Monitor the integration
Observe service health, API traffic and event delivery with signals that have a clear source.
Monitor the parts of the system your customer journey depends on: your own backend, the assigned Partner API, the wallet or transaction provider, and the configured vault data sources.
The monitoring workspace
Open Thesauros Monitor for vault balances, TVL-weighted APY, provider exposure, asset rate history and recent deposit, rebalance and withdrawal events. Filter by asset, network, provider or operating status; open a vault for its details or export the selected rows as CSV.
The preview shares the developer console's sandbox data. Its Sandbox indicator identifies the source. Rates and balances in this environment are simulated. Asset history covers all networks; the vault metrics and provider distribution follow the selected filters.
The workspace refreshes every 30 seconds while visible. An unavailable feed retains the previous snapshot with an error state; a timestamp older than two minutes receives a stale-data indicator. Paused vaults and capacity usage of 90% or more appear in the attention panel. These are operational signals, not a complete risk assessment.
A connected deployment can use a normalized snapshot feed configured on the server. The implementation and feed contract are documented in developer-portal/docs/MONITORING.md. Credentials stay on the server. The original monitoring backend and indexer remain separate services.
Service health
The Partner API's public /status method reports this process, database and configured monitoring service. It returns service states, available chain context and an update timestamp.
If monitoring is not configured, that component is degraded. If an upstream request fails, it can be down while the API process and database remain operational. Handle the component state that affects your flow rather than reducing all outcomes to a single green indicator.
API traffic
GET /usage provides the calling partner's request counts, errors, latency percentiles and most-used endpoints. Its series is based on request logs. The sandbox has a separate simulated usage model.
Track changes over time and connect elevated errors to the actual response codes. A 403 from the wrong credential requires a different response from an upstream timeout.
Event delivery
Use webhook delivery records to detect failed attempts and support recovery. The current service does not provide a durable automatic redelivery queue. Your operating workflow should reconcile important state even when a notification fails.
Incident handoff
Preserve the environment, affected method or transaction, time window and available identifiers. Identify whether customers can continue reading balances, beginning deposits or withdrawing under the actual deployed configuration. Coordinate the next action through the agreed partner support contacts.