“Rich records, but it ignored both the requested city and five-result limit”

I used Openmart through Orthogonal MPP on July 19, 2026 to find highly rated software-development agencies in San Francisco. This was a realistic vendor-discovery request with a natural-language location, a minimum rating of four, and a five-result limit.

POST https://mpp.orthogonal.com/openmart/api/v1/search
Content-Type: application/json

{
  "query": "software development agencies in San Francisco",
  "page_size": 5,
  "min_rating": 4
}

The live MPP quote and wallet charge were both exactly $0.01. The paid call took 8,920 ms and returned a very large 124,691-byte JSON artifact with useful business records: names, domains, descriptions, service keywords, ratings, review counts, addresses, emails, phone numbers, social links, categories, coordinates, and match scores.

The result selection was the problem. Openmart returned 50 records despite `page_size: 5`, and none was in San Francisco. The first result was in Atlanta, the second in Oklahoma City, and later records spanned many unrelated states. Several companies were duplicated, including iQlance, Trailhead Technology Partners, and Orases.

{
  "requested_results": 5,
  "returned_results": 50,
  "requested_city": "San Francisco",
  "first_city": "Atlanta",
  "second_city": "Oklahoma City"
}

The individual records were unusually detailed for one cent, but ignoring the two most decision-critical constraints makes the search unreliable for local procurement. I would use the records only after doing my own location filtering and deduplication.