Browse the docsDefault

Image

POSThttps://labs-api.petpooja.com/v1/extract/chat/completions

Identical wire format to OpenWorld - so the OpenAI SDK works unchanged, just with base_url set to …/v1/extract. The difference is server-side: your text and system messages are discarded and the fixed extraction prompt runs against each image on its own.

Text is ignored

Sending "text": "only give me the invoice number" here does nothing - it is dropped before inference. If you need instruction-guided output, use OpenWorld.

import base64, mimetypes
from openai import OpenAI

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

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

IMAGE_PATH = "<IMAGE_PATH>"

mime = mimetypes.guess_type(IMAGE_PATH)[0] or "image/jpeg"
img_b64 = base64.b64encode(open(IMAGE_PATH, "rb").read()).decode()
image_data_uri = f"data:{mime};base64,{img_b64}"

response = client.chat.completions.create(
  model=MODEL,
  messages=[{
    "role": "user",
    "content": [
      {"type": "text", "text": ""},
      {"type": "image_url", "image_url": {"url": image_data_uri}},
    ],
  }],
)
print(response)

import fs from "node:fs";

const API_KEY = "Your API Key";
const BASE_URL = "https://labs-api.petpooja.com/v1/extract/chat/completions";

const image = fs.readFileSync("<IMAGE_PATH>").toString("base64");

const res = await fetch(`${BASE_URL}`, {
  method: "POST",
  headers: {
    "Content-Type": "application/json",
    Authorization: `Bearer ${API_KEY}`,
  },
  body: JSON.stringify({
    model: "chitragupta",
    messages: [{
      role: "user",
      content: [
        { type: "text", text: "Extract the invoice information as JSON." },
        { type: "image_url", image_url: { url: `data:image/jpeg;base64,${image}` } },
      ],
    }],
  }),
});

const data = await res.json();
console.log(data.choices[0].message.content);

package main

import (
  "encoding/base64"
  "fmt"
  "io"
  "net/http"
  "os"
  "strings"
)

const (
  key   = "Your API Key"
  url   = "https://labs-api.petpooja.com/v1/extract/chat/completions"
  model = "chitragupta"
  path  = "<IMAGE_PATH>"
)

func main() {
  data, err := os.ReadFile(path)
  if err != nil {
    panic(err)
  }
  b64 := base64.StdEncoding.EncodeToString(data)

  body := fmt.Sprintf(`{"model":"%s","messages":[{"role":"user","content":[
{"type":"text","text":"Extract the invoice information from this image as JSON."},
{"type":"image_url","image_url":{"url":"data:image/jpeg;base64,%s"}}
]}]}`, model, b64)

  req, _ := http.NewRequest("POST", url, strings.NewReader(body))
  req.Header.Set("Authorization", "Bearer "+key)
  req.Header.Set("Content-Type", "application/json")

  res, err := http.DefaultClient.Do(req)
  if err != nil {
    panic(err)
  }
  defer res.Body.Close()
  out, _ := io.ReadAll(res.Body)

  fmt.Println(res.StatusCode)
  fmt.Println(string(out))
}

KEY="Your API Key"
BASE="https://labs-api.petpooja.com/v1/extract/chat/completions"
IMAGE="<IMAGE_PATH>"
IMAGE_URI="data:image/jpeg;base64,$(base64 -w0 "$IMAGE")"
cat > /tmp/pp_extract_image.json <<EOF
{
  "model": "chitragupta",
  "max_tokens": 4000,
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Extract all visible text and details from this image."
        },
        {
          "type": "image_url",
          "image_url": {
            "url": "$IMAGE_URI",
            "detail": "high"
          }
        }
      ]
    }
  ]
}
EOF

curl -s -i "$BASE" \
  -H "Authorization: Bearer $KEY" \
  -H "Content-Type: application/json" \
  -d @/tmp/pp_extract_image.json

PDF

POSThttps://labs-api.petpooja.com/v1/extract/chat/completions

Each page of each PDF becomes its own model call, dispatched concurrently, and the per-page results are concatenated in order. A large batch costs roughly the wall-clock of a single page, and one unreadable page never poisons the rest of the document.

Fan-out limit

Limit	Value	On breach
Pages + images per request	25	413 `page_limit_exceeded`

import base64
from openai import OpenAI

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

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

PDF_PATH = "<PDF_PATH>"

pdf_b64 = base64.b64encode(open(PDF_PATH, "rb").read()).decode()
pdf_data_uri = f"data:application/pdf;base64,{pdf_b64}"

response = client.chat.completions.create(
  model=MODEL,
  messages=[{
    "role": "user",
    "content": [
      {"type": "text", "text": "Extract the invoice information from this document as JSON."},
      {"type": "file", "file": {"filename": "invoice.pdf", "file_data": pdf_data_uri}},
    ],
  }],
)

print(response)

import fs from "node:fs";

const apiKey = "Your API Key";
const baseURL = "https://labs-api.petpooja.com/v1/extract/chat/completions";
const model = "chitragupta";
const pdfPath = "<PDF_PATH>";

async function main() {
  const b64 = fs.readFileSync(pdfPath).toString("base64");

  const body = {
    model,
    messages: [{
      role: "user",
      content: [
        { type: "text", text: "Extract information of last page from this document as JSON." },
        { type: "file", file: { filename: "invoice.pdf", file_data: `data:application/pdf;base64,${b64}` } }
      ]
    }]
  };

  const res = await fetch(baseURL, {
    method: "POST",
    headers: {
      "Authorization": `Bearer ${apiKey}`,
      "Content-Type": "application/json"
    },
    body: JSON.stringify(body)
  });

  console.log("HTTP Status:", res.status);
  console.log(await res.text());
}

main();

package main

import (
  "encoding/base64"
  "fmt"
  "io"
  "net/http"
  "os"
  "strings"
)

const (
  apiKey  = "Your API Key"
  baseURL = "https://labs-api.petpooja.com/v1/extract/chat/completions"
  model   = "chitragupta"
  pdfPath = "<PDF_PATH>"
)

func main() {
  data, err := os.ReadFile(pdfPath)
  if err != nil {
    panic(err)
  }
  b64 := base64.StdEncoding.EncodeToString(data)

  body := fmt.Sprintf(`{"model":"%s","messages":[{"role":"user","content":[
{"type":"text","text":"Extract information of last page from this document as JSON."},
{"type":"file","file":{"filename":"invoice.pdf","file_data":"data:application/pdf;base64,%s"}}
]}]}`, model, b64)

  req, _ := http.NewRequest("POST", baseURL, strings.NewReader(body))
  req.Header.Set("Authorization", "Bearer "+apiKey)
  req.Header.Set("Content-Type", "application/json")

  res, err := http.DefaultClient.Do(req)
  if err != nil {
    panic(err)
  }
  defer res.Body.Close()
  out, _ := io.ReadAll(res.Body)

  fmt.Println("HTTP Status:", res.StatusCode)
  fmt.Println(string(out))
}

KEY="Your API Key"
BASE="https://labs-api.petpooja.com/v1/extract/chat/completions"
PDF="<PDF_PATH>"
PDF_URI="data:application/pdf;base64,$(base64 -w0 "$PDF")"
cat > /tmp/pp_chat_pdf.json <<EOF
{
  "model": "chitragupta",
  "max_tokens": 8000,
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "text",
          "text": "Extract all details from this document and return the extracted information."
        },
        {
          "type": "file",
          "file": {
            "file_data": "$PDF_URI",
            "filename": "document.pdf"
          }
        }
      ]
    }
  ]
}
EOF
curl -s -i "$BASE"\
  -H "Authorization: Bearer $KEY" \
  -H "Content-Type: application/json" \
    -d @/tmp/pp_chat_pdf.json

Model response

Both endpoints return a standard chat.completion object. Default packs its per-page results into the single assistant message, each block prefixed with its source tag so you can trace any extracted value back to the exact page it came from.

Field	Type	Description
id	string	Completion id.
model	string	Always the logical name `chitragupta`. The underlying provider is never exposed.
choices[0].message.content	string	The extracted content. For Default, blocks tagged `pdf1_page1:`, `image1:`, in request order.
choices[0].finish_reason	string	`stop` - the model finished. `length` - output was cut short at a token limit.
usage	object	`prompt_tokens`, `completion_tokens`, `total_tokens` - reported for observability. Billing is per page.

finish_reason: length

Because max_tokens already defaults to the 40000 ceiling, a length finish almost always means the context window filled up, not the output cap. Raising max_tokens will not fix it - send fewer pages per request.


request_id   : "value"
usage        : CompletionUsage(completion_tokens="value", prompt_tokens="value", total_tokens="value")
==============================================================================
pdf1_page1: ```json
{"SELLER_DETAILS":{"COMPANY_NAMES" : ["Acme Corp"],"ADDRESS": ["123 Business Rd, Tech Park, New Delhi-110001, India", "Acme...
```

pdf1_page2: ```json
{"SELLER_DETAILS":{"COMPANY_NAMES" : ["Acme Corp"],"ADDRESS": ["123 Business Rd, Tech Park, New Delhi-110001, India", "Acme...
```

pdf1_page3: ```json
{"SELLER_DETAILS":{"COMPANY_NAMES" : ["Acme Corp"],"ADDRESS": ["123 Business Rd, Tech Park, New Delhi-110001, India", "Acme...
```

...