@agentled/cli/dist/commands/do.js

CLI for Agentled — manage workflows, apps, and knowledge from the command line. Zero context-window cost for AI agents.

/**
 * `agentled do` — Semantic Intent Router
 *
 * Describe what you want in natural language. Agentled finds the best matching
 * workflow, extracts inputs from your intent, and optionally executes it.
 *
 * Examples:
 *   agentled do "find the CEO's email for stripe.com"
 *   agentled do "research acme corp and score them" --execute
 *   agentled do "scrape https://example.com" --execute --no-confirm
 */
import { AgentledClient } from '../client.js';
import { printOutput, printError } from '../utils/output.js';
export function registerDoCommand(program) {
    program
        .command('do <intent>')
        .description('Describe what you want — Agentled routes to the best matching workflow')
        .option('--execute', 'Auto-execute the matched workflow', false)
        .option('--no-confirm', 'Skip confirmation when executing with missing inputs')
        .option('--format <fmt>', 'Output format: json, table, minimal', 'json')
        .action(async (intent, opts) => {
        try {
            const client = new AgentledClient();
            if (!client.isAuthenticated) {
                printError('Not authenticated. Run "agentled auth login" or set AGENTLED_API_KEY.');
                return;
            }
            const result = await client.resolveIntent(intent, {
                execute: opts.execute,
                confirm: opts.confirm,
            });
            // Pretty-print based on action
            if (opts.format === 'json') {
                printOutput(result, 'json');
                return;
            }
            // Human-friendly output for table/minimal
            if (!result.match) {
                console.log(result.message || 'No matching workflow found.');
                return;
            }
            const m = result.match;
            console.log(`\nMatch: ${m.name} (${m.confidence}% confidence)`);
            console.log(`  Goal: ${m.goal}`);
            console.log(`  Reasoning: ${m.reasoning}`);
            if (m.extractedInputs && Object.keys(m.extractedInputs).length > 0) {
                console.log(`  Extracted inputs:`);
                for (const [k, v] of Object.entries(m.extractedInputs)) {
                    console.log(`    ${k}: ${v}`);
                }
            }
            if (m.missingInputs && m.missingInputs.length > 0) {
                console.log(`  Missing required inputs: ${m.missingInputs.map((f) => f.name).join(', ')}`);
            }
            if (result.alternatives?.length > 0) {
                console.log(`\nAlternatives:`);
                for (const alt of result.alternatives) {
                    console.log(`  - ${alt.name} (${alt.confidence}% confidence): ${alt.reasoning}`);
                }
            }
            if (result.action === 'executed') {
                console.log(`\nExecution started:`);
                console.log(`  Execution input ID: ${result.execution?.executionInputId}`);
                console.log(`  Workflow ID: ${result.execution?.pipelineId}`);
            }
            if (result.action === 'confirmation_needed') {
                console.log(`\nTo execute, provide the missing inputs:`);
                console.log(`  agentled workflows start ${m.workflowId} --input '${JSON.stringify(m.extractedInputs || {})}'`);
            }
            if (result.action === 'low_confidence') {
                console.log(`\nLow confidence match. To execute anyway:\n  agentled workflows start ${m.workflowId} --input '...'`);
            }
            console.log('');
        }
        catch (e) {
            printError(e.message);
        }
    });
}
//# sourceMappingURL=do.js.map