Elastic Stack (ELK Stack) ml trained model API

The ml trained model API from Elastic Stack (ELK Stack) — 12 operation(s) for ml trained model.

Documentation GitHub OpenAPI

Business capability

Artificial Intelligence Management BC-610.60

Operations 15

Each operation below carries the questions people ask an LLM about it and the instructions they give an agent to run it. Generated by API Evangelist overlay

POST /_ml/trained_models/{model_id}/deployment/cache/_clear Clear a trained model deployment's inference cache · Clear trained model deployment cache #

Ask an LLM

“Can I clear the inference cache of a deployed model without restarting the deployment?”

“How do I flush cached model responses on every node a model is assigned to?”

Tell an agent

Clear the deployment inference cache for model {model_id}.

Flush cached responses for deployed model {model_id} on all nodes.

GET /_ml/trained_models/{model_id} Get trained model configs by ID · Get trained model configuration info #

Ask an LLM

“What inference config and metadata does a particular trained model have?”

“Can I include the decompressed model definition when fetching a trained model?”

Tell an agent

Show the configuration of trained model {model_id}.

Get trained model {model_id} including {include}.

PUT /_ml/trained_models/{model_id} Upload a trained model · Create a trained model #

Ask an LLM

“How do I bring in a model that wasn't created by data frame analytics?”

“Can I tag a trained model and give it a description when I create it?”

Tell an agent

Create trained model {model_id} with inference config {inference_config}.

Upload model {model_id} of type {model_type} tagged {tags}.

DELETE /_ml/trained_models/{model_id} Delete an unreferenced trained model · Delete an unreferenced trained model #

Ask an LLM

“How do I delete a trained model that no ingest pipeline uses anymore?”

“Can I force-delete a trained model?”

Tell an agent destructive · confirm first

Delete trained model {model_id}.

Force delete the trained model {model_id} within {timeout}.

PUT /_ml/trained_models/{model_id}/model_aliases/{model_alias} Create or reassign a trained model alias · Create or update a trained model alias #

Ask an LLM

“Can I give a trained model a friendly alias to use in inference processors?”

“How do I move an alias from an old model to a newly trained one?”

Tell an agent

Point alias {model_alias} at trained model {model_id}.

Reassign alias {model_alias} to model {model_id}.

DELETE /_ml/trained_models/{model_id}/model_aliases/{model_alias} Delete a trained model alias · Delete a trained model alias #

Ask an LLM

“How do I remove an alias from a trained model?”

“What happens if I delete an alias that points to a different model than the one I name?”

Tell an agent destructive · confirm first

Delete alias {model_alias} from trained model {model_id}.

Remove the model alias {model_alias} that refers to {model_id}.

GET /_ml/trained_models List all trained models · Get trained model configuration info #

Ask an LLM

“What trained models are available on my cluster?”

“Can I list only the trained models with certain tags?”

Tell an agent

List all trained models.

List trained models tagged {tags}.

GET /_ml/trained_models/{model_id}/_stats Get usage stats for specific trained models · Get trained models usage info #

Ask an LLM

“How many inference calls has a specific trained model served?”

“Can I get deployment and ingest stats for a few models using a wildcard?”

Tell an agent

Show usage stats for trained model {model_id}.

Get inference and deployment statistics for models matching {model_id}.

GET /_ml/trained_models/_stats Get usage stats for all trained models · Get trained models usage info #

Ask an LLM

“Which of my trained models are actually being used?”

“Can I see usage statistics across every trained model on the cluster?”

Tell an agent

Show usage stats for all trained models.

Get stats for the first {size} trained models.

POST /_ml/trained_models/{model_id}/_infer Run a trained model on documents · Evaluate a trained model #

Ask an LLM

“Can I test a deployed trained model by sending it a few documents directly?”

“How do I override the inference config when evaluating a trained model on sample docs?”

Tell an agent

Run trained model {model_id} on these documents: {docs}.

Infer {docs} with model {model_id} using config {inference_config}.

PUT /_ml/trained_models/{model_id}/definition/{part} Upload one part of a model definition · Create part of a trained model definition #

Ask an LLM

“How do I upload a large model definition in several chunks?”

“What do I need to tell Elasticsearch about the total parts and total length when uploading a definition piece?”

Tell an agent

Upload part {part} of the definition for model {model_id}.

Store definition chunk {part} of {total_parts} for model {model_id}.

PUT /_ml/trained_models/{model_id}/vocabulary Upload an NLP model's vocabulary · Create a trained model vocabulary #

Ask an LLM

“How do I add the tokenizer vocabulary for an NLP model I uploaded?”

“Can I include BPE merges when storing a trained model vocabulary?”

Tell an agent

Upload vocabulary {vocabulary} for NLP model {model_id}.

Store the vocabulary and merges {merges} for model {model_id}.

POST /_ml/trained_models/{model_id}/deployment/_start Deploy a trained model to ML nodes · Start a trained model deployment #

Ask an LLM

“How do I deploy a trained model so it can serve inference?”

“Can I choose the number of allocations and threads per allocation when starting a deployment?”

Tell an agent

Start a deployment of trained model {model_id}.

Deploy {model_id} with {number_of_allocations} allocations and {threads_per_allocation} threads each.

POST /_ml/trained_models/{model_id}/deployment/_stop Stop a trained model deployment · Stop a trained model deployment #

Ask an LLM

“How do I undeploy a trained model to free up ML node memory?”

“Can I force-stop a model deployment that ingest pipelines still reference?”

Tell an agent

Stop the deployment of trained model {model_id}.

Force stop the deployed model {model_id}.

POST /_ml/trained_models/{model_id}/deployment/_update Scale a trained model deployment · Update a trained model deployment #

Ask an LLM

“Can I change the number of allocations on a model that's already deployed?”

“How do I turn on adaptive allocations for a running model deployment?”

Tell an agent

Scale the deployment of {model_id} to {number_of_allocations} allocations.

Enable adaptive allocations {adaptive_allocations} on deployed model {model_id}.

Elastic Stack (ELK Stack) ml trained model API

Operations 15

Documentation

Specifications

Other Resources

Work with this as data

MCP server

Tools for apis

Call it yourself

Get an API key

OpenAPI Specification