Migrate tools and enhance commit message generation with tests - #10
Merged
Conversation
…11 CE + 5 vscode-bridged), and all 11 LLM providers onto the plugin system's factory pattern, wiring them into real FeatureModule.register() bodies backed by buildRuntime(). Found and fixed two real bugs along the way (a git_diff duplicate-registration bug already latent in production, documented rather than silently patched; a mentioned-file/mentioned-files id mismatch in my own new code, caught before it shipped). Resolved a genuine open question — whether interfaces/llm/LlmProvider and kernel/llm/types.ts's LlmProvider could be reconciled — by verifying structural compatibility rather than forcing a cast. Added three new parity tests (tools, context sources, providers) that all pass against hand-derived ground truth from ThunderController's real registration code. Deliberately deferred, by explicit user choice: the actual ThunderController/HeadlessAgentHost swap-over to call buildRuntime(), and de-globalizing runtime state — both flagged as too risky to attempt without live VS Code testing, which isn't available in this environment. Both the plan doc and the repo-committed docs/architecture/enterprise-migration-plan.md now record this as a deliberate stopping point, not an oversight.
…lity - Added support for disabling reasoning in OpenAiCompatibleProvider to prevent excessive output budget usage for Qwen models. - Refactored commit message generation logic to handle multiple attempts and improved error handling for empty responses. - Updated MicroTaskExecutor to log detailed information during commit message generation, including timing and attempt details. - Enhanced validation logic for commit messages to ensure empty responses are handled gracefully. - Adjusted GitService to allow fetching detailed changed files based on staged or unstaged states. - Improved commit message prompt handling to ensure accurate subject and body extraction. - Added tests to verify new functionality, including handling of empty model responses and retry logic for commit message generation.
…and evidence redaction
- Added tests for successful git restore commands in AgentTaskState. - Updated Git permission policy tests to require explicit approval for dangerous commands. - Modified ToolExecutor tests to hard-block writes in plan mode and ensure proper error handling. - Enhanced pipeline routing tests to correctly classify build restoration as a bugfix. - Refined plan mode tests to filter tools to read-only capabilities and count repository-read MCP tools. - Introduced comprehensive tests for skill management, including duplicate ID detection and draft validation. - Implemented new tests for skill engine functionality, ensuring compatibility and safety constraints. - Updated benchmark task manifests with new generation timestamps.
- Added tests for distinct routing of component, visual design, and React performance skills in act-skill-routing. - Implemented approval continuation routing with active plans in act-orchestration. - Recognized retries of identical commands as failures in agent-loop tests. - Validated bundled skill manifests for unique IDs and referenced files in bundled-skills tests. - Ensured broad project repairs are treated as bugfix work in task-analyzer-routing. - Maintained bugfix workflow priority during structured planning in plan-skill-routing. - Improved plan mode orchestration to handle duplicate reproduction steps and diagnostic captures. - Normalized diagnostic build-capture steps and ensured internal phases are hidden in plan views. - Enhanced verification command discovery for pnpm workspaces and non-JS test commands. - Stripped trailing command wrappers to run real verification commands in run_command exit code tests.
…ment, and test-driven development - Introduced `performance-optimization` skill for measuring and optimizing performance regressions and Core Web Vitals. - Added `planning-and-task-breakdown` skill to create implementation plans for complex tasks. - Created `release-management` skill for managing staged releases, including version bumps and GitHub releases. - Implemented `test-driven-development` skill to guide users through TDD practices for implementing and fixing code. - Updated `react-next-performance` skill description for clarity and specificity. - Enhanced `using-agent-skills` to assist in resolving ambiguity between multiple skills. - Improved parsing logic in tests to handle multi-line descriptions in skill markdown files.
- Updated path matching to only consider task-evidence paths, excluding ambient repository file listings. - Adjusted pin rule matching to reflect task references accurately. - Improved input normalization for file scope candidates to handle various input formats. feat: Implement build diagnostics reading and parsing - Added functionality to read and parse TypeScript build diagnostics from a JSON file. - Introduced methods to filter build errors based on file paths and freshness of the diagnostics data. fix: Coerce input for propose_file_scope tool - Enhanced input normalization to repair common misspellings and field name variations for the propose_file_scope tool. test: Add comprehensive tests for new diagnostics and skill ranking features - Created tests for build diagnostics parsing and reading. - Added tests to validate skill ranking logic based on project context and task references. - Included tests for input coercion in propose_file_scope tool. chore: Update benchmark task manifests with new generation timestamps
- Add tool concurrency management in `toolConcurrency.ts` to determine safe parallel execution of tools. - Update `SkillCatalogService.ts` to include manifest path and hash for skills. - Modify `SkillEngine.ts` to propagate manifest path and hash in skill reports. - Enhance `SkillTelemetry.ts` to include manifest version, path, hash, and supported modes in telemetry data. - Revise `SKILL.md` for the planning-and-task-breakdown skill to clarify usage and rules. - Update `skill.json` for the planning-and-task-breakdown skill with new descriptions, intents, and tests. - Add tests for tool concurrency in `tool-concurrency.test.ts` to validate parallel execution logic. - Introduce `no-progress-detector.test.ts` to evaluate tool call fingerprints and no progress scenarios. - Enhance existing tests to cover new functionality and ensure robust behavior in the agent loop and plan execution.
- Implemented `diagnosticFileExtractor.ts` to parse TypeScript and ESLint error outputs, extracting relevant file and diagnostic information. - Enhanced error handling in `analyzeJsonl.ts` to correctly classify skipped tool calls. - Updated `ToolExecutor.ts` to ensure skipped commands are not marked as successful. - Introduced `integrityGuardrail.ts` to prevent harmful requests that weaken safeguards. - Improved file reading in `builtinTools.ts` to handle paths formatted as single strings. - Modified CLI exit behavior to ensure proper process termination. - Added tests for diagnostic extraction and persistence in `diagnostics-store.test.ts`. - Enhanced log analysis tests to verify correct handling of skipped calls. - Removed outdated support and TODO files from benchmark fixtures. - Updated verification logic in `verify.mjs` to accurately assess model output.
…fiers - Introduced INTENT_CATALOG for standardized task definitions including bugfix, feature, refactor, and more. - Implemented LlmIntentClassifier for classifying intents using LLMs, including message normalization and JSON response parsing. - Developed RuleIntentClassifier for deterministic intent classification based on predefined patterns. - Created patterns for various intents to enhance classification accuracy. - Established constants for task intents and confidence thresholds. - Added schemas for intent classification and candidate validation using Zod. - Exported necessary modules and types for integration into the broader system.
- Add TaskTargetExtractor for extracting targets from user messages and referenced artifacts. - Introduce TaskAnalyzerInput and TaskAnalysis types for structured input and output. - Create constants for task analysis patterns including complexity, clarity, scope, constraints, and risks. - Develop RulewiseTaskAnalyzer to analyze tasks based on extracted targets, constraints, and signals. - Implement various analysis signals for clarity, complexity, risk, and scope. - Establish a modular structure for task analysis with separate files for each component.
…s, index, Embeddings
- Implemented PackageJsonReader to parse package.json files, extracting relevant project information such as dependencies, scripts, and entry files. - Created MavenProjectReader for handling Maven project manifests. - Introduced content hashing utilities for consistent snapshot ID generation. - Established workspace ignore policies to manage file and directory exclusions during scans. - Defined schemas for project catalogs and workspace snapshots to enforce data integrity. - Enhanced workspace scanning capabilities with improved statistics and snapshot generation.
…features/ce to v8 - Move Intent system (IntentRouter, classifiers, policy, resolution, schema) to src/v8/intent/ - Move Repository system (catalog, readers, manifests, workspace scanner) to src/v8/repository/ - Add new RepoMap module with builder, renderer, ranker, pageRank, and data-source under src/v8/repository/map/ - Move Task Analyzer (analyzers, classifiers, types) to src/v8/task-analyzer/ - Delete orphaned src/features/ce/repository/index.ts This consolidates core modules under the v8 architecture for better organization.
…emas - Added SQLite code index adapter and constants for configuration. - Implemented schemas for code index files, symbols, imports, and references using Zod for validation. - Created types for code index entities and operations, ensuring type safety across the codebase. - Enhanced RepoMapRanker to integrate with the new code index, allowing for improved ranking and consistency checks. - Removed deprecated data source files and refactored related types to streamline the codebase.
- Introduced RepoGraphBuilder for constructing a repository graph from a code index. - Added constants and types for graph nodes, edges, and builder options. - Created schemas for validating graph structure and statistics. - Removed the CodeIndexReadPort interface as it is no longer needed. - Updated RepoMapBuilder to remove unnecessary imports. - Enhanced the index files for graph and map modules to export necessary components.
- Add RepoMapRanker class for ranking files based on various criteria. - Introduce PageRank algorithm for scoring files based on their connections. - Create schemas for validating RepoMap data structures using Zod. - Implement types for RepoMap files, symbols, and ranking context. - Establish constants for default values and scoring weights. - Add utility functions for path normalization and tokenization. - Introduce budget management for limiting entries and symbols in RepoMap. - Create a new module for handling RepoMap rendering options.
- Add SourceParserRegistry for managing source parsers. - Implement TreeSitterSourceParser for parsing with Tree-sitter. - Implement TypeScriptSourceParser for parsing TypeScript files. - Define schemas for source analysis using Zod for validation. - Create types for source analysis, including symbols, imports, and references.
…schema, constants, and types - Added constants for code indexing including schema version, default values, and SQL migration scripts. - Created a schema definition using Zod for validating code index documents and their components. - Defined types for code indexing, including file versions, document symbols, imports, and references. - Implemented a metadata change detection function to compare desired and current file states. - Established a module structure for code indexing, exporting necessary components for integration.
…ious chunking strategies - Added NodeSha256ChunkHasher for hashing chunk content using SHA256. - Introduced chunking constants and types for configuration and chunk management. - Created chunking schema for validation of chunking inputs and results. - Developed chunking strategies including CodeChunker, MarkdownChunker, and TextChunker. - Implemented ChunkingStrategyRegistry for managing and resolving chunking strategies. - Added tests for chunking service to ensure correct chunking behavior and validation.
…nctionality - Added constants for text index schema version, IDs, defaults, patterns, and SQL statements. - Created schema definitions using Zod for text index documents, search results, and changes. - Implemented the main text index module with necessary exports for coordination, searching, and chunk management. - Developed tests for indexing, searching, and handling changes in text index documents. - Introduced SQLite adapter for managing text index operations and migrations.
- Implemented unit tests for the EmbeddingGenerator and EmbeddingFactory, ensuring correct vector generation and synchronization with the Text Index. - Added a DeterministicProvider for predictable embedding outputs. - Created MemoryTextIndex and MemoryVectorWriter for testing purposes. - Enhanced the Embedding types with detailed interfaces for embedding generation and synchronization. - Updated SQL queries in SqliteTextIndexReader to optimize change retrieval. - Increased MAXIMUM_CHUNK_QUERY_SIZE in constants for better performance. - Added tests to ensure changes are not split across pages during retrieval.
- Added LanceDbVectorIndexWriter for handling embedding index write operations. - Introduced LanceDbTableManager, LanceDbRowMapper, LanceDbFilterBuilder, and LanceDbProfileGuard for managing database interactions and data transformations. - Created constants for vector index configurations, including limits, patterns, and messages. - Developed schemas for validating vector search inputs and results using Zod. - Implemented tests for the vector index writer, ensuring correct handling of state and vector rows, as well as revision mismatches. - Established a new index module for organizing and exporting vector index components.
- Added RepoMapRetrievalSource, TextIndexRetrievalSource, and VectorIndexRetrievalSource to handle different retrieval strategies. - Introduced a new index file to export all retrieval sources for easier access. - Developed a comprehensive testing suite for hybrid retrieval, covering various scenarios including successful retrieval, partial results, and error handling. - Enhanced type definitions for hybrid retrieval requests and results to improve type safety and clarity. - Updated the main repository index to include hybrid retrieval functionalities.
… tests - Introduced a new context selection module with constants defining various parameters and limits. - Created schemas for validating context selection inputs and outputs using Zod. - Defined types for context selection, including modes, references, budgets, and results. - Implemented the ContextSelector class for selecting context based on provided queries and references. - Added comprehensive tests to ensure correct functionality of context selection logic. - Updated repository index to include the new context selection module.
…es, and tests - Introduced context assembly module with core functionalities for assembling context blocks. - Added schemas for context assembly input, output, and various related structures using Zod for validation. - Implemented context assembly factory and error handling mechanisms. - Created comprehensive tests for context assembly functionalities, covering various scenarios including sensitive path handling, content sanitization, and representation fallbacks. - Updated repository index to export the new context assembly module. - Refactored workspace schema imports for consistency.
- Introduced conversation carry functionality to build and manage conversation history. - Added support for resolving plan handoff based on engine mode and pending plans. - Enhanced sidebar provider to save plans to workspace upon approval. - Created new approval presets for UI modes: safe, guided, and pilot. - Updated ComposerControls to reflect new approval options and their descriptions. - Refactored icons for approval actions and updated styles for dropdown options. - Added tests for conversation carry and plan store functionalities to ensure reliability. - Updated SDK contracts to include conversation message schema and approved plan handling. - Enhanced agent engine pipeline to support host-carried plans and adjust reason codes accordingly.
- Added run budget settings to the UI, allowing users to configure limits on model calls, tool calls, loop iterations, and wall time. - Updated the `RunBudgetSettingsSnapshot` interface to include new properties for budget management. - Enhanced the `compactEvent` function to include selected and omitted IDs in tool events for better tracking. - Improved error handling and logging in the `AgentEnginePipeline` to provide clearer feedback on verification failures and output truncation recovery. - Implemented text continuation recovery for truncated answers, ensuring smoother user experience during long responses. - Updated tests to cover new functionality and ensure robustness of the agent engine's behavior with the new budget settings.
- Implemented FileChangesCard component to display file changes with options to open, review, and undo changes. - Enhanced MarkdownMessage component to support inline file references for opening files directly from markdown content. - Updated MessageList component to render FileChangesCard when file changes are present in chat turns. - Introduced fileLinks utility functions for parsing and detecting workspace file references. - Added modeColors utility for managing UI colors based on agent modes. - Extended protocol definitions to include file change views and related types. - Updated styles for file changes summary and card components for improved UI. - Created tests for file changes functionality and file link parsing. - Enhanced planStore tests to include timestamp formatting for file names.
- Implemented , , and functionalities in the NodeFileSystemAdapter. - Updated WorkspaceFileSystemPort interface to include new methods for directory removal and file renaming. - Enhanced ToolCatalog with input and output schemas for the new mutation tools. - Developed MutationTransactionRegistry methods to handle file and directory deletions and moves, including rollback capabilities. - Created comprehensive tests for the new filesystem mutation tools, ensuring correct functionality and error handling. - Updated decision policy constants and pipeline to incorporate new mutation tools. - Adjusted development setup scripts for improved native module handling.
…nostics and manifest reading
…verification outcomes
…omponents for context and activity display
…and public API tests - Updated v8-module-boundaries test to include 'createWorkspaceIndexRuntime' and 'createWorkspaceRetrievalRuntime'. - Enhanced public API smoke test to verify exports of workspace index and retrieval runtime functions.
… and adding typecheck evidence from build scripts
…r default behavior
- Implemented for OpenAI compatible embeddings and metadata management. - Created for host-injected SQLite database handling. - Added TypeScript configuration for the host package. - Introduced for durable run-checkpoint storage. - Enhanced with retry logic for HTTP errors. - Updated model provider support matrix to reflect shipped adapters. - Added tests for and retry logic in . - Updated various package dependencies and scripts for improved development setup.
Combine file-system skills catalog with workspace checkpoint store wiring. Co-authored-by: Cursor <cursoragent@cursor.com>
…and diagnostics filtering
- Add function to load project instructions from AGENTS.md, .mitii/rules, and MITTII.local.md. - Introduce for optional web search functionality, including API integration. - Enhance skills catalog to support bundled skills and improve file system handling. - Create workspace snapshot functionality to capture repository state without full indexing. - Introduce new diagnostics filtering mechanism to identify newly introduced issues after patch applications. - Implement memory reinjection and dropped turns summarization in message compaction logic. - Update decision policy to conditionally grant web search based on host capabilities. - Add comprehensive tests for new features and ensure existing functionality remains intact.
- Removed default host skills from the SDK, allowing hosts to manage skills through the filesystem. - Introduced new skills: `ask-concise`, `bugfix-localize`, `planning-default`, and `safety-always` with detailed descriptions and intents. - Added support for path-scoped skills, enabling skills to be gated based on repository paths. - Implemented a `FileDiffPreviewStore` for managing proposed file changes without mutating original files. - Enhanced planning actions to utilize selected skill phases, improving task management and execution. - Updated tests to cover new functionalities and ensure path-scoped skills are correctly selected based on evidence.
…i/v8 feat: add LICENSE file to @mitii/sdk package docs: update README.md for @mitii/sdk with installation instructions and quick start guide fix: update package.json for @mitii/sdk and @mitii/v8 to reflect new version and repository details docs: enhance README.md for @mitii/v8 with installation, quick start, and development instructions chore: update bump-version script message to reflect Mitii branding
- Updated extension name from to in package.json and related files. - Enhanced README.md with detailed installation instructions, features, and settings. - Added new commands and clarified usage for cloud providers and local setups. - Improved platform notes and development setup instructions. fix: ensure Node ABI compatibility for better-sqlite3 - Introduced scripts to ensure better-sqlite3 matches the current Node ABI. - Updated rebuild scripts to handle Electron and Node ABI separately. - Enhanced error handling and logging for rebuild processes. chore: update package.json scripts and dependencies - Modified test scripts to include ABI checks before running tests. - Adjusted preflight checks and postinstall messages for clarity. - Updated package.json files across multiple packages to include repository and homepage links. test: update tests for renamed extension - Updated architecture tests to reflect the new extension name.
- Added functions to detect empty and transitional assistant turns. - Implemented logic to recover from incomplete answers by prompting for final responses or tool calls. - Enhanced the agent engine pipeline to handle empty and transitional turns, ensuring a complete user-facing answer is provided. - Updated decision policy to route past-tense follow-ups about prior work correctly. - Introduced tests for new functionalities, ensuring robustness and correctness of the recovery mechanism. - Updated documentation to reflect changes in behavior and usage patterns.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This pull request makes significant improvements to the project's build, release, and development workflows, focusing on clarity, maintainability, and aligning automation with the current product release strategy. Key updates include deferring jobs that are not currently supported, modernizing build and test scripts, and improving VS Code development configurations.
Workflow and Release Process Improvements:
/mitiiagent command and npm publishing, providing clear in-file documentation about their status and instructions for future enablement. [1] [2]Development Tooling and Automation:
These changes collectively improve the developer experience, automate repetitive tasks, and ensure the release process matches the current state of the product.