Merge branch 'justinmilner1-MakeArrowsScroll' into preview

.gitignore

@@ -163,4 +163,7 @@ continue_server.dist
 Icon
 Icon?
 
-.continue
+.continue
+
+# Sandbox env default file
+extensions/vscode/manual-testing-sandbox/example.ts

CONTRIBUTING.md

@@ -79,7 +79,7 @@ nvm use
       1. The new VS Code window with the extension is referred to as the _Host VS Code_
       2. The window you started debugging from is referred to as the _Main VS Code_
 
-4. To package the extension, run `npm package` in the `extensions/vscode` directory. This will generate `extensions/vscode/build/continue-patch.vsix`, which you can install by right-clicking and selecting "Install Extension VSIX".
+4. To package the extension, run `npm run package` in the `extensions/vscode` directory. This will generate `extensions/vscode/build/continue-{VERSION}.vsix`, which you can install by right-clicking and selecting "Install Extension VSIX".
 
 ##### Debugging
 

core/.eslintrc.json

@@ -1,27 +1,30 @@
 {
-    "root": true,
-    "parser": "@typescript-eslint/parser",
-    "parserOptions": {
-      "ecmaVersion": 6,
-      "sourceType": "module"
-    },
-    "extends": ["plugin:require-extensions/recommended"],
-    "plugins": ["@typescript-eslint", "import", "require-extensions"],
-    "rules": {
-      "quotes": ["warn", "double", {}],
-      "import/extensions": ["error", "always", {
-          "js": "never",
-          "jsx": "never",
-          "ts": "never",
-          "tsx": "never"
-        }],
-        "@typescript-eslint/naming-convention": "warn",
-        "@typescript-eslint/semi": "warn",
-        "curly": "warn",
-        "eqeqeq": "warn",
-        "no-throw-literal": "warn",
-        "semi": "off"
-      },
-    "ignorePatterns": ["out", "dist", "**/*.d.ts"]
-  }
-
+  "root": true,
+  "parser": "@typescript-eslint/parser",
+  "parserOptions": {
+    "ecmaVersion": 6,
+    "sourceType": "module"
+  },
+  "extends": ["plugin:require-extensions/recommended"],
+  "plugins": ["@typescript-eslint", "import", "require-extensions"],
+  "rules": {
+    "quotes": ["warn", "double", {}],
+    "import/extensions": [
+      "error",
+      "always",
+      {
+        "js": "never",
+        "jsx": "never",
+        "ts": "never",
+        "tsx": "never"
+      }
+    ],
+    "@typescript-eslint/naming-convention": "warn",
+    "@typescript-eslint/semi": "warn",
+    "curly": "warn",
+    "eqeqeq": "warn",
+    "no-throw-literal": "warn",
+    "semi": "off"
+  },
+  "ignorePatterns": ["out", "dist", "**/*.d.ts"]
+}

core/config/load.ts

@@ -28,7 +28,7 @@ import {
   SlashCommand,
 } from "../index.js";
 import TransformersJsEmbeddingsProvider from "../indexing/embeddings/TransformersJsEmbeddingsProvider.js";
-import { AllEmbeddingsProviders } from "../indexing/embeddings/index.js";
+import { allEmbeddingsProviders } from "../indexing/embeddings/index.js";
 import { BaseLLM } from "../llm/index.js";
 import CustomLLMClass from "../llm/llms/CustomLLM.js";
 import FreeTrial from "../llm/llms/FreeTrial.js";
@@ -336,7 +336,7 @@ async function intermediateToFinalConfig(
     | undefined;
   if (embeddingsProviderDescription?.provider) {
     const { provider, ...options } = embeddingsProviderDescription;
-    const embeddingsProviderClass = AllEmbeddingsProviders[provider];
+    const embeddingsProviderClass = allEmbeddingsProviders[provider];
     if (embeddingsProviderClass) {
       if (
         embeddingsProviderClass.name === "_TransformersJsEmbeddingsProvider"

core/config/types.ts

@@ -562,6 +562,7 @@ declare global {
     // Gemini
     | "gemini-pro"
     | "gemini-1.5-pro-latest"
+    | "gemini-1.5-flash-latest"
     // Mistral
     | "mistral-tiny"
     | "mistral-small"

core/index.d.ts

@@ -688,7 +688,8 @@ export type EmbeddingsProviderName =
   | "ollama"
   | "openai"
   | "cohere"
-  | "free-trial";
+  | "free-trial"
+  | "gemini";
 
 export interface EmbedOptions {
   apiBase?: string;

core/indexing/embeddings/BaseEmbeddingsProvider.ts

@@ -1,24 +1,60 @@
-import { EmbedOptions, EmbeddingsProvider, FetchFunction } from "../../index.js";
+import {
+  EmbedOptions,
+  EmbeddingsProvider,
+  FetchFunction,
+} from "../../index.js";
 
-class BaseEmbeddingsProvider implements EmbeddingsProvider {
+export interface IBaseEmbeddingsProvider extends EmbeddingsProvider {
   options: EmbedOptions;
   fetch: FetchFunction;
-  static defaultOptions: Partial<EmbedOptions> | undefined = undefined;
+  defaultOptions?: EmbedOptions;
+  maxBatchSize?: number;
+}
 
-  get id(): string {
-    throw new Error("Method not implemented.");
-  }
+abstract class BaseEmbeddingsProvider implements IBaseEmbeddingsProvider {
+  static maxBatchSize: IBaseEmbeddingsProvider["maxBatchSize"];
+  static defaultOptions: IBaseEmbeddingsProvider["defaultOptions"];
+
+  options: IBaseEmbeddingsProvider["options"];
+  fetch: IBaseEmbeddingsProvider["fetch"];
+  id: IBaseEmbeddingsProvider["id"];
 
-  constructor(options: EmbedOptions, fetch: FetchFunction) {
+  constructor(
+    options: IBaseEmbeddingsProvider["options"],
+    fetch: IBaseEmbeddingsProvider["fetch"],
+  ) {
+    // Overwrite default options with any runtime options
     this.options = {
       ...(this.constructor as typeof BaseEmbeddingsProvider).defaultOptions,
       ...options,
     };
     this.fetch = fetch;
+    this.id = this.options.model || this.constructor.name;
   }
 
-  embed(chunks: string[]): Promise<number[][]> {
-    throw new Error("Method not implemented.");
+  abstract embed(chunks: string[]): Promise<number[][]>;
+
+  static getBatchedChunks(chunks: string[]): string[][] {
+    if (!this.maxBatchSize) {
+      console.warn(
+        `${this.getBatchedChunks.name} should only be called if 'maxBatchSize' is defined`,
+      );
+
+      return [chunks];
+    }
+
+    if (chunks.length > this.maxBatchSize) {
+      return [chunks];
+    }
+
+    const batchedChunks = [];
+
+    for (let i = 0; i < chunks.length; i += this.maxBatchSize) {
+      const batchSizedChunk = chunks.slice(i, i + this.maxBatchSize);
+      batchedChunks.push(batchSizedChunk);
+    }
+
+    return batchedChunks;
   }
 }
 

core/indexing/embeddings/CohereEmbeddingsProvider.ts

@@ -11,10 +11,6 @@ class CohereEmbeddingsProvider extends BaseEmbeddingsProvider {
     model: "embed-english-v3.0",
   };
 
-  get id(): string {
-    return this.options.model ?? "cohere";
-  }
-
   async embed(chunks: string[]) {
     if (!this.options.apiBase?.endsWith("/")) {
       this.options.apiBase += "/";

core/indexing/embeddings/DeepInfraEmbeddingsProvider.ts

@@ -7,10 +7,6 @@ class DeepInfraEmbeddingsProvider extends BaseEmbeddingsProvider {
     model: "sentence-transformers/all-MiniLM-L6-v2",
   };
 
-  get id(): string {
-    return this.options.model ?? "deepinfra";
-  }
-
   async embed(chunks: string[]) {
     const fetchWithBackoff = () =>
       withExponentialBackoff<Response>(() =>

core/indexing/embeddings/FreeTrialEmbeddingsProvider.ts

@@ -7,14 +7,11 @@ import BaseEmbeddingsProvider from "./BaseEmbeddingsProvider.js";
 
 class FreeTrialEmbeddingsProvider extends BaseEmbeddingsProvider {
   static maxBatchSize = 128;
+
   static defaultOptions: Partial<EmbedOptions> | undefined = {
     model: "voyage-code-2",
   };
 
-  get id(): string {
-    return FreeTrialEmbeddingsProvider.defaultOptions?.model!;
-  }
-
   async embed(chunks: string[]) {
     const batchedChunks = [];
     for (

core/indexing/embeddings/GeminiEmbeddingsProvider.ts

@@ -0,0 +1,72 @@
+import { Response } from "node-fetch";
+import { withExponentialBackoff } from "../../util/withExponentialBackoff.js";
+import BaseEmbeddingsProvider, {
+  IBaseEmbeddingsProvider,
+} from "./BaseEmbeddingsProvider.js";
+import {
+  EmbedContentRequest,
+  EmbedContentResponse,
+} from "@google/generative-ai";
+
+/**
+ * [View the Gemini Text Embedding docs.](https://ai.google.dev/gemini-api/docs/models/gemini#text-embedding-and-embedding)
+ */
+class GeminiEmbeddingsProvider extends BaseEmbeddingsProvider {
+  static maxBatchSize = 2048;
+
+  static defaultOptions = {
+    apiBase: "https://generativelanguage.googleapis.com/v1/",
+    model: "models/text-embedding-004",
+  };
+
+  get urlPath(): string {
+    return `${this.options.model}:embedContent`;
+  }
+
+  async getSingleBatchEmbedding(batch: string[]) {
+    const body: EmbedContentRequest = {
+      content: {
+        /**
+         * Listed as optional in the [docs](https://ai.google.dev/api/rest/v1/Content)
+         * but is required in the interface.
+         */
+        role: "user",
+        parts: batch.map((part) => ({ text: part })),
+      },
+    };
+
+    const fetchWithBackoff = () =>
+      withExponentialBackoff<Response>(() =>
+        this.fetch(new URL(this.urlPath, this.options.apiBase), {
+          method: "POST",
+          body: JSON.stringify(body),
+          headers: {
+            // eslint-disable-next-line @typescript-eslint/naming-convention
+            "x-goog-api-key": this.options.apiKey,
+            // eslint-disable-next-line @typescript-eslint/naming-convention
+            "Content-Type": "application/json",
+          },
+        }),
+      );
+
+    const resp = await fetchWithBackoff();
+
+    if (!resp.ok) {
+      throw new Error(await resp.text());
+    }
+
+    const data = (await resp.json()) as EmbedContentResponse;
+
+    return data.embedding.values;
+  }
+
+  async embed(chunks: string[]) {
+    const batches = GeminiEmbeddingsProvider.getBatchedChunks(chunks);
+
+    return await Promise.all(
+      batches.map((batch) => this.getSingleBatchEmbedding(batch)),
+    );
+  }
+}
+
+export default GeminiEmbeddingsProvider;

core/indexing/embeddings/OllamaEmbeddingsProvider.ts

@@ -1,6 +1,8 @@
 import { EmbedOptions, FetchFunction } from "../../index.js";
 import { withExponentialBackoff } from "../../util/withExponentialBackoff.js";
-import BaseEmbeddingsProvider from "./BaseEmbeddingsProvider.js";
+import BaseEmbeddingsProvider, {
+  IBaseEmbeddingsProvider,
+} from "./BaseEmbeddingsProvider.js";
 
 async function embedOne(
   chunk: string,
@@ -18,6 +20,7 @@ async function embedOne(
       }),
     );
   const resp = await fetchWithBackoff();
+
   if (!resp.ok) {
     throw new Error(`Failed to embed chunk: ${await resp.text()}`);
   }
@@ -26,14 +29,11 @@ async function embedOne(
 }
 
 class OllamaEmbeddingsProvider extends BaseEmbeddingsProvider {
-  static defaultOptions: Partial<EmbedOptions> | undefined = {
+  static defaultOptions: IBaseEmbeddingsProvider["defaultOptions"] = {
     apiBase: "http://localhost:11434/",
+    model: "nomic-embed-text",
   };
 
-  get id(): string {
-    return this.options.model ?? "ollama";
-  }
-
   async embed(chunks: string[]) {
     const results: any = [];
     for (const chunk of chunks) {

core/indexing/embeddings/OpenAIEmbeddingsProvider.ts

@@ -13,10 +13,6 @@ class OpenAIEmbeddingsProvider extends BaseEmbeddingsProvider {
     model: "text-embedding-3-small",
   };
 
-  get id(): string {
-    return this.options.model ?? "openai";
-  }
-
   async embed(chunks: string[]) {
     if (!this.options.apiBase?.endsWith("/")) {
       this.options.apiBase += "/";

core/indexing/embeddings/TransformersJsEmbeddingsProvider.ts

@@ -37,17 +37,10 @@ class EmbeddingsPipeline {
 }
 
 export class TransformersJsEmbeddingsProvider extends BaseEmbeddingsProvider {
-  static MaxGroupSize: number = 4;
+  static maxGroupSize: number = 4;
 
-  constructor(modelPath?: string) {
+  constructor() {
     super({ model: "all-MiniLM-L2-v6" }, () => Promise.resolve(null));
-    if (modelPath) {
-      // env.localModelPath = modelPath;
-    }
-  }
-
-  get id(): string {
-    return EmbeddingsPipeline.model;
   }
 
   async embed(chunks: string[]) {
@@ -65,11 +58,11 @@ export class TransformersJsEmbeddingsProvider extends BaseEmbeddingsProvider {
     for (
       let i = 0;
       i < chunks.length;
-      i += TransformersJsEmbeddingsProvider.MaxGroupSize
+      i += TransformersJsEmbeddingsProvider.maxGroupSize
     ) {
       const chunkGroup = chunks.slice(
         i,
-        i + TransformersJsEmbeddingsProvider.MaxGroupSize,
+        i + TransformersJsEmbeddingsProvider.maxGroupSize,
       );
       const output = await extractor(chunkGroup, {
         pooling: "mean",

core/indexing/embeddings/TransformersJsWorkerThread.js

@@ -42,11 +42,11 @@ parentPort.on("message", async (chunks) => {
     for (
       let i = 0;
       i < chunks.length;
-      i += TransformersJsEmbeddingsProvider.MaxGroupSize
+      i += TransformersJsEmbeddingsProvider.maxGroupSize
     ) {
       const chunkGroup = chunks.slice(
         i,
-        i + TransformersJsEmbeddingsProvider.MaxGroupSize,
+        i + TransformersJsEmbeddingsProvider.maxGroupSize,
       );
       const output = await extractor(chunkGroup, {
         pooling: "mean",