Azure Cosmos DB Nodes

Azure Cosmos DB is Microsoft's globally distributed NoSQL database. MaestroHub provides seven nodes for its NoSQL API: point and bulk reads, SQL queries, whole-item writes, partial patches, deletes, and transactional batches — all against the database the connection profile names.

Items cross the boundary as plain JSON in both directions, and a downstream node reads result.item.celsius directly. Every result reports the Request Units the call consumed as result.requestCharge.

For the connection, the function types and their fields, see the Azure Cosmos DB connection guide. To run a pipeline when an item changes rather than reading one on demand, see the Azure Cosmos DB trigger.

Configuration Quick Reference

Field	What you choose	Details
Parameters	Connection, Function, Function Parameters, Timeout Override	Select the connection profile (which fixes the account and database), the function, values for any ((parameter)) the function uses, and optionally a timeout override.
Settings	Description, Timeout (seconds), Retry on Timeout, Retry on Fail, On Error	Node description, maximum execution time, retry behavior on timeout or failure, and error handling strategy. All execution settings default to pipeline-level values.

Zoom

Azure Cosmos DB Read Item Node

Azure Cosmos DB Read Item Node

Read one item by its ID and partition key — the cheapest read Cosmos DB offers.

An item that does not exist is not an error. The node succeeds with result.found false and a null result.item, so a pipeline branches on the lookup instead of routing it through failure handling.

Supported Function Types:

Function Name	Purpose	Common Use Cases
Read Item	Read a single item by ID and partition key	Enrichment lookups, machine state reads, existence checks before a write

---

Zoom

Azure Cosmos DB Read Many Items Node

Azure Cosmos DB Read Many Items Node

Read up to 1,000 items by ID and partition key in one call.

Cosmos DB groups the items by partition and reads each group in one request, which costs far less than a Read Item per entry in a loop. The items that do not exist come back in result.missing.

Supported Function Types:

Function Name	Purpose	Common Use Cases
Read Many Items	Read a known set of items at once	Looking up every device an upstream node named, batch enrichment

---

Zoom

Azure Cosmos DB Query Items Node

Azure Cosmos DB Query Items Node

Run a Cosmos DB SQL query with named parameters.

The limit is sent to Cosmos DB as the page size, so it bounds what is read and billed. Scope the query to one partition with Partition Key whenever you can: it is cheaper, and ORDER BY, TOP, DISTINCT, GROUP BY and aggregates need it.

Supported Function Types:

Function Name	Purpose	Common Use Cases
Query Items	Read the items a SQL query selects	Readings above a threshold, a device's latest readings, paging through a container

---

Zoom

Azure Cosmos DB Write Item Node

Azure Cosmos DB Write Item Node

Create, upsert or replace a whole item.

The partition key is read from the item's own field. result.created says whether the write made a new item or replaced one, and result.etag is the item's new ETag — pass it as If-Match to make the next write conditional on nobody changing the item in between.

Supported Function Types:

Function Name	Purpose	Common Use Cases
Write Item	Store an item, with create, upsert or replace semantics	Persisting readings, upserting machine records, create-only order intake

---

Zoom

Azure Cosmos DB Patch Item Node

Azure Cosmos DB Patch Item Node

Change parts of an existing item: set, add, replace or remove fields, or increment a counter on the server.

Up to ten operations apply atomically, optionally only when a condition on the stored item holds. result.item is the item after the patch, which is how an increment's new value is read.

Supported Function Types:

Function Name	Purpose	Common Use Cases
Patch Item	Partial, atomic updates to a known item	Status flips, cycle counters, appending to an alarm list

---

Zoom

Azure Cosmos DB Delete Item Node

Azure Cosmos DB Delete Item Node

Delete one item by its ID and partition key.

Deleting an item that is not there succeeds with result.deleted false, so the node is safe to replay.

Supported Function Types:

Function Name	Purpose	Common Use Cases
Delete Item	Remove an item	Retention cleanup, erasure requests, compensating actions

---

Zoom

Azure Cosmos DB Transactional Batch Node

Azure Cosmos DB Transactional Batch Node

Run up to 100 creates, upserts, replaces, patches, reads and deletes on one partition as a single transaction.

Either every operation applies or none does. When one fails, the node fails with the operation that caused it and nothing is written; when the batch commits, result.results has one entry per operation, in order.

Supported Function Types:

Function Name	Purpose	Common Use Cases
Transactional Batch	Several writes to one partition, all or nothing	Writing a reading and updating its device summary together

---

Paging through a query

When more results may follow, result.truncated is true and result.continuationToken carries the position. Feed it into the next run's Continuation Token, with the same query and partition key:

run 1: limit 100                         → result.continuationToken "…", result.truncated true
run 2: continuationToken from run 1      → result.continuationToken "…", result.truncated true
run 3: continuationToken from run 2      → result.truncated false, no token

Output

Every result carries result.requestCharge: the Request Units the call consumed. Wherever a result names a partition key, result.partitionKey is written the way the Partition Key field accepts it — a plain string, or a JSON array for a number, boolean, null or hierarchical key — so it passes straight into another node.

Items are plain JSON. The resource-link properties Cosmos DB adds (_rid, _self, _attachments, _lsn) are removed; _etag and _ts stay.

Read Item

Key	Description
result.container	The container the read addressed
result.id	The id the read was made against
result.partitionKey	The partition key the read was made against
result.found	Whether the item exists. A miss is a normal result, so branch on this
result.item	The item, or null when found is false
result.requestCharge	Request Units consumed

Read Many Items

Key	Description
result.container	The container the read addressed
result.items	The items that exist, in no particular order
result.count	How many items came back
result.missing	The requested items that do not exist, each as {id, partitionKey} — empty when every one was found
result.requestCharge	Request Units consumed

Query Items

Key	Description
result.container	The container the query ran against
result.items	The results — whole items, or whatever the SELECT projects
result.count	How many results came back
result.truncated	Whether more may follow — true exactly when result.continuationToken is present
result.continuationToken	Feed it back as the next run's Continuation Token to resume. Absent on the last page
result.requestCharge	Request Units consumed, across every page the node read

Write Item

Key	Description
result.container	The container written to
result.id	The item's id
result.partitionKey	The partition the item was written to, read from its own field
result.etag	The item's new ETag
result.created	true when the write made a new item, false when it replaced one
result.requestCharge	Request Units consumed

Patch Item

Key	Description
result.container	The container the item is in
result.id	The item's id
result.partitionKey	The item's partition key
result.etag	The item's new ETag
result.item	The item as it is after the patch
result.requestCharge	Request Units consumed

Delete Item

Key	Description
result.container	The container the item was in
result.id	The item's id
result.partitionKey	The item's partition key
result.deleted	false when the item was already gone — still a success
result.requestCharge	Request Units consumed

Transactional Batch

Key	Description
result.container	The container the batch ran against
result.partitionKey	The partition every operation ran against
result.count	How many operations ran — all of them, since a batch that fails applies none and fails the node
result.results	One entry per operation, in the order sent: op, id, statusCode, and etag and item where the service returned them
result.requestCharge	Request Units consumed