idempotent-processing

Idempotent Processing

A webhook endpoint occasionally receives the same event twice due to network retries. Processing it twice would create duplicate records. The idempotent processing pipeline needs to compute a deterministic idempotency key for each event, check if it has already been processed, skip duplicates, and return the cached result.

Pipeline

[idp_check_processed]
     |
     v
     <SWITCH>
       |-- unprocessed -> [idp_process]
       |-- processed -> [idp_skip]
       +-- default -> [idp_process]
     |
     v
[idp_record]

Workflow inputs: messageId, payload

Workers

CheckProcessedWorker (task: idp_check_processed)

Checks if a message has been previously processed by looking it up in the in-memory {@link DedupStore}.

• Reads messageId. Writes processingState, previousResult

ProcessWorker (task: idp_process)

Processes a message that has not been seen before. Outputs are deterministic — the resultHash is derived from the messageId so repeated invocations with the same input always produce the same output.

• Computes sha-256 hashes, formats output strings
• Reads messageId, payload. Writes success, resultHash

RecordWorker (task: idp_record)

Records a successfully processed message in the {@link DedupStore} so that future runs with the same messageId are detected as duplicates.

• Reads messageId, resultHash. Writes recorded

SkipWorker (task: idp_skip)

Skips processing for a message that was already handled. Returns the reason for skipping and the previous result hash so callers can verify the original processing outcome.

• Reads messageId, previousResult. Writes skipped, reason, messageId, previousResult

---

15 tests | Workflow: idp_idempotent_processing | Timeout: 60s

See RUNNING.md for setup and usage.