English comment classification
NAWA’s API works for both Arabic and English comments. The same endpoints, same pricing, same response shape. The difference is under the hood: Arabic routes to ALLaM, English routes to Claude.How routing works
| Input language | Model | Dialect detection | Pricing |
|---|---|---|---|
| Arabic | HUMAIN ALLaM | Gulf, Egyptian, Levantine, MSA | Same |
| English | Claude | null (not applicable) | Same |
| Mixed | Claude | Detected if Arabic portion is sufficient | Same |
What changes in the response
For English comments, thedialect and dialect_confidence fields return null, and the language field returns "en". Everything else is identical.
Multilingual channels
If your channel has both Arabic and English comments, you can send all comments through the same/v1/classify endpoint. NAWA detects the language automatically and routes accordingly.
Reply generation
The/v1/comments/:id/reply endpoint also handles English. Arabic replies match the detected dialect. English replies are natural and platform-appropriate.
Set language: "auto" (default) to let NAWA match the commenter’s language, or force a language with language: "en" or language: "ar".
Pricing
Pricing is identical for Arabic and English. No per-language surcharge.| Endpoint | Cost |
|---|---|
| Classify | $0.006 |
| Reply | $0.008 |
| Detect | $0.002 |
| Moderate | $0.004 |