Inline ml-distance functions due to type errors, fix type errors

deno.json

@@ -22,7 +22,6 @@
     "zod": "npm:/zod",
     "zod-to-json-schema": "npm:/zod-to-json-schema",
     "node-llama-cpp": "npm:/node-llama-cpp",
-    "ml-distance": "npm:/ml-distance",
     "pdf-parse": "npm:/pdf-parse",
     "peggy": "npm:/peggy",
     "readline": "https://deno.land/x/[email protected]/mod.ts",

docs/core_docs/docs/integrations/vectorstores/memory.mdx

@@ -34,6 +34,18 @@ import ExampleLoader from "@examples/indexes/vector_stores/memory_fromdocs.ts";
 
 ### Use a custom similarity metric
 
-import ExampleCustom from "@examples/indexes/vector_stores/memory_custom_similarity.ts";
-
-<CodeBlock language="typescript">{ExampleCustom}</CodeBlock>
+```ts
+import { MemoryVectorStore } from "langchain/vectorstores/memory";
+import { OpenAIEmbeddings } from "@langchain/openai";
+import { similarity } from "ml-distance";
+
+const vectorStore = await MemoryVectorStore.fromTexts(
+  ["Hello world", "Bye bye", "hello nice world"],
+  [{ id: 2 }, { id: 1 }, { id: 3 }],
+  new OpenAIEmbeddings(),
+  { similarity: similarity.pearson }
+);
+
+const resultOne = await vectorStore.similaritySearch("hello world", 1);
+console.log(resultOne);
+```

environment_tests/docker-compose.yml

@@ -152,6 +152,8 @@ services:
         condition: service_completed_successfully
       test-exports-esm:
         condition: service_completed_successfully
+      test-exports-tsc:
+        condition: service_completed_successfully
       test-exports-cjs:
         condition: service_completed_successfully
       test-exports-cf:

examples/package.json

@@ -91,7 +91,6 @@
     "js-yaml": "^4.1.0",
     "langchain": "workspace:*",
     "langsmith": "^0.1.30",
-    "ml-distance": "^4.0.0",
     "mongodb": "^6.3.0",
     "pg": "^8.11.0",
     "pickleparser": "^0.2.1",

langchain-core/package.json

@@ -46,7 +46,6 @@
     "decamelize": "1.2.0",
     "js-tiktoken": "^1.0.12",
     "langsmith": "~0.1.39",
-    "ml-distance": "^4.0.0",
     "mustache": "^4.2.0",
     "p-queue": "^6.6.2",
     "p-retry": "4",

langchain-core/src/language_models/chat_models.ts

@@ -123,17 +123,6 @@ export type LangSmithParams = {
   ls_stop?: Array<string>;
 };
 
-interface ChatModelGenerateCachedParameters<
-  T extends BaseChatModel<CallOptions>,
-  CallOptions extends BaseChatModelCallOptions = BaseChatModelCallOptions
-> {
-  messages: BaseMessageLike[][];
-  cache: BaseCache<Generation[]>;
-  llmStringKey: string;
-  parsedOptions: T["ParsedCallOptions"];
-  handledOptions: RunnableConfig;
-}
-
 /**
  * Base class for chat models. It extends the BaseLanguageModel class and
  * provides methods for generating chat based on input messages.
@@ -449,9 +438,14 @@ export abstract class BaseChatModel<
     llmStringKey,
     parsedOptions,
     handledOptions,
-  }: ChatModelGenerateCachedParameters<typeof this>): Promise<
-    LLMResult & { missingPromptIndices: number[] }
-  > {
+  }: {
+    messages: BaseMessageLike[][];
+    cache: BaseCache<Generation[]>;
+    llmStringKey: string;
+    // eslint-disable-next-line @typescript-eslint/no-explicit-any
+    parsedOptions: any;
+    handledOptions: RunnableConfig;
+  }): Promise<LLMResult & { missingPromptIndices: number[] }> {
     const baseMessages = messages.map((messageList) =>
       messageList.map(coerceMessageLikeToMessage)
     );

langchain-core/src/language_models/llms.ts

@@ -43,18 +43,6 @@ export interface BaseLLMParams extends BaseLanguageModelParams {
 
 export interface BaseLLMCallOptions extends BaseLanguageModelCallOptions {}
 
-interface LLMGenerateCachedParameters<
-  T extends BaseLLM<CallOptions>,
-  CallOptions extends BaseLLMCallOptions = BaseLLMCallOptions
-> {
-  prompts: string[];
-  cache: BaseCache<Generation[]>;
-  llmStringKey: string;
-  parsedOptions: T["ParsedCallOptions"];
-  handledOptions: RunnableConfig;
-  runId?: string;
-}
-
 /**
  * LLM Wrapper. Takes in a prompt (or prompts) and returns a string.
  */
@@ -351,9 +339,15 @@ export abstract class BaseLLM<
     parsedOptions,
     handledOptions,
     runId,
-  }: LLMGenerateCachedParameters<typeof this>): Promise<
-    LLMResult & { missingPromptIndices: number[] }
-  > {
+  }: {
+    prompts: string[];
+    cache: BaseCache<Generation[]>;
+    llmStringKey: string;
+    // eslint-disable-next-line @typescript-eslint/no-explicit-any
+    parsedOptions: any;
+    handledOptions: RunnableConfig;
+    runId?: string;
+  }): Promise<LLMResult & { missingPromptIndices: number[] }> {
     const callbackManager_ = await CallbackManager.configure(
       handledOptions.callbacks,
       this.callbacks,

langchain-core/src/utils/math.ts

@@ -1,7 +1,6 @@
-import {
-  similarity as ml_distance_similarity,
-  distance as ml_distance,
-} from "ml-distance";
+import { cosine } from "./ml-distance/similarities.js";
+import { innerProduct as innerProductDistance } from "./ml-distance/distances.js";
+import { euclidean } from "./ml-distance-euclidean/euclidean.js";
 
 type VectorFunction = (xVector: number[], yVector: number[]) => number;
 
@@ -65,15 +64,15 @@ export function normalize(M: number[][], similarity = false): number[][] {
  * @returns {number[][] | [[]]} A matrix where each row represents the cosine similarity values between the corresponding rows of X and Y.
  */
 export function cosineSimilarity(X: number[][], Y: number[][]): number[][] {
-  return matrixFunc(X, Y, ml_distance_similarity.cosine);
+  return matrixFunc(X, Y, cosine);
 }
 
 export function innerProduct(X: number[][], Y: number[][]): number[][] {
-  return matrixFunc(X, Y, ml_distance.innerProduct);
+  return matrixFunc(X, Y, innerProductDistance);
 }
 
 export function euclideanDistance(X: number[][], Y: number[][]): number[][] {
-  return matrixFunc(X, Y, ml_distance.euclidean);
+  return matrixFunc(X, Y, euclidean);
 }
 
 /**

langchain-core/src/utils/ml-distance-euclidean/LICENSE

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2015 ml.js
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

langchain-core/src/utils/ml-distance-euclidean/euclidean.ts

@@ -0,0 +1,11 @@
+export function squaredEuclidean(p: number[], q: number[]) {
+  let d = 0;
+  for (let i = 0; i < p.length; i++) {
+    d += (p[i] - q[i]) * (p[i] - q[i]);
+  }
+  return d;
+}
+
+export function euclidean(p: number[], q: number[]) {
+  return Math.sqrt(squaredEuclidean(p, q));
+}

langchain-core/src/utils/ml-distance/LICENSE

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2014 ml.js
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

langchain-core/src/utils/ml-distance/distances.ts

@@ -0,0 +1,14 @@
+/**
+ *Returns the Inner Product similarity between vectors a and b
+ * @link [Inner Product Similarity algorithm](https://www.naun.org/main/NAUN/ijmmas/mmmas-49.pdf)
+ * @param a - first vector
+ * @param b - second vector
+ *
+ */
+export function innerProduct(a: number[], b: number[]): number {
+  let ans = 0;
+  for (let i = 0; i < a.length; i++) {
+    ans += a[i] * b[i];
+  }
+  return ans;
+}

langchain-core/src/utils/ml-distance/similarities.ts

@@ -0,0 +1,17 @@
+/**
+ * Returns the average of cosine distances between vectors a and b
+ * @param a - first vector
+ * @param b - second vector
+ *
+ */
+export function cosine(a: number[], b: number[]): number {
+  let p = 0;
+  let p2 = 0;
+  let q2 = 0;
+  for (let i = 0; i < a.length; i++) {
+    p += a[i] * b[i];
+    p2 += a[i] * a[i];
+    q2 += b[i] * b[i];
+  }
+  return p / (Math.sqrt(p2) * Math.sqrt(q2));
+}

langchain-core/src/utils/testing/index.ts

@@ -2,7 +2,6 @@
 /* eslint-disable @typescript-eslint/no-explicit-any */
 /* eslint-disable @typescript-eslint/no-unused-vars */
 
-import { similarity as ml_distance_similarity } from "ml-distance";
 import { z } from "zod";
 import {
   BaseCallbackConfig,
@@ -46,6 +45,7 @@ import {
   StructuredOutputMethodOptions,
 } from "../../language_models/base.js";
 import { VectorStore } from "../../vectorstores.js";
+import { cosine } from "../ml-distance/similarities.js";
 
 /**
  * Parser for comma-separated values. It splits the input text by commas
@@ -750,7 +750,7 @@ interface MemoryVector {
  * function.
  */
 export interface FakeVectorStoreArgs {
-  similarity?: typeof ml_distance_similarity.cosine;
+  similarity?: typeof cosine;
 }
 
 /**
@@ -763,7 +763,7 @@ export class FakeVectorStore extends VectorStore {
 
   memoryVectors: MemoryVector[] = [];
 
-  similarity: typeof ml_distance_similarity.cosine;
+  similarity: typeof cosine;
 
   _vectorstoreType(): string {
     return "memory";
@@ -775,7 +775,7 @@ export class FakeVectorStore extends VectorStore {
   ) {
     super(embeddings, rest);
 
-    this.similarity = similarity ?? ml_distance_similarity.cosine;
+    this.similarity = similarity ?? cosine;
   }
 
   /**

langchain/package.json

@@ -942,7 +942,6 @@
     "js-yaml": "^4.1.0",
     "jsonpointer": "^5.0.1",
     "langsmith": "~0.1.40",
-    "ml-distance": "^4.0.0",
     "openapi-types": "^12.1.3",
     "p-retry": "4",
     "uuid": "^10.0.0",

langchain/src/evaluation/embedding_distance/base.ts

@@ -1,4 +1,3 @@
-import { distance, similarity } from "ml-distance";
 import type { EmbeddingsInterface } from "@langchain/core/embeddings";
 import { ChainValues } from "@langchain/core/utils/types";
 import { OpenAIEmbeddings } from "@langchain/openai";
@@ -13,6 +12,9 @@ import {
   StringEvaluator,
   StringEvaluatorArgs,
 } from "../base.js";
+import { cosine } from "../../util/ml-distance/similarities.js";
+import { chebyshev, manhattan } from "../../util/ml-distance/distances.js";
+import { euclidean } from "../../util/ml-distance-euclidean/euclidean.js";
 
 /**
  *
@@ -58,10 +60,10 @@ export function getDistanceCalculationFunction(
 ): VectorFunction {
   const distanceFunctions: { [key in EmbeddingDistanceType]: VectorFunction } =
     {
-      cosine: (X: number[], Y: number[]) => 1.0 - similarity.cosine(X, Y),
-      euclidean: distance.euclidean,
-      manhattan: distance.manhattan,
-      chebyshev: distance.chebyshev,
+      cosine: (X: number[], Y: number[]) => 1.0 - cosine(X, Y),
+      euclidean: euclidean,
+      manhattan: manhattan,
+      chebyshev: chebyshev,
     };
 
   return distanceFunctions[distanceType];

langchain/src/util/ml-distance-euclidean/LICENSE

@@ -0,0 +1,21 @@
+The MIT License (MIT)
+
+Copyright (c) 2015 ml.js
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

langchain/src/util/ml-distance-euclidean/euclidean.ts

@@ -0,0 +1,11 @@
+export function squaredEuclidean(p: number[], q: number[]) {
+  let d = 0;
+  for (let i = 0; i < p.length; i++) {
+    d += (p[i] - q[i]) * (p[i] - q[i]);
+  }
+  return d;
+}
+
+export function euclidean(p: number[], q: number[]) {
+  return Math.sqrt(squaredEuclidean(p, q));
+}