feat: 'Tell the AI more' free-text brief for AI names (v0.5.0)

Appended to the shared prompt (lib/ai-prompt.ts) on both AI paths; capped at
300 chars client- and server-side; shown only when AI is on; kept in
sessionStorage so it survives 'Show me more'.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XStQKxPfEjrTvWo83KFCxG

README.md

@@ -70,6 +70,13 @@ holds a per-app gateway token, never a real Anthropic key, and the route refuses
 anonymous requests and caps each account at roughly 20 requests an hour. The
 platform's monthly AI cap is the backstop.
 
+There's no chat: the AI is briefed from the same keywords box and Style chip
+as the built-in generator (`src/lib/ai-prompt.ts` is the rulebook), and its
+names are merged in front of the built-in list, tagged **AI**, and checked like
+everything else. When AI is on, a **Tell the AI more** field appears — tone,
+words to include or avoid, audience — which is appended to that brief (capped
+at 300 chars, enforced client- and server-side). The built-in generator ignores it.
+
 Sign-in is OTM SSO: `/sso/authorize?aud=app:names` on the control plane hands
 back a 60-second single-use ticket to `/api/auth/otm-sso`, which mints a small
 HMAC-signed cookie (`names.session`). There are no user accounts in this app.

package.json

@@ -1,6 +1,6 @@
 {
   "name": "names-poweredbyotm",
-  "version": "0.4.1",
+  "version": "0.5.0",
   "private": true,
   "scripts": {
     "dev": "next dev --port 3000",

src/app/api/ai-names/route.ts

@@ -13,7 +13,7 @@
 import { NextResponse } from "next/server";
 import type { NextRequest } from "next/server";
 import { readSession, platformAiEnabled } from "@/lib/session";
-import { SYSTEM_PROMPT, userPrompt, parseNames } from "@/lib/ai-prompt";
+import { SYSTEM_PROMPT, userPrompt, parseNames, BRIEF_MAX } from "@/lib/ai-prompt";
 
 export const runtime = "nodejs";
 export const dynamic = "force-dynamic";
@@ -47,6 +47,7 @@ interface Body {
   keywords?: unknown;
   style?: unknown;
   count?: unknown;
+  brief?: unknown;
 }
 
 export async function POST(req: NextRequest) {
@@ -79,6 +80,7 @@ export async function POST(req: NextRequest) {
     : [];
   const style = typeof body.style === "string" ? body.style.slice(0, 20) : "balanced";
   const count = Math.max(4, Math.min(MAX_COUNT, typeof body.count === "number" ? Math.floor(body.count) : 16));
+  const brief = typeof body.brief === "string" ? body.brief.slice(0, BRIEF_MAX) : "";
 
   const base = process.env.ANTHROPIC_BASE_URL!.replace(/\/+$/, "");
   let upstream: Response;
@@ -94,7 +96,7 @@ export async function POST(req: NextRequest) {
         model: MODEL,
         max_tokens: 1024,
         system: SYSTEM_PROMPT,
-        messages: [{ role: "user", content: userPrompt(keywords.length ? keywords : ["a new business"], style, count) }],
+        messages: [{ role: "user", content: userPrompt(keywords.length ? keywords : ["a new business"], style, count, brief) }],
       }),
       signal: AbortSignal.timeout(30_000),
     });

src/app/globals.css

@@ -398,6 +398,10 @@ h2 {
   text-decoration: underline;
 }
 
+.field.brief {
+  margin-top: 4px;
+}
+
 .signin {
   font-size: 14px;
   text-decoration: none;

src/components/Finder.tsx

@@ -10,6 +10,7 @@ import { DEFAULT_TLDS, TLDS, UNVERIFIABLE_TLDS } from "@/lib/tlds";
 import { DEFAULT_REGISTRAR, REGISTRARS } from "@/lib/registrars";
 import { generateWithAi, loadAiSettings, type AiSettings } from "@/lib/ai";
 import { generateWithPlatformAi } from "@/lib/platform-ai";
+import { BRIEF_MAX } from "@/lib/ai-prompt";
 import { BATCH, domainsFor, fetchAvailability, type DomainCheck } from "@/lib/availability-client";
 import { useOtmSession } from "@/hooks/useOtmSession";
 import { useWishlist } from "@/hooks/useWishlist";
@@ -43,6 +44,9 @@ export default function Finder() {
   const [ai, setAi] = useState<AiSettings | null>(null);
   const [useAi, setUseAi] = useState(false);
   const [usePlatformAi, setUsePlatformAi] = useState(true);
+  // Free-text guidance for the AI ("premium, no puns"). Kept in sessionStorage
+  // so it survives "Show me more" and a reload, but not forever.
+  const [brief, setBrief] = useState("");
   const { account, platformAi, signinNote, signOut } = useOtmSession();
   // How many names the last run got from AI vs the built-in generator — shown
   // as a one-line summary so it's obvious whether AI actually contributed
@@ -59,11 +63,21 @@ export default function Finder() {
     setAi(loadAiSettings());
     try {
       setHideTaken(window.localStorage.getItem("names.hideTaken") === "1");
+      setBrief(window.sessionStorage.getItem("names.brief") ?? "");
     } catch {
       /* private mode */
     }
   }, []);
 
+  const onBrief = (v: string) => {
+    setBrief(v);
+    try {
+      window.sessionStorage.setItem("names.brief", v);
+    } catch {
+      /* private mode */
+    }
+  };
+
   const toggleHideTaken = (v: boolean) => {
     setHideTaken(v);
     try {
@@ -137,8 +151,8 @@ export default function Finder() {
       if (wantsOwnKey || wantsPlatform) {
         try {
           const aiNames = wantsOwnKey
-            ? await generateWithAi(ai, parsedKeywords, style, 16)
+            ? await generateWithAi(ai, parsedKeywords, style, 16, brief)
-            : await generateWithPlatformAi(parsedKeywords, style, 16);
+            : await generateWithPlatformAi(parsedKeywords, style, 16, brief);
           if (runId.current !== myRun) return;
           const seen = new Set(list.map((i) => i.name));
           const extra: Idea[] = aiNames
@@ -159,7 +173,7 @@ export default function Finder() {
       setBusy(false);
       await lookup(domainsFor(list, tlds), myRun);
     },
-    [account, ai, lookup, parsedKeywords, platformAi, style, tlds, useAi, usePlatformAi],
+    [account, ai, brief, lookup, parsedKeywords, platformAi, style, tlds, useAi, usePlatformAi],
   );
 
   const onGenerate = () => {
@@ -173,6 +187,7 @@ export default function Finder() {
   };
 
   const showsUnverifiable = tlds.some((t) => UNVERIFIABLE_TLDS.includes(t));
+  const aiOn = ai ? useAi : Boolean(platformAi && account && usePlatformAi);
 
   /** A name is "all taken" once every domain we checked for it resolved taken.
    *  Still-checking or unknown names stay visible — hiding is for certainty. */
@@ -273,6 +288,27 @@ export default function Finder() {
         />
       </div>
 
+      {aiOn ? (
+        <div className="field brief">
+          <label htmlFor="brief">Tell the AI more (optional)</label>
+          <input
+            id="brief"
+            type="text"
+            value={brief}
+            maxLength={BRIEF_MAX}
+            placeholder="sounds premium, no puns, avoid the word auto, could work in Spanish too"
+            onChange={(e) => onBrief(e.target.value)}
+            onKeyDown={(e) => {
+              if (e.key === "Enter") onGenerate();
+            }}
+          />
+          <p className="hint">
+            Tone, words to include or avoid, who it&rsquo;s for &mdash; anything the keywords don&rsquo;t say.
+            Only the AI sees this; the built-in generator ignores it.
+          </p>
+        </div>
+      ) : null}
+
       {tlds.length === 0 ? (
         <p className="notice info">Pick at least one extension to check.</p>
       ) : null}

src/lib/ai-prompt.ts

@@ -17,12 +17,20 @@ export const SYSTEM_PROMPT = [
   'Respond with ONLY a JSON array of strings. No prose, no code fence.',
 ].join("\n");
 
-export function userPrompt(keywords: string[], style: string, count: number): string {
+/** Free-text guidance the visitor can add ("premium, no puns, avoid 'auto'").
-  return [
+ *  Hard cap so a pasted essay can't balloon the request — enforced here AND in
+ *  the server route, since the own-key path calls this straight from the browser. */
+export const BRIEF_MAX = 300;
+
+export function userPrompt(keywords: string[], style: string, count: number, brief = ""): string {
+  const lines = [
     `Business description / keywords: ${keywords.join(", ")}`,
     `Preferred style: ${style}`,
-    `Return exactly ${count} names as a JSON array of lowercase strings.`,
+  ];
-  ].join("\n");
+  const extra = brief.trim().slice(0, BRIEF_MAX);
+  if (extra) lines.push(`Extra guidance from the person naming the business: ${extra}`);
+  lines.push(`Return exactly ${count} names as a JSON array of lowercase strings.`);
+  return lines.join("\n");
 }
 
 /** Pull a JSON array of names out of a model response that may still have prose

src/lib/ai.ts

@@ -98,8 +98,9 @@ export async function generateWithAi(
   keywords: string[],
   style: string,
   count = 24,
+  brief = "",
 ): Promise<string[]> {
-  const prompt = userPrompt(keywords, style, count);
+  const prompt = userPrompt(keywords, style, count, brief);
   const text =
     settings.provider === "anthropic"
       ? await callAnthropic(settings.key, prompt)

src/lib/platform-ai.ts

@@ -1,11 +1,16 @@
 // Client helper for the platform-gateway AI path: POST /api/ai-names and get
 // names back. Requires the OTM sign-in cookie; the route does the gating.
 
-export async function generateWithPlatformAi(keywords: string[], style: string, count = 16): Promise<string[]> {
+export async function generateWithPlatformAi(
+  keywords: string[],
+  style: string,
+  count = 16,
+  brief = "",
+): Promise<string[]> {
   const res = await fetch("/api/ai-names", {
     method: "POST",
     headers: { "content-type": "application/json" },
-    body: JSON.stringify({ keywords, style, count }),
+    body: JSON.stringify({ keywords, style, count, brief }),
   });
   const body = (await res.json().catch(() => ({}))) as { names?: string[]; error?: string };
   if (!res.ok) throw new Error(body.error ?? `AI request failed (${res.status}).`);

src/lib/version.ts

@@ -2,7 +2,7 @@
 // /changelog. Bump VERSION and prepend an entry when shipping a user-facing
 // change. Same convention as rpo.
 
-export const VERSION = "0.4.1";
+export const VERSION = "0.5.0";
 
 export interface ChangelogEntry {
   version: string;
@@ -11,6 +11,13 @@ export interface ChangelogEntry {
 }
 
 export const CHANGELOG: ChangelogEntry[] = [
+  {
+    version: "0.5.0",
+    date: "2026-08-16",
+    changes: [
+      "Tell the AI more: when AI names are on, an extra line lets you steer them — tone, words to include or avoid, who it's for. It's added to the AI brief (the keywords and style still lead); the built-in generator ignores it.",
+    ],
+  },
   {
     version: "0.4.1",
     date: "2026-08-16",