Browse the docsReference

Parameters

Every tuning parameter is optional. Send one and your value is used after validation; omit it, or send null, and the deployment default applies. Omission never means “send nothing” - it means “use the default”.

Parameter	Default	Accepted range	Forwarded	Notes
max_tokensinteger	40000	1 - 40000	yes	The cap equals the default, so `40000` is both the fallback and the maximum. Asking for more is a `422`, not a silent clamp.
temperaturenumber	0	0 - 2	yes	Leave it at `0` for extraction. Higher values make the model paraphrase fields instead of copying them.
top_pnumber	-	-	ignored	Accepted so OpenAI SDKs do not break. Has no effect.
reasoning_effortstring	-	-	ignored	An OpenAI reasoning-tier concept that does not apply to this model.
streamboolean	false	-	ignored	Responses are always returned whole. `true` is treated as `false`.

Unknown fields

Any field we do not recognise is dropped at the edge rather than rejected, so a newer OpenAI SDK will not break against this API - and arbitrary client input never reaches the model.

import os
from openai import OpenAI

API_KEY  = "Your API Key"
BASE_URL = "https://labs-api.petpooja.com/v1"
MODEL    = "chitragupta"

client = OpenAI(api_key=API_KEY, base_url=BASE_URL)

r = client.chat.completions.create(
    model=MODEL,
    messages=[{"role": "user", "content": "Extract this invoice as JSON."}],
  temperature=0,       # 0..2   (omit -> default 0)
  max_tokens=200,      # 1..40000 (omit -> default 40000)
)

Limits

Rate limits apply per API key. Everything else is a per-request structural limit, enforced at the edge before any inference is billed.

Limit	OpenWorld	Default
Pages + images per request	10	25
Model calls per request	1all pages together	1 per pagerun concurrently
Your text instructions	passed text	ignored
Output structure	Based on your prompt	Fixed schema, tagged per page
Max tokens	40,000	40,000
Requests per second	10	10
Requests running at the same timehow many of your calls we process in parallel	20	20

Compress scans to JPEG before encoding - it is the single cheapest latency win available to you.

Error codes

Every failure returns the same envelope: a stable machine-readable code, a human-safe message, and the request_id to quote in a support ticket. Branch on code, never on the message text.

HTTP	Code	What went wrong	How to fix it
401	invalid_api_key	Missing, malformed, or revoked key, or an inactive account.	Fix or replace the key.
402	insufficient_credits	The account is out of credits.	Top up, then resend.
403	permission_denied	This key is not allowed to perform the action.	Check the key's scope and workspace.
404	model_not_found	Unknown model name in the request.	Use `chitragupta`.
405	method_not_allowed	Wrong HTTP method for this route.	All completion routes are `POST`.
409	idempotency_conflict	An `Idempotency-Key` was reused with a different body.	Use a new key, or resend the exact original body.
413	file_too_large	The uploaded file exceeds the size limit.	Compress the file before encoding.
413	page_limit_exceeded	Too many pages and images combined.	Split into smaller requests.
422	invalid_request	Malformed body, or `max_tokens` above the cap.	Fix the payload.
422	system_message_not_allowed	A `system` message was sent; the model uses a fixed prompt.	Move the instruction into the `user` message.
422	images_not_allowed	An image was sent to a text-only model.	Use a vision-capable model.
429	rate_limit_exceeded	Too many requests per second, or too many in flight on this key.	Honor the `Retry-After` header.
429	quota_exceeded	The key or account hit its usage quota.	Wait for the reset or raise the quota.
500	internal_error	Unexpected server-side failure.	Retry with exponential backoff.
502	inference_failed	The model returned an invalid response.	Retry once; if it persists, log the `request_id`.
503	provider_unavailable	Inference capacity is temporarily unavailable.	Retry with exponential backoff.
504	inference_timeout	The model took too long.	Retry with backoff, or send fewer pages.

# See the request_id even on an error:
try:
    r = client.chat.completions.create(model=MODEL, messages=[{"role": "user", "content": "hi"}])
    print(r.id.replace("chatcmpl-", ""))
except Exception as e:
    print(type(e).__name__, getattr(e, "status_code", "?"), getattr(e, "body", None))