Troubleshooting

If something isn't working, scan this page — most issues fall into one of a handful of buckets.

"No API key configured" toast

What it means: The active provider doesn't have a key saved.

Fix: Go to Settings, paste a key into the active provider's card, and click Test. If you want to switch providers instead, click Set as active provider on the card with a key.

"Rate limit exceeded" / 429 errors

What it means: The provider is throttling your account.

Fixes:

• Wait 60 seconds and try again
• Switch to a different model on the same provider (limits are often per-model)
• Switch to a different provider — Groq has the most generous free limits

"Invalid API key"

What it means: The key is wrong, revoked, or for a different provider.

Fixes:

• Copy-paste the key again — invisible spaces at the start/end break validation
• Generate a fresh key in the provider console
• Make sure the key is in the right provider's card (Groq keys start with gsk_, OpenAI with sk-, Anthropic with sk-ant-)

"Model not found" / "You do not have access to this model"

What it means: The selected model isn't enabled on your account.

Fixes:

• Pick a different model from the dropdown
• Upgrade your usage tier with the provider (common for GPT-4 on new OpenAI accounts)
• Update lib/store.ts to only list models your account can access

Blank or truncated responses

What it means: Usually a rate-limit error that happened mid-stream in older versions. Newer ContentAI uses generateText which surfaces these as proper errors.

Fixes:

• Update to the latest version
• Clear site data and retry
• Open DevTools → Network → /api/generate — the response JSON will contain the real error

Dev server won't start

Error: Cannot find module 'next'

rm -rf node_modules .next
pnpm install

Port 3000 already in use

pnpm dev -- --port 3001

SyntaxError: Unexpected token '?'

Your Node version is too old. Install Node 20+:

nvm install 20 && nvm use 20

Build errors

Type error: Object literal may only specify known properties

The AI SDK updated its parameter names. Common renames:

• maxTokens → maxOutputTokens
• toDataStreamResponse() → toTextStreamResponse()

Pull the latest ContentAI or apply these renames manually.

Fumadocs / MDX build errors (docs site only)

If you're hacking on the docs site, delete .source and .next:

rm -rf .source .next
pnpm dev

Zustand migration warnings

If you upgraded ContentAI and see "unknown property 'activeModel'" warnings, the store schema changed. Either:

• Clear site data and re-enter your keys, or
• Ensure you're running the version with the migrate function in lib/store.ts (v1.0.1+)

AI output quality is poor

• Change models. Llama 3.3 70B, Claude Sonnet, and GPT-4o outperform cheaper models dramatically.
• Fill in every field. "Tone" and "Audience" have a huge effect.
• Regenerate 2–3 times and pick the best.
• Use the Editor's AI actions to polish output rather than regenerating from scratch.

Data is gone after switching browsers

The local-first build stores data in localStorage, which is per-browser and per-device. For cross-device sync and accounts, configure Authentication & database (NextAuth + Supabase). See also Add Supabase for other patterns.

Stuck on something not listed?

Check FAQ or reach out through your CodeCanyon support channel.