API Playground

Call the APIs. Right here.

Every ZenithAI capability is also an API. Paste your API key below and test five of them against the live endpoint, straight from this page: image to structured JSON, OCR on scanned documents, PDF text extraction, audio transcription and document summaries. Nothing you upload is retained — files are processed and discarded immediately.

Your API key

Verify your key to unlock the playground. Don’t have one? Request a demonstration and we’ll set you up.

Nothing is retainedFiles are processed in memory and discarded the moment your response is produced. No uploads, prompts or results are stored.

Your key stays with youOnce verified, the key is saved only in this browser for 24 hours so you can keep testing, and is sent only with the calls you run. Clear it any time with the Clear key button.

Demo size limitsImages up to 512 KB · PDFs up to 3 MB · audio up to 2 MB · documents up to 200 KB. These are demo limits, checked in your browser before anything is sent — the API itself accepts far larger files.

Fair-use rate limitUp to 5 API calls per minute per visitor. The counter above the demos shows what you have left; it resets on a rolling minute.

The playground unlocks after your API key is verified. Paste your key above and press Verify key.

5 of 5 calls left this minute

POST /generate

Send an image with a JSON Schema and get back structured data that conforms to it — extraction, description or classification, decided by your prompt and schema.

Image (PNG or JPG, up to 512 KB)

Choose or drop an image…

Click to browse, or drop a file here

Prompt

JSON schema for the reply

Model: Smart

The same call as curl

response

Structured JSON will appear here.

POST /ocr · GET /ocr/jobs/<id>

Recognises text in scanned, image-only documents. A PDF starts an OCR job which this page polls until it finishes; a photo or screenshot is read directly by the vision model via /generate.

Scanned PDF or image

Choose or drop a scan…

PDF up to 3 MB; images up to 512 KB

PDFs that already contain selectable text don’t need OCR — the PDF → Text tab reads those instantly, without a model.

The same call as curl

response

Recognised text will appear here.

POST /pdf

Deterministic extraction of text and tables from PDFs that have a text layer — statements, invoices, reports. No model involved, so it answers in milliseconds.

PDF (up to 3 MB)

Choose or drop a PDF…

Click to browse, or drop a file here

What to extract

The same call as curl

response

Extracted text and tables will appear here.

POST /chat

Attach an audio clip and ask for a transcript — or minutes, action items, a summary. Meeting recordings and voice notes become text your systems can use.

Audio clip (MP3 or WAV, up to 2 MB)

Choose or drop an audio file…

MP3 and WAV work best; convert voice notes (M4A/OPUS) to MP3 first

Instruction

The same call as curl

response

The transcript will appear here.

POST /chat

Attach a Word document, text file or PDF and instruct the model: summarise it, pull out decisions and dates, or answer a question about it. The text is extracted server-side.

Document (DOCX, TXT or PDF, up to 200 KB)

Choose or drop a document…

Click to browse, or drop a file here

Instruction

The same call as curl

response

The summary will appear here.

How every call works

One shape, five capabilities.

01Authenticate with a header — X-API-Key: YOUR_API_KEY — on every request.
02Files travel as base64 inside the JSON body: images in images, documents and audio in attachments.
03Fast calls answer in the response itself; heavy OCR returns a job_id you poll until it’s done.

The same key drives all endpoints, with per-key usage visible to your administrator. Model tiers — Instant, Fast, Smart, Genius — are selected per call with the "model" field.

example · image to structured JSON

curl -X POST https://llm.dil.in/generate \
  -H "Content-Type: application/json" \
  -H "X-API-Key: YOUR_API_KEY" \
  -d '{
    "model": "Smart",
    "prompt": "Describe this image.",
    "images": ["BASE64_IMAGE_DATA"],
    "format": {
      "type": "object",
      "additionalProperties": false,
      "required": ["description"],
      "properties": {"description": {"type": "string"}}
    }
  }'

Bring it inside your boundary

The same APIs, on your infrastructure.

Everything on this page runs identically on a ZenithAI deployment inside your own data centre — your documents, your models, your keys, with no token billing on your own deployment.

Request a Demonstration Explore the Platform