Added new examples for JavaScript (#953)

SilasMarvin · web-flow · commit 2521a4362013 · 2023-08-25T12:57:02.000-07:00
diff --git a/pgml-sdks/rust/pgml/javascript/examples/README.md b/pgml-sdks/rust/pgml/javascript/examples/README.md
@@ -1,7 +1,13 @@
-## Javascript Examples
+## Examples
 
-Here we have a set of examples of different use cases of the pgml javascript SDK.
+### [Semantic Search](./semantic_search.js)
+This is a basic example to perform semantic search on a collection of documents. Embeddings are created using `intfloat/e5-small` model. The results are semantically similar documemts to the query. Finally, the collection is archived.
 
-## Examples:
+### [Question Answering](./question_answering.js)
+This is an example to find documents relevant to a question from the collection of documents. The query is passed to vector search to retrieve documents that match closely in the embeddings space. A score is returned with each of the search result.
 
-1. [Getting Started](./getting-started/) - Simple project that uses the pgml SDK to create a collection, add a pipeline, upsert documents, and run a vector search on the collection.
+### [Question Answering using Instructore Model](./question_answering_instructor.js)
+In this example, we will use `hknlp/instructor-base` model to build text embeddings instead of the default `intfloat/e5-small` model.
+
+### [Extractive Question Answering](./extractive_question_answering.js)
+In this example, we will show how to use `vector_recall` result as a `context` to a HuggingFace question answering model. We will use `Builtins.transform()` to run the model on the database.
diff --git a/pgml-sdks/rust/pgml/javascript/examples/extractive_question_answering.js b/pgml-sdks/rust/pgml/javascript/examples/extractive_question_answering.js
@@ -0,0 +1,62 @@
+const pgml = require("pgml");
+require("dotenv").config();
+
+pgml.js_init_logger();
+
+const main = async () => {
+  // Initialize the collection
+  const collection = pgml.newCollection("my_javascript_eqa_collection_2");
+
+  // Add a pipeline
+  const model = pgml.newModel();
+  const splitter = pgml.newSplitter();
+  const pipeline = pgml.newPipeline(
+    "my_javascript_eqa_pipeline_1",
+    model,
+    splitter,
+  );
+  await collection.add_pipeline(pipeline);
+
+  // Upsert documents, these documents are automatically split into chunks and embedded by our pipeline
+  const documents = [
+    {
+      id: "Document One",
+      text: "PostgresML is the best tool for machine learning applications!",
+    },
+    {
+      id: "Document Two",
+      text: "PostgresML is open source and available to everyone!",
+    },
+  ];
+  await collection.upsert_documents(documents);
+
+  const query = "What is the best tool for machine learning?";
+
+  // Perform vector search
+  const queryResults = await collection
+    .query()
+    .vector_recall(query, pipeline)
+    .limit(1)
+    .fetch_all();
+
+  // Construct context from results
+  const context = queryResults
+    .map((result) => {
+      return result[1];
+    })
+    .join("\n");
+
+  // Query for answer
+  const builtins = pgml.newBuiltins();
+  const answer = await builtins.transform("question-answering", [
+    JSON.stringify({ question: query, context: context }),
+  ]);
+
+  // Archive the collection
+  await collection.archive();
+  return answer;
+};
+
+main().then((results) => {
+  console.log("Question answer: \n", results);
+});
diff --git a/pgml-sdks/rust/pgml/javascript/examples/getting-started/README.md b/pgml-sdks/rust/pgml/javascript/examples/getting-started/README.md
diff --git a/pgml-sdks/rust/pgml/javascript/examples/package-lock.json b/pgml-sdks/rust/pgml/javascript/examples/package-lock.json
diff --git a/pgml-sdks/rust/pgml/javascript/examples/package.json b/pgml-sdks/rust/pgml/javascript/examples/package.json
diff --git a/pgml-sdks/rust/pgml/javascript/examples/question_answering.js b/pgml-sdks/rust/pgml/javascript/examples/question_answering.js
@@ -0,0 +1,55 @@
+const pgml = require("pgml");
+require("dotenv").config();
+
+const main = async () => {
+  // Initialize the collection
+  const collection = pgml.newCollection("my_javascript_qa_collection");
+
+  // Add a pipeline
+  const model = pgml.newModel();
+  const splitter = pgml.newSplitter();
+  const pipeline = pgml.newPipeline(
+    "my_javascript_qa_pipeline",
+    model,
+    splitter,
+  );
+  await collection.add_pipeline(pipeline);
+
+  // Upsert documents, these documents are automatically split into chunks and embedded by our pipeline
+  const documents = [
+    {
+      id: "Document One",
+      text: "PostgresML is the best tool for machine learning applications!",
+    },
+    {
+      id: "Document Two",
+      text: "PostgresML is open source and available to everyone!",
+    },
+  ];
+  await collection.upsert_documents(documents);
+
+  // Perform vector search
+  const queryResults = await collection
+    .query()
+    .vector_recall("What is the best tool for machine learning?", pipeline)
+    .limit(1)
+    .fetch_all();
+
+  // Convert the results to an array of objects
+  const results = queryResults.map((result) => {
+    const [similarity, text, metadata] = result;
+    return {
+      similarity,
+      text,
+      metadata,
+    };
+  });
+
+  // Archive the collection
+  await collection.archive();
+  return results;
+};
+
+main().then((results) => {
+  console.log("Vector search Results: \n", results);
+});
diff --git a/pgml-sdks/rust/pgml/javascript/examples/question_answering_instructor.js b/pgml-sdks/rust/pgml/javascript/examples/question_answering_instructor.js
@@ -0,0 +1,60 @@
+const pgml = require("pgml");
+require("dotenv").config();
+
+const main = async () => {
+  // Initialize the collection
+  const collection = pgml.newCollection("my_javascript_qai_collection");
+
+  // Add a pipeline
+  const model = pgml.newModel("hkunlp/instructor-base", "pgml", {
+    instruction: "Represent the Wikipedia document for retrieval: ",
+  });
+  const splitter = pgml.newSplitter();
+  const pipeline = pgml.newPipeline(
+    "my_javascript_qai_pipeline",
+    model,
+    splitter,
+  );
+  await collection.add_pipeline(pipeline);
+
+  // Upsert documents, these documents are automatically split into chunks and embedded by our pipeline
+  const documents = [
+    {
+      id: "Document One",
+      text: "PostgresML is the best tool for machine learning applications!",
+    },
+    {
+      id: "Document Two",
+      text: "PostgresML is open source and available to everyone!",
+    },
+  ];
+  await collection.upsert_documents(documents);
+
+  // Perform vector search
+  const queryResults = await collection
+    .query()
+    .vector_recall("What is the best tool for machine learning?", pipeline, {
+      instruction:
+        "Represent the Wikipedia question for retrieving supporting documents: ",
+    })
+    .limit(1)
+    .fetch_all();
+
+  // Convert the results to an array of objects
+  const results = queryResults.map((result) => {
+    const [similarity, text, metadata] = result;
+    return {
+      similarity,
+      text,
+      metadata,
+    };
+  });
+
+  // Archive the collection
+  await collection.archive();
+  return results;
+};
+
+main().then((results) => {
+  console.log("Vector search Results: \n", results);
+});
diff --git a/pgml-sdks/rust/pgml/javascript/examples/semantic_search.js b/pgml-sdks/rust/pgml/javascript/examples/semantic_search.js
@@ -27,7 +27,10 @@ const main = async () => {
   // Perform vector search
   const queryResults = await collection
     .query()
-    .vector_recall("Some user query that will match document one first", pipeline)
+    .vector_recall(
+      "Some user query that will match document one first",
+      pipeline,
+    )
     .limit(2)
     .fetch_all();
 
@@ -41,6 +44,7 @@ const main = async () => {
     };
   });
 
+  // Archive the collection
   await collection.archive();
   return results;
 };
diff --git a/pgml-sdks/rust/pgml/python/examples/README.md b/pgml-sdks/rust/pgml/python/examples/README.md
@@ -1,7 +1,7 @@
 ## Examples
 
 ### [Semantic Search](./semantic_search.py)
-This is a basic example to perform semantic search on a collection of documents. It loads the Quora dataset, creates a collection in a PostgreSQL database, upserts documents, generates chunks and embeddings, and then performs a vector search on a query. Embeddings are created using `intfloat/e5-small` model. The results are are semantically similar documemts to the query. Finally, the collection is archived.
+This is a basic example to perform semantic search on a collection of documents. It loads the Quora dataset, creates a collection in a PostgreSQL database, upserts documents, generates chunks and embeddings, and then performs a vector search on a query. Embeddings are created using `intfloat/e5-small` model. The results are semantically similar documemts to the query. Finally, the collection is archived.
 
 ### [Question Answering](./question_answering.py)
 This is an example to find documents relevant to a question from the collection of documents. It loads the Stanford Question Answering Dataset (SQuAD) into the database, generates chunks and embeddings. Query is passed to vector search to retrieve documents that match closely in the embeddings space. A score is returned with each of the search result.
