Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.graphorlm.com/llms.txt

Use this file to discover all available pages before exploring further.

The Get Page Screenshot endpoint renders one page of a source file (PDF, image, or Office document) as a base64-encoded PNG. Use it to lazily fetch the visual preview of a citation returned by /ask-sources without paying the payload cost of inlining base64 in the answer.

​
Endpoint overview

HTTP Method

GET

Endpoint URL

https://sources.graphorlm.com/\{file_id\}/pages/\{page_number\}/screenshot

​
Authentication

This endpoint requires authentication using an API token. Include your API token as a Bearer token in the Authorization header.
Learn how to create and manage API tokens in the API Tokens guide.

​
When to use this endpoint

This is the recommended way to display page previews for citations returned by /ask-sources.
  • The default /ask-sources response does not include image_base64 — clients should call this endpoint per citation, only when the user actually needs to see the page (e.g. on hover, on click, or when rendering a side panel).
  • Each render is keyed by (file_id, page_number) and returns a Cache-Control: public, max-age=3600 hint so browsers and CDNs can cache the image bytes.
  • Prefer this endpoint over passing include_citation_images=true to /ask-sources whenever the answer may cite many pages, or when latency / payload size matters.

​
Request format

​
Headers

HeaderValueRequired
AuthorizationBearer YOUR_API_TOKENYes

​
Path parameters

ParameterTypeRequiredDescription
file_idstringYesUUID of the source file. Same value returned in citations[].file_id from /ask-sources.
page_numberintegerYes1-based page number. For image-type files, must be 1.

​
Query parameters

ParameterTypeRequiredDescription
max_widthintegerNoPixel width cap for the rendered image. Clamped to 300-1600. Default: 900.

​
Supported file types

TypeExtensionsNotes
PDFspdfAny 1-based page number.
Imagespng, jpg, jpeg, webp, gif, etc.page_number must be 1.
Office documentsdoc, docx, ppt, pptx, odtRendered from the auto-converted PDF.
Plain-text and other non-visual formats are not supported and will return 404.

​
Response format

​
Success response (200 OK)

{
  "file_id": "a1b2c3d4-e5f6-7890-abcd-ef1234567890",
  "file_name": "annual-report-2025.pdf",
  "page_number": 42,
  "mime_type": "image/png",
  "width": 900,
  "height": 1163,
  "image_base64": "iVBORw0KGgoAAAANSUhEUgAA…"
}
FieldTypeDescription
file_idstringUUID of the source file.
file_namestringDisplay name of the source file. May be null.
page_numberinteger1-based page number that was rendered.
mime_typestringMIME type of the encoded image (always "image/png").
widthintegerPixel width of the rendered image.
heightintegerPixel height of the rendered image.
image_base64stringBase64-encoded PNG image bytes.

​
Error responses

StatusDescription
401Unauthorized — invalid or missing API token.
404File not found, unsupported file type, or invalid page number.
500Unexpected internal error while rendering.

​
Examples

​
cURL

curl "https://sources.graphorlm.com/a1b2c3d4-e5f6-7890-abcd-ef1234567890/pages/42/screenshot?max_width=1200" \
  -H "Authorization: Bearer YOUR_API_TOKEN"

​
Python (rendering and saving a citation)

import base64
import requests

API_TOKEN = "YOUR_API_TOKEN"
BASE_URL = "https://sources.graphorlm.com"

# 1. Ask a question
ask = requests.post(
    f"{BASE_URL}/ask-sources",
    headers={"Authorization": f"Bearer {API_TOKEN}"},
    json={"question": "What was the revenue in 2025?"},
).json()

# 2. For each citation the user wants to inspect, fetch the page image
for citation in (ask.get("citations") or []):
    file_id = citation["file_id"]
    page_number = citation["page_number"]

    img = requests.get(
        f"{BASE_URL}/{file_id}/pages/{page_number}/screenshot",
        headers={"Authorization": f"Bearer {API_TOKEN}"},
        params={"max_width": 1200},
    ).json()

    png_bytes = base64.b64decode(img["image_base64"])
    with open(f"page_{file_id}_{page_number}.png", "wb") as f:
        f.write(png_bytes)

​
JavaScript (decoding and rendering inline)

const API_URL = "https://sources.graphorlm.com";
const API_TOKEN = "YOUR_API_TOKEN";

async function fetchCitationImage(fileId, pageNumber) {
  const url = `${API_URL}/${fileId}/pages/${pageNumber}/screenshot?max_width=900`;
  const res = await fetch(url, {
    headers: { Authorization: `Bearer ${API_TOKEN}` },
  });
  if (!res.ok) throw new Error(`HTTP ${res.status}`);
  const data = await res.json();
  // Use directly as <img src="data:image/png;base64,...">
  return `data:${data.mime_type};base64,${data.image_base64}`;
}

// Usage from a citation returned by /ask-sources
const src = await fetchCitationImage(citation.file_id, citation.page_number);
imgEl.src = src;

​
Best practices

  1. Lazy-load on user interaction — call this endpoint only when the user hovers, clicks, or expands a citation. Most citations are never inspected.
  2. Cache by (file_id, page_number) — the response is deterministic for a given key. The endpoint returns Cache-Control: public, max-age=3600; respect it on the client to avoid re-rendering the same page.
  3. Pick max_width for the surface you’re rendering — 500 for tooltips/popovers, 900 (default) for inline cards, 1200+ only when the user opens a full preview.
  4. Combine with /ask-sources citations — pair this endpoint with the structured citations array from /ask-sources rather than parsing the inline [N] markup.

Chat API

Ask questions and receive grounded answers with citations

List Sources

Discover the file_id values you can pass here