Break down BPMSoft sales stages straight from OData

The agent works with BPMSoft over OData 4: requests go to your server address into the /0/odata section, which is where contacts, accounts, leads, opportunities, activities, invoices, orders, cases, products and documents are read from. On that data it totals amounts and probabilities by sales stage, deals sitting longer than usual, duplicates in the directories by phone and email and overdue receivables per account.

The first thing the agent reads is $metadata, which spells out the schema of your particular implementation: the set of entities and fields in BPMSoft is extended per customer, and the columns added during implementation often matter more for a report than the standard ones. Relations come from there too: reference fields arrive as identifiers, and without expanding the relation a table shows a GUID where the account name should be. A server response is capped at 20 000 rows, so large selections are fetched page by page.

By default the agent works read-only: it counts, reconciles and reports. It can write — create and update a record, run a business process by schema name — but it does that on your request and while spelling out exactly what will change. Running a process is riskier than an ordinary edit: notifications, tasks and stage transitions follow it, so the agent names the schema and the object it applies to before the start, not after.

The agent's boundaries are the permissions of the integration user: an entity closed by a role shows up in no report at all, and the agent names that cause instead of putting an empty table on screen. Filtering and sorting are handed to the server through $filter, $orderby and $select so that spare columns are never pulled, and dates are cast into the OData format, otherwise the period filters to the wrong range. The request rate stays moderate: paged selections run with pauses instead of taking the server over.