just-vibeML deployment · apply by default
/ml-batch
Build resumable batch inference and output tracking
Use for resumable batch inference; ml-serving handles request/response service behavior.
Make it your own.
In Claude Code, use the slash command and add your context. In Codex, select ml-batch from the just-vibe skill picker, then send the same brief.
Version 0.11.0 also supports /jv ml-batch, /just-vibe ml-batch and /jv:ml-batch in Claude. See shortcut setup and context examples.
/just-vibe:ml-batch Implement resumable batch inference with stable output keys and model-version tracking./just-vibe:ml-batch Resume inference after output writes succeeded but checkpointing failed./just-vibe:ml-batch Plan batch prediction without scanning the full dataset or launching unbounded compute.What the agent does
- Freeze model and input snapshot identity, validate schemas, and partition by stable row keys.
- Stage outputs and commit a manifest or checkpoint only after durable complete partitions, tracking failures and model versions.
- Test resume and replay on controlled input.
Inputs
- model/data versions, partitioning, output keys, checkpoint destination, and run budget.
Optional context: scope, references, constraints, successCriteria, environment, mode, budget.
Scope
- Reads
- Resumable batch inference; real dataset execution must be part of the request.
- Writes
- Apply: only the requested local changes and relevant isolated verification. Inspect/plan requests remain inspection/planning. External actions require their exact action and target in session authorization.
- Mode
- Apply for implementation; model/data versions, partitioning, output keys, checkpoint destination, and run budget.
- Prerequisites
- Versioned model and preprocessing artifacts, input/output schema, runtime/dependencies, operating targets, and authorized environment. Validate artifact trust before loading formats that can execute code. Packaging or writing monitoring configuration does not deploy a model or enable a hosted service.
Expected output
- Batch job with its partition/model manifest, progress and output reconciliation, error policy, failure counts and resume evidence.
How the work is checked
- Restart does not duplicate completed outputs; invalid records are accounted for rather than silently lost.
When to stop or clarify
- No unbounded full-dataset inference. Do not mix predictions from incompatible model versions in one unlabeled output.
Handling missing context
- Infer
- Read artifact format/trust, preprocessing schema, serving runtime, compatibility and existing rollout controls.
- Assume
- Prepare packaging/configuration and isolated checks without treating them as a live deployment.
- Ask
- Resolve the target, rollback compatibility and operating limits before rollout or load generation; missing production access does not block packaging.
Technical guidance
- Evidence
- Identify input snapshot, stable record IDs, model version, partitioning and output commit/checkpoint policy.
- Method
- Write resumable idempotent partitions with provenance and reconcile partial output before advancing progress.
- Pitfall
- Restarting with a different model under the same output partition silently mixes incompatible predictions.
- Check
- Interrupt before/after output commit and retry; verify full membership, no duplicate records and consistent model/data identity.
Situational decisions
When a restart finds partial output or a different model version: Reconcile or isolate it before resuming; never silently mix incompatible predictions.
The coding agent follows this workflow using its available tools. Installation does not grant service access or guarantee an outcome. Read the compatibility notes.