Change openSearchAPI file

New module accepts no arguments in its default function

OpenSearch Client is now imported from architect-plugin-search

openSearchAPI.js

@@ -0,0 +1,277 @@
+/*
+This file contains API calls to the OpenSearch cluster to register and deploy an ml model to the opensearch cluster.
+It also creates a neural ingest pipeline to allow for ingesting of documents into a knn index.
+*/
+import { search as getSearch } from 'architect-functions-search'
+
+export default async function () {
+  const client = await getSearch()
+
+  //Set cluster settings
+  const cluster_settings_request = {
+    method: 'PUT',
+    path: '/_cluster/settings',
+    body: {
+      persistent: {
+        plugins: {
+          ml_commons: {
+            only_run_on_ml_node: 'false',
+            model_access_control_enabled: 'true',
+            native_memory_threshold: '99',
+          },
+        },
+      },
+    },
+  }
+
+  try {
+    const resp = await client.transport.request(cluster_settings_request)
+
+    if (resp && resp.statusCode == 200) {
+      console.log('Updated ML-related cluster settings.')
+    } else {
+      console.log(
+        'Error. Could not update cluster settings. Returned with response: ',
+        resp
+      )
+      return
+    }
+  } catch (e) {
+    console.log('Error: ', e)
+    return
+  }
+
+  //Register model group
+  const register_model_group_request = {
+    method: 'POST',
+    path: '/_plugins/_ml/model_groups/_register',
+    body: {
+      name: 'NLP_model_group',
+      description: 'A model group for NLP models',
+    },
+  }
+
+  let model_group_id
+  try {
+    const resp = await client.transport.request(register_model_group_request)
+
+    if (resp && resp.statusCode == 200) {
+      model_group_id = resp.body.model_group_id
+      console.log(`Registered model group with id: ${model_group_id}`)
+    } else {
+      console.log(
+        'Error. Could not register model group. Returned with response: ',
+        resp
+      )
+      return
+    }
+  } catch (e) {
+    console.log('Error: ', e)
+    return
+  }
+
+  //Register model to model group
+  const register_model_request = {
+    method: 'POST',
+    path: '/_plugins/_ml/models/_register',
+    body: {
+      name: 'huggingface/sentence-transformers/all-MiniLM-L6-v2',
+      version: '1.0.1',
+      model_group_id,
+      model_format: 'TORCH_SCRIPT',
+    },
+  }
+
+  let task_id
+  try {
+    const resp = await client.transport.request(register_model_request)
+
+    if (resp && resp.statusCode == 200) {
+      task_id = resp.body.task_id
+      console.log('Registering model to model group. Task ID is: ', task_id)
+    } else {
+      console.log(
+        'Error. Could not register model to model group. Returned with response: ',
+        resp
+      )
+      return
+    }
+  } catch (e) {
+    console.log('Error: ', e)
+    return
+  }
+
+  //Check status of model registration
+  const check_model_registration_request = {
+    method: 'GET',
+    path: `_plugins/_ml/tasks/${task_id}`,
+  }
+
+  let model_id = ''
+  try {
+    let resp
+    do {
+      resp = await client.transport.request(check_model_registration_request)
+
+      if (resp && resp.statusCode == 200) {
+        console.log('Checking model registration status...')
+
+        if (resp.body.state === 'COMPLETED') {
+          model_id = resp.body.model_id
+          console.log('Model registration completed. Model ID: ', model_id)
+        } else {
+          await new Promise((resolve) => setTimeout(resolve, 2000)) // Wait for 2 seconds before checking again
+        }
+      } else {
+        console.log(
+          'Error. Could not check model registration status. Returned with response: ',
+          resp
+        )
+        return
+      }
+    } while (resp.body.state !== 'COMPLETED')
+  } catch (e) {
+    console.log('Error: ', e)
+    return
+  }
+
+  //Deploy model
+  const deploy_model_request = {
+    method: 'POST',
+    path: `/_plugins/_ml/models/${model_id}/_deploy`,
+  }
+
+  try {
+    const resp = await client.transport.request(deploy_model_request)
+
+    if (resp && resp.statusCode == 200) {
+      task_id = resp.body.task_id
+      console.log('Deploying model. Task ID is: ', task_id)
+    } else {
+      console.log(
+        'Error. Could not deploy model. Returned with response: ',
+        resp
+      )
+      return
+    }
+  } catch (e) {
+    console.log('Error: ', e)
+    return
+  }
+
+  //Check status of model deployment
+  const check_model_deployment_request = {
+    method: 'GET',
+    path: `_plugins/_ml/tasks/${task_id}`,
+  }
+
+  try {
+    let resp
+    do {
+      resp = await client.transport.request(check_model_deployment_request)
+
+      if (resp && resp.statusCode == 200) {
+        console.log('Checking model deployment status...')
+
+        if (resp.body.state === 'COMPLETED') {
+          model_id = resp.body.model_id
+          console.log('Model deployment completed. Model ID: ', model_id)
+        } else {
+          await new Promise((resolve) => setTimeout(resolve, 2000)) // Wait for 2 seconds before checking again
+        }
+      } else {
+        console.log(
+          'Error. Could not check model deployment status. Returned with response: ',
+          resp
+        )
+        return
+      }
+    } while (resp.body.state !== 'COMPLETED')
+  } catch (e) {
+    console.log('Error: ', e)
+    return
+  }
+
+  //Create neural ingest pipeline
+  const pipeline_name = 'nlp-ingest-pipeline'
+  const create_ingest_pipeline_request = {
+    method: 'PUT',
+    path: `/_ingest/pipeline/${pipeline_name}`,
+    body: {
+      description: 'An NLP ingest pipeline',
+      processors: [
+        {
+          text_embedding: {
+            model_id,
+            field_map: {
+              body: 'circular_embedding',
+            },
+          },
+        },
+      ],
+    },
+  }
+
+  try {
+    const resp = await client.transport.request(create_ingest_pipeline_request)
+
+    if (resp && resp.statusCode == 200) {
+      console.log('Successfully created neural ingest pipeline.')
+    } else {
+      console.log(
+        'Error. Could not create neural ingest pipeline. Returned with response: ',
+        resp
+      )
+      return
+    }
+  } catch (e) {
+    console.log('Error: ', e)
+    return
+  }
+
+  //Create knn index
+  await client.indices.create({
+    index: 'circulars',
+    body: {
+      settings: {
+        'index.knn': true,
+        default_pipeline: pipeline_name,
+      },
+      mappings: {
+        properties: {
+          subject: {
+            type: 'text',
+          },
+          submittedHow: {
+            type: 'text',
+          },
+          bibcode: {
+            type: 'text',
+          },
+          createdOn: {
+            type: 'long',
+          },
+          circularId: {
+            type: 'integer',
+          },
+          submitter: {
+            type: 'text',
+          },
+          circular_embedding: {
+            type: 'knn_vector',
+            dimension: 384,
+            method: {
+              engine: 'lucene',
+              space_type: 'l2',
+              name: 'hnsw',
+              parameters: {},
+            },
+          },
+          body: {
+            type: 'text',
+          },
+        },
+      },
+    },
+  })
+}

openSearchApi.js

@@ -2,10 +2,10 @@
 This file contains API calls to the OpenSearch cluster to register and deploy an ml model to the opensearch cluster.
 It also creates a neural ingest pipeline to allow for ingesting of documents into a knn index.
 */
-import { Client } from '@opensearch-project/opensearch'
+import { search as getSearch } from 'architect-functions-search'
 
-export default async function (opts) {
-  const client = new Client(opts)
+export default async function () {
+  const client = await getSearch()
 
   //Set cluster settings
   const cluster_settings_request = {