Skip to content

Latest commit

 

History

History
123 lines (92 loc) · 10.4 KB

CHANGELOG.md

File metadata and controls

123 lines (92 loc) · 10.4 KB

Changelog

All notable changes to this repository are documented here.

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

[2.5.0] - 2026-02-23

Added

  • Business Logic Persistence — ReverseEngineeringProcess and ChunkedReverseEngineeringProcess now persist extracted BusinessLogic records to a new business_logic SQLite table via IMigrationRepository.SaveBusinessLogicAsync. Added GetBusinessLogicAsync and DeleteBusinessLogicAsync to IMigrationRepository, SqliteMigrationRepository, and HybridMigrationRepository.
  • Business Logic Injection into Conversion Prompts — All four converter agents (JavaConverterAgent, CSharpConverterAgent, ChunkAwareJavaConverter, ChunkAwareCSharpConverter) now receive extracted BusinessLogic records via SetBusinessLogicContext() (new method on ICodeConverterAgent). In full-pipeline runs, SmartMigrationOrchestrator wires RE output directly into conversion; --reuse-re loads the same context from a previous persisted RE run. A shared FormatBusinessLogicContext() helper in AgentBase formats the context for all four converters.
  • --reuse-re CLI flag — When combined with --skip-reverse-engineering, loads business logic from the latest persisted RE run and injects it into conversion prompts. doctor.sh convert-only now prompts interactively for this choice.
  • REST API: GET/DELETE /api/runs/{runId}/business-logic — Returns per-file business logic summary (story/feature/rule counts); DELETE removes persisted results to allow re-running RE for that run.
  • Portal: per-run 🔬 RE Results button — Shows the business logic summary table for a run and allows deletion of persisted results directly from the UI.
  • RE Results in Portal Chat — Chat endpoint injects business purpose, user stories, features, and business rules from the business_logic table into the AI prompt context. Updated AI system prompt accordingly.

Fixed

  • Empty Technical Analysis in RE output — ReverseEngineeringProcess and ChunkedReverseEngineeringProcess now fall back to rendering RawAnalysisData when structured CobolAnalysis fields are unpopulated.
  • Total Features always 0 — BusinessLogicExtractorAgent.ExtractFeatures() now matches ### Use Case N: and ### Operation headings in addition to ### Feature:, reflecting the actual AI prompt output.

Changed

  • Dependency mapping runs once per full run — RE processes (ReverseEngineeringProcess, ChunkedReverseEngineeringProcess) now include a dedicated dependency mapping step (step 4/5) and store the result on ReverseEngineeringResult.DependencyMap. MigrationProcess and ChunkedMigrationProcess accept a SetDependencyMap() call and skip AnalyzeDependenciesAsync when a map is already provided. SmartMigrationOrchestrator.RunAsync threads existingDependencyMap through to both migration paths. Dependency output files (dependency-map.json, dependency-diagram.md) are now generated in the RE output folder as well as the migration output folder.
  • doctor.sh — Updated convert-only to prompt for --reuse-re; corrected portal navigation references to match current UI ('📄 Reverse Engineering Results').

[2.4.0] - 2026-02-16

Added

  • Automated Documentation Checker — New GitHub Actions workflow (documentation-updater) that reviews code changes on every push and PR to main, identifies missing or outdated documentation, and notifies the responsible author via PR comments or issues.
  • Speed Profile Selection - New interactive prompt in doctor.sh lets you choose between four speed profiles before running migrations, reverse engineering, or conversion-only:
    • TURBO — Low reasoning on ALL files with no exceptions. 65K token ceiling, parallel file conversion (4 workers), 200ms stagger delay. Designed for testing and smoke runs where speed matters more than quality.
    • FAST — Low reasoning on most files, medium only on the most complex ones. 32K token cap, parallel conversion (3 workers), 500ms stagger. Good for quick iterations and proof-of-concept runs.
    • BALANCED (default) — Uses the three-tier content-aware reasoning system. Simple files get low effort, complex files get high effort. Parallel conversion (2 workers), 1s stagger.
    • THOROUGH — Maximum reasoning on all files regardless of complexity. Parallel conversion (2 workers), 1.5s stagger. Best for critical codebases where accuracy matters more than speed.
  • Shared select_speed_profile() function — Called from run_migration(), run_reverse_engineering(), and run_conversion_only(). Sets CODEX_* environment variables that are picked up by Program.cs OverrideSettingsFromEnvironment() at startup — no C# changes needed.
  • Adaptive Re-Chunking on Output Exhaustion — When reasoning exhaustion retries fail (all escalation attempts exhausted), AgentBase now automatically splits the COBOL source at the best semantic boundary (DIVISION > SECTION > paragraph > midpoint) and processes each half independently with a 50-line context window (second half begins 50 lines before the split point for continuity). Results are merged with duplicate package/import/class removal and validated for truncation signals. This solves the TURBO/FAST paradox where small output token caps caused repeated exhaustion failures rather than triggering the existing input-size-based chunking.
  • Parallel File Conversion — All 4 converter agents (ChunkAwareJavaConverter, ChunkAwareCSharpConverter, JavaConverterAgent, CSharpConverterAgent) now support parallel file conversion via SemaphoreSlim-based concurrency control. Controlled by MaxParallelConversion setting (default: 2). TURBO uses 4 workers, FAST uses 3, BALANCED/THOROUGH use 2.
  • Environment Variable Overrides for Timing — New env vars CODEX_STAGGER_DELAY_MS, CODEX_MAX_PARALLEL_CONVERSION, and CODEX_RATE_LIMIT_SAFETY_FACTOR allow fine-tuning of parallelism and rate limiting without code changes.

Fixed

  • Settings Injection Bug — All agent constructors in MigrationProcess.cs, ChunkedMigrationProcess.cs, and Program.cs were missing the settings parameter, causing AppSettings to always be null inside agents. As a result, runtime configuration (including environment variable overrides such as CODEX_MAX_PARALLEL_CONVERSION) could not be applied, and agents fell back to the default MaxParallelConversion value of 1 (sequential). All 10 constructor call sites now pass settings correctly so both static config and env var overrides take effect as intended.
  • Hardcoded Rate Limit Safety Margin — RateLimitTracker.SafetyMargin was hardcoded at 0.90, ignoring the configurable RateLimitSafetyFactor from ChunkingSettings. Now accepts a safetyMargin parameter wired from settings (TURBO=0.85, default=0.70).

Changed

  • README.md — Added Speed Profile documentation with profile comparison table
  • doctor.sh — Added select_speed_profile() function and integrated into all three run commands. TURBO/FAST profiles now export parallel conversion and stagger delay env vars.
  • TokenHelper.cs — CalculateRequestDelay delay floor lowered from hardcoded 15s to configurable (default 2s, minimum 500ms)
  • ChunkingSettings.cs — Added MaxParallelConversion property (default 1)

[2.3.1] - 2026-02-12

Fixed

  • Line-based chunking fallback for data-only copybooks (no DIVISION/SECTION/PARAGRAPH)
  • SemaphoreSlim disposal (using var) and over-release prevention (lockHeld flag)
  • Config script injection: eval → envsubst in load-config.sh
  • Port cleanup: lsof -sTCP:LISTEN to avoid killing client connections

Added

  • Chunking stress test for line-based fallback on large copybooks

[2.3.0] - 2026-02-06

Changed

  • Removed "Spec-Driven Migration" workflow; focused on "Deep Code Analysis" pipeline
  • Updated architecture diagrams for Deep SQL Analysis flow (Regex → SQLite → Portal)
  • Cleaned up deprecated doctor.sh functions

[2.2.1 – 2.2.2] - 2025-12-16

Fixed

  • BusinessLogicExtractorAgent auth: switched to ResponsesApiClient (HTTP 401 fix)
  • Strict regex for class extraction, preventing AI comment artifacts (e.g., Completes.java)

[2.2.0] - 2025-12-15

Added

  • Smart Chunking - Semantic chunking for large files (>3K lines), parallel processing (6 workers), cross-chunk SignatureRegistry
  • Portal chunks tab with real-time progress; doctor.sh chunking-health command
  • DB tables: chunk_metadata, forward_references, signatures, type_mappings

Fixed

  • 88% code loss on files >50K LOC (now routed through chunked process)
  • Stale run status, duplicate DB paths, portal port conflicts

Configuration

  • MaxLinesPerChunk: 1500, OverlapLines: 300, MaxParallelAnalysis: 6, TokenBudgetPerMinute: 300K

[2.1.0 – 2.1.1] - 2025-11-13 to 2025-11-24

Added

  • C# .NET Support - Dual-language output (Java Quarkus or C# .NET) via CSharpConverterAgent
  • Migration Reports - Portal, CLI, or API (/api/runs/{runId}/report)
  • Mermaid Diagrams - Interactive flowcharts, sequence, class, and ER diagrams
  • Enhanced dependency tracking (CALL, COPY, PERFORM, EXEC SQL, READ/WRITE)

Changed

  • Unified output/ directory; renamed cobol-source/ → source/
  • GPT-5 Mini (32K tokens) configuration

[2.0.0] - 2025-11-11

Added

  • Reverse Engineering - reverse-engineer command, BusinessLogicExtractorAgent, glossary support
  • Hybrid Database - SQLite + Neo4j via HybridMigrationRepository
  • Portal UI - Three-panel dashboard with run selector, graphs, AI chat (port 5028)
  • REST API - /api/runinfo, /api/runs/all, /api/graph, /api/chat
  • DevContainer auto-start, 9 MCP resources per run

Changed

  • Port standardization: 5028 / 7474 / 7687
  • doctor.sh auto-fixes, .NET 9 detection, Windows compatibility

[1.0.0 – 1.3.0] - 2025-10-01 to 2025-10-23

Core (1.0.0)

  • Initial release: COBOL → Java Quarkus migration with AI agents (CobolAnalyzer, JavaConverter, DependencyMapper)
  • SQLite persistence, MCP server, doctor.sh CLI, Azure OpenAI (GPT-4), Dev container

Incremental (1.1.0 – 1.3.0)

  • Neo4j integration → hybrid database (SQLite + Neo4j), dependency graph visualization
  • McpChatWeb portal (three-panel dashboard, 9 MCP resources, run selector, dynamic graphs)
  • .NET 9 standardization, multi-run query support