Atomic Field Upsert

Atomic find-or-create with atomic field update (upsert mode).

This endpoint atomically:

1. Ensures the context exists with correct unique_keys configuration
2. Acquires an advisory lock on the unique key values (prevents race on first insert)
3. Finds an existing log by unique_keys or creates it with initial_data
4. Applies an atomic operation to the specified field
5. If add_to_all_context=true, mirrors the log to the All/* archive context

Required body fields for upsert mode:

• project: Name of the project
• context: Context path for the log
• unique_keys: Key name to type mapping (e.g., {"_assistant_id": "str", "month": "str"})
• initial_data: Data for new log entry (must include all unique key values)
• operation: Atomic operation (+N, -N, *N, /N)

This is race-safe for concurrent first inserts - only one request will create the log, and all concurrent requests will correctly increment the field.

{
  "operation": "+1",
  "field": "cumulative_spend",
  "project": "Assistants",
  "context": "42/7/Spending/Monthly",
  "unique_keys": {
    "_assistant_id": "str",
    "month": "str"
  },
  "initial_data": {
    "_assistant_id": "123",
    "_org_id": 456,
    "month": "2026-01"
  }
}