Releases: corbat-tech/coco
Releases · corbat-tech/coco
v1.5.0
What's Changed
- feat: v1.5.0 - Context-Aware Stack Detection by @victormartingil in #37
Full Changelog: v1.4.0...v1.5.0
v1.4.0
What's Changed
- Release v1.4.0: UX Improvements & COCO Default ON by @victormartingil in #36
Full Changelog: v1.3.0...v1.4.0
v1.3.0
What's New
Added
/openskill andopen_filetool — open files with system default app or execute scripts with auto-detected interpreters/shipskill and GitHub tools (gh_check_auth,gh_repo_info,gh_pr_create,gh_pr_merge,gh_pr_checks,gh_pr_list,gh_release_create)ShipConfigSchemafor configurable release pipelines- Bilingual intent patterns (EN/ES) for open/ship commands
- Comprehensive test suites for clipboard, maturity, trust, allow-path, commands, coco-mode, version-check, github tools, and ship skill
Changed
- .gitignore updated to exclude build artifacts and coverage temp files
- CLAUDE.md simplified
Removed
- Old audit documents, coverage-temp artifacts, and roadmap files
Full Changelog: v1.2.3...v1.3.0
What's Changed
- feat(open): add /open skill, open_file tool, and ship skill for v1.3.0 by @victormartingil in #35
Full Changelog: v1.2.3...v1.3.0
v1.2.3
What's Changed
- feat(v1.2.3): thinking feedback, git tools fix, authorize_path by @victormartingil in #34
Full Changelog: v1.2.2...v1.2.3
v1.2.2
What's Changed
Bug Fixes
- Input line-wrap: Fixed extra blank line appearing when user input exactly fills the first terminal row
- Header box alignment: Replaced custom
visualWidth()with industry-standardstring-widthpackage + space-padding approach (same asboxen/ink). Right border now aligns perfectly regardless of terminal - Bottom separator persistence: Fixed separator disappearing after pressing Enter (caused by
eraseDownclearing too much)
Improvements
- New header design: Renamed to "COCO" with tagline "code that converges to quality" and color hierarchy (bold white title, magenta tagline, dim attribution)
- Version bump: 1.2.1 → 1.2.2
Dependencies
- Added
string-widthfor reliable terminal string width measurement
What's Changed
- fix(v1.2.1): input line-wrap bug, live bottom separator, restore header by @victormartingil in #33
Full Changelog: v1.2.0...v1.2.2
v1.2.0
What's Changed
- feat: v1.1.0 pre-release — 12-dimension quality system & multi-agent foundation by @victormartingil in #31
- chore(ci): bump actions/upload-artifact from 4 to 6 by @dependabot[bot] in #25
- chore(deps): bump @types/node from 25.2.1 to 25.2.2 in the development group by @dependabot[bot] in #26
- chore(deps): bump ora from 9.2.0 to 9.3.0 by @dependabot[bot] in #27
- chore(deps): bump openai from 6.17.0 to 6.18.0 by @dependabot[bot] in #28
- chore(deps): bump @anthropic-ai/sdk from 0.73.0 to 0.74.0 by @dependabot[bot] in #29
- chore(deps): bump oxlint from 0.16.12 to 1.43.0 by @dependabot[bot] in #30
- fix(v1.2.0): 25 bug fixes, README rewrite, version bump by @victormartingil in #32
Full Changelog: v1.1.0...v1.2.0
v1.1.0 — Pre-Release Quality Improvements
🥥 Corbat-Coco v1.1.0
Highlights
- 12-dimension quality scoring system with real analyzers
- 4,350+ tests across 171 test files
- 80%+ test coverage (up from ~55%)
- 19 previously untested modules now fully covered
- Multi-agent coordination with weighted scoring and tool-use
- Interactive onboarding with multi-provider support (Anthropic, OpenAI, Google, LM Studio)
- Zero lint warnings, zero type errors, all CI checks passing
What's New
Added
- 12-dimension quality scoring system with real analyzers
- Comprehensive test suite: 4,350+ tests across 171 test files
- Multi-agent coordination with weighted scoring and tool-use
- Interactive onboarding with multi-provider support
- REPL with image attachment, diff preview, and smart suggestions
- Budget tracking and cost estimation for LLM operations
- Convergence analyzer for quality iteration loops
- Build verifier with real compilation and lint checks
- Import analyzer with circular dependency detection
- Recovery system with automatic LLM provider fallback cycling
Changed
- Test coverage increased from ~55% to 80%+ across all metrics
- Quality thresholds raised to 80% (lines, functions, branches, statements)
- Removed all excluded files from vitest.config.ts coverage exclusion list
Fixed
- Recovery system now correctly classifies "overloaded" and "capacity" as LLM errors
- Cost estimator partial model matching uses longest-match-first ordering
- Smart suggestions empty catch block detection accuracy
- Test failure analyzer stack trace parsing and root cause categorization
- Import analyzer circular dependency detection with .js→.ts extension mapping
- Onboarding LM Studio tests properly mock fetch
- All 9 lint warnings resolved
Full Changelog: v1.0.2...v1.1.0
What's Changed
- chore: rename package to @corbat-tech/coco by @victormartingil in #24
Full Changelog: v1.0.2...v1.1.0
v1.0.2
Note: Versions 1.0.0 and 1.0.1 were published with a bug where
coco --versionreported0.3.0. v1.0.2 is the first correct post-0.3.0 release.
🚀 Features
14 New Agent Tools (#19)
web_fetch,web_search— Web interaction with HTML→markdown conversioncodebase_map,semantic_search,generate_diagram— Codebase intelligencecheckpoint,memory— Cross-session persistenceread_pdf,read_image,query_database— Document processingview_diff,code_review,permissions,diff— Code quality
MCP Enhancements (#19)
- SSE transport with auto-reconnect and exponential backoff
readResource(),getPrompt()client methodsjsonSchemaToZodconverter- Server lifecycle manager (start/stop/restart, health checks)
Enhanced REPL (#19)
- Coco-mode, diff renderer, syntax highlighting
- Bash pattern detection, recommended permissions system
Real Quality Scoring System (#21)
- Coverage analyzer (c8/v8 instrumentation)
- Security scanner (npm audit + OWASP static analysis)
- Complexity analyzer (AST-based cyclomatic complexity)
- Duplication analyzer (token-based similarity detection)
- Unified evaluator with weighted scoring
AST-Aware Generation Pipeline (#21)
- Pre-edit syntax validation with auto-fix (max 3 attempts)
- Import analysis with circular dependency detection
- Build verification before tests (zero broken builds guarantee)
Smart Iteration Loop (#21)
- Test failure root cause analysis with LLM-powered diagnosis
- Targeted fix generator (fixes specific issues, not full rewrites)
- Convergence detection (6 patterns: target reached, stuck, oscillating, diminishing returns, stable, max iterations)
Multi-Agent Coordination (#21)
- Agent executor with autonomous loops
- Parallel execution coordinator with dependency resolution
- 6 predefined roles: researcher, coder, tester, reviewer, optimizer, planner
Multi-Provider Auth & Security
- Allowed-paths system
- Enhanced OAuth flow
🔒 Security
- Fix
globvulnerability (#16) - Fix polynomial ReDoS in
ast-validator.ts
🐛 Bug Fixes
- Version now read dynamically from
package.json—coco --versionreports the correct installed version (#23)
📦 Dependencies
- Zod 3→4, @clack/prompts 0.11→1.0 (#18)
- glob 11→13 (#8)
- @anthropic-ai/sdk 0.39→0.72 (#9)
- CI actions updated (#17)
📊 Stats
- 228 files changed, ~44K lines added
- 3,915 tests passing, 15 skipped
v0.3.0
🚀 Features
Interactive REPL UX
- Autocomplete with ghost-text suggestions
- Multi-column menu layout with keyboard navigation
- Context compaction for long sessions
- Intent recognition for natural language commands
- Enhanced streaming with proper abort handling
Agent System
- Background agent system with task management
- Checkpoint/restore system for long-running tasks
Skills System
- Extensible command framework via skills
MCP Integration
- Model Context Protocol support for external tool servers
Trust Store
- Directory-level permission management
New Commands
/init— Initialize new projects with templates/memory— Manage conversation memory/build— Run build commands/status— Show project and git status/trust— Manage directory permissions/model— Interactive provider/model selection
New Tools
tree— Directory tree visualizationcopy_file— File copy operationsmove_file— File move/rename operations
Provider Improvements
- Interactive provider configuration and model selection
📊 Stats
- 271 files changed, ~38.7K lines added
v0.2.0
See CHANGELOG.md