Skip to content

An AI-powered task-management system you can drop into Cursor, Lovable, Windsurf, Roo, and others.

fix(config): Improve config-manager.js for MCP server integration

Eyal Toledano205a11e82c25
diff --git a/.cursor/mcp.json b/.cursor/mcp.json
index e0ce8ae3..e322a13b 100644
--- a/.cursor/mcp.json
+++ b/.cursor/mcp.json
@@ -5,7 +5,7 @@
"args": ["./mcp-server/server.js"],
"env": {
"ANTHROPIC_API_KEY": "sk-ant-apikeyhere",
- "PERPLEXITY_API_KEY": "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",
+ "PERPLEXITY_API_KEY": "pplx-apikeyhere",
"OPENAI_API_KEY": "sk-proj-1234567890",
"GOOGLE_API_KEY": "AIzaSyB1234567890",
"GROK_API_KEY": "gsk_1234567890",
diff --git a/mcp-server/src/core/direct-functions/parse-prd.js b/mcp-server/src/core/direct-functions/parse-prd.js
index c3220962..47e6973c 100644
--- a/mcp-server/src/core/direct-functions/parse-prd.js
+++ b/mcp-server/src/core/direct-functions/parse-prd.js
@@ -12,41 +12,21 @@ import {
enableSilentMode,
disableSilentMode
} from '../../../../scripts/modules/utils.js';
-import {
- getAnthropicClientForMCP,
- getModelConfig
-} from '../utils/ai-client-utils.js';
/**
* Direct function wrapper for parsing PRD documents and generating tasks.
*
- * @param {Object} args - Command arguments containing input, numTasks or tasks, and output options.
+ * @param {Object} args - Command arguments containing projectRoot, input, output, numTasks options.
* @param {Object} log - Logger object.
* @param {Object} context - Context object containing session data.
* @returns {Promise<Object>} - Result object with success status and data/error information.
*/
export async function parsePRDDirect(args, log, context = {}) {
- const { session } = context; // Only extract session, not reportProgress
+ const { session } = context; // Only extract session
try {
log.info(`Parsing PRD document with args: ${JSON.stringify(args)}`);
- // Initialize AI client for PRD parsing
- let aiClient;
- try {
- aiClient = getAnthropicClientForMCP(session, log);
- } catch (error) {
- log.error(`Failed to initialize AI client: ${error.message}`);
- return {
- success: false,
- error: {
- code: 'AI_CLIENT_ERROR',
- message: `Cannot initialize AI client: ${error.message}`
- },
- fromCache: false
- };
- }
-
// Validate required parameters
if (!args.projectRoot) {
const errorMessage = 'Project root is required for parsePRDDirect';
@@ -57,7 +37,6 @@ export async function parsePRDDirect(args, log, context = {}) {
fromCache: false
};
}
-
if (!args.input) {
const errorMessage = 'Input file path is required for parsePRDDirect';
log.error(errorMessage);
@@ -67,7 +46,6 @@ export async function parsePRDDirect(args, log, context = {}) {
fromCache: false
};
}
-
if (!args.output) {
const errorMessage = 'Output file path is required for parsePRDDirect';
log.error(errorMessage);
@@ -137,58 +115,35 @@ export async function parsePRDDirect(args, log, context = {}) {
success: (message, ...args) => log.info(message, ...args) // Map success to info
};
- // Get model config from session
- const modelConfig = getModelConfig(session);
-
// Enable silent mode to prevent console logs from interfering with JSON response
enableSilentMode();
try {
- // Make sure the output directory exists
- const outputDir = path.dirname(outputPath);
- if (!fs.existsSync(outputDir)) {
- log.info(`Creating output directory: ${outputDir}`);
- fs.mkdirSync(outputDir, { recursive: true });
+ // Execute core parsePRD function - It now handles AI internally
+ const tasksDataResult = await parsePRD(inputPath, outputPath, numTasks, {
+ mcpLog: logWrapper,
+ session
+ });
+
+ // Check the result from the core function (assuming it might return data or null/undefined)
+ if (!tasksDataResult || !tasksDataResult.tasks) {
+ throw new Error(
+ 'Core parsePRD function did not return valid task data.'
+ );
}
- // Execute core parsePRD function with AI client
- await parsePRD(
- inputPath,
- outputPath,
- numTasks,
- {
- mcpLog: logWrapper,
- session
- },
- aiClient,
- modelConfig
+ log.info(
+ `Successfully parsed PRD and generated ${tasksDataResult.tasks?.length || 0} tasks`
);
- // Since parsePRD doesn't return a value but writes to a file, we'll read the result
- // to return it to the caller
- if (fs.existsSync(outputPath)) {
- const tasksData = JSON.parse(fs.readFileSync(outputPath, 'utf8'));
- log.info(
- `Successfully parsed PRD and generated ${tasksData.tasks?.length || 0} tasks`
- );
-
- return {
- success: true,
- data: {
- message: `Successfully generated ${tasksData.tasks?.length || 0} tasks from PRD`,
- taskCount: tasksData.tasks?.length || 0,
- outputPath
- },
- fromCache: false // This operation always modifies state and should never be cached
- };
- } else {
- const errorMessage = `Tasks file was not created at ${outputPath}`;
- log.error(errorMessage);
- return {
- success: false,
- error: { code: 'OUTPUT_FILE_NOT_CREATED', message: errorMessage },
- fromCache: false
- };
- }
+ return {
+ success: true,
+ data: {
+ message: `Successfully generated ${tasksDataResult.tasks?.length || 0} tasks from PRD`,
+ taskCount: tasksDataResult.tasks?.length || 0,
+ outputPath
+ },
+ fromCache: false // This operation always modifies state
+ };
} finally {
// Always restore normal logging
disableSilentMode();
@@ -201,7 +156,7 @@ export async function parsePRDDirect(args, log, context = {}) {
return {
success: false,
error: {
- code: 'PARSE_PRD_ERROR',
+ code: error.code || 'PARSE_PRD_ERROR', // Use error code if available
message: error.message || 'Unknown error parsing PRD'
},
fromCache: false
diff --git a/scripts/modules/ai-services-unified.js b/scripts/modules/ai-services-unified.js
index 0e1c88af..baf089fe 100644
--- a/scripts/modules/ai-services-unified.js
+++ b/scripts/modules/ai-services-unified.js
@@ -209,7 +209,7 @@ async function _unifiedServiceRunner(serviceType, params) {
let providerName, modelId, apiKey, roleParams, providerFnSet, providerApiFn;
try {
- log('info', `Attempting service call with role: ${currentRole}`);
+ log('info', `New AI service call with role: ${currentRole}`);
// --- Corrected Config Fetching ---
// 1. Get Config: Provider, Model, Parameters for the current role
diff --git a/scripts/modules/config-manager.js b/scripts/modules/config-manager.js
index f7b8a392..a3746702 100644
--- a/scripts/modules/config-manager.js
+++ b/scripts/modules/config-manager.js
@@ -73,7 +73,8 @@ const DEFAULTS = {
};
// --- Internal Config Loading ---
-let loadedConfig = null; // Cache for loaded config
+let loadedConfig = null;
+let loadedConfigRoot = null; // Track which root loaded the config
// Custom Error for configuration issues
class ConfigurationError extends Error {
@@ -84,25 +85,20 @@ class ConfigurationError extends Error {
}
function _loadAndValidateConfig(explicitRoot = null) {
- // Determine the root path to use
- const rootToUse = explicitRoot || findProjectRoot();
const defaults = DEFAULTS; // Use the defined defaults
- let configExists = false;
- if (!rootToUse) {
- console.warn(
- chalk.yellow(
- 'Warning: Could not determine project root. Using default configuration.'
- )
- );
- // Allow proceeding with defaults if root finding fails, but validation later might trigger error
- // Or perhaps throw here? Let's throw later based on file existence check.
+ // If no explicit root is provided (e.g., during initial server load),
+ // return defaults immediately and silently.
+ if (!explicitRoot) {
+ return defaults;
}
- const configPath = rootToUse ? path.join(rootToUse, CONFIG_FILE_NAME) : null;
- let config = defaults; // Start with defaults
+ // --- Proceed with loading from the provided explicitRoot ---
+ const configPath = path.join(explicitRoot, CONFIG_FILE_NAME);
+ let config = { ...defaults }; // Start with a deep copy of defaults
+ let configExists = false;
- if (configPath && fs.existsSync(configPath)) {
+ if (fs.existsSync(configPath)) {
configExists = true;
try {
const rawData = fs.readFileSync(configPath, 'utf-8');
@@ -125,59 +121,55 @@ function _loadAndValidateConfig(explicitRoot = null) {
global: { ...defaults.global, ...parsedConfig?.global }
};
- // --- Validation (Only warn if file exists but content is invalid) ---
- // Validate main provider/model
+ // --- Validation (Warn if file content is invalid) ---
+ // Only use console.warn here, as this part runs only when an explicitRoot *is* provided
if (!validateProvider(config.models.main.provider)) {
console.warn(
chalk.yellow(
- `Warning: Invalid main provider "${config.models.main.provider}" in ${CONFIG_FILE_NAME}. Falling back to default.`
+ `Warning: Invalid main provider "${config.models.main.provider}" in ${configPath}. Falling back to default.`
)
);
config.models.main = { ...defaults.models.main };
}
-
- // Validate research provider/model
if (!validateProvider(config.models.research.provider)) {
console.warn(
chalk.yellow(
- `Warning: Invalid research provider "${config.models.research.provider}" in ${CONFIG_FILE_NAME}. Falling back to default.`
+ `Warning: Invalid research provider "${config.models.research.provider}" in ${configPath}. Falling back to default.`
)
);
config.models.research = { ...defaults.models.research };
}
-
- // Validate fallback provider if it exists
if (
config.models.fallback?.provider &&
!validateProvider(config.models.fallback.provider)
) {
console.warn(
chalk.yellow(
- `Warning: Invalid fallback provider "${config.models.fallback.provider}" in ${CONFIG_FILE_NAME}. Fallback model configuration will be ignored.`
+ `Warning: Invalid fallback provider "${config.models.fallback.provider}" in ${configPath}. Fallback model configuration will be ignored.`
)
);
config.models.fallback.provider = undefined;
config.models.fallback.modelId = undefined;
}
} catch (error) {
+ // Use console.error for actual errors during parsing
console.error(
chalk.red(
`Error reading or parsing ${configPath}: ${error.message}. Using default configuration.`
)
);
- config = defaults; // Reset to defaults on parse error
- // Do not throw ConfigurationError here, allow fallback to defaults if file is corrupt
+ config = { ...defaults }; // Reset to defaults on parse error
}
} else {
- // Config file doesn't exist
- // **Changed: Log warning instead of throwing error**
+ // Config file doesn't exist at the provided explicitRoot.
+ // Use console.warn because an explicit root *was* given.
console.warn(
chalk.yellow(
- `Warning: ${CONFIG_FILE_NAME} not found at project root (${rootToUse || 'unknown'}). Using default configuration. Run 'task-master models --setup' to configure.`
+ `Warning: ${CONFIG_FILE_NAME} not found at provided project root (${explicitRoot}). Using default configuration. Run 'task-master models --setup' to configure.`
)
);
- // Return defaults instead of throwing error
- return defaults;
+ // Keep config as defaults
+ config = { ...defaults };
}
return config;
@@ -185,18 +177,32 @@ function _loadAndValidateConfig(explicitRoot = null) {
/**
* Gets the current configuration, loading it if necessary.
+ * Handles MCP initialization context gracefully.
* @param {string|null} explicitRoot - Optional explicit path to the project root.
* @param {boolean} forceReload - Force reloading the config file.
* @returns {object} The loaded configuration object.
*/
function getConfig(explicitRoot = null, forceReload = false) {
- if (!loadedConfig || forceReload) {
- loadedConfig = _loadAndValidateConfig(explicitRoot);
- }
- // If an explicitRoot was provided for a one-off check, don't cache it permanently
- if (explicitRoot && !forceReload) {
- return _loadAndValidateConfig(explicitRoot);
+ // Determine if a reload is necessary
+ const needsLoad =
+ !loadedConfig ||
+ forceReload ||
+ (explicitRoot && explicitRoot !== loadedConfigRoot);
+
+ if (needsLoad) {
+ const newConfig = _loadAndValidateConfig(explicitRoot); // _load handles null explicitRoot
+
+ // Only update the global cache if loading was forced or if an explicit root
+ // was provided (meaning we attempted to load a specific project's config).
+ // We avoid caching the initial default load triggered without an explicitRoot.
+ if (forceReload || explicitRoot) {
+ loadedConfig = newConfig;
+ loadedConfigRoot = explicitRoot; // Store the root used for this loaded config
+ }
+ return newConfig; // Return the newly loaded/default config
}
+
+ // If no load was needed, return the cached config
return loadedConfig;
}
diff --git a/scripts/modules/task-manager/models.js b/scripts/modules/task-manager/models.js
index 40c5c5bd..fb88ba9a 100644
--- a/scripts/modules/task-manager/models.js
+++ b/scripts/modules/task-manager/models.js
@@ -79,14 +79,14 @@ async function getModelConfiguration(options = {}) {
// Check API keys
const mainCliKeyOk = isApiKeySet(mainProvider);
- const mainMcpKeyOk = getMcpApiKeyStatus(mainProvider);
+ const mainMcpKeyOk = getMcpApiKeyStatus(mainProvider, projectRoot);
const researchCliKeyOk = isApiKeySet(researchProvider);
- const researchMcpKeyOk = getMcpApiKeyStatus(researchProvider);
+ const researchMcpKeyOk = getMcpApiKeyStatus(researchProvider, projectRoot);
const fallbackCliKeyOk = fallbackProvider
? isApiKeySet(fallbackProvider)
: true;
const fallbackMcpKeyOk = fallbackProvider
- ? getMcpApiKeyStatus(fallbackProvider)
+ ? getMcpApiKeyStatus(fallbackProvider, projectRoot)
: true;
// Get available models to find detailed info
diff --git a/tests/unit/ai-services-unified.test.js b/tests/unit/ai-services-unified.test.js
index 3d7a4351..ae0733cf 100644
--- a/tests/unit/ai-services-unified.test.js
+++ b/tests/unit/ai-services-unified.test.js
@@ -162,7 +162,7 @@ describe('Unified AI Services', () => {
expect(mockLog).toHaveBeenNthCalledWith(
2,
'info',
- 'Attempting service call with role: main'
+ 'New AI service call with role: main'
);
expect(mockLog).toHaveBeenNthCalledWith(
3,
@@ -229,7 +229,7 @@ describe('Unified AI Services', () => {
);
expect(mockLog).toHaveBeenCalledWith(
'info',
- 'Attempting service call with role: main'
+ 'New AI service call with role: main'
);
expect(mockLog).toHaveBeenCalledWith('info', 'Retrieved AI client', {
provider: mockClient.provider,
@@ -277,7 +277,7 @@ describe('Unified AI Services', () => {
// Check subsequent fallback attempts (which also fail)
expect(mockLog).toHaveBeenCalledWith(
'info',
- 'Attempting service call with role: fallback'
+ 'New AI service call with role: fallback'
);
expect(mockLog).toHaveBeenCalledWith(
'error',
@@ -285,7 +285,7 @@ describe('Unified AI Services', () => {
);
expect(mockLog).toHaveBeenCalledWith(
'info',
- 'Attempting service call with role: research'
+ 'New AI service call with role: research'
);
expect(mockLog).toHaveBeenCalledWith(
'error',
@@ -349,7 +349,7 @@ describe('Unified AI Services', () => {
);
expect(mockLog).toHaveBeenCalledWith(
'info',
- 'Attempting service call with role: fallback'
+ 'New AI service call with role: fallback'
);
expect(mockLog).toHaveBeenCalledWith(
'info',
@@ -431,7 +431,7 @@ describe('Unified AI Services', () => {
);
expect(mockLog).toHaveBeenCalledWith(
'info',
- 'Attempting service call with role: research'
+ 'New AI service call with role: research'
);
expect(mockLog).toHaveBeenCalledWith(
'info',
@@ -509,7 +509,7 @@ describe('Unified AI Services', () => {
);
expect(mockLog).toHaveBeenCalledWith(
'info',
- 'Attempting service call with role: research'
+ 'New AI service call with role: research'
);
expect(mockLog).toHaveBeenCalledWith(
'info',
@@ -554,7 +554,7 @@ describe('Unified AI Services', () => {
// Check logs for sequence
expect(mockLog).toHaveBeenCalledWith(
'info',
- 'Attempting service call with role: fallback'
+ 'New AI service call with role: fallback'
);
expect(mockLog).toHaveBeenCalledWith(
'error',
@@ -568,7 +568,7 @@ describe('Unified AI Services', () => {
);
expect(mockLog).toHaveBeenCalledWith(
'info',
- 'Attempting service call with role: research'
+ 'New AI service call with role: research'
);
expect(mockLog).toHaveBeenCalledWith(
'info',
@@ -613,7 +613,7 @@ describe('Unified AI Services', () => {
// Check logs for sequence
expect(mockLog).toHaveBeenCalledWith(
'info',
- 'Attempting service call with role: research'
+ 'New AI service call with role: research'
);
expect(mockLog).toHaveBeenCalledWith(
'error',
@@ -627,7 +627,7 @@ describe('Unified AI Services', () => {
);
expect(mockLog).toHaveBeenCalledWith(
'info',
- 'Attempting service call with role: fallback'
+ 'New AI service call with role: fallback'
);
expect(mockLog).toHaveBeenCalledWith(
'info',