Onboarding: add USER profile interview with LLM/template generation

CHANGELOG.md

@@ -36,6 +36,7 @@ Status: beta.
 - Docs: add LINE channel guide. Thanks @thewilloftheshadow.
 - Docs: credit both contributors for Control UI refresh. (#1852) Thanks @EnzeD.
 - Onboarding: add Venice API key to non-interactive flow. (#1893) Thanks @jonisjongithub.
+- Onboarding: add interactive USER profile interview with LLM-powered markdown generation and template fallback.
 - Onboarding: strengthen security warning copy for beta + access control expectations.
 - Tlon: format thread reply IDs as @ud. (#1837) Thanks @wca4a.
 - Gateway: prefer newest session metadata when combining stores. (#1823) Thanks @emanuelst.

src/commands/onboard-user-interview.ts

@@ -0,0 +1,234 @@
+import fs from "node:fs/promises";
+import os from "node:os";
+import path from "node:path";
+
+import { runEmbeddedPiAgent } from "../agents/pi-embedded.js";
+import { DEFAULT_USER_FILENAME } from "../agents/workspace.js";
+import { resolveConfiguredModelRef } from "../agents/model-selection.js";
+import { DEFAULT_MODEL, DEFAULT_PROVIDER } from "../agents/defaults.js";
+import type { MoltbotConfig } from "../config/config.js";
+import type { RuntimeEnv } from "../runtime.js";
+import { shortenHomePath } from "../utils.js";
+import type { WizardPrompter } from "../wizard/prompts.js";
+
+export type UserInterviewAnswers = {
+  name: string;
+  preferredName?: string;
+  interests?: string;
+  riskPreference?: "always-ask" | "low-risk-auto" | "trust-me";
+};
+
+export async function conductUserInterview(
+  prompter: WizardPrompter,
+): Promise<UserInterviewAnswers> {
+  const name = await prompter.text({
+    message: "What's your name?",
+    placeholder: "Alex",
+    validate: (value) => {
+      const trimmed = value.trim();
+      return !trimmed ? "Name is required" : undefined;
+    },
+  });
+
+  const preferredNameRaw = await prompter.text({
+    message: "What should I call you?",
+    placeholder: name,
+    initialValue: name,
+  });
+  const preferredName = preferredNameRaw.trim() || name;
+
+  const interestsRaw = await prompter.text({
+    message: "What are your main interests or focus areas?",
+    placeholder: "software development, AI, productivity tools",
+  });
+  const interests = interestsRaw.trim() || undefined;
+
+  await prompter.note(
+    [
+      "Risk preference helps me understand when to ask for approval.",
+      "Examples of non-reversible actions:",
+      "- Sending emails or messages",
+      "- Making payments or purchases",
+      "- Deleting files permanently",
+      "- Publishing content publicly",
+    ].join("\n"),
+    "Risk Preference",
+  );
+
+  const riskPreference = (await prompter.select({
+    message: "Should I ask before non-reversible actions?",
+    options: [
+      {
+        value: "always-ask",
+        label: "Always ask (safest)",
+        hint: "I'll confirm before any risky action",
+      },
+      {
+        value: "low-risk-auto",
+        label: "Auto-approve low-risk actions",
+        hint: "I'll handle routine tasks but ask for critical ones",
+      },
+      {
+        value: "trust-me",
+        label: "Trust me with everything",
+        hint: "I'll make decisions autonomously (requires strong oversight)",
+      },
+    ],
+    initialValue: "always-ask",
+  })) as "always-ask" | "low-risk-auto" | "trust-me";
+
+  return { name, preferredName, interests, riskPreference };
+}
+
+function generateUserMarkdownSimple(answers: UserInterviewAnswers): string {
+  const riskPreferenceText = {
+    "always-ask":
+      "Prefers maximum safety: always ask before any non-reversible action (sending messages, making payments, deleting files, publishing content).",
+    "low-risk-auto":
+      "Comfortable with routine automation: auto-approve low-risk tasks, but always ask before critical actions like payments, permanent deletions, or public publishing.",
+    "trust-me":
+      "Trusts autonomous decision-making: can proceed with most actions independently, but still values transparency and explanations.",
+  };
+
+  const riskContext = answers.riskPreference
+    ? riskPreferenceText[answers.riskPreference]
+    : "Risk preference not specified.";
+  const interestsNote = answers.interests
+    ? `Interested in ${answers.interests}.`
+    : "Interests to be discovered over time.";
+  const interestsContext = answers.interests
+    ? `${answers.name} is interested in ${answers.interests}. The assistant should be prepared to help with tasks, questions, and projects related to these areas.`
+    : `${answers.name}'s specific interests will become clearer through conversation. The assistant should actively learn and adapt to their needs.`;
+
+  return `# USER.md - About Your Human
+
+*Learn about the person you're helping. Update this as you go.*
+
+- **Name:** ${answers.name}
+- **What to call them:** ${answers.preferredName || answers.name}
+- **Notes:** ${interestsNote}
+
+## Context
+
+${interestsContext}
+
+**Risk Preference:** ${riskContext}
+
+The assistant should tailor its approach based on this information, always prioritizing ${answers.preferredName || answers.name}'s preferences and safety guidelines.
+
+---
+
+The more you know, the better you can help. But remember — you're learning about a person, not building a dossier. Respect the difference.`;
+}
+
+function buildUserProfilePrompt(answers: UserInterviewAnswers): string {
+  const riskPreferenceDescription = {
+    "always-ask": "Always ask for approval before any non-reversible action (safest)",
+    "low-risk-auto":
+      "Auto-approve low-risk routine tasks, but ask for critical actions like payments or deletions",
+    "trust-me": "Full autonomy - make decisions independently (requires strong oversight)",
+  };
+
+  const riskDesc = answers.riskPreference
+    ? riskPreferenceDescription[answers.riskPreference]
+    : undefined;
+  const templateContent = `# USER.md - About Your Human
+
+*Learn about the person you're helping. Update this as you go.*
+
+- **Name:** ${answers.name}
+- **What to call them:** ${answers.preferredName || answers.name}
+- **Notes:** [Brief one-line summary]
+
+## Context
+
+[2-3 paragraphs with interests and risk preference guidance]
+
+---
+
+The more you know, the better you can help. But remember — you're learning about a person, not building a dossier. Respect the difference.`;
+
+  return `You are helping to create a USER.md profile for a personal AI assistant.
+
+Based on the following interview answers, generate a well-structured markdown document.
+
+Interview Answers:
+- Name: ${answers.name}
+- Preferred Name: ${answers.preferredName || answers.name}
+${answers.interests ? `- Interests: ${answers.interests}` : ""}
+${riskDesc ? `- Risk Preference: ${riskDesc}` : ""}
+
+Requirements:
+1. Follow the USER.md template structure
+2. In "Notes" section: brief summary of interests if provided
+3. In "Context" section, write 2-3 paragraphs covering:
+   - Their interests and focus areas
+   - IMPORTANT: Their risk preference and when to ask for approval
+   - How the assistant should help them
+
+Template Structure:
+${templateContent}
+
+Output ONLY the markdown content. Start directly with "# USER.md - About Your Human".`;
+}
+
+export async function generateUserMarkdown(
+  answers: UserInterviewAnswers,
+  config: MoltbotConfig,
+  workspaceDir: string,
+): Promise<{ markdown: string; usedLLM: boolean }> {
+  let tempSessionFile: string | null = null;
+
+  try {
+    const prompt = buildUserProfilePrompt(answers);
+    const tempDir = await fs.mkdtemp(path.join(os.tmpdir(), "moltbot-user-profile-"));
+    tempSessionFile = path.join(tempDir, "session.jsonl");
+
+    const modelRef = resolveConfiguredModelRef({
+      cfg: config,
+      defaultProvider: DEFAULT_PROVIDER,
+      defaultModel: DEFAULT_MODEL,
+    });
+
+    const result = await runEmbeddedPiAgent({
+      sessionId: `onboard-user-interview-${Date.now()}`,
+      sessionKey: "temp:user-interview",
+      sessionFile: tempSessionFile,
+      workspaceDir,
+      config,
+      prompt,
+      provider: modelRef.provider,
+      model: modelRef.model,
+      timeoutMs: 30_000,
+      runId: `user-profile-gen-${Date.now()}`,
+    });
+
+    if (result.payloads && result.payloads.length > 0) {
+      const text = result.payloads[0]?.text;
+      if (text?.trim()) {
+        return { markdown: text.trim(), usedLLM: true };
+      }
+    }
+
+    return { markdown: generateUserMarkdownSimple(answers), usedLLM: false };
+  } catch (err) {
+    return { markdown: generateUserMarkdownSimple(answers), usedLLM: false };
+  } finally {
+    if (tempSessionFile) {
+      try {
+        await fs.rm(path.dirname(tempSessionFile), { recursive: true, force: true });
+      } catch {}
+    }
+  }
+}
+
+export async function saveUserProfile(
+  workspaceDir: string,
+  markdown: string,
+  runtime: RuntimeEnv,
+): Promise<string> {
+  const userPath = path.join(workspaceDir, DEFAULT_USER_FILENAME);
+  await fs.writeFile(userPath, markdown, "utf-8");
+  runtime.log(`✓ USER profile saved: ${shortenHomePath(userPath)}`);
+  return userPath;
+}

src/wizard/onboarding.ts

@@ -20,6 +20,11 @@ import {
 import { promptRemoteGatewayConfig } from "../commands/onboard-remote.js";
 import { setupSkills } from "../commands/onboard-skills.js";
 import { setupInternalHooks } from "../commands/onboard-hooks.js";
+import {
+  conductUserInterview,
+  generateUserMarkdown,
+  saveUserProfile,
+} from "../commands/onboard-user-interview.js";
 import type {
   GatewayAuthChoice,
   OnboardMode,
@@ -426,6 +431,52 @@ export async function runOnboardingWizard(
     skipBootstrap: Boolean(nextConfig.agents?.defaults?.skipBootstrap),
   });
 
+  // USER Profile Setup
+  const wantsUserProfile = await prompter.confirm({
+    message: "Set up your USER profile? (helps the assistant understand you better)",
+    initialValue: true,
+  });
+
+  if (wantsUserProfile) {
+    await prompter.note(
+      [
+        "Let's set up your profile so I can assist you better.",
+        "This will only take a minute.",
+      ].join("\n"),
+      "USER Profile Setup",
+    );
+
+    try {
+      // Conduct interview
+      const userAnswers = await conductUserInterview(prompter);
+
+      // Generate markdown (tries LLM, falls back to template)
+      const progress = prompter.progress("Generating profile...");
+      const { markdown: userMarkdown, usedLLM } = await generateUserMarkdown(
+        userAnswers,
+        nextConfig,
+        workspaceDir,
+      );
+
+      // Save to USER.md
+      await saveUserProfile(workspaceDir, userMarkdown, runtime);
+
+      if (usedLLM) {
+        progress.stop("✓ Profile created (AI-generated)");
+      } else {
+        progress.stop("✓ Profile created (template-based)");
+        await prompter.note(
+          "Profile created from template. AI generation unavailable (auth not configured yet).",
+          "Profile Setup",
+        );
+      }
+    } catch (err) {
+      // Interview was cancelled or failed
+      const errorMsg = err instanceof Error ? err.message : String(err);
+      runtime.error(`User interview failed: ${errorMsg}`);
+    }
+  }
+
   if (opts.skipSkills) {
     await prompter.note("Skipping skills setup.", "Skills");
   } else {