Choosing an Effort Level

Every /extract and /generate request accepts an optional effort parameter. It controls the tradeoff between speed and capability; specifically, how hard Tabstack works to render and process the target page before extracting data.

---

The three levels

Level	Page fetch	What it does
min	1-5s	Fetches raw HTML. No JavaScript execution. Use for static, server-rendered pages.
standard	3-15s	Fetches with enhanced reliability and light JS handling. Default for all requests.
max	15-60s	Full headless browser rendering. Executes JavaScript, waits for dynamic content to load.

Default is standard when the parameter is omitted.

Effort is not your whole response time

The times above cover fetching and rendering the page. They do not cover extracting your data out of it, and for /extract/json and /generate/json that second step usually dominates. Total response time is roughly page fetch plus extraction, and extraction scales with how much you ask for.

Measured on the same page at the same standard effort, varying only the schema:

Schema	Observed total
2 flat string/number fields	2-6s
4 fields across an array of ~30 items	18-120s

Same URL, same effort level, same day. The only variable was the number of values requested.

What that means for you: if you are extracting a handful of fields, effort is a good predictor of latency and the table above holds. If you are extracting an array (every product on a listing page, every story on a feed, every row of a table), expect to run well past 30 seconds regardless of effort, and expect a long tail. Do not size a timeout, a serverless function, or a retry budget off the fetch column alone.

To keep array extractions faster, narrow what you ask for: fewer properties per item, and a description on the array that bounds it (for example, “the first 10 stories” rather than “all stories”). See Schema Design.

/extract/markdown is not affected. It has no schema, so it stays inside the fetch times above.

For deployment implications, see Serverless and edge functions.

---

When to use each level

min: Fastest, static pages only

Use when:

• The page is server-rendered HTML (blogs, news sites, documentation)
• Speed matters and the content doesn’t require JavaScript to appear
• You’re running high-volume extractions and want lowest latency

Don’t use when:

• The page uses a JavaScript framework (React, Vue, Next.js, Angular)
• Content is loaded asynchronously after the initial HTML response
• You see empty fields or missing data with standard

TypeScript

const result = await client.extract.json({
  url: 'https://news.ycombinator.com', // Static HTML, min works fine
  effort: 'min',
  json_schema: { /* ... */ }
})

Python

result = client.extract.json(
    url="https://news.ycombinator.com",  # Static HTML, min works fine
    effort="min",
    json_schema={...},
)

CLI

tabstack extract json https://news.ycombinator.com \
  --schema @schema.json \
  --effort min

curl

curl -X POST https://api.tabstack.ai/v1/extract/json \
  -H "Authorization: Bearer $TABSTACK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://news.ycombinator.com",
    "effort": "min",
    "json_schema": { "type": "object" }
  }'

standard: The default, works for most pages

Use when:

• You’re not sure what the page renders like
• You want reliable results without committing to full browser rendering time
• The page may have some JavaScript but its primary content is in the initial HTML

This is the right starting point. If results feel incomplete, move to max.

TypeScript

const result = await client.extract.json({
  url: 'https://example.com/products',
  // effort: 'standard' is the default, can be omitted
  json_schema: { /* ... */ }
})

Python

result = client.extract.json(
    url="https://example.com/products",
    # effort="standard" is the default, can be omitted
    json_schema={...},
)

CLI

# effort defaults to standard, can be omitted
tabstack extract json https://example.com/products \
  --schema @schema.json

curl

curl -X POST https://api.tabstack.ai/v1/extract/json \
  -H "Authorization: Bearer $TABSTACK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://example.com/products",
    "json_schema": { "type": "object" }
  }'

max: Full rendering, JS-heavy sites

Use when:

• The page is a Single Page Application (React, Vue, Angular, Next.js client-side)
• Content loads lazily or after user interaction
• Pricing tables, product listings, or data grids are rendered by JavaScript
• You’re getting empty fields with standard
• The page is behind a login with JS-rendered content after auth

TypeScript

const result = await client.extract.json({
  url: 'https://app.example.com/dashboard',
  effort: 'max', // Wait for JS to fully render
  json_schema: { /* ... */ }
})

Python

result = client.extract.json(
    url="https://app.example.com/dashboard",
    effort="max",  # Wait for JS to fully render
    json_schema={...},
)

CLI

tabstack extract json https://app.example.com/dashboard \
  --schema @schema.json \
  --effort max

curl

curl -X POST https://api.tabstack.ai/v1/extract/json \
  -H "Authorization: Bearer $TABSTACK_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "url": "https://app.example.com/dashboard",
    "effort": "max",
    "json_schema": { "type": "object" }
  }'

---

Decision flow

Is the page server-rendered HTML with no JavaScript?
  └─ Yes → try 'min'
  └─ No or unsure → use 'standard' (default)

Are you getting empty fields or incomplete data?
  └─ Yes → upgrade to 'max'

Is the page a React/Vue/Angular SPA or loads content dynamically?
  └─ Yes → use 'max' from the start

---

Cost and timeout notes

max uses full headless browser rendering. It consumes more compute than min or standard. If your target page consistently needs max, factor this into your per-request cost expectations.

Default request timeout is 60 seconds. max requests on complex SPAs can approach this limit on very slow pages. If you’re hitting timeouts, consider whether the page can be targeted more specifically with a starting URL or nocache.