Log in, then extract JSON

Requires a Pro license

Sign into ZRM, switch to the APAC region, open the sales report, and have the formatter return it as JSON, saved to sales_report.json. This example uses Anthropic for both the navigator and the formatter with the key in ANTHROPIC_API_KEY, but any supported provider works.

Run locally

zshot -t llm_json -f sales_report.json \
  --navigate "Login, change the region to APAC, and click on Sales Report" \
  --navigator-credential username=success@zshot-app.com \
  --navigator-credential password=password \
  --navigator-type "anthropic://$ANTHROPIC_API_KEY@" \
  --llm-format-model "anthropic://$ANTHROPIC_API_KEY@" \
  --llm-format-prompt "The sales report as JSON" \
  --llm-format-expects '{"type":"array","items":{"type":"object","properties":{"company":{"type":"string"},"contact":{"type":"string"},"value":{"type":"number"}}}}' \
  "https://zshot-cli.com/example_assets/zrm/"

{
  "url": "https://zshot-cli.com/example_assets/zrm/",
  "output_type": "llm_json",
  "navigate": "Login, change the region to APAC, and click on Sales Report",
  "navigator_type": "anthropic://THE_ANTHROPIC_API_KEY@",
  "navigator_credential": {
    "username": "success@zshot-app.com",
    "password": "password"
  },
  "llm_format_model": "anthropic://THE_ANTHROPIC_API_KEY@",
  "llm_format_prompt": "The sales report as JSON",
  "llm_format_expects": {
    "type": "array",
    "items": {
      "type": "object",
      "properties": {
        "company": { "type": "string" },
        "contact": { "type": "string" },
        "value": { "type": "number" }
      }
    }
  }
}

curl -X POST http://127.0.0.1:3000/ \
  -H "Content-Type: application/json" \
  -d '{"url":"https://zshot-cli.com/example_assets/zrm/","output_type":"llm_json","navigate":"Login, change the region to APAC, and click on Sales Report","navigator_type":"anthropic://'$ANTHROPIC_API_KEY'@","navigator_credential":{"username": "success@zshot-app.com","password": "password"}, "llm_format_model": "anthropic://'$ANTHROPIC_API_KEY'@", "llm_format_prompt": "The sales report as JSON", "llm_format_expects": { "type": "array", "items": { "type": "object", "properties": { "company": { "type": "string" }, "contact": { "type": "string" }, "value": { "type": "number" } } } }}' \
  -o sales_report.json

The navigator opens the APAC sales report; the formatter returns it as JSON. zshot ≥ 0.17.0

Expected output

[
  {
    "company": "Acme Staffing",
    "contact": "Kyle Saito",
    "value": 1272000
  },
  {
    "company": "Globex Technologies",
    "contact": "Chad Petrov",
    "value": 1227000
  },
  {
    "company": "Abstergo Procurement",
    "contact": "Lena Nakamura",
    "value": 1211000
  },
  {
    "company": "Spectre Staffing",
    "contact": "Roman Whitaker",
    "value": 1135000
  },
  {
    "company": "Spectre Staffing",
    "contact": "Cheryl Ferreira",
    "value": 1119000
  },
  {
    "company": "Globex Technologies",
… 304 more lines

The navigator drives the app with --navigate; the formatter shapes the rendered page into JSON with --llm-format-prompt constrained by the JSON Schema in --llm-format-expects. See LLM JSON.