Great Work Labs
Recipes

Small models, chained

Most jobs take two to four models. Each recipe below lists the steps, what it costs per item at today's prices, the code, and a prompt you can hand to your coding agent.

Triage a support inbox

$0.0001 per ticket

Every incoming ticket gets a language, a queue, a mood and a PII-free copy you can send to your LLM.

Lingo lingo-1 detect the language to pick the right team
Veil veil-1 redact emails, phones and card numbers before anything else sees the text
Sort sort-1 route to billing, bug report, feature request or account access
Tone tone-1 flag angry customers for a faster reply
const run = (model, body) => labs("/v1/run", { model, ...body });
const lang  = await run("lingo-1", { input: ticket.body });
const clean = await run("veil-1",  { input: ticket.body, redact: "numbered" });
const queue = await run("sort-1",  { input: clean.output.redacted, labels: ["billing", "bug report", "feature request", "account access"] });
const tone  = await run("tone-1",  { input: ticket.body });
// send clean.output.redacted (not ticket.body) to your LLM
Prompt for your coding agent
Build "triage a support inbox" into this project with Great Work Labs (https://labs.greatwork.company).
Steps: lingo-1 (detect the language to pick the right team); veil-1 (redact emails, phones and card numbers before anything else sees the text); sort-1 (route to billing, bug report, feature request or account access); tone-1 (flag angry customers for a faster reply).
Read https://labs.greatwork.company/llms-full.txt first, and each model's page (https://labs.greatwork.company/models/<id>.md) before using its options. API key from GWL_API_KEY. Batch arrays where possible; handle 402 (out of credit), 429 and 503 (retry with backoff).

Answer from your docs without making things up

$0.000425 per question (20 candidates, 3-sentence answer)

Screen retrieved chunks for injected instructions, rerank them, and check the final answer against its sources.

Atlas atlas-1 embed your documents once and the question each time
Ladder ladder-1 rerank the top 20 candidates
Moat moat-1 drop chunks that contain instructions aimed at your model
Ground ground-1 flag sentences in the answer the sources don't support
const q = await run("atlas-1", { input: question, kind: "query" });
const top = vectorSearch(q.output.embedding, 20);               // your index
const ranked = await run("ladder-1", { input: question, documents: top.map(d => d.text), top_n: 5 });
const chunks = ranked.output.results.map(r => top[r.index].text);
const safe = (await run("moat-1", { input: chunks, threshold: 0.9 })).output;
const context = chunks.filter((_, i) => !safe[i].injection).join("\n\n");
const answer = await yourLLM(question, context);
const check = await run("ground-1", { input: answer, context });   // check.output.sentences[i].supported
Prompt for your coding agent
Build "answer from your docs without making things up" into this project with Great Work Labs (https://labs.greatwork.company).
Steps: atlas-1 (embed your documents once and the question each time); ladder-1 (rerank the top 20 candidates); moat-1 (drop chunks that contain instructions aimed at your model); ground-1 (flag sentences in the answer the sources don't support).
Read https://labs.greatwork.company/llms-full.txt first, and each model's page (https://labs.greatwork.company/models/<id>.md) before using its options. API key from GWL_API_KEY. Batch arrays where possible; handle 402 (out of credit), 429 and 503 (retry with backoff).

Meeting notes with who said what

$0.032 per 30-minute meeting (in 5-minute parts)

A recording becomes a speaker-labeled transcript, a summary and a title.

Scribe scribe-1 transcribe with speakers: true
Brief brief-1 summarize the transcript
Headline headline-1 title it for your notes app
const t = await run("scribe-1", { input: audioUrl, speakers: true });
const lines = t.output.segments.map(s => `${s.speaker}: ${s.text}`).join("\n");
const summary = await run("brief-1", { input: lines, length: "long" });
const title = await run("headline-1", { input: summary.output.summary });
Prompt for your coding agent
Build "meeting notes with who said what" into this project with Great Work Labs (https://labs.greatwork.company).
Steps: scribe-1 (transcribe with `speakers: true`); brief-1 (summarize the transcript); headline-1 (title it for your notes app).
Read https://labs.greatwork.company/llms-full.txt first, and each model's page (https://labs.greatwork.company/models/<id>.md) before using its options. API key from GWL_API_KEY. Batch arrays where possible; handle 402 (out of credit), 429 and 503 (retry with backoff).

Moderate a community

$0.00015 per post (comment + one image)

Comments, DMs and uploads each get the check that fits them, so a person only reviews what's flagged.

Sieve sieve-1 insults, threats and slurs in comments
Warden warden-1 self-harm, weapons, drugs and other policy categories
Bait bait-1 scam and spam DMs
Curtain curtain-1 nudity in uploaded images
const [tox, policy] = await Promise.all([
  run("sieve-1",  { input: post.text }),
  run("warden-1", { input: post.text }),
]);
const img = post.imageUrl ? await run("curtain-1", { input: post.imageUrl }) : null;
const hold = tox.output.flagged || policy.output.flagged || img?.output.nsfw;
Prompt for your coding agent
Build "moderate a community" into this project with Great Work Labs (https://labs.greatwork.company).
Steps: sieve-1 (insults, threats and slurs in comments); warden-1 (self-harm, weapons, drugs and other policy categories); bait-1 (scam and spam DMs); curtain-1 (nudity in uploaded images).
Read https://labs.greatwork.company/llms-full.txt first, and each model's page (https://labs.greatwork.company/models/<id>.md) before using its options. API key from GWL_API_KEY. Batch arrays where possible; handle 402 (out of credit), 429 and 503 (retry with backoff).

Turn receipt photos into expenses

$0.0009 per receipt

Read the text, then fill your expense schema. Ask Glance when you only need one number.

Ink ink-1 read every line of the receipt
Mold mold-1 fill merchant, date, total, tax and currency
Glance glance-1 or ask a single question about the image
const text = await run("ink-1", { input: photoUrl });
const expense = await run("mold-1", { input: text.output.text, schema: {
  type: "object", properties: {
    merchant: { type: "string" }, date: { type: "string", description: "YYYY-MM-DD" },
    total: { type: "number" }, tax: { type: "number" }, currency: { type: "string" } } } });
Prompt for your coding agent
Build "turn receipt photos into expenses" into this project with Great Work Labs (https://labs.greatwork.company).
Steps: ink-1 (read every line of the receipt); mold-1 (fill merchant, date, total, tax and currency); glance-1 (or ask a single question about the image).
Read https://labs.greatwork.company/llms-full.txt first, and each model's page (https://labs.greatwork.company/models/<id>.md) before using its options. API key from GWL_API_KEY. Batch arrays where possible; handle 402 (out of credit), 429 and 503 (retry with backoff).

Voice commands for your app

$0.00047 per 10-second command

Speech becomes a function call your app can run, with arguments that match your schema.

Scribe scribe-1 transcribe the voice note
Cue cue-1 pick the function and fill its arguments
const said = await run("scribe-1", { input: audioDataUrl, timestamps: "none" });
const call = await run("cue-1", { input: said.output.text, tools: myTools });
if (call.output.tool) await handlers[call.output.tool](call.output.arguments);
Prompt for your coding agent
Build "voice commands for your app" into this project with Great Work Labs (https://labs.greatwork.company).
Steps: scribe-1 (transcribe the voice note); cue-1 (pick the function and fill its arguments).
Read https://labs.greatwork.company/llms-full.txt first, and each model's page (https://labs.greatwork.company/models/<id>.md) before using its options. API key from GWL_API_KEY. Batch arrays where possible; handle 402 (out of credit), 429 and 503 (retry with backoff).

Send each prompt to the cheapest model that can do it

$0.00005 per prompt

Check the prompt is safe, then pick a small or large LLM by task and complexity.

Warden warden-1 refuse unsafe prompts before they cost anything
Route route-1 task type and complexity score
const [safety, route] = await Promise.all([run("warden-1", { input: prompt }), run("route-1", { input: prompt })]);
if (safety.output.flagged) return refuse();
const model = route.output.task === "code_generation" || route.output.complexity > 0.35 ? BIG_MODEL : SMALL_MODEL;
Prompt for your coding agent
Build "send each prompt to the cheapest model that can do it" into this project with Great Work Labs (https://labs.greatwork.company).
Steps: warden-1 (refuse unsafe prompts before they cost anything); route-1 (task type and complexity score).
Read https://labs.greatwork.company/llms-full.txt first, and each model's page (https://labs.greatwork.company/models/<id>.md) before using its options. API key from GWL_API_KEY. Batch arrays where possible; handle 402 (out of credit), 429 and 503 (retry with backoff).

Enrich a product catalog

$0.00363 per product

Every product photo gets a clean cutout, alt text, tags and search keywords.

Cutout cutout-1 transparent PNG of the product
Glance glance-1 alt text
Prism prism-1 your own style tags
Pluck pluck-1 keywords from the description
const [cut, alt, tags, kw] = await Promise.all([
  run("cutout-1", { input: photoUrl }),
  run("glance-1", { input: photoUrl }),
  run("prism-1",  { input: photoUrl, labels: ["minimal", "rustic", "sporty", "luxury", "something else"] }),
  run("pluck-1",  { input: product.description, top_k: 6 }),
]);
Prompt for your coding agent
Build "enrich a product catalog" into this project with Great Work Labs (https://labs.greatwork.company).
Steps: cutout-1 (transparent PNG of the product); glance-1 (alt text); prism-1 (your own style tags); pluck-1 (keywords from the description).
Read https://labs.greatwork.company/llms-full.txt first, and each model's page (https://labs.greatwork.company/models/<id>.md) before using its options. API key from GWL_API_KEY. Batch arrays where possible; handle 402 (out of credit), 429 and 503 (retry with backoff).

Route voicemails

$0.00116 per 1-minute voicemail

Know the language and the reason for every voicemail before anyone listens to it.

Ear ear-1 spoken language from the first 30 seconds
Scribe scribe-1 transcript in that language
Sort sort-1 reason for the call
const lang = await run("ear-1", { input: voicemailUrl });
const text = await run("scribe-1", { input: voicemailUrl, language: lang.output.language, timestamps: "none" });
const why = await run("sort-1", { input: text.output.text, labels: ["new order", "complaint", "billing question", "appointment"] });
Prompt for your coding agent
Build "route voicemails" into this project with Great Work Labs (https://labs.greatwork.company).
Steps: ear-1 (spoken language from the first 30 seconds); scribe-1 (transcript in that language); sort-1 (reason for the call).
Read https://labs.greatwork.company/llms-full.txt first, and each model's page (https://labs.greatwork.company/models/<id>.md) before using its options. API key from GWL_API_KEY. Batch arrays where possible; handle 402 (out of credit), 429 and 503 (retry with backoff).