Browse the docsModel response
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...
```
...