Merge pull request #2723 from continuedev/pe/jb-codebase-transformers

bugfix(jb): no embeddings w/ transformers.js

.gitignore

@@ -166,3 +166,5 @@ extensions/intellij/.idea/**
 **/.idea/workspace.xml
 **/.idea/usage.statistics.xml
 **/.idea/shelf/
+
+extensions/intellij/bin

.vscode/settings.json

@@ -17,7 +17,8 @@
     "**/core/vendor/**": true,
     "**/gui/dist": true,
     "**/extensions/vscode/gui/**": true,
-    "**/extensions/vscode/out/**": true
+    "**/extensions/vscode/out/**": true,
+    "**/extensions/intellij/build/**": true
   },
   "eslint.workingDirectories": ["./core"]
 }

core/context/retrieval/pipelines/BaseRetrievalPipeline.ts

@@ -14,13 +14,13 @@ export interface RetrievalPipelineOptions {
   llm: ILLM;
   config: ContinueConfig;
   ide: IDE;
-
   input: string;
   nRetrieve: number;
   nFinal: number;
   tags: BranchAndDir[];
   pathSep: string;
   filterDirectory?: string;
+  includeEmbeddings?: boolean; // Used to handle JB w/o an embeddings model
 }
 
 export interface IRetrievalPipeline {
@@ -29,6 +29,7 @@ export interface IRetrievalPipeline {
 
 export default class BaseRetrievalPipeline implements IRetrievalPipeline {
   private lanceDbIndex: LanceDbIndex;
+
   constructor(protected readonly options: RetrievalPipelineOptions) {
     this.lanceDbIndex = new LanceDbIndex(
       options.config.embeddingsProvider,

core/context/retrieval/pipelines/NoRerankerRetrievalPipeline.ts

@@ -5,7 +5,7 @@ import BaseRetrievalPipeline from "./BaseRetrievalPipeline.js";
 
 export default class NoRerankerRetrievalPipeline extends BaseRetrievalPipeline {
   async run(): Promise<Chunk[]> {
-    const { input, nFinal, filterDirectory } = this.options;
+    const { input, nFinal, filterDirectory, includeEmbeddings } = this.options;
 
     // We give 1/4 weight to recently edited files, 1/4 to full text search,
     // and the remaining 1/2 to embeddings
@@ -17,10 +17,9 @@ export default class NoRerankerRetrievalPipeline extends BaseRetrievalPipeline {
 
     const ftsChunks = await this.retrieveFts(input, ftsNFinal);
 
-    const embeddingsChunks = await this.retrieveEmbeddings(
-      input,
-      embeddingsNFinal,
-    );
+    const embeddingsChunks = includeEmbeddings
+      ? await this.retrieveEmbeddings(input, embeddingsNFinal)
+      : [];
 
     const recentlyEditedFilesChunks =
       await this.retrieveAndChunkRecentlyEditedFiles(recentlyEditedNFinal);

core/context/retrieval/pipelines/RerankerRetrievalPipeline.ts

@@ -6,12 +6,15 @@ import BaseRetrievalPipeline from "./BaseRetrievalPipeline.js";
 
 export default class RerankerRetrievalPipeline extends BaseRetrievalPipeline {
   private async _retrieveInitial(): Promise<Chunk[]> {
-    const { input, nRetrieve, filterDirectory } = this.options;
+    const { input, nRetrieve, filterDirectory, includeEmbeddings } =
+      this.options;
 
     let retrievalResults: Chunk[] = [];
 
     const ftsChunks = await this.retrieveFts(input, nRetrieve);
-    const embeddingsChunks = await this.retrieveEmbeddings(input, nRetrieve);
+    const embeddingsChunks = includeEmbeddings
+      ? await this.retrieveEmbeddings(input, nRetrieve)
+      : [];
     const recentlyEditedFilesChunks =
       await this.retrieveAndChunkRecentlyEditedFiles(nRetrieve);
 

core/context/retrieval/retrieval.ts

@@ -1,8 +1,10 @@
 import path from "path";
+
 import { BranchAndDir, ContextItem, ContextProviderExtras } from "../../";
 import TransformersJsEmbeddingsProvider from "../../indexing/embeddings/TransformersJsEmbeddingsProvider";
 import { resolveRelativePathInWorkspace } from "../../util/ideUtils";
 import { INSTRUCTIONS_BASE_ITEM } from "../providers/utils";
+
 import { RetrievalPipelineOptions } from "./pipelines/BaseRetrievalPipeline";
 import NoRerankerRetrievalPipeline from "./pipelines/NoRerankerRetrievalPipeline";
 import RerankerRetrievalPipeline from "./pipelines/RerankerRetrievalPipeline";
@@ -21,15 +23,17 @@ export async function retrieveContextItemsFromEmbeddings(
   // transformers.js not supported in JetBrains IDEs right now
 
   const isJetBrainsAndTransformersJs =
-    extras.embeddingsProvider.id === TransformersJsEmbeddingsProvider.model &&
+    extras.embeddingsProvider.providerName ===
+      TransformersJsEmbeddingsProvider.providerName &&
     (await extras.ide.getIdeInfo()).ideType === "jetbrains";
 
   if (isJetBrainsAndTransformersJs) {
-    throw new Error(
-      "The 'transformers.js' context provider is not currently supported in JetBrains. " +
-        "For now, you can use Ollama to set up local embeddings, or use our 'free-trial' " +
-        "embeddings provider. See here to learn more: " +
-        "https://docs.continue.dev/walkthroughs/codebase-embeddings#embeddings-providers",
+    void extras.ide.showToast(
+      "warning",
+      "Codebase retrieval is limited when `embeddingsProvider` is empty or set to `transformers.js` in JetBrains. " +
+        "You can use Ollama to set up local embeddings, use our 'free-trial', " +
+        "or configure your own. See here to learn more: " +
+        "https://docs.continue.dev/customize/model-types/embeddings",
     );
   }
 
@@ -85,6 +89,7 @@ export async function retrieveContextItemsFromEmbeddings(
     input: extras.fullInput,
     llm: extras.llm,
     config: extras.config,
+    includeEmbeddings: !isJetBrainsAndTransformersJs,
   };
 
   const pipeline = new pipelineType(pipelineOptions);

docs/docs/customize/deep-dives/codebase.md

@@ -5,7 +5,11 @@ keywords: [talk, embeddings, reranker, codebase, experimental]
 
 # @Codebase
 
-Continue indexes your codebase so that it can later automatically pull in the most relevant context from throughout your workspace. This is done via a combination of embeddings-based retrieval and keyword search. By default, all embeddings are calculated locally with `all-MiniLM-L6-v2` and stored locally in `~/.continue/index`.
+Continue indexes your codebase so that it can later automatically pull in the most relevant context from throughout your workspace. This is done via a combination of embeddings-based retrieval and keyword search. By default, all embeddings are calculated locally using `transformers.js` and stored locally in `~/.continue/index`.
+
+:::info[`transformers.js cannot be used in JetBrains`]
+Currently, `transformers.js` cannot be used in JetBrains IDEs. However, you can select a differet embeddings model from [the list here](../model-types/embeddings.md).
+:::
 
 Currently, the codebase retrieval feature is available as the "codebase" and "folder" context providers. You can use them by typing `@Codebase` or `@Folder` in the input box, and then asking a question. The contents of the input box will be compared with the embeddings from the rest of the codebase (or folder) to determine relevant files.
 

docs/docs/customize/model-types/embeddings.md

@@ -9,6 +9,10 @@ An "embeddings model" is trained to convert a piece of text into a vector, which
 
 In Continue, embeddings are generated during indexing and then used by [@Codebase](../deep-dives/codebase.md) to perform similarity search over your codebase.
 
+:::info[Default model (VS Code only)]
+`transformers.js` is used as a default embeddings model in VS Code. In JetBrains, there currently is no default.
+:::
+
 ## Recommended embedding models
 
 If you have the ability to use any model, we recommend `voyage-code-2`, which is listed below along with the rest of the options for embeddings models.
@@ -35,7 +39,7 @@ See [here](../model-providers/top-level/ollama.md#embeddings-model) for instruct
 
 ### Transformers.js (currently VS Code only)
 
-[Transformers.js](https://huggingface.co/docs/transformers.js/index) is a JavaScript port of the popular [Transformers](https://huggingface.co/transformers/) library. It allows embeddings to be calculated entirely locally. The model used is `all-MiniLM-L6-v2`, which is shipped alongside the Continue extension and used as the default when you have not explicitly configured an embeddings provider.
+[Transformers.js](https://huggingface.co/docs/transformers.js/index) is a JavaScript port of the popular [Transformers](https://huggingface.co/transformers/) library. It allows embeddings to be calculated entirely locally. The model used is `all-MiniLM-L6-v2`, which is shipped alongside the Continue extension.
 
 ```json title="config.json"
 {
@@ -72,4 +76,4 @@ See [here](../model-providers/top-level/gemini.md#embeddings-model) for instruct
 
 ### Vertex
 
-See [here](../model-providers/more/vertex.md#embeddings-model) for instructions on how to use Vertex for embeddings.
+See [here](../model-providers/more/vertex.md#embeddings-model) for instructions on how to use Vertex for embeddings.

extensions/intellij/src/main/resources/config_schema.json

@@ -2425,8 +2425,8 @@
         },
         "embeddingsProvider": {
           "title": "Embeddings Provider",
-          "markdownDescription": "The method that will be used to generate codebase embeddings. The default is transformers.js, which will run locally in the browser. Learn about the other options [here](https://docs.continue.dev/features/codebase-embeddings#embeddings-providers).",
-          "x-intellij-html-description": "The method that will be used to generate codebase embeddings. The default is transformers.js, which will run locally in the browser. Learn about the other options <a href='https://docs.continue.dev/features/codebase-embeddings#embeddings-providers'>here</a>.",
+          "markdownDescription": "The method that will be used to generate codebase embeddings. The default is `transformers.js`, which will run locally in the browser. Learn about the other options [here](https://docs.continue.dev/features/codebase-embeddings#embeddings-providers).\n\n **Note**: `transformers.js` currently cannot be used in JetBrains.",
+          "x-intellij-html-description": "The method that will be used to generate codebase embeddings. The default is <code>transformers.js</code>, which will run locally in the browser. Learn about the other options <a href='https://docs.continue.dev/features/codebase-embeddings#embeddings-providers'>here</a>.<br><br><strong>Note</strong>: <code>transformers.js</code> currently cannot be used in JetBrains.",
           "type": "object",
           "properties": {
             "provider": {

extensions/vscode/config_schema.json

@@ -2425,8 +2425,8 @@
         },
         "embeddingsProvider": {
           "title": "Embeddings Provider",
-          "markdownDescription": "The method that will be used to generate codebase embeddings. The default is transformers.js, which will run locally in the browser. Learn about the other options [here](https://docs.continue.dev/features/codebase-embeddings#embeddings-providers).",
-          "x-intellij-html-description": "The method that will be used to generate codebase embeddings. The default is transformers.js, which will run locally in the browser. Learn about the other options <a href='https://docs.continue.dev/features/codebase-embeddings#embeddings-providers'>here</a>.",
+          "markdownDescription": "The method that will be used to generate codebase embeddings. The default is `transformers.js`, which will run locally in the browser. Learn about the other options [here](https://docs.continue.dev/features/codebase-embeddings#embeddings-providers).",
+          "x-intellij-html-description": "The method that will be used to generate codebase embeddings. The default is <code>transformers.js</code>, which will run locally in the browser. Learn about the other options <a href='https://docs.continue.dev/features/codebase-embeddings#embeddings-providers'>here</a>.<br><br><strong>Note</strong>: <code>transformers.js</code> currently cannot be used in JetBrains.",
           "type": "object",
           "properties": {
             "provider": {