Skip to content

CHANGELOG.md

Latest commit

 

History

History
160 lines (131 loc) · 7.84 KB

CHANGELOG.md

File metadata and controls

160 lines (131 loc) · 7.84 KB
Raw

Changelog

All notable changes to GitVoyant are documented in this file.

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

[0.3.0] — 2026-03-29

Multi-Language Temporal Analysis

  • Extended temporal analysis to Python, JavaScript/TypeScript, Java, and Go
  • Introduced Analyzer protocol (ABC) defining the interface for language-specific code analysis
  • Created PythonAnalyzer by extracting _extract_metrics() and _cyclomatic_complexity() from TemporalEvaluator
  • Created JavaScriptAnalyzer using tree-sitter-javascript (handles .js, .jsx, .ts)
  • Created JavaAnalyzer using tree-sitter-java (handles .java)
  • Created GoAnalyzer using tree-sitter-go (handles .go)
  • Central analyzer registry mapping file extensions to analyzer instances
  • TemporalEvaluator delegates to the resolved analyzer instead of calling ast.parse directly
  • Automatic language detection by file extension
  • --language CLI flag for restricting analysis to a single language
  • GITVOYANT_SUPPORTED_LANGUAGES configuration option for limiting enabled languages
  • Per-language directory exclusions: node_modules, dist, .next, target, .gradle, vendor, build
  • File discovery in TemporalEvaluatorService and langchain_bindings now uses registry lookups instead of hardcoded .py checks

Multi-Model Support

  • Default model updated to claude-sonnet-4-5-20250929
  • Verified support for claude-opus-4-6 and claude-sonnet-4-6
  • Model selection via CLAUDE_MODEL environment variable
  • Agent runtime now consumes the full config contract via get_claude_config(), wiring model, temperature, and max_tokens through to ChatAnthropic
  • Default max_tokens increased from 4000 to 8192 to match current model capabilities

Dependency Upgrades

  • anthropic minimum version raised from 0.18.0 to 0.80.0 (required for current model IDs and models.list() endpoint)
  • langchain-anthropic minimum version raised from 0.1.0 to 1.0.0
  • langchain and langchain-core minimum versions raised from 0.1.0 to 0.3.0
  • Added tree-sitter>=0.24.0, tree-sitter-javascript>=0.23.0, tree-sitter-java>=0.23.0, tree-sitter-go>=0.23.0

Schema and Output Changes

  • Added language field to TemporalEvaluation domain entity
  • Added language field to EvaluationResponse DTO
  • Added language column to CLI Rich table output
  • Updated all tool descriptions from "Python file" to "source file"
  • CLI indicators changed from emoji to text markers: [+] (improving), [-] (declining), [~] (stable)
  • Agent prompt strings rewritten: branded as "GitVoyant" instead of vendor name

Bug Fixes

  • max_tokens was configured in settings but never passed to ChatAnthropic in agent runtime; now wired through
  • Corrected stale model identifiers that were returning HTTP 404 from the Anthropic API

Documentation

  • All markdown documentation (README, USER_GUIDE, CONTRIBUTING, CHANGELOG, COMPLEXITY_REQUIREMENTS, TEMPORAL_ANALYSIS_EXPLAINED) rewritten for declarative register, no speculative roadmaps, no performative language
  • CITATION.cff rewritten: updated abstract, keywords, and version to reflect multi-language scope

Removed

  • All emoji characters across the entire codebase (source, docs, CI, CLI output, banner, agent prompts)
  • All vendor attribution from documentation, user-facing strings, and CITATION.cff
  • docs/index.html static architecture diagram (stale, duplicative of README)
  • Speculative roadmap sections from CONTRIBUTING.md and CHANGELOG.md
  • Badge wall from README.md

Testing

  • 63+ tests total
  • 13 unit tests for JavaScriptAnalyzer
  • 14 unit tests for JavaAnalyzer
  • 12 unit tests for GoAnalyzer
  • All three supported models verified end-to-end against the live Anthropic API through the full agent runtime

[0.2.0] — 2025-07-15

Architecture Transformation

  • Complete rewrite using Domain-Driven Design (DDD) principles
  • Clean Architecture with separation of presentation, application, domain, and infrastructure layers
  • Rich domain entities: TemporalEvaluation, Repository with embedded business logic
  • Value objects: ComplexityTrend, ConfidenceRank, TimeTable with validation and business rules
  • Use cases: AnalyzeFileUseCase, AnalyzeRepoUseCase orchestrating business workflows

AI Agent Platform

  • LLM integration via LangChain and Anthropic agent runtime
  • Conversational interface for natural language temporal analysis through AI agents
  • Specialized tools: temporal_analysis_tool, repo_evaluation_tool for agent consumption
  • LangGraph ReAct agent runtime with GitVoyant tools
  • Context management with output suppression and response formatting

Enhanced Temporal Intelligence

  • Confidence scoring based on commit history depth (0.4-0.9 scale)
  • Progressive disclosure: higher confidence with more temporal data points
  • Low confidence warnings as reliability indicators for limited data
  • Improved quality decay forecasting with confidence bounds
  • Proper handling of insufficient data with graceful degradation

CLI

  • Rich formatting with color-coded metrics and styled tables
  • Command structure: gitvoyant analyze temporal and gitvoyant analyze agent
  • Comprehensive error messages and user guidance
  • Smart resolution of local paths and remote Git URLs
  • Structured rendering with risk-ranked file listings

Quality

  • 55%+ test coverage with unit, integration, and CLI tests
  • Full type annotations throughout codebase
  • Robust error recovery and partial result handling
  • Structured logging with configurable levels
  • Environment-based configuration with .env support

Developer Experience

  • UV package manager, Ruff linting and formatting
  • Makefile with build automation
  • Proper CLI installation with gitvoyant command
  • Google-style docstrings throughout

Technical Implementation

  • Git integration via GitPython for commit history processing
  • NumPy/Pandas for linear regression and trend analysis
  • Python AST parsing for cyclomatic complexity calculation
  • AsyncIO integration for service layer operations
  • Immutable value objects and rich domain entities via dataclasses

Removed (v0.1.0 Research Artifacts)

  • Flask-specific research examples and results
  • Research-oriented demonstration scripts
  • Illustrative graphs and sample data
  • REST API (simplified to CLI and agent interfaces)

Migration from v0.1.0

  • Core temporal evaluation algorithm preserved in full
  • Research algorithms restructured into production architecture
  • Same linear regression and complexity analysis foundation, now with confidence scoring
  • IMPROVING/DECLINING/STABLE classification with statistical rigor

[0.1.0] — 2025-07-01

Initial Research Implementation

  • First implementation of Git history-based quality analysis
  • Flask repository analysis as discovery case
  • 878-line research prototype demonstrating temporal intelligence viability
  • Basic IMPROVING/DECLINING/STABLE quality pattern classification

Core Algorithms

  • AST-based cyclomatic complexity measurement across Git commits
  • Statistical trend analysis via linear regression for complexity evolution
  • Commit-by-commit code analysis and metrics extraction
  • Quality pattern detection for positive and negative trends

Research Validation

  • Analysis of Flask, Django, and other open-source repositories
  • Identification of quality engineering signatures in successful projects
  • Demonstrated that temporal analysis provides insights unavailable to static tools
  • Established temporal intelligence as a viable analytical approach

Key Findings

  • Historical analysis yields information inaccessible to point-in-time evaluation
  • Deliberate complexity reduction patterns are observable in mature projects
  • Quality decay forecasting is feasible from temporal data
  • Approach validated as suitable for production AI agent platform