)
async def save_memory(self, **kwargs: Any) -> Any:
content = kwargs["content"]
self.saved_contents.append(content)
return SimpleNamespace(
id=f"mem-{self.db_path}-{len(self.saved_contents)}",
content=content,
metadata={},
)
async def delete_memory(self, memory_id: str) -> bool:
return True
@pytest.fixture(autouse=True)
def patch_backend(monkeypatch: pytest.MonkeyPatch) -> None:
"""Swap LocalBackend out at every import site the handler/router uses."""
_FakeBackend.instances.clear()
monkeypatch.setattr(sr_mod, "LocalBackend", _FakeBackend)
# The handler's _init_backend_locked imports LocalBackend locally;
# patch the same target there too. The route below is the canonical
# import path used by the handler.
import headroom.memory.backends.local as _local_mod
monkeypatch.setattr(_local_mod, "LocalBackend", _FakeBackend)
@pytest.fixture
def handler(tmp_path: Path) -> MemoryHandler:
cfg = MemoryConfig(
enabled=True,
backend="local",
db_path=str(tmp_path / "memory.db"),
inject_context=True,
mode=MemoryMode.AUTO_TAIL,
)
h = MemoryHandler(cfg, agent_type="test")
return h
def _ctx_for_cwd(cwd: str, user_id: str = "alice") -> Any:
return sr_mod.RequestContext(
headers={"x-headroom-cwd": cwd},
system_prompt="",
base_user_id=user_id,
)
def test_two_cwds_route_to_two_backends(handler: MemoryHandler) -> None:
"""Saves under different cwds must land on different backends."""
async def run() -> None:
await handler._ensure_initialized()
ctx_a = _ctx_for_cwd("/Users/me/code/project-a")
ctx_b = _ctx_for_cwd("/Users/me/code/project-b")
await handler._execute_save(
{"content": "Note about Project A's redis config"},
"alice",
"anthropic",
request_context=ctx_a,
)
await handler._execute_save(
{"content": "Note about Project B's auth setup"},
"alice",
"anthropic",
request_context=ctx_b,
)
# The router must have created at least 3 backends:
# 1 legacy (during init), 1 for project-a, 1 for project-b.
backends_with_content = [b for b in _FakeBackend.instances if b.saved_contents]
assert len(backends_with_content) == 2
contents_per_backend = {tuple(b.saved_contents) for b in backends_with_content}
assert ("Note about Project A's redis config",) in contents_per_backend
assert ("Note about Project B's auth setup",) in contents_per_backend
asyncio.run(run())
def test_search_returns_only_current_workspace_memories(handler: MemoryHandler) -> None:
"""Project A's search must not see Project B's memories."""
async def run() -> None:
await handler._ensure_initialized()
ctx_a = _ctx_for_cwd("/Users/me/code/project-a")
ctx_b = _ctx_for_cwd("/Users/me/code/project-b")
# Save under A and B.
await handler._execute_save(
{"content": "A memory"}, "alice", "anthropic", request_context=ctx_a
)
await handler._execute_save(
{"content": "B memory"}, "alice", "anthropic", request_context=ctx_b
)
# Seed each fake backend's search return so we can detect bleed.
backend_a, _, _ = handler._resolve_for_request("alice", ctx_a)
backend_b, _, _ = handler._resolve_for_request("alice", ctx_b)
backend_a.search_results = [ # type: ignore[attr-defined]
SimpleNamespace(
memory=SimpleNamespace(id="a1", content="A memory", metadata={}),
score=0.9,
related_entities=[],
)
]
backend_b.search_results = [ # type: ignore[attr-defined]
SimpleNamespace(
memory=SimpleNamespace(id="b1", content="B memory", metadata={}),
score=0.9,
related_entities=[],
)
]
msgs_a = [{"role": "user", "content": "what's the redis config?"}]
msgs_b = [{"role": "user", "content": "what's the auth setup?"}]
ctx_block_a = await handler.search_and_format_context(
"alice", msgs_a, request_context=ctx_a
)
ctx_block_b = await handler.search_and_format_context(
"alice", msgs_b, request_context=ctx_b
)
assert ctx_block_a is not None
assert ctx_block_b is not None
# The block from Project A must contain A's memory and *not* B's.
assert "A memory" in ctx_block_a
assert "B memory" not in ctx_block_a
# Symmetric assertion for B.
assert "B memory" in ctx_block_b
assert "A memory" not in ctx_block_b
# Provenance header (Fix C) must include the workspace name.
assert "workspace: project-a" in ctx_block_a
assert "workspace: project-b" in ctx_block_b
asyncio.run(run())
def test_legacy_callers_without_ctx_hit_legacy_backend(handler: MemoryHandler) -> None:
"""Callers that don't pass a RequestContext keep the pre-fix shape.
Verifies the backward-compatibility seam: legacy tests / mocks that
call ``search_and_format_context(user, messages)`` get the same
behaviour they had before — the legacy single-DB backend, no scope
header. This is the path tests + qdrant deployments take.
"""
async def run() -> None:
await handler._ensure_initialized()
legacy_backend = handler._backend
legacy_backend.search_results = [ # type: ignore[attr-defined]
SimpleNamespace(
memory=SimpleNamespace(id="g1", content="Global memory", metadata={}),
score=0.9,
related_entities=[],
)
]
block = await handler.search_and_format_context(
"alice", [{"role": "user", "content": "anything"}]
)
assert block is not None
assert "Global memory" in block
# No provenance suffix in legacy header.
assert block.startswith("## Relevant Memories for This User")
asyncio.run(run())
def test_user_mode_partitions_by_user_id(tmp_path: Path) -> None:
"""``--memory-storage=user`` opens one DB per base_user_id."""
cfg = MemoryConfig(
enabled=True,
backend="local",
db_path=str(tmp_path / "memory.db"),
storage_mode=sr_mod.MemoryStorageMode.USER,
)
h = MemoryHandler(cfg, agent_type="test")
async def run() -> None:
await h._ensure_initialized()
ctx_alice = sr_mod.RequestContext(headers={}, system_prompt="", base_user_id="alice")
ctx_bob = sr_mod.RequestContext(headers={}, system_prompt="", base_user_id="bob")
backend_alice, scope_a, _ = h._resolve_for_request("alice", ctx_alice)
backend_bob, scope_b, _ = h._resolve_for_request("bob", ctx_bob)
assert scope_a.mode is sr_mod.MemoryStorageMode.USER
assert scope_b.mode is sr_mod.MemoryStorageMode.USER
assert backend_alice is not backend_bob
asyncio.run(run())
# ---------------------------------------------------------------------------
# Unresolved-project fail-closed (incident 2026-05-26).
#
# When `mode=PROJECT` and `unresolved_project_fallback="empty"` (the new
# default), an inbound request with no project-resolution signal
# (x-headroom-project-id / x-headroom-cwd / x-headroom-project /
# system-prompt cwd:) must return None from search_and_format_context —
# NOT silently pool the
# request's memory into the GLOBAL bucket. The old GLOBAL fallback was
# what surfaced a memory from a prior unrelated TAM-550 session into
# a live PR-review thread, where the agent misread it as a new command.
# ---------------------------------------------------------------------------
@pytest.mark.parametrize("preinitialized", [False, True])
def test_unresolved_project_returns_no_context(tmp_path: Path, preinitialized: bool) -> None:
"""No project signals + PROJECT mode + empty fallback → no memory injection."""
cfg = MemoryConfig(
enabled=True,
backend="local",
db_path=str(tmp_path / "memory.db"),
inject_context=True,
mode=MemoryMode.AUTO_TAIL,
storage_mode=sr_mod.MemoryStorageMode.PROJECT, # PROJECT mode triggers resolution.
# unresolved_project_fallback="empty" — the new default applied
# by MemoryHandler when building the BackendRouterConfig.
)
handler = MemoryHandler(cfg, agent_type="test")
async def run() -> None:
if preinitialized:
await handler._ensure_initialized()
backend_count = len(_FakeBackend.instances)
# Request with NO project-resolution signal: no header, no cwd,
# no parseable system-prompt cwd: line.
ctx_unresolved = sr_mod.RequestContext(
headers={}, # No x-headroom-* headers.
system_prompt="You are helpful.", # No env block.
base_user_id="alice",
)
# Seed a backend so search WOULD return something — to prove the
# gate is at the scope-resolution layer, not just an empty store.
for backend in _FakeBackend.instances:
backend.search_results = [
SimpleNamespace(
memory=SimpleNamespace(
id="should-not-leak", content="Stale prior content", metadata={}
),
score=0.99,
related_entities=[],
)
]
msgs = [{"role": "user", "content": "Just a friendly hello"}]
context = await handler.search_and_format_context("alice", msgs, ctx_unresolved)
assert len(_FakeBackend.instances) == backend_count
assert handler.initialized is preinitialized
# Fail-closed: no memory injected even though backends have data.
assert context is None, (
"Unresolved project in PROJECT mode must skip injection — "
"incident on 2026-05-26 (TAM-550) was caused by the GLOBAL "
"fallback pooling prior-session content into a fresh thread."
)
asyncio.run(run())
@pytest.mark.parametrize(
("tool_name", "input_data"),
[
("memory_save", {"content": "must not persist"}),
("memory_search", {"query": "must not read"}),
("memory_update", {"memory_id": "m1", "new_content": "must not update"}),
("memory_delete", {"memory_id": "m1"}),
("memory_list", {}),
],
)
def test_unresolved_project_memory_tools_fail_closed(
tmp_path: Path, tool_name: str, input_data: dict[str, Any]
) -> None:
cfg = MemoryConfig(
enabled=True,
backend="local",
db_path=str(tmp_path / "memory.db"),
mode=MemoryMode.TOOL,
storage_mode=sr_mod.MemoryStorageMode.PROJECT,
)
handler = MemoryHandler(cfg, agent_type="test")
async def run() -> None:
ctx = sr_mod.RequestContext(
headers={}, system_prompt="You are helpful.", base_user_id="alice"
)
result = await handler._execute_memory_tool(
tool_name, input_data, "alice", request_context=ctx
)
payload = __import__("json").loads(result)
assert payload == {"status": "skipped", "reason": "project_unresolved"}
assert _FakeBackend.instances == []
asyncio.run(run())
def test_unresolved_project_native_memory_tool_fails_closed(tmp_path: Path) -> None:
handler = MemoryHandler(
MemoryConfig(
enabled=True,
backend="local",
db_path=str(tmp_path / "memory.db"),
storage_mode=sr_mod.MemoryStorageMode.PROJECT,
use_native_tool=True,
native_memory_dir=str(tmp_path / "native"),
),
agent_type="test",
)
ctx = sr_mod.RequestContext(headers={}, system_prompt="You are helpful.", base_user_id="alice")
results = asyncio.run(
handler.handle_memory_tool_calls(
{
"content": [
{
"type": "tool_use",
"id": "native-create",
"name": "memory",
"input": {
"command": "create",
"path": "/memories/canary.txt",
"file_text": "must not persist",
},
},
{
"type": "tool_use",
"id": "custom-save",
"name": "memory_save",
"input": {"content": "must not persist"},
},
]
},
"alice",
"anthropic",
request_context=ctx,
tool_names=frozenset({"memory", "memory_save"}),
)
)
assert all(
__import__("json").loads(result["content"])
== {"status": "skipped", "reason": "project_unresolved"}
for result in results
)
assert _FakeBackend.instances == []