How do you build Arabic AI apps with Plugsky?

Key facts

Arabic script	Right-to-left UTF-8 text handled as-is; mixed-direction spans need explicit markers
Tokenisation	Clitics and diacritics can split into extra subwords — measure with the Plugsky token calculator
Embeddings	plugsky-embed-multilingual for Arabic and cross-language RAG
API compatibility	OpenAI-compatible POST https://api.plugsky.com/v1/chat/completions — chat, streaming, JSON mode and function calling
Models	30+ models behind one endpoint, from free tiers to frontier reasoning
Free tier	Free plan with plugsky-micro and plugsky-lite, no card required
Deployment	Plugsky cloud, your VPC, on-prem and air-gapped options
Product status	Chat, streaming, JSON mode, function calling, embeddings, RAG and agents live; audio, images, batch and fine-tuning coming soon

TL;DR

• Change one base_url and send Arabic text unchanged — no separate endpoint.
• Count tokens on real Arabic prompts; clitics and tashkeel inflate counts.
• Pin the Arabic variety (MSA or dialect) in the system prompt.
• Use plugsky-embed-multilingual for Arabic and cross-language retrieval.
• Start free with plugsky-micro and plugsky-lite, then scale as quality demands.

How it works, step by step

1. Create a Plugsky API key on the free plan (no card) and set base_url to https://api.plugsky.com/v1.
2. Send a small set of real Arabic prompts to /v1/chat/completions and compare output across two or three models.
3. Count tokens for those prompts with the Plugsky token calculator and set chunk sizes that fit your model context.
4. Normalise text before indexing or prompting: unify alef variants, remove tatweel and bidi controls, and keep a display copy untouched.
5. For RAG, embed with plugsky-embed-multilingual and test cross-language queries alongside Arabic-only queries.
6. Score candidate models on a Arabic gold set with native-speaker review, then cut production traffic over.

Original data

Try it yourself

Open the RAG sandbox →

How Plugsky handles Arabic text

Arabic is written right-to-left in a cursive Arabic script, with optional diacritics (tashkeel) that readers normally infer from context.

Decide early between Modern Standard Arabic and a dialect such as Gulf, Egyptian or Levantine, then state the expected variety in the system prompt, because the same request produces different tone and vocabulary in each.

Evaluate with a small gold set written by Arabic speakers: check grammar, register and direction handling in mixed Arabic and Latin answers, because automated metrics miss those errors.

Tokenisation and cost in Arabic

Token counts for Arabic rarely match word counts. Clitics such as و, ال and ب, plus suffixes, frequently split into several subword tokens, and fully vocalised text inflates counts further.

• Count tokens on your real corpus before sizing context windows.
• Drop tashkeel unless it changes meaning; it adds tokens and readers infer it.
• Unify alef variants and digits for retrieval, but keep the display copy untouched.
• Budget extra for mixed Arabic and Latin text: code, URLs and product names break the script run.

Arabic retrieval and RAG

Cross-language retrieval is the main RAG decision: index Arabic documents with plugsky-embed-multilingual and query in Arabic or English against the same collection.

• Use plugsky-embed-multilingual for Arabic and mixed-language corpora.
• Strip bidi control characters and tatweel before indexing; both pollute vectors.
• Store the original text for display and a normalised copy for retrieval.
• Test English-query/Arabic-passage retrieval as a first-class eval.

Code example: a Arabic request

Point your existing OpenAI client at https://api.plugsky.com/v1 and pass Arabic text in the content field — no language flag and no separate endpoint. Streaming, JSON mode and function calling keep the same request shapes.

client = OpenAI(base_url="https://api.plugsky.com/v1", api_key=os.environ["PLUGSKY_API_KEY"])

client.chat.completions.create(model="plugsky-pro", messages=[{"role": "user", "content": "اكتب ملخّصاً من ثلاث نقاط لهذا العقد بالعربية الفصحى."}])

Start on the free plan with plugsky-micro and plugsky-lite, then compare paid models on a Arabic gold set before cutover. See the docs for request details.

For production, log the model name and your normalisation settings with each request, and re-run the Arabic gold set whenever either changes — language quality regressions usually come from prompt or preprocessing drift, not from the model alone.

Honest comparison

Capability	Plugsky	Arabic workflow today	Building in-house
API compatibility	OpenAI-compatible — change base_url and model name	Varies by provider and SDK	Full rewrite
Arabic text handling	UTF-8 RTL passes through; you control normalisation and bidi hygiene	Depends on provider tokeniser and prompt hygiene	You build normalisation, segmentation and evals
Token budget	Fixed tokeniser per model; measure with the Plugsky token calculator and chunk to fit	Varies by provider and model	You host and tune each tokeniser
Multilingual retrieval	plugsky-embed-multilingual available for cross-language RAG	Often needs a separate embedding vendor	You serve and maintain embeddings
Sovereignty	Cloud, VPC, on-prem and air-gapped with residency options	Usually US/EU public endpoints	You own the full stack

Frequently asked questions

Can Plugsky handle Arabic text?

Yes. The API accepts UTF-8 Arabic input on the OpenAI-compatible chat endpoint; output quality depends on the model, so compare two or three on your own prompts before choosing.

How do I estimate token usage for Arabic?

Token counts for Arabic rarely match word counts. Use the token calculator at /tools/llm-token-calculator before sizing context windows or chunk lengths.

Should I use MSA or a dialect?

Pick one per feature and state it in the system prompt. MSA is the safer default for formal content; Gulf or Egyptian dialect fits conversational products.

Does the API handle right-to-left rendering?

The API transports Arabic as UTF-8 text; rendering is your UI's responsibility. Wrap code, URLs and numbers in bidi isolation characters and test in your own front end.

Is there a multilingual embedding model?

Yes — plugsky-embed-multilingual is part of the 30+ model catalogue and is built for cross-language retrieval. Keep one embedding model per vector collection.

Can I keep data in my region?

Plugsky supports cloud, VPC, on-prem and air-gapped deployment with data-residency options; confirm your requirements with the docs and the enterprise team.

How do I migrate an existing app?

Change base_url to https://api.plugsky.com/v1 and map the model name. Streaming, JSON mode, function calling and embeddings keep the same request shapes.

Is there a free plan?

Yes — the free plan includes two free models, plugsky-micro and plugsky-lite, with no card. A 14-day full-access trial unlocks the paid catalogue.