fix(P1+P2): canonicalize project names at every trust boundary

Three findings from codex's review of the previous P1+P2 fix. The
earlier commit (f2372ef) only fixed alias resolution at the context
builder. Codex correctly pointed out that the same fragmentation
applies at every other place a project name crosses a boundary —
project_state writes/reads, interaction capture/listing/filtering,
memory create/queries, and reinforcement's downstream queries. Plus
a real bug in the interaction `since` filter where the storage
format and the documented ISO format don't compare cleanly.

The fix is one helper used at every boundary instead of duplicating
the resolution inline.

New helper: src/atocore/projects/registry.py::resolve_project_name
---------------------------------------------------------------
- Single canonicalization boundary for project names
- Returns the canonical project_id when the input matches any
  registered id or alias
- Returns the input unchanged for empty/None and for unregistered
  names (preserves backwards compat with hand-curated state that
  predates the registry)
- Documented as the contract that every read/write at the trust
  boundary should pass through

P1 — Trusted Project State endpoints
------------------------------------
src/atocore/context/project_state.py: set_state, get_state, and
invalidate_state now all canonicalize project_name through
resolve_project_name BEFORE looking up or creating the project row.

Before this fix:
- POST /project/state with project="p05" called ensure_project("p05")
  which created a separate row in the projects table
- The state row was attached to that alias project_id
- Later context builds canonicalized "p05" -> "p05-interferometer"
  via the builder fix from f2372ef and never found the state
- Result: trusted state silently fragmented across alias rows

After this fix:
- The alias is resolved to the canonical id at every entry point
- Two captures (one via "p05", one via "p05-interferometer") write
  to the same row
- get_state via either alias or the canonical id finds the same row

Fixes the highest-priority gap codex flagged because Trusted Project
State is supposed to be the most dependable layer in the AtoCore
trust hierarchy.

P2.a — Interaction capture project canonicalization
----------------------------------------------------
src/atocore/interactions/service.py: record_interaction now
canonicalizes project before storing, so interaction.project is
always the canonical id regardless of what the client passed.

Downstream effects:
- reinforce_from_interaction queries memories by interaction.project
  -> previously missed memories stored under canonical id
  -> now consistent because interaction.project IS the canonical id
- the extractor stamps candidates with interaction.project
  -> previously created candidates in alias buckets
  -> now creates candidates in the canonical bucket
- list_interactions(project=alias) was already broken, now fixed by
  canonicalizing the filter input on the read side too

Memory service applied the same fix:
- src/atocore/memory/service.py: create_memory and get_memories
  both canonicalize project through resolve_project_name
- This keeps stored memory.project consistent with the
  reinforcement query path

P2.b — Interaction `since` filter format normalization
------------------------------------------------------
src/atocore/interactions/service.py: new _normalize_since helper.

The bug:
- created_at is stored as 'YYYY-MM-DD HH:MM:SS' (no timezone, UTC by
  convention) so it sorts lexically and compares cleanly with the
  SQLite CURRENT_TIMESTAMP default
- The `since` parameter was documented as ISO 8601 but compared as
  a raw string against the storage format
- The lexically-greater 'T' separator means an ISO timestamp like
  '2026-04-07T12:00:00Z' is GREATER than the storage form
  '2026-04-07 12:00:00' for the same instant
- Result: a client passing ISO `since` got an empty result for any
  row from the same day, even though those rows existed and were
  technically "after" the cutoff in real-world time

The fix:
- _normalize_since accepts ISO 8601 with T, optional Z suffix,
  optional fractional seconds, optional +HH:MM offsets
- Uses datetime.fromisoformat for parsing (Python 3.11+)
- Converts to UTC and reformats as the storage format before the
  SQL comparison
- The bare storage format still works (backwards compat path is a
  regex match that returns the input unchanged)
- Unparseable input is returned as-is so the comparison degrades
  gracefully (rows just don't match) instead of raising and
  breaking the listing endpoint

builder.py refactor
-------------------
The previous P1 fix had inline canonicalization. Now it uses the
shared helper for consistency:
- import changed from get_registered_project to resolve_project_name
- the inline lookup is replaced with a single helper call
- the comment block now points at representation-authority.md for
  the canonicalization contract

New shared test fixture: tests/conftest.py::project_registry
------------------------------------------------------------
- Standardizes the registry-setup pattern that was duplicated
  across test_context_builder.py, test_project_state.py,
  test_interactions.py, and test_reinforcement.py
- Returns a callable that takes (project_id, [aliases]) tuples
  and writes them into a temp registry file with the env var
  pointed at it and config.settings reloaded
- Used by all 12 new regression tests in this commit

Tests (12 new, all green on first run)
--------------------------------------
test_project_state.py:
- test_set_state_canonicalizes_alias: write via alias, read via
  every alias and the canonical id, verify same row id
- test_get_state_canonicalizes_alias_after_canonical_write
- test_invalidate_state_canonicalizes_alias
- test_unregistered_project_state_still_works (backwards compat)

test_interactions.py:
- test_record_interaction_canonicalizes_project
- test_list_interactions_canonicalizes_project_filter
- test_list_interactions_since_accepts_iso_with_t_separator
- test_list_interactions_since_accepts_z_suffix
- test_list_interactions_since_accepts_offset
- test_list_interactions_since_storage_format_still_works

test_reinforcement.py:
- test_reinforcement_works_when_capture_uses_alias (end-to-end:
  capture under alias, seed memory under canonical, verify
  reinforcement matches)
- test_get_memories_filter_by_alias

Full suite: 174 passing (was 162), 1 warning. The +12 is the
new regression tests, no existing tests regressed.

What's still NOT canonicalized (and why)
----------------------------------------
- _rank_chunks's secondary substring boost in builder.py — the
  retriever already does the right thing via its own
  _project_match_boost which calls get_registered_project. The
  redundant secondary boost still uses the raw hint but it's a
  multiplicative factor on top of correct retrieval, not a
  filter, so it can't drop relevant chunks. Tracked as a future
  cleanup but not a P1.
- update_memory's project field (you can't change a memory's
  project after creation in the API anyway).
- The retriever's project_hint parameter on direct /query calls
  — same reasoning as the builder boost, plus the retriever's
  own get_registered_project call already handles aliases there.
This commit is contained in:
2026-04-07 08:29:33 -04:00
parent f2372eff9e
commit fb6298a9a1
9 changed files with 391 additions and 24 deletions

View File

@@ -1,5 +1,6 @@
"""pytest configuration and shared fixtures."""
import json
import os
import sys
import tempfile
@@ -29,6 +30,45 @@ def tmp_data_dir(tmp_path):
return tmp_path
@pytest.fixture
def project_registry(tmp_path, monkeypatch):
"""Stand up an isolated project registry pointing at a temp file.
Returns a callable that takes one or more (project_id, [aliases])
tuples and writes them into the registry, then forces the in-process
settings singleton to re-resolve. Use this when a test needs the
canonicalization helpers (resolve_project_name, get_registered_project)
to recognize aliases.
"""
registry_path = tmp_path / "test-project-registry.json"
def _set(*projects):
payload = {"projects": []}
for entry in projects:
if isinstance(entry, str):
project_id, aliases = entry, []
else:
project_id, aliases = entry
payload["projects"].append(
{
"id": project_id,
"aliases": list(aliases),
"description": f"test project {project_id}",
"ingest_roots": [
{"source": "vault", "subpath": f"incoming/projects/{project_id}"}
],
}
)
registry_path.write_text(json.dumps(payload), encoding="utf-8")
monkeypatch.setenv("ATOCORE_PROJECT_REGISTRY_PATH", str(registry_path))
from atocore import config
config.settings = config.Settings()
return registry_path
return _set
@pytest.fixture
def sample_markdown(tmp_path) -> Path:
"""Create a sample markdown file for testing."""

View File

@@ -209,3 +209,96 @@ def test_list_interactions_endpoint_returns_summaries(tmp_data_dir):
assert body["interactions"][0]["response_chars"] == 50
# The list endpoint never includes the full response body
assert "response" not in body["interactions"][0]
# --- alias canonicalization on interaction capture/list -------------------
def test_record_interaction_canonicalizes_project(project_registry):
"""Capturing under an alias should store the canonical project id.
Regression for codex's P2 finding: reinforcement and extraction
query memories by interaction.project; if the captured project is
a raw alias they would silently miss memories stored under the
canonical id.
"""
init_db()
project_registry(("p05-interferometer", ["p05", "interferometer"]))
interaction = record_interaction(
prompt="quick capture", response="response body", project="p05", reinforce=False
)
assert interaction.project == "p05-interferometer"
fetched = get_interaction(interaction.id)
assert fetched.project == "p05-interferometer"
def test_list_interactions_canonicalizes_project_filter(project_registry):
init_db()
project_registry(("p06-polisher", ["p06", "polisher"]))
record_interaction(prompt="a", response="ra", project="p06-polisher", reinforce=False)
record_interaction(prompt="b", response="rb", project="polisher", reinforce=False)
record_interaction(prompt="c", response="rc", project="atocore", reinforce=False)
# Query by an alias should still find both p06 captures
via_alias = list_interactions(project="p06")
via_canonical = list_interactions(project="p06-polisher")
assert len(via_alias) == 2
assert len(via_canonical) == 2
assert {i.prompt for i in via_alias} == {"a", "b"}
# --- since filter format normalization ------------------------------------
def test_list_interactions_since_accepts_iso_with_t_separator(tmp_data_dir):
init_db()
record_interaction(prompt="early", response="r", reinforce=False)
time.sleep(1.05)
pivot = record_interaction(prompt="late", response="r", reinforce=False)
# pivot.created_at is in storage format 'YYYY-MM-DD HH:MM:SS'.
# Build the equivalent ISO 8601 with 'T' that an external client
# would naturally send.
iso_with_t = pivot.created_at.replace(" ", "T")
items = list_interactions(since=iso_with_t)
assert any(i.id == pivot.id for i in items)
# The early row must also be excluded if its timestamp is strictly
# before the pivot — since is inclusive on the cutoff
early_ids = {i.id for i in items if i.prompt == "early"}
assert early_ids == set() or len(items) >= 1
def test_list_interactions_since_accepts_z_suffix(tmp_data_dir):
init_db()
pivot = record_interaction(prompt="pivot", response="r", reinforce=False)
time.sleep(1.05)
after = record_interaction(prompt="after", response="r", reinforce=False)
iso_with_z = pivot.created_at.replace(" ", "T") + "Z"
items = list_interactions(since=iso_with_z)
ids = {i.id for i in items}
assert pivot.id in ids
assert after.id in ids
def test_list_interactions_since_accepts_offset(tmp_data_dir):
init_db()
pivot = record_interaction(prompt="pivot", response="r", reinforce=False)
time.sleep(1.05)
after = record_interaction(prompt="after", response="r", reinforce=False)
iso_with_offset = pivot.created_at.replace(" ", "T") + "+00:00"
items = list_interactions(since=iso_with_offset)
assert any(i.id == after.id for i in items)
def test_list_interactions_since_storage_format_still_works(tmp_data_dir):
"""The bare storage format must still work for backwards compatibility."""
init_db()
pivot = record_interaction(prompt="pivot", response="r", reinforce=False)
items = list_interactions(since=pivot.created_at)
assert any(i.id == pivot.id for i in items)

View File

@@ -131,3 +131,68 @@ def test_format_project_state():
def test_format_empty():
"""Test formatting empty state."""
assert format_project_state([]) == ""
# --- Alias canonicalization regression tests --------------------------------
def test_set_state_canonicalizes_alias(project_registry):
"""Writing state via an alias should land under the canonical project id.
Regression for codex's P1 finding: previously /project/state with
project="p05" created a separate alias row that later context builds
(which canonicalize the hint) would never see.
"""
project_registry(("p05-interferometer", ["p05", "interferometer"]))
set_state("p05", "status", "next_focus", "Wave 2 ingestion")
# The state must be reachable via every alias AND the canonical id
via_alias = get_state("p05")
via_canonical = get_state("p05-interferometer")
via_other_alias = get_state("interferometer")
assert len(via_alias) == 1
assert len(via_canonical) == 1
assert len(via_other_alias) == 1
# All three reads return the same row id (no fragmented duplicates)
assert via_alias[0].id == via_canonical[0].id == via_other_alias[0].id
assert via_canonical[0].value == "Wave 2 ingestion"
def test_get_state_canonicalizes_alias_after_canonical_write(project_registry):
"""Reading via an alias should find state written under the canonical id."""
project_registry(("p04-gigabit", ["p04", "gigabit"]))
set_state("p04-gigabit", "status", "phase", "Phase 1 baseline")
via_alias = get_state("gigabit")
assert len(via_alias) == 1
assert via_alias[0].value == "Phase 1 baseline"
def test_invalidate_state_canonicalizes_alias(project_registry):
"""Invalidating via an alias should hit the canonical row."""
project_registry(("p06-polisher", ["p06", "polisher"]))
set_state("p06-polisher", "decision", "frame", "kinematic mounts")
success = invalidate_state("polisher", "decision", "frame")
assert success is True
active = get_state("p06-polisher")
assert len(active) == 0
def test_unregistered_project_state_still_works(project_registry):
"""Hand-curated state for an unregistered project must still round-trip.
Backwards compatibility with state created before the project
registry existed: resolve_project_name returns the input unchanged
when the registry has no record, so the raw name is used as-is.
"""
project_registry() # empty registry
set_state("orphan-project", "status", "phase", "Standalone")
entries = get_state("orphan-project")
assert len(entries) == 1
assert entries[0].value == "Standalone"

View File

@@ -314,3 +314,62 @@ def test_api_post_interactions_accepts_reinforce_false(tmp_data_dir):
reloaded = [m for m in get_memories(memory_type="preference", limit=20) if m.id == mem.id][0]
assert reloaded.confidence == 0.5
assert reloaded.reference_count == 0
# --- alias canonicalization end-to-end -------------------------------------
def test_reinforcement_works_when_capture_uses_alias(project_registry):
"""End-to-end: capture under an alias, seed memory under canonical id,
verify reinforcement still finds and bumps the memory.
Regression for codex's P2 finding: previously interaction.project
was stored verbatim and reinforcement queried memories using that
raw value, so capturing under "p05" while memories live under
"p05-interferometer" silently missed everything.
"""
init_db()
project_registry(("p05-interferometer", ["p05", "interferometer"]))
# Seed an active memory under the CANONICAL id
mem = create_memory(
memory_type="project",
content="the lateral support pads use GF-PTFE for thermal stability",
project="p05-interferometer",
confidence=0.5,
)
# Capture an interaction under the ALIAS — this is the bug case
record_interaction(
prompt="status update",
response=(
"Quick note: the lateral support pads use GF-PTFE for thermal "
"stability and that's still the current selection."
),
project="p05",
)
# The seeded memory should have been reinforced
reloaded = [
m
for m in get_memories(memory_type="project", project="p05-interferometer", limit=20)
if m.id == mem.id
][0]
assert reloaded.confidence > 0.5
assert reloaded.reference_count == 1
def test_get_memories_filter_by_alias(project_registry):
"""Filtering memories by an alias should find rows stored under canonical."""
init_db()
project_registry(("p04-gigabit", ["p04", "gigabit"]))
create_memory(memory_type="project", content="m1", project="p04-gigabit")
create_memory(memory_type="project", content="m2", project="gigabit")
via_alias = get_memories(memory_type="project", project="p04")
via_canonical = get_memories(memory_type="project", project="p04-gigabit")
assert len(via_alias) == 2
assert len(via_canonical) == 2
assert {m.content for m in via_alias} == {"m1", "m2"}