Logs and metrics

The ERPBridge SDK REST surface mirrors the server REST endpoints. All methods use the same typed error hierarchy.

Logs

Use client.logs to access the server's structured JSON logs.

const recent = await client.logs.recent() // LogRecord[] — the last buffered records

Use stream() to consume the server's SSE log stream (data: <json>\n\n). The stream reconnects automatically, including after failed connection attempts, until you abort it:

const controller = new AbortController()

for await (const record of client.logs.stream({ signal: controller.signal })) {
  console.log(record.level, record.msg)
  if (enough) controller.abort() // the iterator finishes cleanly
}

Metrics

Use client.metrics to read the server's Prometheus endpoint.

const text = await client.metrics.text()       // raw Prometheus text format
const parsed = await client.metrics.parsed()   // { families, skipped }

parsed() parses counters, gauges, and histograms, including _bucket, _sum, and _count series, into typed families. It reports unsupported summary and untyped families in skipped instead of throwing.

Health

const status = await client.health() // { status: 'ok' }

Cache

const stats = await client.cache.stats()        // { exactKeys, redisMemory }
const flushed = await client.cache.flush({ all: true }) // { deleted, status }

flush() targets a tool, a module, or the whole cache. The endpoints work with the configured Redis backend and the bounded in-memory backend. In-memory mode returns an empty string for redisMemory. A configured but unreachable Redis backend remains selected and returns a typed ServerError. The ERPBridge SDK does not switch to memory mode.

The typed error hierarchy

Every REST failure is an ErpbridgeError subclass, such as NotFoundError, RateLimitError (with retryAfter), ServerError (with status and body), or ProtocolError (with a numeric code for malformed server responses). See the quickstart for the full table.