Adds a new CLI command and MCP tool to reorganize tasks and subtasks within the hierarchy. Features include:
- Moving tasks between different positions in the task list
- Converting tasks to subtasks and vice versa
- Moving subtasks between parents
- Moving multiple tasks at once with comma-separated IDs
- Creating placeholder tasks when moving to new IDs
- Validation to prevent accidental data loss
This is particularly useful for resolving merge conflicts when multiple team members create tasks on different branches.
Enhance analyze-complexity to support analyzing specific tasks by ID or range:
- Add --id option for comma-separated task IDs
- Add --from/--to options for analyzing tasks within a range
- Implement intelligent merging with existing reports
- Update CLI, MCP tools, and direct functions for consistent support
- Add changeset documenting the feature
This commit implements automatic tasks.json file creation when it doesn't exist:
- When tasks.json is missing or invalid, create a new one with { tasks: [] }
- Allows adding tasks immediately after initializing a project without parsing a PRD
- Replaces error with informative feedback about file creation
- Enables smoother workflow for new projects or directories
This change improves user experience by removing the requirement to parse a PRD
before adding the first task to a newly initialized project. Closes#494
This commit significantly improves the functionality by implementing
fuzzy semantic search to find contextually relevant dependencies:
- Add Fuse.js for powerful fuzzy search capability with weighted multi-field matching
- Implement score-based relevance ranking with high/medium relevance tiers
- Enhance context generation to include detailed information about similar tasks
- Fix context shadowing issue that prevented detailed task information from
reaching the AI model
- Add informative CLI output showing semantic search results and dependency patterns
- Improve formatting of dependency information in prompts with task titles
The result is that newly created tasks are automatically placed within the correct
dependency structure without manual intervention, with the AI having much better
context about which tasks are most relevant to the new one being created.
This significantly improves the user experience by reducing the need to manually
update task dependencies after creation, all without increasing token usage or costs.
This commit introduces several improvements to AI interactions and
task management functionalities:
- AI Provider Enhancements (for Telemetry & Robustness):
- :
- Added a check in to ensure
is a string, throwing an error if not. This prevents downstream
errors (e.g., in ).
- , , :
- Standardized return structures for their respective
and functions to consistently include /
and fields. This aligns them with other providers (like
Anthropic, Google, Perplexity) for consistent telemetry data
collection, as part of implementing subtask 77.14 and similar work.
- Task Expansion ():
- Updated to be more explicit
about using an empty array for empty to
better guide AI output.
- Implemented a pre-emptive cleanup step in
to replace malformed with
before JSON parsing. This improves resilience to AI output quirks,
particularly observed with Perplexity.
- Adjusts issue in commands.js where successfulRemovals would be undefined. It's properly invoked from the result variable now.
- Updates supported models for Gemini
These changes address issues observed during E2E tests, enhance the
reliability of AI-driven task analysis and expansion, and promote
consistent telemetry data across multiple AI providers.
This commit updates the Perplexity AI provider () to ensure its functions return data in a structure consistent with other providers and the expectations of the unified AI service layer ().
Specifically:
- now returns an object instead of only the text string.
- now returns an object instead of only the result object.
These changes ensure that can correctly extract both the primary AI-generated content and the token usage data for telemetry purposes when Perplexity models are used. This resolves issues encountered during E2E testing where complexity analysis (which can use Perplexity for its research role) failed due to unexpected response formats.
The function was already compliant.
This commit introduces two key improvements:
1. **Google Provider Telemetry:**
- Updated to include token usage data (, ) in the responses from and .
- This aligns the Google provider with others for consistent AI usage telemetry.
2. **Robust AI Object Response Handling:**
- Modified to more flexibly handle responses from .
- The add-task module now check for the AI-generated object in both and , improving compatibility with different AI provider response structures (e.g., Gemini).
These changes enhance the reliability of AI interactions, particularly with the Google provider, and ensure accurate telemetry collection.
This commit applies the standard telemetry pattern to the analyze-task-complexity command and its corresponding MCP tool.
Key Changes:
1. Core Logic (scripts/modules/task-manager/analyze-task-complexity.js):
- The call to generateTextService now includes commandName: 'analyze-complexity' and outputType.
- The full response { mainResult, telemetryData } is captured.
- mainResult (the AI-generated text) is used for parsing the complexity report JSON.
- If running in CLI mode (outputFormat === 'text'), displayAiUsageSummary is called with the telemetryData.
- The function now returns { report: ..., telemetryData: ... }.
2. Direct Function (mcp-server/src/core/direct-functions/analyze-task-complexity.js):
- The call to the core analyzeTaskComplexity function now passes the necessary context for telemetry (commandName, outputType).
- The successful response object now correctly extracts coreResult.telemetryData and includes it in the data.telemetryData field returned to the MCP client.
This commit applies the standard telemetry pattern to the update-subtask command and its corresponding MCP tool.
Key Changes:
1. Core Logic (scripts/modules/task-manager/update-subtask-by-id.js):
- The call to generateTextService now includes commandName: 'update-subtask' and outputType.
- The full response { mainResult, telemetryData } is captured.
- mainResult (the AI-generated text) is used for the appended content.
- If running in CLI mode (outputFormat === 'text'), displayAiUsageSummary is called with the telemetryData.
- The function now returns { updatedSubtask: ..., telemetryData: ... }.
2. Direct Function (mcp-server/src/core/direct-functions/update-subtask-by-id.js):
- The call to the core updateSubtaskById function now passes the necessary context for telemetry (commandName, outputType).
- The successful response object now correctly extracts coreResult.telemetryData and includes it in the data.telemetryData field returned to the MCP client.
This commit applies the standard telemetry pattern to the update-tasks command and its corresponding MCP tool.
Key Changes:
1. Core Logic (scripts/modules/task-manager/update-tasks.js):
- The call to generateTextService now includes commandName: 'update-tasks' and outputType.
- The full response { mainResult, telemetryData } is captured.
- mainResult (the AI-generated text) is used for parsing the updated task JSON.
- If running in CLI mode (outputFormat === 'text'), displayAiUsageSummary is called with the telemetryData.
- The function now returns { success: true, updatedTasks: ..., telemetryData: ... }.
2. Direct Function (mcp-server/src/core/direct-functions/update-tasks.js):
- The call to the core updateTasks function now passes the necessary context for telemetry (commandName, outputType).
- The successful response object now correctly extracts coreResult.telemetryData and includes it in the data.telemetryData field returned to the MCP client.
This commit applies the standard telemetry pattern to the command and its corresponding MCP tool.
Key Changes:
1. **Core Logic ():**
- The call to now includes and .
- The full response is captured.
- (the AI-generated text) is used for parsing the updated task JSON.
- If running in CLI mode (), is called with the .
- The function now returns .
2. **Direct Function ():**
- The call to the core function now passes the necessary context for telemetry (, ).
- The successful response object now correctly extracts and includes it in the field returned to the MCP client.
This commit implements AI usage telemetry for the `expand-all-tasks` command/tool and refactors its CLI output for clarity and consistency.
Key Changes:
1. **Telemetry Integration for `expand-all-tasks` (Subtask 77.8):**\n - The `expandAllTasks` core logic (`scripts/modules/task-manager/expand-all-tasks.js`) now calls the `expandTask` function for each eligible task and collects the individual `telemetryData` returned.\n - A new helper function `_aggregateTelemetry` (in `utils.js`) is used to sum up token counts and costs from all individual expansions into a single `telemetryData` object for the entire `expand-all` operation.\n - The `expandAllTasksDirect` wrapper (`mcp-server/src/core/direct-functions/expand-all-tasks.js`) now receives and passes this aggregated `telemetryData` in the MCP response.\n - For CLI usage, `displayAiUsageSummary` is called once with the aggregated telemetry.
2. **Improved CLI Output for `expand-all`:**\n - The `expandAllTasks` core function now handles displaying a final "Expansion Summary" box (showing Attempted, Expanded, Skipped, Failed counts) directly after the aggregated telemetry summary.\n - This consolidates all summary output within the core function for better flow and removes redundant logging from the command action in `scripts/modules/commands.js`.\n - The summary box border is green for success and red if any expansions failed.
3. **Code Refinements:**\n - Ensured `chalk` and `boxen` are imported in `expand-all-tasks.js` for the new summary box.\n - Minor adjustments to logging messages for clarity.
This commit integrates AI usage telemetry for the `expand-task` command/tool and resolves issues related to incorrect return type handling and logging.
Key Changes:
1. **Telemetry Integration for `expand-task` (Subtask 77.7):**\n - Applied the standard telemetry pattern to the `expandTask` core logic (`scripts/modules/task-manager/expand-task.js`) and the `expandTaskDirect` wrapper (`mcp-server/src/core/direct-functions/expand-task.js`).\n - AI service calls now pass `commandName` and `outputType`.\n - Core function returns `{ task, telemetryData }`.\n - Direct function correctly extracts `task` and passes `telemetryData` in the MCP response `data` field.\n - Telemetry summary is now displayed in the CLI output for the `expand` command.
2. **Fix AI Service Return Type Handling (`ai-services-unified.js`):**\n - Corrected the `_unifiedServiceRunner` function to properly handle the return objects from provider-specific functions (`generateText`, `generateObject`).\n - It now correctly extracts `providerResponse.text` or `providerResponse.object` into the `mainResult` field based on `serviceType`, resolving the "text.trim is not a function" error encountered during `expand-task`.
3. **Log Cleanup:**\n - Removed various redundant or excessive `console.log` statements across multiple files (as indicated by recent changes) to reduce noise and improve clarity, particularly for MCP interactions.
Implements AI usage telemetry capture and propagation for the command and MCP tool, following the established telemetry pattern.
Key changes:
- **Core ():**
- Modified the call to include and .
- Updated to receive from .
- Adjusted to return an object .
- Added a call to to show telemetry data in the CLI output when not in MCP mode.
- **Direct Function ():**
- Updated the call to the core function to pass , , and .
- Modified to correctly handle the new return structure from the core function.
- Ensures received from the core function is included in the field of the successful MCP response.
- **MCP Tool ():**
- No changes required; existing correctly passes through the object containing .
- **CLI Command ():**
- The command's action now relies on the core function to handle CLI success messages and telemetry display.
This ensures that AI usage for the functionality is tracked and can be displayed or logged as appropriate for both CLI and MCP interactions.
This commit introduces a standardized pattern for capturing and propagating AI usage telemetry (cost, tokens, model used) across the Task Master stack and applies it to the 'add-task' functionality.
Key changes include:
- **Telemetry Pattern Definition:**
- Added defining the integration pattern for core logic, direct functions, MCP tools, and CLI commands.
- Updated related rules (, ,
Usage: mcp [OPTIONS] COMMAND [ARGS]...
MCP development tools
╭─ Options ──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ --help Show this message and exit. │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯
╭─ Commands ─────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╮
│ version Show the MCP version. │
│ dev Run a MCP server with the MCP Inspector. │
│ run Run a MCP server. │
│ install Install a MCP server in the Claude desktop app. │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯, , ) to reference the new telemetry rule.
- **Core Telemetry Implementation ():**
- Refactored the unified AI service to generate and return a object alongside the main AI result.
- Fixed an MCP server startup crash by removing redundant local loading of and instead using the imported from for cost calculations.
- Added to the object.
- ** Integration:**
- Modified (core) to receive from the AI service, return it, and call the new UI display function for CLI output.
- Updated to receive from the core function and include it in the payload of its response.
- Ensured (MCP tool) correctly passes the through via .
- Updated to correctly pass context (, ) to the core function and rely on it for CLI telemetry display.
- **UI Enhancement:**
- Added function to to show telemetry details in the CLI.
- **Project Management:**
- Added subtasks 77.6 through 77.12 to track the rollout of this telemetry pattern to other AI-powered commands (, , , , , , ).
This establishes the foundation for tracking AI usage across the application.
This commit introduces several improvements and refactorings across MCP tools, core logic, and configuration.
**Major Changes:**
1. **Refactor updateSubtaskById:**
- Switched from generateTextService to generateObjectService for structured AI responses, using a Zod schema (subtaskSchema) for validation.
- Revised prompts to have the AI generate relevant content based on user request and context (parent/sibling tasks), while explicitly preventing AI from handling timestamp/tag formatting.
- Implemented **local timestamp generation (new Date().toISOString()) and formatting** (using <info added on ...> tags) within the function *after* receiving the AI response. This ensures reliable and correctly formatted details are appended.
- Corrected logic to append only the locally formatted, AI-generated content block to the existing subtask.details.
2. **Consolidate MCP Utilities:**
- Moved/consolidated the withNormalizedProjectRoot HOF into mcp-server/src/tools/utils.js.
- Updated MCP tools (like update-subtask.js) to import withNormalizedProjectRoot from the new location.
3. **Refactor Project Initialization:**
- Deleted the redundant mcp-server/src/core/direct-functions/initialize-project-direct.js file.
- Updated mcp-server/src/core/task-master-core.js to import initializeProjectDirect from its correct location (./direct-functions/initialize-project.js).
**Other Changes:**
- Updated .taskmasterconfig fallback model to claude-3-7-sonnet-20250219.
- Clarified model cost representation in the models tool description (taskmaster.mdc and mcp-server/src/tools/models.js).
Problem: The MCP tool previously handled project root acquisition and path resolution within its method, leading to potential inconsistencies and repetition.
Solution: Refactored the tool () to utilize the new Higher-Order Function (HOF) from .
Specific Changes:
- Imported HOF.
- Updated the Zod schema for the parameter to be optional, as the HOF handles deriving it from the session if not provided.
- Wrapped the entire function body with the HOF.
- Removed the manual call to from within the function body.
- Destructured the from the object received by the wrapped function, ensuring it's the normalized path provided by the HOF.
- Used the normalized variable when calling and when passing arguments to .
This change standardizes project root handling for the tool, simplifies its method, and ensures consistent path normalization. This serves as the pattern for refactoring other MCP tools.
Modified update-subtask-by-id core, direct function, and tool to pass projectRoot for .env API key fallback. Removed check preventing appending details to completed subtasks.
Problem:
- Task Master model configuration wasn't properly checking for API keys in the project's .env file when running through MCP
- The isApiKeySet function was only checking session.env and process.env but not inspecting the .env file directly
- This caused incorrect API key status reporting in MCP tools even when keys were properly set in .env
Solution:
- Modified resolveEnvVariable function in utils.js to properly read from .env file at projectRoot
- Updated isApiKeySet to correctly pass projectRoot to resolveEnvVariable
- Enhanced the key detection logic to have consistent behavior between CLI and MCP contexts
- Maintains the correct precedence: session.env → .env file → process.env
Testing:
- Verified working correctly with both MCP and CLI tools
- API keys properly detected in .env file in both contexts
- Deleted .cursor/mcp.json to confirm introspection of .env as fallback works
- Refactors the core `removeTask` function (`task-manager/remove-task.js`) to accept and iterate over comma-separated task/subtask IDs.
- Updates dependency cleanup and file regeneration logic to run once after processing all specified IDs.
- Adjusts the `remove-task` CLI command (`commands.js`) description and confirmation prompt to handle multiple IDs correctly.
- Fixes a bug in the CLI confirmation prompt where task/subtask titles were not being displayed correctly.
- Updates the `remove_task` MCP tool description to reflect the new multi-ID capability.
This addresses the previously known issue where only the first ID in a comma-separated list was processed.
Closes#140
Implements the ability to filter subtasks displayed by the `task-master show <id>` command using the `--status` (or `-s`) flag. This is also available in the MCP context.
- Modified `commands.js` to add the `--status` option to the `show` command definition.
- Updated `utils.js` (`findTaskById`) to handle the filtering logic and return original subtask counts/arrays when filtering.
- Updated `ui.js` (`displayTaskById`) to use the filtered subtasks for the table, display a summary line when filtering, and use the original subtask list for the progress bar calculation.
- Updated MCP `get_task` tool and `showTaskDirect` function to accept and pass the `status` parameter.
- Added changeset entry.
Integrates the OpenRouter AI provider using the Vercel AI SDK adapter (@openrouter/ai-sdk-provider). This allows users to configure and utilize models available through the OpenRouter platform.
- Added src/ai-providers/openrouter.js with standard Vercel AI SDK wrapper functions (generateText, streamText, generateObject).
- Updated ai-services-unified.js to include the OpenRouter provider in the PROVIDER_FUNCTIONS map and API key resolution logic.
- Verified config-manager.js handles OpenRouter API key checks correctly.
- Users can configure OpenRouter models via .taskmasterconfig using the task-master models command or MCP models tool. Requires OPENROUTER_API_KEY.
- Enhanced error handling in ai-services-unified.js to provide clearer messages when generateObjectService fails due to lack of underlying tool support in the selected model/provider endpoint.
Adds the ability for users to specify custom model IDs for Ollama and OpenRouter providers, bypassing the internal supported model list.
- Introduces --ollama and --openrouter flags for the 'task-master models --set-<role>' command.
- Updates the interactive 'task-master models --setup' to include options for entering custom Ollama/OpenRouter IDs.
- Implements live validation against the OpenRouter API when a custom OpenRouter ID is provided.
- Refines the model setting logic to prioritize explicit provider flags/choices.
- Adds warnings when custom models are set.
- Updates the changeset file.
Integrates the xAI provider into the unified AI service layer, allowing the use of Grok models (e.g., grok-3, grok-3-mini).
Changes include:
- Added dependency.
- Created with implementations for generateText, streamText, and generateObject (stubbed).
- Updated to include the xAI provider in the function map.
- Updated to recognize the 'xai' provider and the environment variable.
- Updated to include known Grok models and their capabilities (object generation marked as likely unsupported).
- Add OpenAI provider implementation using @ai-sdk/openai.\n- Update `models` command/tool to display API key status for configured providers.\n- Implement model-specific `maxTokens` override logic in `config-manager.js` using `supported-models.json`.\n- Improve AI error message parsing in `ai-services-unified.js` for better clarity.
The add-task command handler in commands.js was incorrectly passing null for the manualTaskData parameter to the core addTask function. This caused the core function to always fall back to the AI generation path, even when only manual flags like --title and --description were provided. This commit updates the call to pass the correctly constructed manualTaskData object, ensuring that manual task creation via the CLI works as intended without unnecessarily calling the AI service.
This commit updates all relevant documentation (READMEs, docs/*, .cursor/rules) to accurately reflect the finalized unified AI service architecture and the new configuration system (.taskmasterconfig + .env/mcp.json). It also includes the final code cleanup steps related to the refactoring.
Key Changes:
1. **Documentation Updates:**
* Revised `README.md`, `README-task-master.md`, `assets/scripts_README.md`, `docs/configuration.md`, and `docs/tutorial.md` to explain the new configuration split (.taskmasterconfig vs .env/mcp.json).
* Updated MCP configuration examples in READMEs and tutorials to only include API keys in the `env` block.
* Added/updated examples for using the `--research` flag in `docs/command-reference.md`, `docs/examples.md`, and `docs/tutorial.md`.
* Updated `.cursor/rules/ai_services.mdc`, `.cursor/rules/architecture.mdc`, `.cursor/rules/dev_workflow.mdc`, `.cursor/rules/mcp.mdc`, `.cursor/rules/taskmaster.mdc`, `.cursor/rules/utilities.mdc`, and `.cursor/rules/new_features.mdc` to align with the new architecture, removing references to old patterns/files.
* Removed internal rule links from user-facing rules (`taskmaster.mdc`, `dev_workflow.mdc`, `self_improve.mdc`).
* Deleted outdated example file `docs/ai-client-utils-example.md`.
2. **Final Code Refactor & Cleanup:**
* Corrected `update-task-by-id.js` by removing the last import from the old `ai-services.js`.
* Refactored `update-subtask-by-id.js` to correctly use the unified service and logger patterns.
* Removed the obsolete export block from `mcp-server/src/core/task-master-core.js`.
* Corrected logger implementation in `update-tasks.js` for CLI context.
* Updated API key mapping in `config-manager.js` and `ai-services-unified.js`.
3. **Configuration Files:**
* Updated API keys in `.cursor/mcp.json`, replacing `GROK_API_KEY` with `XAI_API_KEY`.
* Updated `.env.example` with current API key names.
* Added `azureOpenaiBaseUrl` to `.taskmasterconfig` example.
4. **Task Management:**
* Marked documentation subtask 61.10 as 'done'.
* Includes various other task content/status updates from the diff summary.
5. **Changeset:**
* Added `.changeset/cuddly-zebras-matter.md` for user-facing `expand`/`expand-all` improvements.
This commit concludes the major architectural refactoring (Task 61) and ensures the documentation accurately reflects the current system.
This commit completes the major refactoring initiative (Task 61) to migrate all AI-interacting task management functions to the unified service layer (`ai-services-unified.js`) and standardized configuration (`config-manager.js`).
Key Changes:
1. **Refactor `update-task-by-id` & `update-subtask-by-id`:**
* Replaced direct AI client logic and config fetching with calls to `generateTextService`.
* Preserved original prompt logic while ensuring JSON output format is requested.
* Implemented robust manual JSON parsing and Zod validation for text-based AI responses.
* Corrected logger implementation (`logFn`/`isMCP`/`report` pattern) for both CLI and MCP contexts.
* Ensured correct passing of `session` context to the unified service.
* Refactored associated direct function wrappers (`updateTaskByIdDirect`, `updateSubtaskByIdDirect`) to remove AI client initialization and call core logic appropriately.
2. **CLI Environment Loading:**
* Added `dotenv.config()` to `scripts/dev.js` to ensure consistent loading of the `.env` file for CLI operations.
3. **Obsolete Code Removal:**
* Deleted unused helper files:
* `scripts/modules/task-manager/get-subtasks-from-ai.js`
* `scripts/modules/task-manager/generate-subtask-prompt.js`
* `scripts/modules/ai-services.js`
* `scripts/modules/ai-client-factory.js`
* `mcp-server/src/core/utils/ai-client-utils.js`
* Removed corresponding imports/exports from `scripts/modules/task-manager.js` and `mcp-server/src/core/task-master-core.js`.
4. **Verification:**
* Successfully tested `update-task` and `update-subtask` via both CLI and MCP after refactoring.
5. **Task Management:**
* Marked subtasks 61.38, 61.39, 61.40, 61.41, and 61.33 as 'done'.
* Includes other task content/status updates as reflected in the diff.
This completes the migration of core AI features to the new architecture, enhancing maintainability and flexibility.
Completes the refactoring of the AI-interacting task management functions by aligning `update-tasks.js` with the unified service architecture and removing now-unused helper files.
Key Changes:
- **`update-tasks.js` Refactoring:**
- Replaced direct AI client calls and AI-specific config fetching with a call to `generateTextService` from `ai-services-unified.js`.
- Preserved the original system and user prompts requesting a JSON array output.
- Implemented manual JSON parsing (`parseUpdatedTasksFromText`) with Zod validation to handle the text response reliably.
- Updated the core function signature to accept the standard `context` object (`{ session, mcpLog }`).
- Corrected logger implementation to handle both MCP (`mcpLog`) and CLI (`consoleLog`) contexts appropriately.
- **Related Component Updates:**
- Refactored `mcp-server/src/core/direct-functions/update-tasks.js` to use the standard direct function pattern (logger wrapper, silent mode, call core function with context).
- Verified `mcp-server/src/tools/update.js` correctly passes arguments and context.
- Verified `scripts/modules/commands.js` (update command) correctly calls the refactored core function.
- **Obsolete File Cleanup:**
- Removed the now-unused `scripts/modules/task-manager/get-subtasks-from-ai.js` file and its export, as its functionality was integrated into `expand-task.js`.
- Removed the now-unused `scripts/modules/task-manager/generate-subtask-prompt.js` file and its export for the same reason.
- **Task Management:**
- Marked subtasks 61.38, 61.39, and 61.41 as complete.
This commit finalizes the alignment of `updateTasks`, `updateTaskById`, `expandTask`, `expandAllTasks`, `analyzeTaskComplexity`, `addTask`, and `parsePRD` with the unified AI service and configuration management patterns.
Refactors the `expandTask` and `expandAllTasks` features to complete subtask 61.38 and enhance functionality based on subtask 61.37's refactor.
Key Changes:
- **Additive Expansion (`expandTask`, `expandAllTasks`):**
- Modified `expandTask` default behavior to append newly generated subtasks to any existing ones.
- Added a `force` flag (passed down from CLI/MCP via `--force` option/parameter) to `expandTask` and `expandAllTasks`. When `force` is true, existing subtasks are cleared before generating new ones.
- Updated relevant CLI command (`expand`), MCP tool (`expand_task`, `expand_all`), and direct function wrappers (`expandTaskDirect`, `expandAllTasksDirect`) to handle and pass the `force` flag.
- **Complexity Report Integration (`expandTask`):**
- `expandTask` now reads `scripts/task-complexity-report.json`.
- If an analysis entry exists for the target task:
- `recommendedSubtasks` is used to determine the number of subtasks to generate (unless `--num` is explicitly provided).
- `expansionPrompt` is used as the primary prompt content for the AI.
- `reasoning` is appended to any additional context provided.
- If no report entry exists or the report is missing, it falls back to default subtask count (from config) and standard prompt generation.
- **`expandAllTasks` Orchestration:**
- Refactored `expandAllTasks` to primarily iterate through eligible tasks (pending/in-progress, considering `force` flag and existing subtasks) and call the updated `expandTask` function for each.
- Removed redundant logic (like complexity reading or explicit subtask clearing) now handled within `expandTask`.
- Ensures correct context (`session`, `mcpLog`) and flags (`useResearch`, `force`) are passed down.
- **Configuration & Cleanup:**
- Updated `.cursor/mcp.json` with new Perplexity/Anthropic API keys (old ones invalidated).
- Completed refactoring of `expandTask` started in 61.37, confirming usage of `generateTextService` and appropriate prompts.
- **Task Management:**
- Marked subtask 61.37 as complete.
- Updated `.changeset/cuddly-zebras-matter.md` to reflect user-facing changes.
These changes finalize the refactoring of the task expansion features, making them more robust, configurable via complexity analysis, and aligned with the unified AI service architecture.
Refactored the `expandTask` feature (`scripts/modules/task-manager/expand-task.js`) and related components (`commands.js`, `mcp-server/src/tools/expand-task.js`, `mcp-server/src/core/direct-functions/expand-task.js`) to integrate with the unified AI service layer (`ai-services-unified.js`) and configuration management (`config-manager.js`).
The refactor involved:
- Removing direct AI client calls and configuration fetching from `expand-task.js`.
- Attempting to use `generateObjectService` for structured subtask generation. This failed due to provider-specific errors (Perplexity internal errors, Anthropic schema translation issues).
- Reverting the core AI interaction to use `generateTextService`, asking the LLM to format its response as JSON containing a "subtasks" array.
- Re-implementing manual JSON parsing and Zod validation (`parseSubtasksFromText`) to handle the text response reliably.
- Updating prompt generation functions (`generateMainSystemPrompt`, `generateMainUserPrompt`, `generateResearchUserPrompt`) to request the correct JSON object structure within the text response.
- Ensuring the `expandTaskDirect` function handles pre-checks (force flag, task status) and correctly passes the `session` context and logger wrapper to the core `expandTask` function.
- Correcting duplicate imports in `commands.js`.
- Validating the refactored feature works correctly via both CLI (`task-master expand --id <id>`) and MCP (`expand_task` tool) for main and research roles.
This aligns the task expansion feature with the new architecture while using the more robust text generation approach due to current limitations with structured output services. Closes subtask 61.37.
Refactored the feature and related components (CLI command, MCP tool, direct function) to integrate with the unified AI service layer ().
Initially, was implemented to leverage structured output generation. However, this approach encountered persistent errors:
- Perplexity provider returned internal server errors.
- Anthropic provider failed with schema type and model errors.
Due to the unreliability of for this specific use case, the core AI interaction within was reverted to use . Basic manual JSON parsing and cleanup logic for the text response were reintroduced.
Key changes include:
- Removed direct AI client initialization (Anthropic, Perplexity).
- Removed direct fetching of AI model configuration parameters.
- Removed manual AI retry/fallback/streaming logic.
- Replaced direct AI calls with a call to .
- Updated wrapper to pass session context correctly.
- Updated MCP tool for correct path resolution and argument passing.
- Updated CLI command for correct path resolution.
- Preserved core functionality: task loading/filtering, report generation, CLI summary display.
Both the CLI command ([INFO] Initialized Perplexity client with OpenAI compatibility layer
[INFO] Initialized Perplexity client with OpenAI compatibility layer
Analyzing task complexity from: tasks/tasks.json
Output report will be saved to: scripts/task-complexity-report.json
Analyzing task complexity and generating expansion recommendations...
[INFO] Reading tasks from tasks/tasks.json...
[INFO] Found 62 total tasks in the task file.
[INFO] Skipping 31 tasks marked as done/cancelled/deferred. Analyzing 31 active tasks.
Skipping 31 tasks marked as done/cancelled/deferred. Analyzing 31 active tasks.
[INFO] Claude API attempt 1/2
[ERROR] Error in Claude API call: 400 {"type":"error","error":{"type":"invalid_request_error","message":"max_tokens: 100000 > 64000, which is the maximum allowed number of output tokens for claude-3-7-sonnet-20250219"}}
[ERROR] Non-overload Claude API error: 400 {"type":"error","error":{"type":"invalid_request_error","message":"max_tokens: 100000 > 64000, which is the maximum allowed number of output tokens for claude-3-7-sonnet-20250219"}}
Claude API error: 400 {"type":"error","error":{"type":"invalid_request_error","message":"max_tokens: 100000 > 64000, which is the maximum allowed number of output tokens for claude-3-7-sonnet-20250219"}}
[ERROR] Error during AI analysis: 400 {"type":"error","error":{"type":"invalid_request_error","message":"max_tokens: 100000 > 64000, which is the maximum allowed number of output tokens for claude-3-7-sonnet-20250219"}}
[ERROR] Error analyzing task complexity: 400 {"type":"error","error":{"type":"invalid_request_error","message":"max_tokens: 100000 > 64000, which is the maximum allowed number of output tokens for claude-3-7-sonnet-20250219"}}) and the MCP tool () have been verified to work correctly with this revised approach.
Resolves persistent 404 'Not Found' errors when calling Anthropic models via the Vercel AI SDK. The primary issue was likely related to incorrect or missing API headers.
- Refactors Anthropic provider (src/ai-providers/anthropic.js) to use the standard 'anthropic-version' header instead of potentially outdated/incorrect beta headers when creating the client instance.
- Updates the default fallback model ID in .taskmasterconfig to 'claude-3-5-sonnet-20241022'.
- Fixes the interactive model setup (task-master models --setup) in scripts/modules/commands.js to correctly filter and default the main model selection.
- Improves the cost display in the 'task-master models' command output to explicitly show 'Free' for models with zero cost.
- Updates description for the 'id' parameter in the 'set_task_status' MCP tool definition for clarity.
- Updates list of models and costs
- Unified Service: Introduced 'scripts/modules/ai-services-unified.js' to centralize AI interactions using provider modules ('src/ai-providers/') and the Vercel AI SDK.
- Provider Modules: Implemented 'anthropic.js' and 'perplexity.js' wrappers for Vercel SDK.
- 'updateSubtaskById' Fix: Refactored the AI call within 'updateSubtaskById' to use 'generateTextService' from the unified layer, resolving runtime errors related to parameter passing and streaming. This serves as the pattern for refactoring other AI calls in 'scripts/modules/task-manager/'.
- Task Status: Marked Subtask 61.19 as 'done'.
- Rules: Added new 'ai-services.mdc' rule.
This centralizes AI logic, replacing previous direct SDK calls and custom implementations. API keys are resolved via 'resolveEnvVariable' within the service layer. The refactoring of 'updateSubtaskById' establishes the standard approach for migrating other AI-dependent functions in the task manager module to use the unified service.
Relates to Task 61.
The interactive model setup triggered by `task-master models --setup` was previously attempting to call non-existent setter functions (`setMainModel`, etc.) in `config-manager.js`, leading to errors and preventing configuration updates.
This commit refactors the `--setup` logic within the `models` command handler in `scripts/modules/commands.js`. It now correctly:
- Loads the current configuration using `getConfig()`." -m "- Updates the appropriate sections of the loaded configuration object based on user selections from `inquirer`." -m "- Saves the modified configuration using the existing `writeConfig()` function from `config-manager.js`." -m "- Handles disabling the fallback model correctly."
This commit focuses on standardizing configuration and API key access patterns across key modules as part of subtask 61.34.
Key changes include:
- Refactored `ai-services.js` to remove global AI clients and use `resolveEnvVariable` for API key checks. Client instantiation now relies on `getAnthropicClient`/`getPerplexityClient` accepting a session object.
- Refactored `task-manager.js` (`analyzeTaskComplexity` function) to use the unified `generateTextService` from `ai-services-unified.js`, removing direct AI client calls.
- Replaced direct `process.env` access for model parameters and other configurations (`PERPLEXITY_MODEL`, `CONFIG.*`) in `task-manager.js` with calls to the appropriate getters from `config-manager.js` (e.g., `getResearchModelId(session)`, `getMainMaxTokens(session)`).
- Ensured `utils.js` (`resolveEnvVariable`) correctly handles potentially undefined session objects.
- Updated function signatures where necessary to propagate the `session` object for correct context-aware configuration/key retrieval.
This moves towards the goal of using `ai-client-factory.js` and `ai-services-unified.js` as the standard pattern for AI interactions and centralizing configuration management through `config-manager.js`.
