API Tools — streaming infrastructure for reliable API and AI data

Configuration

v1 Config Fields

Used with DDStream (v1 GET timeseries). All fields map directly to DDStreamConfig in the C++ API.

Field	Type	Default	Description
`api_key`	`string`	`—`	Datadog API key (required)
`app_key`	`string`	`—`	Datadog Application key (required)
`query`	`string`	`—`	Datadog metric query, e.g. avg:system.cpu.user{*}
`range_start`	`int64_t`	`—`	Start of time range, Unix epoch seconds
`range_end`	`int64_t`	`—`	End of time range, Unix epoch seconds
`chunk_seconds`	`int64_t`	`14400`	Window size per HTTP request (4 hours)
`granularity`	`int32_t`	`60`	Metric interval in seconds
`connect_timeout_ms`	`int32_t`	`5000`	TCP connection timeout in milliseconds
`request_timeout_ms`	`int32_t`	`30000`	Full HTTP request timeout in milliseconds
`max_retries`	`int32_t`	`3`	Maximum retries for 429/5xx responses
`point_reserve`	`int32_t`	`50000`	Pre-allocated points per batch buffer
`enable_prefetch`	`bool`	`true`	Background-thread prefetch of the next chunk
`enable_arrow`	`bool`	`false`	Return Arrow RecordBatch from C API instead of DataPoint[]
`log_level`	`LogLevel`	`NONE`	0=NONE 1=ERROR 2=WARN 3=INFO 4=DEBUG

v2 Config Fields

Used with DDStreamV2Timeseries and DDStreamV2Scalar. The v2 config accepts multiple named queries and optional formula expressions.

Field	Type	Default	Description
`api_key`	`string`	`—`	Datadog API key (required)
`app_key`	`string`	`—`	Datadog Application key (required)
`site`	`string`	`datadoghq.com`	Datadog site, e.g. datadoghq.eu
`range_start`	`int64_t`	`—`	Start of time range, Unix epoch seconds
`range_end`	`int64_t`	`—`	End of time range, Unix epoch seconds
`chunk_seconds`	`int64_t`	`14400`	Window size per HTTP request
`interval_ms`	`int64_t`	`60000`	Metric granularity in milliseconds
`queries`	`V2Query[]`	`—`	Named query variables, e.g. [{name:"q0", query:"avg:..."}]
`formulas`	`V2Formula[]`	`[]`	Optional formula expressions combining query variables
`connect_timeout_ms`	`int32_t`	`5000`	TCP connection timeout in milliseconds
`request_timeout_ms`	`int32_t`	`30000`	Full HTTP request timeout in milliseconds
`max_retries`	`int32_t`	`3`	Maximum retries for 429/5xx responses
`enable_prefetch`	`bool`	`true`	Background-thread prefetch of the next chunk
`log_level`	`int32_t`	`0`	0=NONE 1=ERROR 2=WARN 3=INFO 4=DEBUG

Chunk Size Guide

The default chunk_seconds=14400 (4 hours) works well for 1-minute granularity. Larger chunks reduce HTTP overhead; smaller chunks reduce memory usage. Use this table as a starting point:

Granularity	Points per chunk	Recommended `chunk_seconds`
60 s (1 min)	240	14400 (4 h) - default
300 s (5 min)	48	86400 (24 h)
3600 s (1 h)	4	604800 (7 days)
Custom	`chunk_seconds / granularity`	Target 100–5000 points per request

If you see HTTP 400 or 503 errors, reduce chunk_seconds. If you see many small batches with fast processing, increase it to reduce overhead.