microsoft/TypeScript · error · Error

Request: ${request.command} not allowed in LanguageServiceMo

Error message

Request: ${request.command} not allowed in LanguageServiceMode.PartialSemantic

What it means

Thrown by Session (session.ts:1084) for commands listed in invalidPartialSemanticModeCommands when tsserver is started in LanguageServiceMode.PartialSemantic. In that mode the server runs as a lightweight syntax+partial-semantic server and refuses heavy semantic operations; the throw is wired into the handler map at session construction so the request never reaches the language service.

Source

Thrown at src/server/session.ts:1084

            allowLocalPluginLoads: opts.allowLocalPluginLoads,
            typesMapLocation: opts.typesMapLocation,
            serverMode: opts.serverMode,
            session: this,
            canUseWatchEvents: opts.canUseWatchEvents,
            incrementalVerifier: opts.incrementalVerifier,
        };
        this.projectService = new ProjectService(settings);
        this.projectService.setPerformanceEventHandler(this.performanceEventHandler.bind(this));
        this.gcTimer = new GcTimer(this.host, /*delay*/ 7000, this.logger);

        // Make sure to setup handlers to throw error for not allowed commands on syntax server
        switch (this.projectService.serverMode) {
            case LanguageServiceMode.Semantic:
                break;
            case LanguageServiceMode.PartialSemantic:
                invalidPartialSemanticModeCommands.forEach(commandName =>
                    this.handlers.set(commandName, request => {
                        throw new Error(`Request: ${request.command} not allowed in LanguageServiceMode.PartialSemantic`);
                    })
                );
                break;
            case LanguageServiceMode.Syntactic:
                invalidSyntacticModeCommands.forEach(commandName =>
                    this.handlers.set(commandName, request => {
                        throw new Error(`Request: ${request.command} not allowed in LanguageServiceMode.Syntactic`);
                    })
                );
                break;
            default:
                Debug.assertNever(this.projectService.serverMode);
        }
    }

    private sendRequestCompletedEvent(requestId: number, performanceData: PerformanceData | undefined): void {
        this.event<protocol.RequestCompletedEventBody>(
            {

View on GitHub (pinned to b465fdbfe1)

Solutions

1. Start tsserver without `--serverMode partialSemantic` (default Semantic mode) when the client needs full language-service features.
2. On the client, gate blocked commands on the negotiated serverMode and degrade gracefully (e.g. skip semantic diagnostics) when running partial-semantic.
3. If partial mode was chosen for memory, raise `maxTsServerMemory`/resources and run full Semantic instead.
4. Cross-check the requested command against invalidPartialSemanticModeCommands (session.ts:908) before sending.

Example fix

// before — server started in partial mode but client sends semantic cmd
spawn('tsserver', ['--serverMode', 'partialSemantic']);
client.send({ command: 'semanticDiagnosticsSync', arguments: { file: 'a.ts' } });
// throws: Request: semanticDiagnosticsSync not allowed in LanguageServiceMode.PartialSemantic

// after — use full semantic mode
spawn('tsserver', []); // default Semantic
client.send({ command: 'semanticDiagnosticsSync', arguments: { file: 'a.ts' } });

// Block-listed commands in PartialSemantic mode (mirror session.ts:908).
const invalidPartialSemanticModeCommands: ReadonlySet<string> = new Set([
  'openExternalProject','openExternalProjects','closeExternalProject','synchronizeProjectList',
  'emitOutput','compileOnSaveAffectedFileList','compileOnSaveEmitFile','compilerOptionsDiagnosticsFull',
  'encodedSemanticClassificationsFull','semanticDiagnosticsSync','suggestionDiagnosticsSync',
  'geterrForProject','reload','reloadProjects','getCodefixes','getCodeFixesFull','getCombinedCodeFix',
  'getCombinedCodeFixFull','applyCodeActionCommand','getSupportedCodeFixes','getApplicableRefactors',
  'getMoveToRefactoringFileSuggestions','getEditsForRefactor','getEditsForRefactorFull',
  'organizeImports','organizeImportsFull','getEditsForFileRename','getEditsForFileRenameFull',
  'prepareCallHierarchy','provideCallHierarchyIncomingCalls','provideCallHierarchyOutgoingCalls',
  'getPasteEdits','copilotRelated',
]);
function isAllowedInPartialSemantic(command: string): boolean {
  return !invalidPartialSemanticModeCommands.has(command);
}

function isPartialSemanticAllowed(command: string, mode: string): boolean {
  if (mode !== 'partialSemantic') return true;
  return !invalidPartialSemanticModeCommands.has(command);
}

When it happens

Trigger: tsserver is launched with `--serverMode partialSemantic` (or the equivalent configuration) and a client sends one of the blocked commands — e.g. SemanticDiagnosticsSync, GetCodeFixes, OrganizeImports, GetEditsForRefactor, ProvideCallHierarchyIncomingCalls, Reload, EmitOutput (see the list at session.ts:908-942).

Common situations: An editor/IDE starts tsserver in partial-semantic mode to save memory and then issues a semantic request (go-to-def refactor, semantic diagnostics); a CI tool sends the same request payload regardless of server mode; mismatch between client expectations and the configured serverMode.

Related errors

• Request: ${request.command} not allowed in LanguageServiceMo
• No Project.
• The project's language service is disabled.
• Project '${project.getProjectName()}' does not contain docum
• Invalid name for template cancellation pipe: it should have

AI-assisted analysis of microsoft/TypeScript@b465fdbfe1 (2026-08-12). Data as JSON: /api/errors/22850bf29f1541d8. Report an issue: GitHub.