Proofreadings

Grammarly-style writing assistance: correct spelling, grammar and punctuation, adjust tone, replace borrowed words with pure-language equivalents, rewrite or paraphrase text, suggest alternatives for a selection, and analyze writing quality.

POST/v1/proofreadings

Request

📋Request

{
  "model": "your-proofreading-model",
  "task": "proofread",
  "input": "Ths sentense have some erors.",
  "language": "en",
  "spell_check": true,
  "grammar": true,
  "explain": true
}

Parameters

model

stringrequired

Proofreading model id (type: proofreading).

input

stringrequired

The text to check. Max length depends on the model (Max Input Characters).

task

string

One of proofread, spelling, grammar, punctuation, suggestions, kurdish_pure, tone, rewrite, paraphrase, simplify, shorten, expand, alternatives, analyze, statistics. Defaults to the model's default task.

language

string

BCP 47 language tag, e.g. ckb (Sorani), kmr (Kurmanji), ar, en. Must be one of the model's supported languages.

spell_check

boolean

Fix spelling mistakes (proofread task).

grammar

boolean

Fix grammar mistakes (proofread task).

punctuation

boolean

Fix punctuation (proofread task).

suggestions

boolean

Include style / clarity / word-choice suggestions.

kurdish_pure

boolean

Pure language: replace borrowed words with native equivalents.

tone

string

Target tone: neutral, formal, informal, professional, friendly, casual, academic, confident, persuasive, diplomatic, empathetic, enthusiastic, simple, journalistic, creative. Required for task: "tone".

explain

boolean

Attach an explanation to each change.

merge_adjacent_changes

boolean

Merge neighbouring edits into a single change.

goal

string

Rewrite goal: clarity, fluency, concise, simplify, expand, paraphrase, formalize, polish.

length

string

shorter, same or longer (rewrite tasks).

n

integer

Number of rewrites / alternatives to return.

selection

{ start, end }

UTF-16 offsets of the span to generate alternatives for (alternatives task).

custom_dictionary

string[]

Words that must never be flagged (names, brands, jargon).

instructions

string

Extra free-form instructions (max 500 chars).

⚠

Model-specific support

Each model declares which properties, tasks, tones and languages it supports (see the model page or GET /v1/models → parameters). Sending an unsupported property or value returns 400 unsupported_parameter. Omitted properties fall back to the model's defaults.

Response

📋200 OK — proofread / rewrite tasks

{
  "id": "prf_abc123",
  "object": "proofreading",
  "created": 1790000000,
  "model": "your-proofreading-model",
  "task": "proofread",
  "choices": [
    {
      "index": 0,
      "text": "This sentence has some errors.",
      "changes": [
        {
          "id": "chg_1",
          "type": "spell-correction",
          "start_index": 0,
          "end_index": 3,
          "old_text": "Ths",
          "new_text": "This",
          "category": "spelling",
          "explanation": "Misspelled word."
        },
        {
          "id": "chg_2",
          "type": "replace",
          "start_index": 13,
          "end_index": 17,
          "old_text": "have",
          "new_text": "has",
          "category": "grammar",
          "explanation": "Subject-verb agreement."
        }
      ],
      "finish_reason": "stop"
    }
  ],
  "usage": { "input_tokens": 24, "output_tokens": 61, "total_tokens": 85 }
}

Each change has a type (insert, delete, replace, spell-correction), UTF-16 start_index/end_index into the original input, and an optional category (spelling, grammar, punctuation, style, clarity, tone, word-choice). The alternatives task returns selection + alternatives[], and statistics returns a statistics object.

📋200 OK — analyze task

{
  "id": "prf_def456",
  "object": "proofreading",
  "task": "analyze",
  "analysis": {
    "detected_language": "en",
    "tone": "informal",
    "tone_scores": [{ "tone": "informal", "score": 0.82 }],
    "formality": "informal",
    "readability": 71.4,
    "grade_level": "intermediate",
    "sentiment": "neutral",
    "quality_score": 64,
    "issue_counts": { "spelling": 2, "grammar": 1, "punctuation": 0, "style": 1 },
    "summary": "Casual message with a few spelling mistakes.",
    "statistics": { "characters": 34, "characters_no_spaces": 28, "words": 7, "sentences": 1, "paragraphs": 1, "average_words_per_sentence": 7, "reading_time_seconds": 2 }
  },
  "usage": { "input_tokens": 20, "output_tokens": 110, "total_tokens": 130 }
}

Examples

💻cURL — professional rewrite

curl https://api.pawan.krd/v1/proofreadings \
  -H "Authorization: Bearer pk-your_key" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "your-proofreading-model",
    "task": "rewrite",
    "input": "hey, can u send me the report asap",
    "tone": "professional",
    "goal": "polish",
    "n": 2
  }'

📒JavaScript — Sorani with pure language

const response = await fetch("https://api.pawan.krd/v1/proofreadings", {
  method: "POST",
  headers: {
    Authorization: "Bearer pk-your_key",
    "Content-Type": "application/json",
  },
  body: JSON.stringify({
    model: "your-proofreading-model",
    task: "proofread",
    input: "ئەم دەقە هەڵەی تێدایە",
    language: "ckb",
    kurdish_pure: true,
  }),
});

const result = await response.json();
for (const change of result.choices[0].changes) {
  console.log(change.category, change.old_text, "→", change.new_text);
}

🐍Python — alternatives for a selection

import requests

response = requests.post(
    "https://api.pawan.krd/v1/proofreadings",
    headers={"Authorization": "Bearer pk-your_key"},
    json={
        "model": "your-proofreading-model",
        "task": "alternatives",
        "input": "The meeting was very good.",
        "selection": {"start": 16, "end": 25},
        "n": 3,
    },
)

for alt in response.json()["alternatives"]:
    print(alt["text"], "-", alt["note"])

ℹ

Streaming

Proofreadings are returned as a single JSON response; stream: true is rejected.

Text Completions

Embeddings