Aidp Mlops

by oracle-samples90b42d6c24d4No licenseListed Oct 8, 2026Updated Oct 8, 2026

Track ML work in AIDP's MLflow-compatible MLOps — experiments, runs, metrics/params, registered models, and model versions. Use when the user wants to log/track experiments, list runs or metrics, register a model, transition a model-version stage, or query the model registry. Primary engine is the official `aidp` CLI (`aidp mlops …`); the same Preview MLflow REST API via `oci raw-request` is the no-CLI fallback. Verify live first.

Instructions onlyDevOps & CloudData & Analytics
AI-generated overview

Tracks ML experiments, runs, metrics and registered models through AIDP's MLflow-compatible MLOps CLI or REST API.

What it does
This skill guides an agent through AIDP's MLflow-compatible MLOps surface, covering experiments, runs, metrics and parameters, registered models and model versions. It documents the preferred aidp mlops CLI commands, including a grouped index of 46 subcommands, and a no-CLI fallback using MLflow REST endpoints via oci raw-request. It also describes a verify-first workflow, payload persistence for mutating operations, and confirmation before impactful stage transitions. It produces command invocations and request bodies rather than files.
When to use it
Use it when a user wants to log or track experiments, list runs or metrics, register a model, transition a model-version stage, or query the model registry in AIDP. It is also relevant when the CLI is unavailable and the MLflow REST surface must be used instead.
Requirements
Requires the aidp CLI with an AIDP instance ID, API-key authentication, a profile and a region; the fallback path requires the OCI CLI for oci raw-request. Network access to the AIDP MLOps endpoints is needed. No scripts are shipped; it is instructions only.

aidp-mlops — MLflow experiments & model registry (Preview)

Interact with AIDP's MLflow-compatible MLOps surface. This is AIDP-native MLOps — distinct from OCI Data Science.

CLI (preferred): aidp mlops <command> --instance-id <DATALAKE_OCID> --auth api_key --profile DEFAULT --region <r>

  • Experiments/runs: aidp mlops list-experiments | create-experiment | list-experiment-runs | log-experiment-run-metric
  • Registry: aidp mlops list-registered-models | create-registered-model | create-model-version | transition-model-version-stage
  • Full surface = 46 subcommands — see the grouped command index below (experiments / runs / registered-models / model-versions / tags / artifacts).

Fallback (no CLI): same MLflow REST surface via oci raw-request (identical endpoint + auth; see references/oci-raw-request.md).

Preview + verify-first (no-fabrication): MLOps is Preview. Probed 2026-06-10: GET …/workspaces/<ws>/mlops/api/2.0/mlflow/experiments/search → 404 — either MLOps isn't provisioned in this tenancy, or (likely) the MLflow endpoints are POST-shaped (experiments/search is a POST in MLflow). Confirm the working path/verb with a live aidp mlops list-experiments (CLI) before relying on the REST surface; record in references/rest-endpoint-map.md. Don't assert unverified endpoints.

When to use

  • "Track/log an experiment or run", "list runs/metrics", "register a model", "promote a model version", "query the model registry".

Workflow

  1. Verify with aidp mlops list-experiments (CLI) — or experiments/search (REST fallback; auth ladder).
  2. Standard MLflow patterns: create/find an experiment → create a run → log params/metrics → register the model → create/transition a model version. Bodies follow the MLflow REST schema.
  3. For training itself, run code in a notebook (aidp-notebooks) and log to MLOps from there.

Mutating ops (create-experiment, run logging, create-registered-model, create-model-version, transition-model-version-stage): persist the body to .aidp/payloads/ and confirm first (references/payloads.md).

Full aidp mlops command index (46 subcommands)

All (Preview). Positional args differ by group: experiment/run/artifact commands take <DLID> <WS> (workspace-scoped); registered-model/model-version commands take <DLID> only (no <WS>). --body <JSON> rows are mutating or POST-search reads — persist the body to .aidp/payloads/ and confirm first (references/payloads.md). Field names below are confirmed from docs/cli/README.md mlops section (lines 2455–4049). Source: CLI README §ML Ops.

Experiments (<DLID> <WS>)

CommandMutates?Body / options (required *)
list-experimentsread (POST)body ListExperimentsDetails: filter · max_results · order_by[] · page_token · view_type (ACTIVE_ONLY|DELETED_ONLY|ALL)
get-experiment-by-idreadopt --experiment-id* (works on deleted experiments)
get-experiment-by-namereadopt --experiment-name* (prefers active on name clash)
create-experimentyesbody CreateExperimentDetails: name* · artifact_location · tags[] (each ExperimentTag {key*,value})
update-experimentyesbody UpdateExperimentDetails: experiment_id* · new_name
delete-experimentyesbody DeleteExperimentDetails: experiment_id*
restore-experimentyesbody RestoreExperimentDetails: experiment_id*

Experiment runs (<DLID> <WS>)

CommandMutates?Body / options (required *)
list-experiment-runsread (POST)body ListExperimentRunsDetails: experiment_ids[] · filter · max_results · order_by[] · page_token · run_view_type (ACTIVE_ONLY|DELETED_ONLY|ALL)
get-experiment-run-by-idreadopt --run-id*
get-experiment-run-metric-historyreadopt --run-id* --metric-key* · --page-token --max-results
list-artifactsreadopt --run-id* · --path --page-token
list-logged-modelsread (POST)body ListLoggedModelsDetails: experiment_ids[] · filter · max_results · order_by[] (each LoggedModelOrder {field_name*,ascending}) · page_token
create-experiment-runyesbody CreateExperimentRunDetails: experiment_id · run_name · start_time · tags[] (ExperimentRunTag {key,value})
update-experiment-runyesbody UpdateExperimentRunDetails: run_id* · end_time · run_name · status (RUNNING|SCHEDULED|FINISHED|FAILED|KILLED|INTERNAL_ERROR)
delete-experiment-runyesbody DeleteExperimentRunDetails: run_id*
restore-experiment-runyesbody RestoreExperimentRunDetails: run_id*
log-experiment-run-metricyesbody LogExperimentRunMetricDetails: key* run_id* timestamp* value* · step
log-experiment-run-paramyesbody LogExperimentRunParamDetails: key* run_id* value*
log-experiment-run-modelyesbody LogExperimentRunModelDetails: model_json* run_id*
log-experiment-run-inputsyesbody LogExperimentRunInputsDetails: run_id* · dataset_inputs[] (DatasetInput {dataset*,tags[]}; Dataset {digest*,name*,source*,source_type*,profile,schema}; InputTag {key*,value*})
log-experiment-run-batchyesbody LogExperimentRunBatchDetails: run_id* · metrics[] (ExperimentRunMetric {key,value,timestamp,step}) · params[] (ExperimentRunParam {key,value}) · tags[] (ExperimentRunTag {key,value})

Experiment / run tags (<DLID> <WS>)

CommandMutates?Body (required *)
set-experiment-tagyesSetExperimentTagDetails: experiment_id* key* value*
delete-experiment-tagyesDeleteExperimentTagDetails: experiment_id* key*
update-experiment-tagsyesUpdateExperimentTagsDetails: experiment_id* · set_tags[] (ExperimentTag {key*,value}) · delete_tags[] (ExperimentTagKey {key*})
set-experiment-run-tagyesSetExperimentRunTagDetails: run_id* key* value*
delete-experiment-run-tagyesDeleteExperimentRunTagDetails: run_id* key*
update-experiment-run-tagsyesUpdateExperimentRunTagsDetails: run_id* · set_tags[] (ExperimentRunTag {key,value}) · delete_tags[] (ExperimentRunTagKey {key*})

Registered models (<DLID> only — no <WS>)

CommandMutates?Body / options (required *)
list-registered-modelsreadopt --filter --max-results (default 100, max 1000) --page-token --order-by
get-registered-modelreadopt --name*
create-registered-modelyesbody CreateRegisteredModelDetails: name* · description · deployment_job_id · tags[] (RegisteredModelTag {key,value})
update-registered-modelyesbody UpdateRegisteredModelDetails: name* · description · deployment_job_id
rename-registered-modelyesbody RenameRegisteredModelDetails: name* · new_name
delete-registered-modelyesbody DeleteRegisteredModelDetails: name*
set-registered-model-tagyesbody SetRegisteredModelTagDetails: name* key* value*
delete-registered-model-tagyesbody DeleteRegisteredModelTagDetails: name* key*
update-registered-model-tagsyesbody UpdateRegisteredModelTagsDetails: name* · set_tags[] (RegisteredModelTag {key,value}) · delete_tags[] (RegisteredModelTagKey {key*})

Model versions (<DLID> only — except create-workspace-model-version which is <DLID> <WS>)

CommandMutates?Body / options (required *)
list-model-versionsreadopt --filter --max-results --page-token --order-by
get-model-versionreadopt --name* --version*
create-model-versionyesbody CreateModelVersionDetails: name* source* · description · model_id · run_id · run_link · tags[] (ModelVersionTag {key,value})
create-workspace-model-versionyes<DLID> <WS> · body CreateModelVersionDetails (same fields as above)
update-model-versionyesbody UpdateModelVersionDetails: name* version* · description
delete-model-versionyesbody DeleteModelVersionDetails: name* version*
transition-model-version-stageyes (impactful)body TransitionModelVersionStageDetails: name* version* stage* archive_existing_versions* — stage is a free string in the SDK (no enum constants); confirm valid stage names live before relying on them
set-model-version-tagyesbody SetModelVersionTagDetails: name* version* key* value*
delete-model-version-tagyesbody DeleteModelVersionTagDetails: name* version* key*
update-model-version-tagsyesbody UpdateModelVersionTagsDetails: name* version* · set_tags[] (ModelVersionTag {key,value}) · delete_tags[] (ModelVersionTagKey {key*})

Timestamps (start_time/end_time/metric timestamp) are Unix epoch milliseconds (per SDK field docs). The 13 tag commands are: set-/delete-/update- × {experiment, experiment-run, model-version, registered-model} tag(s) — note update-*-tags (plural) takes batched set_tags/delete_tags arrays.

Fallback (no CLI) — REST endpoints (workspace-scoped MLflow; Preview, try 20260430 → 20240831)

Base: …/workspaces/{ws}/mlops/api/2.0/mlflow/…

  • Experiments: experiments/create|search|get
  • Runs: runs/create|update|search|get · runs/log-metric|log-parameter|log-batch
  • Registered models: registered-models/create|get|search
  • Model versions: model-versions/create|get|search|transition-stage

Guardrails

  • Stage transitions (e.g. to Production) are impactful — confirm.
  • Preview surface may shift — cite live status; don't assert unverified endpoints.

References

  • references/aidp-cli-map.md (skill → CLI map) · references/payloads.md (persist + confirm bodies) · references/oci-raw-request.md (base URL + auth ladder) · references/no-mcp-rest-map.md (control-plane runs MCP-free) · references/rest-endpoint-map.md (record verified MLflow paths here) · pairs with aidp-notebooks for the training run itself

Source and attribution

Source:oracle-samples/oracle-aidp-samplesinai/claude-code-plugins/oracle-ai-data-platform-workbench-engineer-agent/skills/aidp-mlopsat commit90b42d6

License: No license

Content belongs to its original authors. SourceWeft indexes it from a public repository.

Report or request removal

More from oracle-samples/oracle-aidp-samples

Aidp Workspace Admin

oracle-samples

Provision and inspect AIDP DataLake instances and workspaces, including private-network workspaces attached to a customer VCN/subnet. Use when the user wants to create/list/get a workspace or DataLake instance, set up a new (e.g. private) AIDP environment, or replicate a customer setup. Create/delete are guarded — confirm before any provisioning.

Awaiting classificationOct 8, 2026

Aidp Volumes

oracle-samples

Work with AIDP volumes — list volumes, browse files inside a volume, upload/download via the PAR flow, and create directories. Use when the user mentions volumes, needs to stage large/binary files, or move data in/out of a volume (distinct from the workspace filesystem). Control-plane via the official `aidp` CLI.

Awaiting classificationOct 8, 2026

Aidp Verified Queries

oracle-samples

Maintains a repository of validated question-to-Spark-SQL pairs so an agent reuses trusted SQL before writing new queries.

Data & AnalyticsOct 8, 2026

Aidp User Settings

oracle-samples

Manage AIDP DataLake user settings and preferences via the aidp CLI or oci raw-request fallback.

Productivity & WorkflowOct 8, 2026

Aidp Spark Optimization

oracle-samples

Guides Apache Spark 3.5.0 performance tuning: partitions, shuffle, joins, skew, memory, file layout, AQE and Delta Lake.

Data & AnalyticsOct 8, 2026

Aidp Semantic Model

oracle-samples

Maintains a .aidp/semantic.md business-meaning layer defining metrics, joins, synonyms and value dictionaries for NL-to-SQL grounding.

Data & AnalyticsOct 8, 2026

More in DevOps & Cloud

M5 Onboard

anthropics

Featured

Provisions M5Stack ESP32 boards by detecting them on USB, flashing UIFlow 2.0 firmware, and installing a MicroPython app bundle.

DevOps & Cloud37Kupdated today

Google Agents Cli Publish

google

Featured

Guides publishing and registering agents with Gemini Enterprise and Agent Registry using the agents-cli publish command.

DevOps & CloudOct 8, 2026

Spanner Basics

google

Featured

Guides Google Cloud Spanner administration, schema design, querying and performance diagnosis.

DevOps & CloudOct 8, 2026

Secops Cases

google

Featured

Manages Google Security Operations SOAR cases across their lifecycle via MCP tools.

DevOps & CloudOct 8, 2026

Managed Airflow Migrations

google

Featured

Guides migration of Apache Airflow DAGs to Airflow 2.11.1 or Airflow 3 in Managed Service for Apache Airflow.

DevOps & CloudOct 8, 2026

Managed Airflow Dag Troubleshooting

google

Featured

Guides deterministic troubleshooting of failed Managed Service for Apache Airflow DAG runs and task instances using gcloud commands.

DevOps & CloudOct 8, 2026