项目文件夹

文件
Ziming Wang ac582c7cf5 fix(memory): parse YAML frontmatter in scan and improve search relevance (#12)
* fix(memory): parse YAML frontmatter in scan and improve search relevance

The memory scanner previously treated the first non-empty line as the
description, returning raw "---" for files with YAML frontmatter.  This
broke search matching because the title and description fields—the only
fields the search function inspected—contained no meaningful content.

Changes:
- Parse name/description/type from YAML frontmatter in scan, consistent
  with the existing skill loader pattern.
- Add body_preview and memory_type fields to MemoryHeader so downstream
  consumers can leverage structured metadata.
- Search now matches against body content in addition to metadata, with
  metadata hits weighted 2x for relevance ordering.
- Tokenizer handles CJK characters for multilingual memory queries.
- Eight new tests covering frontmatter parsing, search relevance ranking,
  body content matching, and CJK tokenization.

* fix: address review feedback on memory scan and search

- Exclude description line from body_preview for non-frontmatter files
  to prevent double-counting in search scoring (meta 2x + body 1x).
- Guard fallback description against '---' delimiters from malformed
  frontmatter, consistent with skills/loader.py behavior.
- Rename CJK→Han in tokenizer docstring and variable names to reflect
  that only Han ideographs are tokenized individually (kana/hangul
  excluded by design as single characters lack lexical meaning).
- Fix CHANGELOG section ordering so existing Added entries stay under
  their original heading.
- Add tests for malformed frontmatter and body_preview exclusion.
2026-04-04 17:57:43 +08:00

202 行
6.8 KiB
Python

"""Tests for memory helpers."""
from __future__ import annotations
from pathlib import Path
from openharness.memory import (
find_relevant_memories,
get_memory_entrypoint,
get_project_memory_dir,
load_memory_prompt,
)
from openharness.memory.scan import _parse_memory_file, scan_memory_files
def test_memory_paths_are_stable(tmp_path: Path, monkeypatch):
monkeypatch.setenv("OPENHARNESS_DATA_DIR", str(tmp_path / "data"))
project_dir = tmp_path / "repo"
project_dir.mkdir()
memory_dir = get_project_memory_dir(project_dir)
entrypoint = get_memory_entrypoint(project_dir)
assert memory_dir.exists()
assert entrypoint.parent == memory_dir
def test_load_memory_prompt_includes_entrypoint(tmp_path: Path, monkeypatch):
monkeypatch.setenv("OPENHARNESS_DATA_DIR", str(tmp_path / "data"))
project_dir = tmp_path / "repo"
project_dir.mkdir()
entrypoint = get_memory_entrypoint(project_dir)
entrypoint.write_text("# Index\n- [Testing](testing.md)\n", encoding="utf-8")
prompt = load_memory_prompt(project_dir)
assert prompt is not None
assert "Persistent memory directory" in prompt
assert "Testing" in prompt
def test_find_relevant_memories(tmp_path: Path, monkeypatch):
monkeypatch.setenv("OPENHARNESS_DATA_DIR", str(tmp_path / "data"))
project_dir = tmp_path / "repo"
project_dir.mkdir()
memory_dir = get_project_memory_dir(project_dir)
(memory_dir / "pytest_tips.md").write_text("Pytest markers and fixtures\n", encoding="utf-8")
(memory_dir / "docker_notes.md").write_text("Docker compose caveats\n", encoding="utf-8")
matches = find_relevant_memories("fix pytest fixtures", project_dir)
assert matches
assert matches[0].path.name == "pytest_tips.md"
# --- Frontmatter parsing tests ---
def test_parse_frontmatter_extracts_fields(tmp_path: Path):
path = tmp_path / "project_auth.md"
path.write_text(
"---\n"
"name: auth-rewrite\n"
"description: Auth middleware driven by compliance\n"
"type: project\n"
"---\n"
"\n"
"Session token storage rework for legal team.\n",
encoding="utf-8",
)
header = _parse_memory_file(path, path.read_text(encoding="utf-8"))
assert header.title == "auth-rewrite"
assert header.description == "Auth middleware driven by compliance"
assert header.memory_type == "project"
assert "Session token storage" in header.body_preview
def test_parse_frontmatter_falls_back_without_frontmatter(tmp_path: Path):
path = tmp_path / "quick_note.md"
path.write_text("Redis cache invalidation strategy\n\nDetails here.\n", encoding="utf-8")
header = _parse_memory_file(path, path.read_text(encoding="utf-8"))
assert header.title == "quick_note"
assert header.description == "Redis cache invalidation strategy"
assert header.memory_type == ""
# Description line must not be duplicated into body_preview.
assert header.body_preview == "Details here."
def test_parse_malformed_frontmatter_does_not_return_delimiter(tmp_path: Path):
"""Unclosed frontmatter must not leak '---' into description."""
path = tmp_path / "broken.md"
path.write_text("---\nname: oops\nActual content here.\n", encoding="utf-8")
header = _parse_memory_file(path, path.read_text(encoding="utf-8"))
# The key invariant: description is never the raw delimiter.
assert header.description != "---"
assert header.description # non-empty
def test_parse_frontmatter_skips_headings_for_description(tmp_path: Path):
path = tmp_path / "notes.md"
path.write_text("# My Heading\n\nActual description here.\n", encoding="utf-8")
header = _parse_memory_file(path, path.read_text(encoding="utf-8"))
assert header.description == "Actual description here."
def test_parse_frontmatter_handles_quoted_values(tmp_path: Path):
path = tmp_path / "quoted.md"
path.write_text(
'---\nname: "my-project"\ndescription: \'A quoted desc\'\ntype: feedback\n---\nBody.\n',
encoding="utf-8",
)
header = _parse_memory_file(path, path.read_text(encoding="utf-8"))
assert header.title == "my-project"
assert header.description == "A quoted desc"
assert header.memory_type == "feedback"
def test_scan_memory_files_with_frontmatter(tmp_path: Path, monkeypatch):
monkeypatch.setenv("OPENHARNESS_DATA_DIR", str(tmp_path / "data"))
project_dir = tmp_path / "repo"
project_dir.mkdir()
memory_dir = get_project_memory_dir(project_dir)
(memory_dir / "topic.md").write_text(
"---\nname: my-topic\ndescription: Important topic\ntype: reference\n---\nContent.\n",
encoding="utf-8",
)
headers = scan_memory_files(project_dir)
assert len(headers) == 1
assert headers[0].title == "my-topic"
assert headers[0].description == "Important topic"
assert headers[0].memory_type == "reference"
# --- Search relevance tests ---
def test_search_prefers_metadata_over_body(tmp_path: Path, monkeypatch):
monkeypatch.setenv("OPENHARNESS_DATA_DIR", str(tmp_path / "data"))
project_dir = tmp_path / "repo"
project_dir.mkdir()
memory_dir = get_project_memory_dir(project_dir)
# File A: "redis" appears in frontmatter description
(memory_dir / "a_redis.md").write_text(
"---\nname: cache-layer\ndescription: Redis caching strategy\n---\nGeneral notes.\n",
encoding="utf-8",
)
# File B: "redis" appears only in body
(memory_dir / "b_infra.md").write_text(
"---\nname: infra-notes\ndescription: Infrastructure overview\n---\nWe use redis for sessions.\n",
encoding="utf-8",
)
matches = find_relevant_memories("redis caching", project_dir)
assert len(matches) == 2
assert matches[0].title == "cache-layer"
def test_search_finds_body_content(tmp_path: Path, monkeypatch):
monkeypatch.setenv("OPENHARNESS_DATA_DIR", str(tmp_path / "data"))
project_dir = tmp_path / "repo"
project_dir.mkdir()
memory_dir = get_project_memory_dir(project_dir)
(memory_dir / "deploy.md").write_text(
"---\nname: deploy\ndescription: Deployment notes\n---\nKubernetes rollout strategy details.\n",
encoding="utf-8",
)
matches = find_relevant_memories("kubernetes rollout", project_dir)
assert matches
assert matches[0].title == "deploy"
def test_search_handles_cjk_queries(tmp_path: Path, monkeypatch):
monkeypatch.setenv("OPENHARNESS_DATA_DIR", str(tmp_path / "data"))
project_dir = tmp_path / "repo"
project_dir.mkdir()
memory_dir = get_project_memory_dir(project_dir)
(memory_dir / "chinese_note.md").write_text(
"---\nname: meeting\ndescription: 项目会议纪要\n---\n讨论了部署计划。\n",
encoding="utf-8",
)
matches = find_relevant_memories("会议", project_dir)
assert matches
assert matches[0].title == "meeting"