careervivid 1.12.48 → 1.12.50

package/dist/commands/interview.d.ts

@@ -0,0 +1,21 @@
+/**
+ * cv interview — interactive AI mock interview with real-time voice audio.
+ *
+ * Usage:
+ *   cv interview                          Prompt for role interactively (voice mode)
+ *   cv interview --role "Sr SWE"          Specify role directly (voice mode)
+ *   cv interview --role "PM" --text       Text-only fallback (no audio required)
+ *   cv interview --role "SDE" --resume <id>  Load resume for context
+ *
+ * Voice mode requires sox (handles both mic input and speaker output):
+ *   macOS:  brew install sox
+ *   Linux:  sudo apt install sox
+ *
+ * AI calls:
+ *   - Token vend: cliGetInterviewToken Cloud Function (validates cv_live_ key, deducts credits)
+ *   - Voice session: gemini-3.1-flash-live-preview via @google/genai Live API (direct WebSocket)
+ *   - Feedback: agentProxy Cloud Function (standard HTTP, existing pattern)
+ */
+import { Command } from "commander";
+export declare function registerInterviewCommand(program: Command): void;
+//# sourceMappingURL=interview.d.ts.map

package/dist/commands/interview.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"interview.d.ts","sourceRoot":"","sources":["../../src/commands/interview.ts"],"names":[],"mappings":"AAAA;;;;;;;;;;;;;;;;;GAiBG;AAEH,OAAO,EAAE,OAAO,EAAE,MAAM,WAAW,CAAC;AA4oBpC,wBAAgB,wBAAwB,CAAC,OAAO,EAAE,OAAO,GAAG,IAAI,CAmG/D"}

package/dist/commands/interview.js

@@ -0,0 +1,690 @@
+/**
+ * cv interview — interactive AI mock interview with real-time voice audio.
+ *
+ * Usage:
+ *   cv interview                          Prompt for role interactively (voice mode)
+ *   cv interview --role "Sr SWE"          Specify role directly (voice mode)
+ *   cv interview --role "PM" --text       Text-only fallback (no audio required)
+ *   cv interview --role "SDE" --resume <id>  Load resume for context
+ *
+ * Voice mode requires sox (handles both mic input and speaker output):
+ *   macOS:  brew install sox
+ *   Linux:  sudo apt install sox
+ *
+ * AI calls:
+ *   - Token vend: cliGetInterviewToken Cloud Function (validates cv_live_ key, deducts credits)
+ *   - Voice session: gemini-3.1-flash-live-preview via @google/genai Live API (direct WebSocket)
+ *   - Feedback: agentProxy Cloud Function (standard HTTP, existing pattern)
+ */
+import chalk from "chalk";
+import readline from "readline";
+import ora from "ora";
+import { spawn } from "child_process";
+import { GoogleGenAI, Modality } from "@google/genai";
+import { getApiKey } from "../config.js";
+import { isApiError, resumeGet } from "../api.js";
+/** Strip ANSI escape codes for accurate string length measurement */
+const stripAnsi = (s) => s.replace(/\x1B\[[0-9;]*m/g, "");
+// ─── Constants ────────────────────────────────────────────────────────────────
+const AGENT_PROXY_URL = process.env.CV_FUNCTIONS_URL
+    ? `${process.env.CV_FUNCTIONS_URL}/agentProxy`
+    : "https://us-west1-jastalk-firebase.cloudfunctions.net/agentProxy";
+const CLI_TOKEN_URL = process.env.CV_FUNCTIONS_URL
+    ? `${process.env.CV_FUNCTIONS_URL}/cliGetInterviewToken`
+    : "https://us-west1-jastalk-firebase.cloudfunctions.net/cliGetInterviewToken";
+const CLI_BILL_URL = process.env.CV_FUNCTIONS_URL
+    ? `${process.env.CV_FUNCTIONS_URL}/cliInterviewBill`
+    : "https://us-west1-jastalk-firebase.cloudfunctions.net/cliInterviewBill";
+const LIVE_MODEL = "gemini-3.1-flash-live-preview";
+const FEEDBACK_MODEL = "gemini-2.5-flash";
+const END_TOKEN = "<END_INTERVIEW>";
+const WRAP_WIDTH = 80;
+// Audio constants (matching tts.py)
+const SEND_SAMPLE_RATE = 16000; // mic → Gemini (16kHz PCM)
+const RECV_SAMPLE_RATE = 24000; // Gemini → speaker (24kHz PCM)
+const CHUNK_MS = 100; // send audio in 100ms chunks
+// ─── Helpers ──────────────────────────────────────────────────────────────────
+function wordWrap(text, width = WRAP_WIDTH) {
+    const lines = [];
+    for (const paragraph of text.split("\n")) {
+        const words = paragraph.split(" ");
+        let current = "";
+        for (const word of words) {
+            if (stripAnsi(current + " " + word).length > width && current.length > 0) {
+                lines.push(current);
+                current = word;
+            }
+            else {
+                current = current.length === 0 ? word : current + " " + word;
+            }
+        }
+        if (current.length > 0)
+            lines.push(current);
+        if (paragraph === "")
+            lines.push("");
+    }
+    return lines.join("\n");
+}
+function printAI(text) {
+    const clean = text.replace(END_TOKEN, "").trim();
+    if (!clean)
+        return;
+    console.log("");
+    console.log(chalk.cyan.bold("  Vivid ❯"));
+    wordWrap(clean).split("\n").forEach(l => console.log(`  ${chalk.cyan(l)}`));
+    console.log("");
+}
+function printUser(text) {
+    if (!text.trim())
+        return;
+    console.log(chalk.dim("\n  [You said] ") + chalk.white(text.trim()));
+}
+function printSystem(msg) {
+    console.log(chalk.dim(`\n  ${msg}\n`));
+}
+function printBanner(role, mode) {
+    const modeLabel = mode === "voice"
+        ? chalk.green("🎙  Voice Mode")
+        : chalk.yellow("⌨  Text Mode");
+    console.log("\n" + chalk.bold.bgHex("#4f46e5").white("  CareerVivid — Interview Studio  "));
+    console.log(chalk.dim(`  Role: ${role}`));
+    console.log(chalk.dim(`  ${modeLabel}`));
+    console.log(chalk.dim("  ─────────────────────────────────────────────────"));
+    if (mode === "voice") {
+        console.log(chalk.dim("  Speak your answers naturally."));
+        console.log(chalk.dim("  Press Ctrl+C to end and generate your feedback report."));
+    }
+    else {
+        console.log(chalk.dim(`  Type your answers. Type ${chalk.white("exit")} or press Ctrl+C to end.`));
+    }
+    console.log("");
+}
+function printReport(report) {
+    const header = chalk.bgHex("#4f46e5").white.bold;
+    console.log("\n" + header("  ═══════════════════════════════════════  "));
+    console.log(header("      📋  Interview Feedback Report         "));
+    console.log(header("  ═══════════════════════════════════════  ") + "\n");
+    const score = (val) => {
+        const color = val >= 80 ? chalk.green : val >= 60 ? chalk.yellow : chalk.red;
+        return color.bold(`${val}/100`);
+    };
+    console.log(chalk.bold("  Scores"));
+    console.log(`    Overall          ${score(report.overallScore)}`);
+    console.log(`    Communication    ${score(report.communicationScore)}`);
+    console.log(`    Confidence       ${score(report.confidenceScore)}`);
+    console.log(`    Relevance        ${score(report.relevanceScore)}`);
+    console.log("\n" + chalk.green.bold("  ✅ Strengths"));
+    wordWrap(report.strengths, 72).split("\n").forEach(l => console.log(`     ${chalk.green(l)}`));
+    console.log("\n" + chalk.yellow.bold("  💡 Areas for Improvement"));
+    wordWrap(report.areasForImprovement, 72).split("\n").forEach(l => console.log(`     ${chalk.yellow(l)}`));
+    console.log("\n" + chalk.dim("  ─────────────────────────────────────────────────────────────"));
+    console.log(chalk.dim("  View full history at: https://careervivid.app/interview"));
+    console.log("");
+}
+// ─── sox audio check ──────────────────────────────────────────────────────────
+/** Check if sox is available on PATH. Returns its path or null. */
+async function findSox() {
+    const candidates = [
+        "/opt/homebrew/bin/sox", // Apple Silicon brew
+        "/usr/local/bin/sox", // Intel brew
+        "/usr/bin/sox", // Linux
+        "sox", // if on $PATH
+    ];
+    for (const p of candidates) {
+        try {
+            await new Promise((resolve, reject) => {
+                const probe = spawn(p, ["--version"]);
+                probe.on("close", (code) => (code === 0 ? resolve() : reject()));
+                probe.on("error", reject);
+            });
+            return p;
+        }
+        catch { /* try next */ }
+    }
+    return null;
+}
+// ─── Sox audio I/O ────────────────────────────────────────────────────────────
+/**
+ * Start microphone recording via sox.
+ * Returns a ChildProcess whose stdout emits raw 16kHz/16-bit/mono PCM.
+ */
+function startMic(soxPath) {
+    // sox -t coreaudio default  → raw PCM 16kHz 16-bit signed mono
+    // Falls back to -t alsa on Linux
+    const inputType = process.platform === "darwin" ? "coreaudio" : "alsa";
+    const inputDevice = process.platform === "darwin" ? "default" : "default";
+    return spawn(soxPath, [
+        "-q",
+        "-t", inputType, inputDevice,
+        "-r", String(SEND_SAMPLE_RATE),
+        "-b", "16",
+        "-e", "signed",
+        "-c", "1",
+        "-t", "raw", "-",
+    ]);
+}
+/**
+ * Start a sox speaker subprocess.
+ * Returns a ChildProcess whose stdin accepts raw 24kHz/16-bit/mono PCM.
+ */
+function startSpeaker(soxPath) {
+    const outputType = process.platform === "darwin" ? "coreaudio" : "alsa";
+    const outputDevice = process.platform === "darwin" ? "default" : "default";
+    return spawn(soxPath, [
+        "-q",
+        "-t", "raw",
+        "-r", String(RECV_SAMPLE_RATE),
+        "-b", "16",
+        "-e", "signed",
+        "-c", "1", "-",
+        "-t", outputType, outputDevice,
+    ]);
+}
+// ─── Token Vend ───────────────────────────────────────────────────────────────
+async function getGeminiToken(role) {
+    const apiKey = getApiKey();
+    if (!apiKey)
+        throw new Error("No API key. Run: cv login");
+    const res = await fetch(CLI_TOKEN_URL, {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({ apiKey, role }),
+    });
+    const data = await res.json();
+    if (res.status === 402) {
+        throw new Error("AI credit limit reached. Upgrade at https://careervivid.app/pricing");
+    }
+    if (!res.ok) {
+        throw new Error(data?.error || `Token vend failed (${res.status})`);
+    }
+    return { geminiKey: data.geminiKey, sessionId: data.sessionId };
+}
+// ─── Duration Billing ─────────────────────────────────────────────────────────
+/** Call cliInterviewBill at session end. Returns credit summary for display. */
+async function billSession(sessionId) {
+    const apiKey = getApiKey();
+    if (!apiKey || !sessionId)
+        return null;
+    try {
+        const res = await fetch(CLI_BILL_URL, {
+            method: "POST",
+            headers: { "Content-Type": "application/json" },
+            body: JSON.stringify({ apiKey, sessionId }),
+        });
+        if (!res.ok)
+            return null;
+        return await res.json();
+    }
+    catch {
+        return null; // billing failure is non-fatal for UX
+    }
+}
+// ─── agentProxy (question gen + feedback) ────────────────────────────────────
+async function callAgentProxy(opts) {
+    const apiKey = getApiKey();
+    if (!apiKey)
+        throw new Error("No API key. Run: cv login");
+    const body = {
+        apiKey,
+        model: FEEDBACK_MODEL,
+        contents: opts.contents,
+    };
+    if (opts.systemInstruction)
+        body.systemInstruction = opts.systemInstruction;
+    if (opts.responseSchema) {
+        body.generationConfig = {
+            responseMimeType: opts.responseMimeType ?? "application/json",
+            responseSchema: opts.responseSchema,
+        };
+    }
+    const res = await fetch(AGENT_PROXY_URL, {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify(body),
+    });
+    const data = await res.json();
+    if (!res.ok)
+        throw new Error(data?.error || `agentProxy error (${res.status})`);
+    const parts = data?.candidates?.[0]?.content?.parts ?? [];
+    return parts.map((p) => p.text ?? "").join("").trim();
+}
+// ─── Build system prompt ──────────────────────────────────────────────────────
+function buildSystemPrompt(role, questions, resumeContext) {
+    let prompt = `You are an expert AI interviewer. Your name is Vivid. If the candidate asks your name, say "My name is Vivid." Do not say Gemini or any other name.
+
+You are conducting a real-time voice interview for the position of: "${role}".
+
+Start with a warm, polished introduction that outlines the role and key responsibilities. Then ask: "Do you have any questions before we begin the interview?" Wait briefly before proceeding.
+
+Ask the questions below one at a time. You may ask one or two follow-up questions when a candidate's answer invites it.
+
+After the final question and the candidate's response:
+1. Give a 2–3 sentence summary of overall performance.
+2. Provide 2–3 short, personalized improvement tips.
+3. End with: "Thank you for your time today! Your feedback report is being generated."
+4. Append the exact token ${END_TOKEN} at the very end (do not narrate this token).
+
+Interview Questions:
+${questions.map((q, i) => `${i + 1}. ${q}`).join("\n")}
+
+**Policies:**
+- Never fabricate company details; suggest candidates verify with their recruiter.
+- Maintain a polite, professional, encouraging tone.
+- Keep responses concise — this is a voice interview, so avoid long monologues.
+- Do not use markdown formatting. Speak naturally.`;
+    if (resumeContext) {
+        prompt += `\n\nCandidate resume (use for targeted follow-ups):\n--- RESUME ---\n${resumeContext}`;
+    }
+    return prompt;
+}
+// ─── Generate Questions (via agentProxy) ─────────────────────────────────────
+async function generateQuestions(role, numQuestions) {
+    const spinner = ora(chalk.dim("Generating interview questions...")).start();
+    try {
+        const prompt = `Based on the following role, generate ${numQuestions} insightful interview questions covering technical skills, behavioral competencies, and role-specific scenarios. Return ONLY a valid JSON array of strings.\n\nRole: "${role}"`;
+        const text = await callAgentProxy({
+            contents: [{ role: "user", parts: [{ text: prompt }] }],
+            responseSchema: { type: "ARRAY", items: { type: "STRING" } },
+            responseMimeType: "application/json",
+        });
+        let clean = text.trim().replace(/^```(?:json)?\n?/, "").replace(/\n?```$/, "");
+        const questions = JSON.parse(clean);
+        spinner.succeed(chalk.dim(`Generated ${questions.length} questions`));
+        return questions;
+    }
+    catch (err) {
+        spinner.fail("Failed to generate questions.");
+        throw err;
+    }
+}
+// ─── Analyze Transcript (via agentProxy) ─────────────────────────────────────
+async function analyzeTranscript(transcript, role) {
+    const spinner = ora(chalk.dim("Generating feedback report...")).start();
+    try {
+        const formatted = transcript
+            .map(e => `${e.speaker === "ai" ? "Interviewer" : "Candidate"}: ${e.text}`)
+            .join("\n\n");
+        const prompt = `You are an expert interview coach. Analyze this interview transcript for the role "${role}" and return a JSON object with: overallScore, communicationScore, confidenceScore, relevanceScore (numbers 0-100), strengths (string), areasForImprovement (string).\n\nTranscript:\n---\n${formatted}\n---`;
+        const text = await callAgentProxy({
+            contents: [{ role: "user", parts: [{ text: prompt }] }],
+            responseSchema: {
+                type: "OBJECT",
+                properties: {
+                    overallScore: { type: "NUMBER" },
+                    communicationScore: { type: "NUMBER" },
+                    confidenceScore: { type: "NUMBER" },
+                    relevanceScore: { type: "NUMBER" },
+                    strengths: { type: "STRING" },
+                    areasForImprovement: { type: "STRING" },
+                },
+                required: ["overallScore", "communicationScore", "confidenceScore", "relevanceScore", "strengths", "areasForImprovement"],
+            },
+        });
+        let clean = text.trim().replace(/^```(?:json)?\n?/, "").replace(/\n?```$/, "");
+        const report = JSON.parse(clean);
+        spinner.succeed(chalk.dim("Feedback report ready"));
+        return report;
+    }
+    catch (err) {
+        spinner.fail("Failed to generate feedback.");
+        throw err;
+    }
+}
+// ─── VOICE SESSION ────────────────────────────────────────────────────────────
+async function runVoiceSession(opts) {
+    const { role, questions, resumeContext, soxPath } = opts;
+    printBanner(role, "voice");
+    // Get Gemini token from Cloud Function
+    const connectSpinner = ora(chalk.dim("Connecting to Vivid...")).start();
+    let geminiKey;
+    let sessionId;
+    try {
+        ({ geminiKey, sessionId } = await getGeminiToken(role));
+    }
+    catch (err) {
+        connectSpinner.fail(chalk.red(err.message));
+        throw err;
+    }
+    const ai = new GoogleGenAI({ apiKey: geminiKey });
+    const systemInstruction = buildSystemPrompt(role, questions, resumeContext);
+    const transcript = [];
+    let ended = false;
+    let outputBuf = ""; // accumulates AI transcription
+    let inputBuf = ""; // accumulates user transcription
+    // Half-duplex mute: stop sending mic audio while Vivid is speaking
+    // to prevent the mic picking up speaker output (echo loop).
+    let vividSpeaking = false;
+    let muteTimer = null;
+    // ── Audio processes ──────────────────────────────────────────────────
+    const micProc = startMic(soxPath);
+    const speakerProc = startSpeaker(soxPath);
+    micProc.stderr.on("data", () => { });
+    speakerProc.stderr.on("data", () => { });
+    // ── Connect to Live API ──────────────────────────────────────────────
+    let session;
+    try {
+        session = await ai.live.connect({
+            model: LIVE_MODEL,
+            callbacks: {
+                onopen: () => {
+                    connectSpinner.succeed(chalk.green("✅ Vivid is live — start speaking!"));
+                    process.stdout.write(chalk.green("\n  ● Listening...\r"));
+                    // Pipe mic PCM → Gemini, muted while Vivid is speaking
+                    micProc.stdout.on("data", (chunk) => {
+                        if (ended || chunk.length === 0 || vividSpeaking)
+                            return;
+                        try {
+                            session.sendRealtimeInput({
+                                audio: {
+                                    data: chunk.toString("base64"),
+                                    mimeType: `audio/pcm;rate=${SEND_SAMPLE_RATE}`,
+                                },
+                            });
+                        }
+                        catch { /* session may be closing */ }
+                    });
+                },
+                onmessage: (msg) => {
+                    // ── Audio output (Vivid speaking) → sox speaker ───
+                    const audioPart = msg.serverContent?.modelTurn?.parts?.[0]?.inlineData?.data;
+                    if (audioPart) {
+                        // Mute mic: cancel any pending unmute and stay muted
+                        vividSpeaking = true;
+                        if (muteTimer) {
+                            clearTimeout(muteTimer);
+                            muteTimer = null;
+                        }
+                        process.stdout.write(chalk.blue("  ◈ Vivid speaking...\r"));
+                        const pcmBuf = Buffer.from(audioPart, "base64");
+                        speakerProc.stdin.write(pcmBuf);
+                    }
+                    // ── Output transcription (what Vivid said) ────────
+                    const outText = msg.serverContent?.outputTranscription?.text;
+                    if (outText)
+                        outputBuf += outText;
+                    // ── Input transcription (what user said) ──────────
+                    const inText = msg.serverContent?.inputTranscription?.text;
+                    if (inText)
+                        inputBuf += inText;
+                    // ── Turn complete ─────────────────────────────────
+                    if (msg.serverContent?.turnComplete) {
+                        if (outputBuf.trim()) {
+                            const aiText = outputBuf.trim();
+                            printAI(aiText);
+                            transcript.push({ speaker: "ai", text: aiText.replace(END_TOKEN, "").trim() });
+                            if (aiText.includes(END_TOKEN))
+                                ended = true;
+                            outputBuf = "";
+                        }
+                        if (inputBuf.trim()) {
+                            printUser(inputBuf.trim());
+                            transcript.push({ speaker: "user", text: inputBuf.trim() });
+                            inputBuf = "";
+                        }
+                        if (!ended) {
+                            // Unmute mic after a short delay so the speaker
+                            // tail doesn't get captured (echo suppression buffer)
+                            muteTimer = setTimeout(() => {
+                                vividSpeaking = false;
+                                muteTimer = null;
+                                process.stdout.write(chalk.green("  ● Listening...\r"));
+                            }, 800);
+                        }
+                    }
+                },
+                onerror: (e) => {
+                    console.log(chalk.red(`\n  Connection error: ${e.message ?? e}`));
+                    ended = true;
+                },
+                onclose: () => { ended = true; },
+            },
+            config: {
+                responseModalities: [Modality.AUDIO],
+                inputAudioTranscription: {},
+                outputAudioTranscription: {},
+                speechConfig: {
+                    voiceConfig: {
+                        prebuiltVoiceConfig: { voiceName: "Zephyr" },
+                    },
+                },
+                systemInstruction: { parts: [{ text: systemInstruction }] },
+                contextWindowCompression: {
+                    triggerTokens: "104857",
+                    slidingWindow: { targetTokens: "52428" },
+                },
+            },
+        });
+    }
+    catch (err) {
+        connectSpinner.fail(chalk.red(`Failed to connect: ${err.message}`));
+        micProc.kill();
+        speakerProc.kill();
+        throw err;
+    }
+    // ── Wait until interview ends or Ctrl+C ─────────────────────────────
+    await new Promise((resolve) => {
+        const check = setInterval(() => {
+            if (ended) {
+                clearInterval(check);
+                resolve();
+            }
+        }, 200);
+        process.once("SIGINT", () => {
+            clearInterval(check);
+            printSystem("Interview ended by user.");
+            ended = true;
+            resolve();
+        });
+    });
+    // ── Cleanup ──────────────────────────────────────────────────────────
+    try {
+        micProc.kill("SIGTERM");
+    }
+    catch { /* ignore */ }
+    try {
+        speakerProc.stdin.end();
+    }
+    catch { /* ignore */ }
+    try {
+        session.close();
+    }
+    catch { /* ignore */ }
+    // Wait a moment for final audio to drain
+    await new Promise(r => setTimeout(r, 800));
+    // ── Bill session (duration-based) ───────────────────────────────────
+    const billSpinner = ora(chalk.dim("Calculating session cost...")).start();
+    const bill = await billSession(sessionId);
+    if (bill) {
+        billSpinner.succeed(chalk.dim(`Session: ${bill.durationMinutes}min · `) +
+            chalk.hex("#4f46e5").bold(`${bill.creditsCharged} credits used`) +
+            chalk.dim(` · ${bill.creditsRemaining} remaining`));
+    }
+    else {
+        billSpinner.stop();
+    }
+    // ── Feedback report ──────────────────────────────────────────────────
+    const userTurns = transcript.filter(t => t.speaker === "user").length;
+    if (userTurns < 1) {
+        printSystem("Not enough conversation to generate a feedback report.");
+        return;
+    }
+    console.log(chalk.dim("\n  Generating your personalized feedback report..."));
+    try {
+        const report = await analyzeTranscript(transcript, role);
+        printReport(report);
+    }
+    catch (err) {
+        console.log(chalk.red(`  Failed to generate feedback: ${err.message}`));
+    }
+}
+// ─── TEXT SESSION (fallback) ──────────────────────────────────────────────────
+async function runTextSession(opts) {
+    const { role, questions, resumeContext } = opts;
+    const systemInstruction = buildSystemPrompt(role, questions, resumeContext);
+    printBanner(role, "text");
+    const history = [];
+    const transcript = [];
+    let ended = false;
+    const rl = readline.createInterface({ input: process.stdin, output: process.stdout, terminal: false });
+    const askUser = () => new Promise(resolve => {
+        if (process.stdin.isTTY)
+            process.stdout.write(chalk.white.bold("\n  you ❯ "));
+        rl.once("line", line => resolve(line.trim()));
+        rl.once("close", () => resolve(null));
+    });
+    const spinner = ora(chalk.dim("Vivid is connecting...")).start();
+    try {
+        const greeting = await callAgentProxy({
+            contents: [{ role: "user", parts: [{ text: "Hello" }] }],
+            systemInstruction,
+        });
+        spinner.stop();
+        history.push({ role: "user", parts: [{ text: "Hello" }] });
+        history.push({ role: "model", parts: [{ text: greeting }] });
+        transcript.push({ speaker: "user", text: "Hello" });
+        transcript.push({ speaker: "ai", text: greeting });
+        if (greeting.includes(END_TOKEN))
+            ended = true;
+        printAI(greeting);
+    }
+    catch (err) {
+        spinner.fail(chalk.red("Failed to connect to AI interviewer."));
+        throw err;
+    }
+    while (!ended) {
+        const input = await askUser();
+        if (input === null || input.toLowerCase() === "exit" || input.toLowerCase() === "q") {
+            printSystem("Interview ended early.");
+            break;
+        }
+        if (input === "")
+            continue;
+        history.push({ role: "user", parts: [{ text: input }] });
+        transcript.push({ speaker: "user", text: input });
+        const aiSpinner = ora({ text: "" }).start();
+        try {
+            const aiResponse = await callAgentProxy({ contents: history, systemInstruction });
+            aiSpinner.stop();
+            history.push({ role: "model", parts: [{ text: aiResponse }] });
+            transcript.push({ speaker: "ai", text: aiResponse.replace(END_TOKEN, "").trim() });
+            if (aiResponse.includes(END_TOKEN))
+                ended = true;
+            printAI(aiResponse);
+        }
+        catch (err) {
+            aiSpinner.stop();
+            console.log(chalk.red(`\n  Error: ${err.message}\n`));
+        }
+    }
+    rl.close();
+    const userTurns = transcript.filter(t => t.speaker === "user").length;
+    if (userTurns < 2) {
+        printSystem("Not enough conversation to generate a feedback report.");
+        return;
+    }
+    console.log(chalk.dim("\n  Generating your personalized feedback report..."));
+    try {
+        const report = await analyzeTranscript(transcript, role);
+        printReport(report);
+    }
+    catch (err) {
+        console.log(chalk.red(`  Failed to generate feedback: ${err.message}`));
+    }
+}
+// ─── Command Registration ─────────────────────────────────────────────────────
+export function registerInterviewCommand(program) {
+    program
+        .command("interview")
+        .description("Start an interactive AI voice interview session in the terminal")
+        .option("-r, --role <role>", "Role or job description to practice for")
+        .option("-q, --questions <n>", "Number of interview questions to generate", "5")
+        .option("--resume <id>", "Load a specific resume ID for context (from cv resumes list)")
+        .option("--text", "Use text-only mode (no audio required)")
+        .addHelpText("after", `
+Examples:
+  cv interview
+  cv interview --role "Senior Software Engineer at Stripe"
+  cv interview --role "Product Manager" --questions 7
+  cv interview --role "Data Scientist" --resume my-resume-id
+  cv interview --role "SWE" --text        (text-only, no sox needed)
+
+Voice mode setup (one-time):
+  macOS:  brew install sox
+  Linux:  sudo apt install sox
+`)
+        .action(async (opts) => {
+        if (!getApiKey()) {
+            console.error(chalk.red("\nNo API key configured.\n\n" +
+                "  Run: cv login     (browser login)\n" +
+                "       cv auth set-key <key>  (API key)\n"));
+            process.exit(1);
+        }
+        // ── Role prompt ──────────────────────────────────────────────────
+        let role = opts.role?.trim();
+        if (!role) {
+            const rl = readline.createInterface({ input: process.stdin, output: process.stdout });
+            role = await new Promise(resolve => {
+                rl.question(chalk.bold("\n  What role are you interviewing for?\n  ❯ "), answer => {
+                    rl.close();
+                    resolve(answer.trim());
+                });
+            });
+        }
+        if (!role) {
+            console.error(chalk.red("  Role is required."));
+            process.exit(1);
+        }
+        const numQuestions = Math.min(Math.max(parseInt(opts.questions, 10) || 5, 1), 12);
+        // ── Optional resume context ──────────────────────────────────────
+        let resumeContext;
+        if (opts.resume) {
+            const spinner = ora(chalk.dim("Loading resume...")).start();
+            try {
+                const result = await resumeGet(opts.resume);
+                if (isApiError(result)) {
+                    spinner.warn(chalk.yellow(`Could not load resume: ${result.message}. Continuing without it.`));
+                }
+                else {
+                    resumeContext = result.cvMarkdown;
+                    spinner.succeed(chalk.dim(`Resume loaded: ${result.title}`));
+                }
+            }
+            catch {
+                spinner.warn(chalk.yellow("Could not load resume. Continuing without it."));
+            }
+        }
+        // ── Generate questions ───────────────────────────────────────────
+        let questions;
+        try {
+            questions = await generateQuestions(role, numQuestions);
+        }
+        catch (err) {
+            console.error(chalk.red(`\n  Failed to generate questions: ${err.message}\n`));
+            process.exit(1);
+        }
+        // ── Determine mode ───────────────────────────────────────────────
+        if (opts.text) {
+            await runTextSession({ role, questions, resumeContext });
+            return;
+        }
+        // Probe for sox
+        const soxPath = await findSox();
+        if (!soxPath) {
+            console.log(chalk.yellow("\n  ⚠  sox not found — falling back to text mode.\n" +
+                "\n  To enable voice, install sox:\n" +
+                "    macOS:  brew install sox\n" +
+                "    Linux:  sudo apt install sox\n" +
+                "\n  Or run in text mode:  cv interview --text\n"));
+            await runTextSession({ role, questions, resumeContext });
+            return;
+        }
+        // Voice mode
+        try {
+            await runVoiceSession({ role, questions, resumeContext, soxPath });
+        }
+        catch (err) {
+            console.error(chalk.red(`\n  Interview error: ${err.message}\n`));
+            process.exit(1);
+        }
+    });
+}

package/dist/index.js

@@ -53,6 +53,7 @@ import { registerJobsCommand } from "./commands/jobs.js";
 import { registerResumesCommand } from "./commands/resumes.js";
 import { registerReferralCommand } from "./commands/referral.js";
 import { registerEvalCommand } from "./commands/eval.js";
+import { registerInterviewCommand } from "./commands/interview.js";
 const __dirname = dirname(fileURLToPath(import.meta.url));
 const pkg = JSON.parse(readFileSync(join(__dirname, "../package.json"), "utf-8"));
 const program = new Command();
@@ -77,6 +78,7 @@ registerJobsCommand(program);
 registerResumesCommand(program);
 registerReferralCommand(program);
 registerEvalCommand(program);
+registerInterviewCommand(program);
 // Shortcuts for whiteboard creation
 registerNewCommand(program);
 registerListTemplatesCommand(program);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "careervivid",
-  "version": "1.12.48",
+  "version": "1.12.50",
   "description": "Official CLI for CareerVivid — publish articles, diagrams, and portfolio updates from your terminal or AI agent",
   "type": "module",
   "bin": {