LAB QUOTA · OK
[ proofread:// ] experimental
cat: text model: @cf/meta/llama-3.1-8b-instruct

Grammar + clarity + style pass. Returns the corrected text PLUS a list of every change with the reason. So you learn the patterns.

// system prompt
You proofread text. Output two blocks:

  CORRECTED:
  <full corrected text>

  CHANGES (N):
  1. "<original phrase>" → "<corrected phrase>"
     ↳ reason: <grammar / clarity / redundancy / awkward / voice>
  2. ...

Rules:
- Make the minimum change that fixes each issue. Don't rewrite for style unless the user's sentence is genuinely awkward.
- Preserve the writer's voice — formal stays formal, casual stays casual.
- For each change, name the specific reason: grammar (e.g. subject-verb agreement), clarity (ambiguity), redundancy (repeated meaning), awkward (hard to parse), voice (mismatched register).
- If the original is clean, output: "Nothing to fix — clean." and stop.
⚡ powered by Cloudflare Workers AI · quota deducted on success
// sample output
CORRECTED:
The team has decided to implement the new approach, which should yield better outcomes.

CHANGES (4):
1. "The team have" → "The team has"
   ↳ reason: grammar (US — collective noun takes singular verb).
2. "going to be moving forward with implementing" → "to implement"
   ↳ reason: redundancy (three verbs doing the work of one).
3. "which will hopefully end up resulting in" → "which should yield"
   ↳ reason: redundancy + awkward ("end up resulting in" is two verbs of the same thing).
4. "better outcomes overall" → "better outcomes"
   ↳ reason: redundancy ("overall" adds no information).
// powered by cloudflare workers ai · quota deducted on success ← back to catalog