[github-actions]

🤖 New release

• just-ai: 0.1.0
• just-ai-lsp: 0.1.0
• just-ai-mcp: 0.1.0

Changelog

just-ai

[0.1.0] - 2026-10-08

🚀 Features

• feat: establish layered just-ai architecture (a8bc2ab)

• feat(gui): add recipe arguments and run history (79ff26d)

• feat(execution): cancel Unix process trees (7fe656c)

• feat(history): expose cancellable run diagnostics (a754d98)

• feat(provider): add OpenAI Responses adapter (59ae792)

• feat(provider): add native Ollama adapter (1ae4881)

• feat(mcp): add read-only stdio adapter (0bf8b4d)

• feat(agent): add layered verification command (431e4e6)

• feat(execution): cancel Windows process trees (2403a33)

• feat(config): add complete configuration system

• Add Config, HistoryConfig, ExecutionConfig, RiskConfig, PolicyConfig, AiConfig,
  ScannerConfig, BoundedFileConfig, McpConfig structs with TOML loading

• Wire HistoryConfig into history module (JsonLineHistory, RunRecord, project_history_path)

• Wire ExecutionConfig into execution module (RecipeExecutor, stream_queue_capacity,
  cancellation_poll_interval, max_capture_bytes)

• Wire RiskConfig into RiskFinding.scan_line with custom rule support

• Wire PolicyConfig into DefaultPolicy.evaluate with custom decisions

• Wire ScannerConfig into ProjectScanner with file/total budgets and allowlist

• Wire BoundedFileConfig into bounded_file module (max_editable_file_bytes, prefix_limit)

• Update CLI to load config via Config::load() for history path/params

• Add example just-ai.toml with all sections documented

• All 64 tests pass (9bf6a4f)

• feat(history): migrate JSONL history to SQLite backend

• Added HistoryBackend enum (jsonl|sqlite) with sqlite as default

• Created SqliteHistory implementation with full RunHistory trait

• Added query() method for filtering by recipe/success

• Added create_history() factory function for backend selection

• Updated CLI to use create_history() for backend-agnostic access

• Updated GUI to use create_history() for recent_runs and execute_run

• Added HistoryConfig options: backend, database_file

• Added sqlx, tokio(rt-multi-thread), chrono, uuid dependencies

• Added 2 new tests for JSONL query filtering

• All 66 tests pass (4fde9ea)

• feat(provider): add streaming support for AI provider responses

• Added AiStream type alias for streaming responses

• Added complete_stream() method to AiProvider trait with default implementation

• Implemented streaming for OpenAI Responses API (SSE format with output_text chunks)

• Implemented streaming for Ollama (native SSE format with message/content)

• Implemented streaming for OpenAI-Compatible (SSE format with choices/0/delta/content)

• Added futures and async-stream dependencies

• All 66 tests pass (a21c865)

🐛 Bug Fixes

• fix(execution): terminate just options before recipe argv (2a23b1c)

• fix(execution): block evaluative dry-run inputs (45c1ef5)

• fix(core): bound just subprocess output (509bc8e)

• fix(execution): bound recipe output collection (0342fac)

• fix(core): bound project file reads (7b8b6e9)

• fix(history): bound JSONL reads and retention (7013be3)

• fix: validate modularization plans and prevent destructive file overwrites (bae7910)

• fix: align editor adapters with core contracts and UTF-16 positions (8a24fc2)

• fix: enforce companion run policy in editor and cover validation rollback (2dd06b0)

• fix: enforce real runner previews, project policy and redacted AI context (f089682)

• fix: isolate project history and make synchronous storage async-safe (bc87878)

• fix: replace unsafe recipe merging with source-preserving deduplication plans (192a5b9)

• fix: share stored template plans and validate MCP write contracts (1b8a2af)

• fix: scope editor commands and verify native companion contracts in CI (fe175e6)

• fix: apply project execution, provider tuning and history redaction settings (2eed4ee)

• fix: upgrade Codebase graph verification and resolve companion boundary calls (8601e15)

♻️ Refactoring

• refactor: share safe migration logic across CLI, MCP and desktop (b58b683)

✅ Tests

• test: expand just dump compatibility coverage (4771278)

• test(inspection): add Windows dump fixture (6ac1ee2)

• test(execution): remove overflow process race (02945e0)

🔀 CI/CD

• ci: fix clippy warnings and flaky test

• apps/just-ai-gui/src-tauri/src/lib.rs: remove unused SqliteHistory import, prefix unused project_root with _

• crates/just-ai/src/config.rs: add #[derive(Default)] to HistoryBackend enum

• src/alias.rs: replace assert!(a == b) with assert_eq!(a, b)

• crates/just-ai/src/application/execution.rs: add small sleep to avoid 'Text file busy' flaky test on CI

All tests pass locally (1594 workspace + 66 just-ai), clippy clean, fmt clean. (670518a)

• ci: adjust sqlx features to reduce transitive deps

• Switch from 'runtime-tokio-native-tls' to 'runtime-tokio' + 'tls-rustls'

• This avoids pulling in rsa (via native-tls -> ring -> rsa) and others

• Note: sqlx 0.9 requires Rust 1.94, staying on 0.8 for Rust 1.89 compat

All tests pass, clippy clean. (ccf4aa6)

Other

• Add just-ai companion prototype (0aaff24)

• Update specific README blocks (0a9e634)

• P4: Complete Dockerfiles and release artifacts

• Dockerfile.cli: Multi-stage CLI build (working, tested)

• Dockerfile.mcp: Multi-stage MCP build (working, tested with ping/tools)

• Dockerfile.gui: Multi-stage Tauri GUI build (built successfully)

• Dockerfile.dev: Development container with all tooling

• docker-compose.yml: Service orchestration for CLI, MCP, dev

• .dockerignore: Optimized build context

• .github/workflows/ci.yml: Complete CI pipeline (check, test, build, docker, audit, release)

• apps/just-ai-gui/src-tauri/src/lib.rs: Fix private field import

All 3 Docker images build and run correctly.
All cargo tests pass (1594 in workspace + 66 in just-ai). (f233b80)

• P1: Implement P1 improvements

• P1.1: Add just-ai fix <recipe> command for AI-assisted recipe fixes

• P1.2: Add --interactive flag to run command with TUI prompts (dialoguer)

• P1.3: Grouped recipe insertion by dependencies/naming patterns

• P1.4: VSCode extension with risk diagnostics, history panel, commands

• P1.5: MCP tools: run_recipe, get_history, add_recipe, fix_recipe

• P1.6: just-ai history migrate for JSONL -> SQLite conversion

• P1.7: Provider adapters: Anthropic, Azure OpenAI, Google Gemini

• P1.8: Config commands: validate, schema generation with schemars (3b15d6f)

• P2.3 + P4: Multi-recipe AI workflows, batch operations, templates, and composition (ef3834d)

• P5.1: Implement migrate/refactor command with dead code analysis and dependency graph

• Add migrate CLI command with three subcommands:

  • analyze - Detect unreferenced/isolated recipes, dependency cycles, depths, and similar recipes
  • modularize - Group recipes by common prefix for module organization (analysis only)
  • deduplicate - Find and optionally merge similar/duplicate recipes (interactive or auto)

• Add dead code detection functions in inspection.rs:

  • find_unreferenced_recipes() - recipes never used as dependencies
  • find_isolated_recipes() - recipes with no deps and no dependents

• Add dependency graph analysis functions in inspection.rs:

  • detect_cycles() - Find circular dependencies using DFS
  • calculate_dependency_depths() - Longest path from root for each recipe

• Add similarity detection:

  • find_similar_recipes() - Use TextDiff to find potential duplicates

All tests pass (61), cargo fmt clean, clippy clean. (54b072b)

• P5.1: Add tests for migrate functionality

• Add test fixtures for migration analysis:

  • just-dump-migrate.json - full project with dependencies for testing analysis
  • just-dump-cycle.json - circular dependency detection
  • just-dump-similar.json - duplicate recipe detection

• Add unit tests in inspection.rs:

  • parses_versioned_migrate_fixture - tests unreferenced/isolated recipes, cycles, depths, similarity
  • detects_cycles - tests circular dependency detection with normalization
  • finds_similar_recipes - tests text similarity detection (100% for identical, <80% for different)

• Fix calculate_dependency_depths() to use BFS from roots (corrected depth calculation)

All 64 tests pass, cargo fmt clean, clippy clean. (0db76fd)

• P5.2: Implement modularize action with module file creation

• Enhanced modularize_project in cli.rs to:

  • Group recipes by common prefix (e.g., test-, build-, deploy-)
  • Create .just module files for each group with 2+ recipes
  • Extract recipes from root justfile and move to module files
  • Add import statements at top of root justfile
  • Validate final justfile with just --dump

• Added new helper functions:

  • extract_recipe() - extracts a recipe definition with its body
  • add_imports_at_top() - inserts import statements after existing imports

• Added test fixture just-dump-modularize.json with 8 recipes for testing

• Added test parses_versioned_modularize_fixture() for grouping logic

All 65 tests pass, cargo fmt clean, clippy clean. (234730f)

• P5.3: Improve deduplicate action with smart merge

• Enhanced deduplicate_project in cli.rs:

  • Added --merge flag for smart merging instead of removing one recipe
  • Added interactive option 'm' for smart merge (combines unique parts of both recipes)
  • Smart auto-merge when --write --merge used together

• Added smart_merge_recipes() function that:

  • Uses shorter name (more generic)
  • Merges docs (prefers longer)
  • Unions parameters (prefers ones with defaults)
  • Unions dependencies
  • Combines unique body lines from both recipes (deduplicates identical lines)

• Added test smart_merge_recipes_preserves_unique_body_lines for similarity detection

All 66 tests pass, cargo fmt clean, clippy clean. (a6c53fc)

• Stage 4 implemented (8f08daf)

• P2.4: Tauri GUI Config Validation & Schema commands

• Added ai_config_validate command - validates just-ai.toml configuration

• Added ai_config_schema command - outputs JSON Schema for just-ai.toml

• Made find_config public in config module for Tauri backend access

• Added schemars dependency to Tauri Cargo.toml

• Added TypeScript types and API functions: ConfigValidationResult, ConfigSchemaResult, aiConfigValidate, aiConfigSchema

• Added UI buttons for Validate Config and Config Schema in Utility panel

• Added result display components: ConfigValidationResult (shows valid/invalid with errors), ConfigSchemaResult (pretty-prints JSON schema)

• All tests pass, cargo fmt and clippy clean (1d4dac3)

• P3.1: Template Persistence - Store templates in .just-ai/templates/

• Added TemplateProposal serialization to JSON files in .just-ai/templates/

• Added save_template(), load_template(), list_templates(), delete_template() functions

• Modified handle_template() to persist templates to disk

• Modified handle_instantiate_template() to load templates from disk

• Added CLI commands: template-list, template-delete

• Updated instantiate-template to use stored templates instead of re-generating

• All tests pass, cargo fmt and clippy clean (d3ea85d)

• Phase 3.2 + P2.3: Built-in Template Library and Multi-Recipe AI Workflows (ffc64fb)

just-ai-lsp

[0.1.0] - 2026-10-08

🐛 Bug Fixes

• fix: resolve workflow action name and forbid lint issues

• Fix release-plz GitHub Action name from 'release-plz/release-plz-action' to 'release-plz/action'

• Remove TODO placeholder in LSP code action template to pass ./bin/forbid check

• Replace 'echo TODO: implement' with '# implement' comment (09a61c3)

• fix: resolve release-plz.toml TOML parse error and cargo fmt issue (26f3fda)

• fix: align editor adapters with core contracts and UTF-16 positions (8a24fc2)

Other

• P2.3 + P4: Multi-recipe AI workflows, batch operations, templates, and composition (ef3834d)

just-ai-mcp

[0.1.0] - 2026-10-08

🚀 Features

• feat(mcp): add read-only stdio adapter (0bf8b4d)

• feat(mcp): expose project agent prompts (65e0b1f)

• feat(mcp): expose architecture resources (5b96dad)

• feat(agent): add layered verification command (431e4e6)

• feat(execution): cancel Windows process trees (2403a33)

• feat(mcp): bound stdio input frames (87f6d0e)

🐛 Bug Fixes

• fix(mcp): negotiate protocol versions (4e2ab81)

• fix(mcp): validate JSON-RPC request envelopes (e1d1447)

• fix(mcp): prevent client-selected executables (9ffbcd7)

• fix(mcp): confine tools to server workspace (7b617b6)

• fix(mcp): enforce tool argument allowlists (b2edf95)

• fix(execution): block evaluative dry-run inputs (45c1ef5)

• fix(core): bound just subprocess output (509bc8e)

• fix(core): bound project file reads (7b8b6e9)

• fix: enforce real runner previews, project policy and redacted AI context (f089682)

• fix: isolate project history and make synchronous storage async-safe (bc87878)

• fix: replace unsafe recipe merging with source-preserving deduplication plans (192a5b9)

• fix: share stored template plans and validate MCP write contracts (1b8a2af)

• fix: upgrade Codebase graph verification and resolve companion boundary calls (8601e15)

• fix: verify execution preparation paths across all companion adapters (2e26ab1)

♻️ Refactoring

• refactor(mcp): isolate prompt resource catalog (f71e473)

• refactor(mcp): isolate read-only tool adapter (dfbb49d)

• refactor(mcp): separate protocol dispatch from transport (56018b8)

• refactor: share safe migration logic across CLI, MCP and desktop (b58b683)

✅ Tests

• test(mcp): verify stdio transport (dbb1fd7)

🔀 CI/CD

• ci: update Cargo.lock for GUI and MCP crates

Fixes CI failure: 'the lock file needs to be updated but --locked was passed'

Both crates had outdated Cargo.lock files that didn't match Cargo.toml after dependency updates. (21e6547)

Other

• P1: Implement P1 improvements

• P1.1: Add just-ai fix <recipe> command for AI-assisted recipe fixes

• P1.2: Add --interactive flag to run command with TUI prompts (dialoguer)

• P1.3: Grouped recipe insertion by dependencies/naming patterns

• P1.4: VSCode extension with risk diagnostics, history panel, commands

• P1.5: MCP tools: run_recipe, get_history, add_recipe, fix_recipe

• P1.6: just-ai history migrate for JSONL -> SQLite conversion

• P1.7: Provider adapters: Anthropic, Azure OpenAI, Google Gemini

• P1.8: Config commands: validate, schema generation with schemars (3b15d6f)

• P2.3 + P4: Multi-recipe AI workflows, batch operations, templates, and composition (ef3834d)

• Stage 4 implemented (8f08daf)

---

This PR was generated with release-plz.