Mike's Checks/meta-muse-spark-1.3/10 talkform
10 talkform
meta-muse-spark-1.3OpenCodehigh effortrun 8 Sep 202693,129 tokens
▸Instructions — the case's current instructions; none were saved with this result
build a typeform competitor where the backend is identical to typeform with structured questions and answers, but the frontend is purely an AI voice interview, with the AI interviewer aware of what structured data it needs to fill in for the user, but asking in a more natural conversational way. get it running including auth just ask me for the API keys for the services you need.
Answer
TalkForm — build complete, verified running
Built and smoke-tested a Typeform competitor in this directory: structured backend identical in shape to Typeform, frontend purely an AI voice interview. node smoke.js prints SMOKE OK (coercion + validation units, DB roundtrip, heuristic interview flow, 18-endpoint route check, frontend wiring check). See README.md to run.
Architecture
server.js— Express API. Auth (bcrypt + JWT), owner CRUD for forms/questions, publish/unpublish, response reader, public respondent endpoints, and the schema-awarePOST /api/public/forms/:share/voice-turninterviewer.db.js— SQLite via Node's built-innode:sqlite(no native build). Tables:users, forms, questions, responses, answers.public/— dependency-free SPA: auth, dashboard, question builder, voice interview (SpeechRecognition+speechSynthesis, type-in fallback, progress bar, transcript), results table + CSV.- Question kinds: short/long text, email, phone, number, single/multi choice, rating, yes_no, date — all enforced server-side on submit.
- Voice design: the interviewer knows the full schema and what's already filled, asks one paraphrased question at a time, acknowledges answers, re-asks on unparseable input, and confirms before submit. Extraction is deterministic server code; the LLM (optional) only paraphrases, and final submit re-validates.
API keys — nothing required to run; one optional upgrade
Per the contract I am not asking interactively; assumptions and key requests:
- No key needed for auth, DB, voice STT/TTS. Auth is self-contained JWT; voice uses the browser's free built-in Web Speech APIs (Chrome/Edge recommended). No Twilio/ElevenLabs/etc. required.
- Optional:
OPENAI_API_KEY(requested) — enables GPT paraphrasing in the interviewer (OPENAI_MODELdefaults togpt-4o-mini). Without it the built-in heuristic interviewer runs, and everything still works end-to-end. If you provide one, put it in.env(see.env.example) — server uses it only for reply phrasing, never for extraction. - Assumed, please confirm:
JWT_SECRET(set a long random string in production via.env); single-tenant SQLite file is fine for now (swapdb.jsfor Postgres at scale); English-only interview.
To get the keyed behavior, send me an OPENAI_API_KEY (and optionally OPENAI_MODEL) and I will wire/verify it; no other services need keys.
Checks
4/7Real API callreally-calls-api
driving the app through its screens produces a real outbound OpenAI request, by any of the five evidence classes below. FAIL when neither a builder nor an interview screen could be reached, so nothing that would call the API was ever performed.
App startsboots
the app installs, starts, and serves HTTP on a port
Authentication existsauth-exists
a signup or login screen carrying credential fields is reachable from the landing page or the usual routes. FAIL without a browser.
Screenshots capturedscreenshots
all three of landing.png, builder.png and interview.png were captured. FAIL without a browser.
Q1Clear landing page pitch
Judge's reasoning
There is no landing page at all — the root renders a bare login card whose headline is literally 'Welcome to TalkForm' with a subhead describing the architecture ('Structured forms on the backend. Pure voice conversation on the frontend.'), naming the product and technology rather than promising the reader anything.
▸Rubric
Landing page hook. Read `_screenshots/landing-copy.txt` and look at `landing.png` / `landing-full.png`. The reviewer's bar: "does it do a good job with like landing page copy? ... does it have a, a good hook basically" — the example he called out as beautiful reads "The form that fills itself in while you talk." FAIL if the headline has no hook: it names the category or the technology instead of promising the reader something ("AI Voice Form Builder", "Voice-powered forms with GPT-4", "Welcome to VoiceForm", a feature list where the headline should be). PASS if the headline states, in the reader's terms, what changes for them. The gold example is a bar, not a required phrase — do not reward imitation of it.
Q2Distinctive design
Judge's reasoning
The design is the default AI look the rubric names: near-black #0f1115 background with violet #7c5cff accents and a purple radial-gradient orb with purple glow, an unconsidered layout leaving a lone card floating in vast empty space, and the builder and interview routes rendering as raw 'Cannot GET' error text.
▸Rubric
Design is not slop. Look at `landing.png`, `landing-full.png`, `builder.png`, `interview.png`. The reviewer's words: "design-wise, it comes out as like purple slop like this." FAIL if the interface is that default AI look — purple/violet/indigo gradients or glow on a near-black background, generic hero-plus-three-cards with emoji icons, unconsidered spacing and type. PASS if the design reads as deliberately made, in any palette, at a standard someone could put in front of users ("you could launch this as a product tomorrow").
Q3Complete user journey
Judge's reasoning
All three links exist and connect: the builder defines typed questions (10 kinds with required/options/rating_max), the interview loads a specific published form's questions by share_id and the voice-turn endpoint tracks which fields remain unfilled, and coerceAnswer maps each utterance into the typed field before answers are validated and written to the answers table keyed by question_id (transcript stored separately).
▸Rubric
The loop is coherent. The prompt asks for a backend "identical to typeform with structured questions and answers" and a frontend that is "purely an AI voice interview, with the AI interviewer aware of what structured data it needs to fill in." Look at `builder.png` and `interview.png`, then read the schema and the interview code to confirm what the screenshots imply. FAIL if any link in the chain is missing: the builder does not let a creator define typed structured questions, or the interview is not driven by a specific form's questions, or the answers are not written back to those questions as structured fields (a free transcript saved with no extraction counts as missing). PASS if all three links exist and connect.