Job Actions

Jobs transform events after the input parses them and before the output delivers them. Use this page to choose actions by outcome, then jump into the DSL reference for exact fields.

Choose by outcome

Outcome	Typical sequence	Edition
Normalize payloads	`json` -> `convert` -> `rename` -> `remove`	Both
Parse unstructured logs	`extract` -> `key-value` -> `assert`	Both
Enrich with lookup data	`enrich` -> `time`	Both
Build LLM features from text	`pdf-text` or `docx-to-text` -> `chunk` -> `tokenize`	Both
Generate model output in-pipeline	`chunk` -> `infer` -> `assert`	`infer` is Enterprise
Score or detect anomalies	`infer` (`anomaly-detect`) -> `scoring` -> `filter`	`infer` is Enterprise

AI pipeline patterns

Structured LLM extraction

actions:
  - chunk:
      input-field: body
      output-field: chunks
  - infer:
      workload:
        llm-completion:
          llm:
            provider: openai-compat
            model: your-model
            input-field: chunks
            response-field: ai_result
            response-format: json
            prompt:
              system: Extract only the requested fields.
              schema: '{"type":"object"}'
      timeout-ms: 15000
      on-error: dlq:ai_failures
  - assert:
      behaviour: drop-onfailure
      schema:
        schema-string: '{"type":"object"}'

Embeddings and clustering

actions:
  - infer:
      workload:
        embedding:
          embedding:
            provider: openai-compat
            model: your-embedding-model
            input-field: text
            response-field: vector
  - cluster:
      input-field: vector
      output-field: cluster_id

`infer` guardrails (recommended defaults)

Use response-format: json plus prompt.schema when downstream systems expect structured output.
Set timeout-ms, rate-limit, and concurrency before production rollout.
Configure cache (namespace, ttl, max-entries) for repeated prompts.
Set on-error explicitly (fail, skip, or dlq:name) instead of relying on implicit behavior.
Store provider credentials in variables (for example ${dyn|OPENAI_API_KEY}), not inline literals.

Common actions

add

add creates or overwrites fields from literals, template placeholders ({{ }}), and runtime expansions (${ }}).

convert

convert normalizes data types (string, number, datetime, boolean) and lets you define failure behavior per conversion.

filter

filter gates events using schema rules, pattern matches, or expressions so invalid data does not reach outputs.

enrich

enrich joins event fields against CSV or SQLite lookup assets and maps matched values back into the event.

Run & Trace checklist for AI jobs

Confirm input-field receives the expected text payload.
Verify response-field shape is stable across multiple samples.
Inspect token usage fields when configured to estimate cost and rate-limit pressure.
Test malformed or empty inputs to validate on-error behavior.
Re-run with representative data volume to validate latency and concurrency settings.

For complete parameter details, use the DSL index and open the linked action pages.