All notable changes to Anchor Engine will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

• HTTP Endpoints for Path Management (Standard 039)

  • POST /v1/system/path-add — Add watched path via watchdog integration
  • POST /v1/system/path-remove — Remove watched path + trigger hot-slotting purge

• MCP Tools with Direct DB Operations

  • anchor_set_path(path) — Adds path to watch list (direct DB insert + HTTP fallback)
  • anchor_remove_path(path) — Full cleanup: atoms, molecules, edges, tags, sources + mirrored files

• Hot-Slotting Purge Enhancement

  • Complete database purge when removing watched paths (was only removing path entry)
  • Removes all associated data: atoms, molecules, edges, tags, sources from PGlite
  • Cleans mirrored brain files and metadata notebooks

• MCP Tool Architecture

  • Tools now use direct database operations as primary method with HTTP fallback
  • More reliable than pure HTTP approach (doesn't require running server)

• Endpoint Registration — Previously /v1/system/path-add returned 404; now properly registered via setupSystemRoutes() in system.ts
• Data Accumulation Bug — Watched path removal no longer leaves orphaned database entries and mirrored files

• Updated doc_policy.md to match reality: 38 standards, flat directory, current docs/ structure
• Removed stale archive-legacy/ references and decisions/ directory (merged into relevant standards)

Standards Flattened:

• Flattened specs/current-standards/ from 14 category subdirectories into single flat directory
• 38 active standards numbered 001–032, ordered foundational → assistive
• Duplicate-numbered standards (014, 018, 019, 027, 028, 029) preserved with distinct names

Architecture Consolidated:

• Merged ARCHITECTURE.md into spec.md — architecture diagrams, web dashboard, engine core modules now in system spec
• Removed ux-ui-recursion-workflow.md (no longer needed)

Docs Restored from Git History:

• Restored: code-patterns.md, design-patterns.md, star-algebra-reference.md
• Restored: integrations/CODE_OF_CONDUCT.md, integrations/CONTRIBUTING.md
• Renamed: -settings-configs.md → settings-configs.md

Root Cleanup:

• Deleted 19 one-off fix scripts cluttering project root
• Removed PM2 configs, .babelrc, config/ directory
• Dropped package-lock.json (pnpm only)
• Cleaned .gitignore: removed embedded JavaScript code, fixed contradictions with tracked files
• Restored missing root package.json + pnpm-workspace.yaml
• Re-added CITATION.cff to tracking

API Routes:

• Added POST /v1/distills route to trigger distillation
• Removed compounds.ts route (always returned 404, table removed per Standard 051)
• Fixed /v1/files/read to allow .md/.txt/.yml extensions (was blocking distill output reads)
• stats.ts: replaced hardcoded placeholder with real implementation (uptime, memory, DB row counts, search counter)

MCP Tools:

• Fixed anchor_distill: was calling non-existent POST /v1/distills → now calls POST /v1/memory/distill
• Fixed anchor_distill body format: maps {source_url, source_text} → proper RadialDistillRequest
• Added anchor_list_distills and anchor_get_distill tools
• callAnchorAPI() now supports GET requests

Distill Manager:

• Rewritten to query PGlite distills table instead of in-memory-only store
• recordDistill() writes metadata pointers to DB on completion
• In-memory cache retained for fast UI polling

UI:

• Fixed output_format from unsupported 'source-grouped' to 'decision-records'
• Added decision-records format handler (parsedContent.records)

• Consolidated docs/testing/LIVE-FIRE-TEST-SUITE.md (571 lines) into specs/current-standards/search-retrieval/014-search-algorithm-testing.md
• Added P5 edge cases section with 10+ test examples
• Added P6 performance benchmarks section with latency validation tests
• Added CI/CD integration examples for GitHub Actions workflow
• Added troubleshooting guide covering common issues and fixes
• Added best practices section with 7 key recommendations
• Standard 014 expanded from 301 to 541 lines (80% increase)

• Deleted 15 deprecated testing patterns from tests/legacy/ directory:
  • comprehensive-test.js, full-sequence-test.js, config-test.js, db-close-test.js
  • db-test.js, distillation-test.js, fixed-startup-test.js, individual-import-test.js
  • minimal-db-test.js, minimal-test.js, route-setup-test.js, test-tool-executor.js
  • test-fixed-engine.js, test-native-module.js, test-server.js, accurate-test.js

• Consolidated docs/testing/API-SURFACE.md references into standard specs
• Deleted docs/testing/LIVE-FIRE-TEST-SUITE.md (merged into Standard 014)
• Removed now-empty docs/testing/ directory

Implemented agent-friendly background startup/shutdown scripts:

• scripts/start-engine-bg.mjs - Node.js background startup with:

  • Build verification (engine/dist/index.js)
  • Port conflict detection (auto-finds available port)
  • File logging to .anchor/logs/start-{timestamp}.log
  • Immediate exit after health check (non-blocking for agents)
  • Health check retries before success confirmation

• scripts/stop-engine-bg.mjs - Node.js background shutdown with:

  • Port-based PID detection (cross-platform, Windows 11 compatible)
  • Graceful SIGTERM shutdown first
  • Force kill if graceful shutdown fails
  • File logging to .anchor/logs/stop-{timestamp}.log

Benefits:

• No blocking console output during startup/shutdown
• Automatic port conflict resolution
• Persistent logging for debugging
• Cross-platform compatibility (Windows 11, Linux, macOS)

New P5 Edge Cases Tests:

• Empty query handling
• Very long query validation
• Special characters: &, |, ", '
• HTML entity/XSS prevention
• Unicode character support
• Single character queries
• Malformed input handling

New P6 Performance Benchmarks:

• Semantic search latency (<3s target)
• Tag-based search latency (<1s target)
• Byte offset search latency (<500ms target)
• FTS search latency (<200ms target)
• Empty query latency (<100ms target)
• Performance validation with 50% margin tolerance

CI/CD Integration Examples:

• GitHub Actions workflow for automated testing
• Watch mode for development iterations
• Phase-specific execution (P0-P4 for CI, P5-P6 for full suites)

Troubleshooting Guide:

• Tests failing immediately (engine not running)
• Logs not being created (permissions/PROJECT_ROOT issues)
• Performance tests exceeding targets (insufficient corpus)
• Pain point detection not working (Standard 027 configuration)

• Updated docs/INDEX.md with recent cleanup notes
• Added version bump: Standard 014 v1.0.0 → v2.0.0
• Updated CHANGELOG.md with testing cleanup history

• Testing Infrastructure

  • Playwright-based e2e UI verification tests (test_suite/anchor_engine_ux_test_suite.py)
  • Comprehensive API documentation with OpenAPI specs and curl examples
  • System metrics endpoint /v1/stats (uptime, memory, request counters)
  • Reliable engine startup/shutdown scripts

• QwenPaw Environment Integration

  • PowerShell environment fix script (Fix-QwenPawEnv.ps1) for PATH ordering and UTF-8 support
  • Git Bash .bashrc with Windows→Linux compatibility aliases
  • PowerShell 7 profile auto-loader (.powershell_profile.ps1)

• Project Structure

  • Moved to coding_projects/anchor-engine-node workspace
  • New infrastructure: agents/, specs/decisions/, specs/INTEGRATIONS/
  • Database initialization scripts (create-db.js, init-db.js)

• Major Cleanup (v5.1.0 preparation)

  • Removed orphaned route: engine/src/routes/v1/ingest-updated.ts
  • Removed orphaned service: engine/src/services/ingest/ingest-atomic-updated.ts
  • Deprecated bright-nodes reference in documentation
  • Cleaned all test artifacts, JSON build outputs, and temporary files (20+ files)
  • Consolidated inbox: old inbox → internal-inbox

• Documentation

  • Added Testing section to README.md with e2e test instructions
  • Created API surface documentation covering all 15 route files + 4 additional endpoints
  • Updated UX/UI recursion workflow docs (require live engine for testing)

• Search route implementation restored from previous working commit
• Removed circular imports and missing dependencies in route architecture
• Cleaned stale TODO/FIXME comments from codebase

• Distillation Endpoint Optimization

  • Added max_molecules parameter to limit corpus-mode queries and prevent timeouts
  • Restored timeout_seconds support with user-configurable timeout (default: 60s) instead of hardcoded 120s
  • Fixed streaming mode to respect max_molecules and use proper default values

• Build Infrastructure

  • Restored build script (scripts/build.ts) that compiles TypeScript → JavaScript for production deployment
  • Added missing devDependencies: @types/node, rimraf, typescript

• GitHub Ingestion Service

  • Fixed import extensions from .ts to .js (NodeNext module resolution requirement)
  • Corrected corrupted method calls: .tson() → .json(), .tsON.parse() → JSON.parse()
  • Removed duplicate entries in file extension mapping

• New radial distillation v2 with semantic aggregation across sources
• Memory typing system for better context management
• Temporal cascade for improved search relevance
• New output formats for distillation summaries

• Updated path configuration to .anchor structure
• Simplified version loading mechanism
• Worker-based GitHub sync for better concurrency

• OOM hardening with better memory management
• Critical bug fixes in search pipeline
• Performance improvements in tag discovery

• Added API key strength validation (Standard 132)
• Fixed auth bypass vulnerability in /v1/test/* endpoints (Standard 131)
• SQL injection fix via parameterized LIMIT clauses (Standard 130)
• Path traversal prevention (Standard 129)

• Streaming distillation with real-time output
• UI improvements for search results
• Watchdog process for memory management