Skip to content

CHANGELOG.md

Latest commit

 

History

History
205 lines (167 loc) · 9.16 KB

CHANGELOG.md

File metadata and controls

205 lines (167 loc) · 9.16 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

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

[Unreleased]

Added

  • PersonaArchitect SEP Integration: Persona and Lesson entities with full CRUD CLI
    • Persona entity: name, role, domain, expertise, CoV/FAP/OV structured sections
    • Lesson entity: title, domain, category, content, difficulty, source_persona
    • CLI: engram persona create|list|show|update|delete, engram lesson create|list|show|update|delete
    • Two-tier storage: embedded defaults + filesystem overrides
    • Migrated 172 persona YAMLs with domain-adapted CoV/FAP/OV sections
  • Health CLI: Git-based workspace diagnostics
    • engram health check — consistency validation and orphan detection
    • Health skill for structured project health assessment
  • Doc Subcommand: engram doc with entity documentation, CLI reference, mdBook build, NLQ integration
  • Task Archiving: Archive completed tasks to reduce noise in active queries
  • NLQ Deep Graph Walk: Extended natural language queries to traverse entity relationships
  • Workflow Command Guards: Enforce task-workflow binding before state transitions
  • Escalation CLI: engram escalation create|list|approve|deny|cancel for permission management
  • Analytics & DORA Metrics: Deployment frequency, lead time, change failure rate, MTTR reporting
    • engram analytics CLI with DORA dashboard
  • StructuredFeedback Trait: Unified feedback interface across validation, testing, and review
  • Transition Side-Effects: Automatic actions on workflow state changes
  • Quality Gates Consolidation: Single quality gate pipeline replacing scattered checks
  • Session Enhancements: Session summaries, zombie session detection, stale task identification, session filters, --scope flags on engram next
  • Locus TUI Screens: Contexts, ADRs, and Theories views with auto-refresh
  • Workflow Engine: Commit policy enforcement, auto-transition fixes
  • Knowledge Types: Typed knowledge entities for better categorization
  • Skills: engram-commit-convention, ask-first refinement, engram-first research protocols

Changed

  • Renamed GitStorage → GitRefsStorage throughout codebase
  • Dead code cleanup, compliance fixes, sysinfo ResourceMonitor

[0.6.3] - 2026-04-08

Added

  • --all and --offset pagination flags on all list commands

[0.6.2] - 2026-04-08

Fixed

  • UTF-8 safe string truncation in truncate() and skills preview

[0.6.1] - 2026-04-07

Added

  • TUI Sync View: Remote listing, sync status table, pull/push actions in Locus TUI
  • Session Skills: session-start and session-end protocols with sync integration

Fixed

  • PATH-first binary resolution in commit-msg hook template

[0.6.0] - 2026-04-07

Added

  • Sync System (8 phases): Multi-agent coordination via git refs
    • Phase 1: Project ID derivation from SHA-512 of root commit
    • Phase 2: Versioned sidecar refs written on every entity write
    • Phase 3: RemoteConfig with project_id and import-git-remotes subcommand
    • Phase 4+5: Version-aware pull/push with refs/engram/* refspec
    • Phase 6: engram sync both (pull-then-push)
    • Phase 7: Conflict resolution CLI with stateless re-derivation
    • Phase 8: engram sync status with per-type table and --json output
  • Locus TUI: Terminal UI for browsing entities, views, events with full test suite
  • Auto-discover Skills: Runtime skill loading from ./skills/ instead of compile-time embedding
  • test-harness-review Skill: Audit test coverage across behavioral, integration, property, scenario, and unit tests
  • mdBook build output with theme assets

Changed

  • Skill path references updated from opencode to engram
  • Prompts stripped of OpenCode-specific content

[0.5.0] - 2026-04-04

Added

  • Screenplay Skills (11): session-start, beat-sheet-builder, outliner, logline-writer, theme-developer, world-builder, character-developer, scene-writer, dialogue-refiner, plot-hole-finder, rewriter
  • Extended engram ask query to search across all entity types

[0.4.0] - 2026-04-03

Added

  • Go-to-Market Skills (3): market-validation, gtm-strategy, launch-execution
    • Corresponding pipelines: 101-market-validation, 102-gtm-strategy, 103-launch-execution
  • Tmux Skill: Named session management for parallel agent isolation
  • NLQ Fulltext Search: Natural language query intent mapping to fulltext search
  • Meta Skills: Extended setup with --force, --dir, --tool flags
  • mdBook branding with logo, favicon, and Engram color theme
  • Install script (install.sh) as release artifact

Fixed

  • Corrected all 14 skill CLI commands
  • Enforced --task-id on reasoning create

[0.3.0] - 2026-03-10

Added

  • Theory Building Entity: Formal mental model capture based on Peter Naur's "Programming as Theory Building" (1985)

    • domain_name: High-level domain identifier
    • conceptual_model: HashMap of concept → definition
    • system_mapping: HashMap of concept → code location
    • design_rationale: HashMap of decision → reason
    • invariants: Vec of must-be-true statements
    • iteration_count: Tracks theory evolution
    • apply_reflection_updates(): Evolve theory from reflections

  • StateReflection Entity: Cognitive dissonance detection for AI agents

    • theory_id: Theory being evaluated
    • observed_state: Raw error/observation
    • cognitive_dissonance: Vec of conflicts detected
    • proposed_theory_updates: How to resolve conflicts
    • dissonance_score: 0.0-1.0 (≥0.7 requires theory mutation)
    • trigger_type: test_failure, runtime_error, unexpected_output, etc.
    • severity: none, low, medium, high, critical

  • Session Enhancements

    • New Reflecting status - blocks code execution until theory updated
    • bind_theory() - injects theory invariants into session metadata
    • active_theory_id, theory_ids, reflection_ids fields

  • CLI Commands

    • engram theory create|list|show|update|delete
    • engram reflect create|list|show|record-dissonance|propose-update|resolve|requires-mutation
    • engram session bind-theory|trigger-reflection|resolve-reflection

  • mdBook Documentation: Full documentation with Theory Building and State Reflection guides

  • Pipeline: prompts/ai/pipelines/00-theory-building.yaml for theory extraction workflows

  • Agent Persona: prompts/agents/168-the-theorist.yaml for reverse-engineering mental models

Fixed

  • Various clippy and formatting issues
  • Syntax errors in theory.rs and session.rs

[0.1.2] - 2026-01-20

Added

  • Agent Sandboxing System: Complete permission escalation workflow
    • EscalationHandler with create/approve/deny/cancel operations
    • Risk assessment and alternative suggestions for blocked operations
    • Integration with SandboxEngine validation pipeline
    • Automatic escalation creation on permission violations
    • Priority inference based on sandbox level and operation type
  • engram next command: AI prompt generation for task-driven development
    • Automatic task prioritization and selection
    • Workflow state-aware prompt interpolation
    • Dynamic prompt templates in workflow states
    • Task relationship and dependency analysis
  • Workflow Validator: State transition validation logic
    • Workflow state progression validation
    • Transition rule enforcement
    • State-based commit policy validation
  • Perkeep Integration: Backup and restore for all entity types
    • Content-addressable blob storage using Perkeep server
    • Backup/restore commands: engram perkeep backup, restore, list, health
    • Configuration via PERKEEP_SERVER and PERKEEP_AUTH_TOKEN environment variables
  • engram info command: Workspace and storage visibility
    • Shows storage backend path and type
    • Displays entity counts by type
    • Shows current agent and workspace path
  • Analytics System: Task duration and workflow reporting
    • TaskDurationReport: Tracks time to complete tasks
    • WorkflowStageReport: Analyzes workflow state progression
    • BottleneckReport: Identifies workflow inefficiencies
  • Vector Search: Optional semantic similarity search
    • SQLite-based vector storage with sqlite-vec
    • FastEmbed provider for embeddings
    • SearchQuery and SearchResult types for similarity search

Fixed

  • Build warnings and compilation errors
  • Feature-gated sandbox module to prevent unused code warnings
  • Test compilation errors in BDD framework
  • Agent sandbox entity serialization in generic conversion
  • Workflow instance persistence to storage (fixed in previous commit)

Changed

  • Improved workflow engine with prompt template support
  • Enhanced workflow state machine with validation rules

[0.1.1] - 2026-01-19

Added

  • Initial engram implementation with core entities
  • Git-backed storage system
  • Basic CLI commands for task, context, and relationship management
  • Workflow engine foundation
  • Commit validation system

Fixed

  • Various stability and reliability improvements

[0.1.0] - 2026-01-17

Added

  • Initial release
  • Core entity types (Task, Context, Reasoning, Relationship)
  • Basic CLI interface
  • Git storage backend