Lark Bitable Integration Guide

Tariffin exposes a webhook at https://tariffin.com/api/lark/classify. It takes a product description + destination country and returns HS code, tariff rate, and landed cost. Lark Bitable’s Automation feature can watch a column, POST to the webhook, and write the response back into other columns.

Add 8 fields. The first 4 are inputs you fill, the last 4 will be auto-populated (leave them empty for now).

In your Bitable, click the ⚙️ Automation icon top-right → New flow → Start from blank.

{
  "product_name": "{{Product}}",
  "destination": "{{Destination}}",
  "declared_value_usd": {{Declared Value (USD)}},
  "origin": "{{Origin}}"
}

POST https://tariffin.com/api/lark/classify

Headers

Authorization: Bearer tfk_xxxxxxxxxxxxxxxxxxxxxxxx
content-type: application/json

Body fields

Full example

curl -X POST https://tariffin.com/api/lark/classify \
  -H "Authorization: Bearer tfk_xxxxxxxxxxxxxxxxxxxx" \
  -H "content-type: application/json" \
  -d '{
    "product_name": "Portable bluetooth speaker, plastic, USB-C, 5W",
    "destination": "US",
    "declared_value_usd": 1000,
    "origin": "CN"
  }'

Sample response

{
  "hs_code": "851829",
  "description_en": "Other loudspeakers, single",
  "description_zh": "其他扬声器,单声道",
  "confidence": 0.92,
  "reasoning": "Portable bluetooth speaker with plastic housing, USB-C, 5W → HS 8518.29.",
  "mfn_rate": 0.024,
  "full_code": "8518.29.80",
  "source": "USITC HTS",
  "source_url": "https://hts.usitc.gov/?query=8518.29",
  "landed_cost_usd": 1117.6,
  "landed_cost_local": 1117.6,
  "currency": "USD",
  "effective_rate_percent": 2.4,
  "detected_language": "en",
  "provider": "deepseek",
  "latency_ms": 1843
}

8.1 Compare across multiple destinations

Want to compare the same product across all 13 countries? Use Lark’s linked-table feature: main table = products, child table = product × destination (13 rows per product). The Automation fires per child row, so each call hits exactly one country — faster and easier to debug.

8.2 Backfilling a large existing table

Already have 1,000 rows to classify? Temporarily switch the Automation trigger to “Scheduled — every day at 8 AM, fill empty HS Code fields”. Note the 60/min limit — 1,000 rows takes ~17 minutes.

8.3 Multiple teammates / use cases

Issue a separate API key per teammate or use case (name them descriptively in the dashboard, e.g. “alice-prod”, “bob-test”). Benefits:

• Per-key usage stats and last-used timestamps
• Revoke a single key when someone leaves — others unaffected
• Rate limits bucketed per key, so they don’t compete

8.4 Use the webhook outside Lark

The endpoint is a plain REST API. Any HTTP-capable tool works: Python requests, n8n, Zapier, shell scripts — just include the Bearer token.

8.5 In-app test

Before wiring up Automation, fire one sample request from Dashboard → Lark integration → Webhook test to confirm your account’s key works end-to-end.