Skip to content

v1.0.0-beta.8

Choose a tag to compare

View all tags
@cx2002302-lang cx2002302-lang released this 25 Jun 10:26
v1.0.0-beta.8
0a9405a

Changelog

v1.0.0-beta.8 — OpenClaw 2026.6.x + Hermes MCP Compatibility

Release Date: 2026-06-25

🔧 Compatibility

  • OpenClaw 2026.6.x support — Declared contracts.tools in plugin manifest for new tool policy
  • Tool policy abstraction — Added scripts/lib/compat.sh to centralize version-specific OpenClaw configuration
  • systemPromptOverride deprecation handling — setup-skill-prompt.sh now skips on OpenClaw >= 2026.6.x
  • Hermes MCP bridge — New src/mcp/http-bridge.ts exposes Zettelkasten tools via Streamable HTTP for Hermes Agent

🧪 Testing

  • 1724 tests passing, 72 test files, 0 failures
  • Replaced hardcoded /test/... paths with temp directories via src/testing/test-fs.ts
  • Added local compatibility matrix runner: scripts/run-compat-matrix.sh
  • Added Hermes mock-LLM E2E test and MiniMax real-LLM E2E test

🛠️ Fixes

  • Audit consistency edge case when totalLinks === 0
  • Glow ranking empty-cache fallback
  • DB ↔ Markdown file system sync
  • Hermes config path (/opt/data/config.yaml)

v1.0.0-beta.7 — Security Hardening + Code Quality + Test Expansion

Release Date: 2026-05-22

🛡️ Security Fixes

  • SQL Injection Whitelist — All raw SQL queries now use strict column-name whitelists to prevent injection via user-controlled orderBy/sortBy parameters
  • Path Traversal Protection — File system operations validate and sanitize paths to block directory traversal attacks
  • Database Connection Safety — Added explicit db.close() calls in all repository and service layers to prevent connection leaks
  • Promise Rejection Handling — Unhandled Promise rejections in async flows are now properly caught and logged

🧹 Code Quality

  • Magic Numbers Eliminated — Extracted 20+ magic numbers into named constants (src/core/constants.ts) for maintainability
  • Console Cleanup — Removed all stray console.log/console.error calls; replaced with structured logging or removed entirely
  • Type Safety Enhancement — Strengthened TypeScript strictness: added explicit return types, fixed any usages, tightened null checks
  • Dead Code Removal — Deleted unused imports and obsolete helper functions across 12 files
  • Utility Reuse — Consolidated duplicated validation logic into shared utility functions (src/core/utils.ts)
  • Promise.allSettled Adoption — Replaced fragile Promise.all + manual error handling with Promise.allSettled for bulk operations

🧪 Test Coverage

  • 862 tests passing, 36 test files, 0 failures (was 689)
  • 173 new test cases added across 10 new test files:
    • src/service/phase6/__tests__/audit-service.test.ts
    • src/service/phase6/__tests__/moc-service.test.ts
    • src/service/phase6/__tests__/serendipity-service.test.ts
    • src/mcp/__tests__/phase6-tools.test.ts
  • Security regression tests for SQL whitelist and path sanitization
  • Boundary case coverage for new Phase 6 MCP tools

📦 Architecture

  • Phase 6 services stabilized: audit-service, moc-service, serendipity-service
  • MCP tool registry updated with Phase 6 endpoints
  • Plugin manifest (openclaw.plugin.json) aligned with latest tool surface

v1.0.0-beta.6 — Auto-Review + Distillation Quality Guard + Memory Format Fix

Release Date: 2026-05-21

✨ New Features

  • Nightly Auto-Review — Plugin cron now automatically reviews Inbox notes at 2:00 AM

    • High-quality notes (confidence ≥ 0.7 + content length ≥ 200): auto-approved → zettels
    • Low-quality notes (quality < 0.4): auto-flagged → stays in inbox with review record
    • Medium-quality notes: kept in inbox for human review
    • ReviewService.autoReviewInbox() processes up to 100 pending notes per run

  • Distillation Quality Guard — 6-layer content validation prevents placeholder notes from entering the database

    • Empty title rejection
    • Minimum 50-character content requirement
    • Placeholder title detection (9 patterns: Memory entry X, Untitled, Note X, etc.)
    • Metadata-only content detection
    • Title-content identical check
    • Meaningful word count validation (≥ 5 words)

  • Markdown Memory Format Support — MemoryParser now supports .md memory logs in addition to .json

    • Auto-detects file extension and selects appropriate parser
    • Falls back from .json to .md when JSON file is missing
    • Parses Markdown sections (## Entry) with Type/Time/Content fields

🔧 Bug Fixes

  • Distillation Empty Content: LLM-generated summaries with only metadata (e.g., > 来源：Professional > 时间：... > 权重：X) are now rejected at creation time instead of polluting the database
  • Inbox Accumulation: 120 unreviewed notes accumulated because autoReview() existed but was never called by the plugin. Fixed by adding auto-review step to nightly cron
  • PR#1 Merged: Accepted kriptoburak's placeholder URL fix (YOUR_USERNAME → cx2002302-lang) after TweetClaw promotional content was removed

🧹 Data Cleanup

  • Cleaned 33 empty/placeholder notes (source=distilled, content only metadata) from production database
  • Post-cleanup state: 129 notes | 0 inbox backlog | 17/17 health checks passing

🧪 Test Coverage

  • 689 tests passing, 26 test files, 0 failures
  • Auto-review verified with 3-tier test: high-quality→approve, medium→flag/skip, low→flag

v1.0.0-beta.5 — Phase 5 Evolution System + Concurrency Safety + Documentation Overhaul

Release Date: 2026-05-16

✨ New Features

  • Phase 5 Evolution System — 10 new MCP tools + 5 new CLI commands

    • Review system: zk_get_review_panel, zk_submit_review, zk_get_review_stats
    • Feedback system: zk_submit_feedback, zk_get_feedback_stats, zk_analyze_feedback_trends
    • Prompt evolution: zk_get_active_prompt, zk_get_prompt_stats
    • Sample curation: zk_get_curation_stats, zk_export_samples
    • CLI: zk review-stats, zk review-pending, zk feedback-stats, zk prompt-stats, zk curation-stats

  • Database Schema — 4 new tables for evolution tracking

    • zettel_prompt_versions — Prompt version history with quality scoring
    • zettel_sample_curations — CEQRC sample quality assessment (6 dimensions)
    • zettel_system_tunings — System parameter tuning history
    • zettel_export_batches — Sample export batch tracking

  • Documentation Overhaul — Complete documentation system rebuilt

    • Full SKILL.md (6551 bytes) + PROMPT.md (7737 bytes) with use cases and examples
    • details/ folder with compact versions for progressive disclosure
    • docs/ folder with 9 feature documents (README + 01-09)

  • Concurrency Safety — Verified under rapid-fire parallel operations

    • 20 notes + 20 feedback + 10 links (bidirectional = 18) + 20 reviews
    • Zero duplicate IDs, zero orphaned records, zero data inconsistency

🔧 Bug Fixes

  • ZK-BUG-003: deploy.sh alsoAllow cleanup incorrectly removed "zettelkasten" plugin ID, causing MCP tools to become invisible to agents. Fixed by adding or x == 'zettelkasten' to the filter condition.
  • Schema Alignment: 3 Phase 5 tables (zettel_prompt_versions, zettel_sample_curations, zettel_system_tunings) had DDL fields mismatched with Repository code. Fixed DDL to match runtime expectations.
  • Old Schema Migration: Auto-detect and recreate tables with stale column definitions on zk init.

🧪 Test Coverage

  • 689 tests passing, 26 test files, 0 failures
  • Agent E2E: 7/7 rounds passed (boundary cases, composition, Phase 5, error recovery, intelligence, Chinese search, concurrency)
  • Production deployment verified: 143 notes, 70 links, 543 feedback, 292 reviews
  • All 28 MCP tools callable via OpenClaw Agent

v1.0.0-beta.4 — Test Suite + Input Validation + Performance Benchmark

Release Date: 2026-05-12

✨ New Features

  • Performance Benchmark Suite (scripts/benchmark.mjs)

    • Tested at 1K / 5K / 10K notes scale
    • All 7 performance thresholds passed at 10K scale
    • FTS search 1.9ms, single note read 0.08ms, glow recalc 1,013ms
    • Full report: plans/PERFORMANCE-BENCHMARK.md

  • E2E Tool Chain Test (scripts/e2e-tool-test.mjs)

    • End-to-end validation of all 16 MCP tool flows
    • Covers create, search, link, glow, path, heatmap, archive, review, feedback

🛡️ Input Validation

  • NoteService: Reject empty title/content, confidence must be 0-1 (create + update)
  • LinkService: Reject self-links, validate link type against 11 allowed values
  • HeatmapService: Clamp glowMin to [0,1], normalize negative limit to 0

🔧 Bug Fixes

  • ZK-BUG-001: tools.alsoAllow 放入 Skill ID（如 "zettelkasten-brain"）导致 OpenClaw Gateway 工具策略验证失败，引发崩溃-重启循环。修复：deploy.sh 自动清理 alsoAllow 中的非 zk_ 前缀条目；AGENTS.md 明确文档说明。
  • ZK-BUG-002: SKILL.md 缺少 YAML frontmatter，可能影响 Skill 加载。修复：添加标准 frontmatter（name, description, version）。
  • archive-service.ts: zombie.folder field did not exist in GlowMetrics, causing already-archived notes to be re-processed. Fixed by querying zettel_notes.folder directly.
  • feedback-repository.ts: source field could be undefined, causing SQLite binding error. Fixed with source ?? null.
  • feedback-repository.ts: rating returned null from DB but type expected undefined. Fixed with row.rating ?? undefined.
  • feedback-service.test.ts: Time race condition in analyzeTrends test — now created before submitFeedback, causing feedback created_at to be later than query end time. Fixed by using tomorrow as end boundary.

🧪 Test Coverage

  • 689 tests passing, 26 test files, 0 failures
  • New test files: feedback-repository.test.ts (25), review-repository.test.ts (20), archive-service.test.ts (14)
  • Heatmap boundary tests: empty DB, empty filters, negative limit, glowMin overflow
  • Input validation tests: empty strings, out-of-range confidence, self-links, invalid types

v1.0.0-beta.3 — Wave 3: Knowledge Heatmap & Network Graph

Release Date: 2026-05-11

✨ New Features

  • Knowledge Heatmap (zk_heatmap)

    • Real-time knowledge base activity analysis
    • Folder distribution, glow score distribution, connection density ranking
    • Isolated notes detection, recent activity tracking
    • CLI: openclaw zk heatmap --days 30

  • Network Graph Visualization (zk_network_graph)

    • Export knowledge graph as JSON with nodes and edges
    • Support folder filter, glow score threshold, node limit
    • Degree calculation and weighted edge rendering
    • CLI: openclaw zk graph-export --limit 200 --glow-min 0.4

  • 18 MCP Tools (9 read-only + 9 read-write)

    • Read-only: zk_search_notes, zk_get_note, zk_get_backlinks, zk_find_path, zk_glow_ranking, zk_find_zombies, zk_search_archived, zk_get_archive_log, zk_knowledge_heatmap, zk_network_graph
    • Read-write: zk_create_note, zk_update_note, zk_archive_note, zk_unarchive_note, zk_run_ceqrc_workflow, zk_distill_memory, zk_get_inbox_queue, zk_review_note

🔧 Bug Fixes

  • Commander.js parseInt/parseFloat NaN bug: Option handler signature (value, previous) caused parseInt("10", 200) → NaN → SQLite datatype mismatch. Fixed with safeParseInt/safeParseFloat wrappers and Number.isFinite() validation.

🧪 Test Coverage

  • Agent E2E: 34/34 passed (CLI + MCP handlers + plugin config)
  • Unit tests: 112/112 passed (heatmap-service, note-service)

v1.0.0-beta.2 — Wave 2: Auto-Archive & Knowledge Health

Release Date: 2026-05-10

✨ New Features

  • Auto-Archive Scheduling: Cron-based zombie note detection and archiving
  • Archive History: Full log of archive/unarchive operations with timestamps
  • Timestamp Preservation: Archived notes retain original creation/update times
  • Archive Service: archive-service.ts with dry-run support

v1.0.0-beta.1 — Wave 1: Knowledge Glow & Path Search

Release Date: 2026-05-09

✨ New Features

  • Knowledge Glow Score: CEQRC engine with confidence, entropy, quality, recency, connections
  • Path Finder: Bidirectional BFS for shortest knowledge path between notes
  • Archive Mechanism: Manual archive/unarchive with folder migration
  • FTS Search: Full-text search with SQLite FTS5
Assets 2
Loading