lhcb-ntuple-wizard-test 1.1.7 → 1.1.9

package/dist/lib/worker.js

@@ -1,5 +1,11 @@
 "use strict";
 
+Object.defineProperty(exports, "__esModule", {
+  value: true
+});
+exports.calcCosSim = calcCosSim;
+exports.embed = embed;
+exports.readJsonFile = readJsonFile;
 var _transformers = require("@xenova/transformers");
 /*****************************************************************************\
 * (c) Copyright 2024 CERN for the benefit of the LHCb Collaboration           *

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "lhcb-ntuple-wizard-test",
-  "version": "1.1.7",
+  "version": "1.1.9",
   "description": "An application to access large-scale open data from LHCb",
   "url": "https://gitlab.cern.ch/lhcb-dpa/wp6-analysis-preservation-and-open-data/lhcb-ntuple-wizard-frontend/issues",
   "private": false,

package/dist/components/semantic.js

@@ -1,95 +0,0 @@
-"use strict";
-
-Object.defineProperty(exports, "__esModule", {
-  value: true
-});
-exports.embedCorpus = embedCorpus;
-exports.getSimilarity = getSimilarity;
-exports.loadModel = loadModel;
-var _transformers = require("@xenova/transformers");
-/*****************************************************************************\
-* (c) Copyright 2024 CERN for the benefit of the LHCb Collaboration           *
-*                                                                             *
-* This software is distributed under the terms of the GNU General Public      *
-* Licence version 3 (GPL Version 3), copied verbatim in the file "COPYING".   *
-*                                                                             *
-* In applying this licence, CERN does not waive the privileges and immunities *
-* granted to it by virtue of its status as an Intergovernmental Organization  *
-* or submit itself to any jurisdiction.                                       *
-\*****************************************************************************/
-
-// To always download the model from huggingface.com
-_transformers.env.allowLocalModels = false;
-_transformers.env.useBrowserCache = false;
-// Create a new worker
-const worker = new Worker(new URL("./worker.jsx", import.meta.url));
-/**
- * @type {function}
- * Promise resolve function for loading the model
- */
-let loadResolve;
-/**
- * @type {function}
- */
-let queryResolve;
-worker.onmessage = function (event) {
-  const message = event.data;
-  switch (message.type) {
-    case "progress":
-      if (message.progress.status === "ready") {
-        loadResolve();
-      }
-      break;
-    case "corpus":
-      // the corpus is embedded
-      queryResolve();
-      break;
-    case "result":
-      queryResolve(message.result);
-      break;
-  }
-};
-/**
- *
- * @param {string} modelname
- * Load the model with the provided model name
- * @returns
- */
-async function loadModel(modelname) {
-  worker.postMessage({
-    type: "init",
-    model: modelname
-  });
-  return new Promise(resolve => {
-    loadResolve = resolve;
-  });
-}
-/**
- * Passes corpus embedding from backend to the worker
- * @returns
- */
-async function embedCorpus(metadata) {
-  worker.postMessage({
-    type: "corpus",
-    kgdoc: metadata.metadata.kgdoc,
-    emb: metadata.metadata.embedding
-  });
-  return new Promise(resolve => {
-    queryResolve = resolve;
-  });
-}
-/**
- *
- * @param {string} query#
- * Calculate the similarity between the query and the corpus
- * @returns
- */
-async function getSimilarity(query) {
-  worker.postMessage({
-    type: "similarity",
-    query: query
-  });
-  return new Promise(resolve => {
-    queryResolve = resolve;
-  });
-}

package/dist/components/worker.js

@@ -1,196 +0,0 @@
-"use strict";
-
-var _transformers = require("@xenova/transformers");
-/*****************************************************************************\
-* (c) Copyright 2024 CERN for the benefit of the LHCb Collaboration           *
-*                                                                             *
-* This software is distributed under the terms of the GNU General Public      *
-* Licence version 3 (GPL Version 3), copied verbatim in the file "COPYING".   *
-*                                                                             *
-* In applying this licence, CERN does not waive the privileges and immunities *
-* granted to it by virtue of its status as an Intergovernmental Organization  *
-* or submit itself to any jurisdiction.                                       *
-\*****************************************************************************/
-
-// This is so that we will download the models from huggingface.co
-_transformers.env.allowLocalModels = false;
-_transformers.env.useBrowserCache = false;
-// store the embeddings in a dict
-let query_embedding;
-/**
- * @type {Object.<string, Array>}
- */
-let embedding_dict = {};
-/**
- * @type {pipeline}
- */
-let embedder;
-
-// these will hold the information to tie variable explanations to TupleTools
-let dataKeys;
-let paths;
-/**
- * @type {Array}
- * */
-let vars;
-/**
- * This function calculates the embedding and stores it in a dict.
- * If the text is already embedded, it will return the embedding from the dict.
- * @param {string} text
- * @param {bool} embedNew
- * @returns embeddingVector
- */
-async function embed(text) {
-  let embedNew = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : true;
-  if (text in embedding_dict) {
-    return embedding_dict[text];
-  }
-  const e0 = await embedder(text, {
-    pooling: "mean",
-    normalize: true
-  });
-  if (embedNew) {
-    embedding_dict[text] = e0.data;
-  }
-  return e0.data;
-}
-/**
- * This function calculates the cosine similarity between two embeddings.
- * @param {Array} corpus_embedding
- * @param {Array} query_embedding
- * @returns cosine similarity
- * */
-function calcCosSim(corpus_embedding, query_embedding) {
-  let dotProduct = 0;
-  let queryMag = 0;
-  let embMag = 0;
-  let loop_length = query_embedding.length;
-  // because the embeddings might have different dimensions
-  if (query_embedding.length > corpus_embedding.length) {
-    loop_length = corpus_embedding.length;
-  }
-  for (let i = 0; i < loop_length; i++) {
-    dotProduct += query_embedding[i] * corpus_embedding[i];
-    queryMag += query_embedding[i] * query_embedding[i];
-    embMag += corpus_embedding[i] * corpus_embedding[i];
-  }
-  const sim = dotProduct / (Math.sqrt(queryMag) * Math.sqrt(embMag));
-  return sim;
-}
-/**This is a helper function to read the nested json file containing the documentation.
- * @returns [vars, paths, dataKeys]
- */
-
-function readJsonFile(data) {
-  // go through nested object
-  vars = [];
-  paths = [];
-  const iterateObject = function (obj) {
-    let path = arguments.length > 1 && arguments[1] !== undefined ? arguments[1] : "";
-    Object.keys(obj).forEach(key => {
-      const currentPath = path ? `${path}.${key}` : key;
-      if (obj[key] instanceof Object) {
-        iterateObject(obj[key], currentPath);
-      } else {
-        vars.push(obj[key]);
-        paths.push(currentPath);
-      }
-    });
-  };
-
-  // go through data and put it into an array
-  const dataArray = Object.values(data);
-  dataKeys = Object.keys(data);
-  iterateObject(dataArray);
-  return [vars, paths, dataKeys];
-}
-
-// worker on message definition
-self.onmessage = async event => {
-  const message = event.data;
-  switch (message.type) {
-    case "init":
-      // load the model and reset the embedding dict
-      embedding_dict = {};
-      embedder = await (0, _transformers.pipeline)("feature-extraction", message.model, {
-        progress_callback: progress => {
-          self.postMessage({
-            type: "progress",
-            progress: progress
-          });
-        }
-      });
-      break;
-    case "corpus":
-      {
-        // embed the corpus or load the embeddings
-        readJsonFile(message.kgdoc); // this is so we can later match variables to tupletools
-
-        Object.keys(message.emb).forEach(key => {
-          let embarr = [];
-          for (const fl of message.emb[key].split(",")) {
-            embarr.push(fl.replace("[", "").replace("]", "").replace("np.float32", "").replace("(", "").replace(")", "").replace(" ", ""));
-          }
-          embedding_dict[key] = embarr;
-        });
-        self.postMessage({
-          type: "corpus"
-        });
-        break;
-      }
-    case "similarity":
-      {
-        //calculate query embedding and then calculate similarity
-        query_embedding = await embed(message.query, false);
-        let sim_dict = {};
-        Object.keys(embedding_dict).forEach(key => {
-          sim_dict[key] = calcCosSim(embedding_dict[key], query_embedding);
-        });
-        // sort by highest similarity (closest to 1)
-        const sortedSimDict = Object.fromEntries(Object.entries(sim_dict).sort((_ref, _ref2) => {
-          let [, a] = _ref;
-          let [, b] = _ref2;
-          return b - a;
-        }));
-        const top5 = Object.fromEntries(Object.entries(sortedSimDict).slice(0, 20));
-        let loki_result = [];
-        let ttool_result = [];
-        let result = {
-          loki: loki_result,
-          ttool: ttool_result
-        };
-        Object.keys(top5).forEach(key => {
-          if (key === "basic,charged: : ") {
-            return;
-          } // This is a workaround until we implement a fix in Analysis (and corresponding DaVinci release)
-          let description = key.substring(key.indexOf(":") + 2);
-          let variable = key.substring(0, key.indexOf(":") - 1);
-          let varpath = paths[vars.indexOf(description)];
-          let tupletool = dataKeys[varpath.split(".")[0]];
-          if (tupletool === "LoKi_functors") {
-            if (top5[key] > 0.86) {
-              if (loki_result.length < 5) {
-                loki_result.push([description, variable, tupletool]);
-              }
-            }
-          } else {
-            if (top5[key] > 0.75) {
-              if (ttool_result.length < 5) {
-                ttool_result.push([description, variable, tupletool]);
-              }
-            }
-          }
-        });
-        if (loki_result.length === 0) {
-          loki_result.push(["No matches found", "No matches found", "No matches found"]);
-        }
-        if (ttool_result.length === 0) {
-          ttool_result.push(["No matches found", "No matches found", "No matches found"]);
-        }
-        self.postMessage({
-          type: "result",
-          result: result
-        });
-      }
-  }
-};