llm-exe 2.3.6 → 2.3.8

package/dist/index.d.mts

@@ -996,6 +996,10 @@ interface AmazonEmbeddingOptions extends GenericEmbeddingOptions {
     awsSecretKey?: string;
     awsAccessKey?: string;
 }
+interface CohereBedrockEmbeddingOptions extends AmazonEmbeddingOptions {
+    inputType?: "search_document" | "search_query" | "classification" | "clustering";
+    truncate?: "NONE" | "START" | "END" | "LEFT" | "RIGHT";
+}
 interface GenericLLm extends BaseLlmOptions {
     model?: string;
     system?: string;
@@ -1059,6 +1063,9 @@ type AllEmbedding = {
     "amazon.embedding.v1": {
         input: AmazonEmbeddingOptions;
     };
+    "amazon:cohere.embedding.v1": {
+        input: CohereBedrockEmbeddingOptions;
+    };
 };
 type AllLlm = {
     "openai.chat.v1": {
@@ -1255,7 +1262,7 @@ interface BaseRequest<_T extends Record<string, any>> {
 interface BaseLlm<_T extends BaseLlCall = BaseLlCall> extends BaseRequest<_T> {
 }
 
-type LlmProvider = "openai.chat" | "openai.embedding" | "google.embedding" | "openai.chat-mock" | "anthropic.chat" | "amazon:anthropic.chat" | "amazon:meta.chat" | "amazon:nova.chat" | "amazon.embedding" | "xai.chat" | "google.chat" | "ollama.chat" | "deepseek.chat";
+type LlmProvider = "openai.chat" | "openai.embedding" | "google.embedding" | "openai.chat-mock" | "anthropic.chat" | "amazon:anthropic.chat" | "amazon:meta.chat" | "amazon:nova.chat" | "amazon.embedding" | "amazon:cohere.embedding" | "xai.chat" | "google.chat" | "ollama.chat" | "deepseek.chat";
 interface Config<Pk = LlmProviderKey> {
     /**
      * Unique identifier for this configuration (e.g., "openai.chat.v1", "anthropic.claude-3-opus")
@@ -1356,7 +1363,13 @@ interface Config<Pk = LlmProviderKey> {
          */
         functions?: (functions: any[], options?: any, currentInput?: Record<string, any>, config?: Config) => Record<string, any>;
     };
-    transformResponse: (result: any, _config?: Config<any>) => OutputResult;
+    /**
+     * Optional response transformer for chat/LLM configs. The LLM call path
+     * (`llm.call.ts`) defaults to `OutputDefault` when this is omitted.
+     * Embedding configs do not use this — their flow dispatches via
+     * `getEmbeddingOutputParser` instead.
+     */
+    transformResponse?: (result: any, _config?: Config<any>) => OutputResult;
 }
 
 declare function enforceResultAttributes<O>(input: any): {

package/dist/index.d.ts

@@ -996,6 +996,10 @@ interface AmazonEmbeddingOptions extends GenericEmbeddingOptions {
     awsSecretKey?: string;
     awsAccessKey?: string;
 }
+interface CohereBedrockEmbeddingOptions extends AmazonEmbeddingOptions {
+    inputType?: "search_document" | "search_query" | "classification" | "clustering";
+    truncate?: "NONE" | "START" | "END" | "LEFT" | "RIGHT";
+}
 interface GenericLLm extends BaseLlmOptions {
     model?: string;
     system?: string;
@@ -1059,6 +1063,9 @@ type AllEmbedding = {
     "amazon.embedding.v1": {
         input: AmazonEmbeddingOptions;
     };
+    "amazon:cohere.embedding.v1": {
+        input: CohereBedrockEmbeddingOptions;
+    };
 };
 type AllLlm = {
     "openai.chat.v1": {
@@ -1255,7 +1262,7 @@ interface BaseRequest<_T extends Record<string, any>> {
 interface BaseLlm<_T extends BaseLlCall = BaseLlCall> extends BaseRequest<_T> {
 }
 
-type LlmProvider = "openai.chat" | "openai.embedding" | "google.embedding" | "openai.chat-mock" | "anthropic.chat" | "amazon:anthropic.chat" | "amazon:meta.chat" | "amazon:nova.chat" | "amazon.embedding" | "xai.chat" | "google.chat" | "ollama.chat" | "deepseek.chat";
+type LlmProvider = "openai.chat" | "openai.embedding" | "google.embedding" | "openai.chat-mock" | "anthropic.chat" | "amazon:anthropic.chat" | "amazon:meta.chat" | "amazon:nova.chat" | "amazon.embedding" | "amazon:cohere.embedding" | "xai.chat" | "google.chat" | "ollama.chat" | "deepseek.chat";
 interface Config<Pk = LlmProviderKey> {
     /**
      * Unique identifier for this configuration (e.g., "openai.chat.v1", "anthropic.claude-3-opus")
@@ -1356,7 +1363,13 @@ interface Config<Pk = LlmProviderKey> {
          */
         functions?: (functions: any[], options?: any, currentInput?: Record<string, any>, config?: Config) => Record<string, any>;
     };
-    transformResponse: (result: any, _config?: Config<any>) => OutputResult;
+    /**
+     * Optional response transformer for chat/LLM configs. The LLM call path
+     * (`llm.call.ts`) defaults to `OutputDefault` when this is omitted.
+     * Embedding configs do not use this — their flow dispatches via
+     * `getEmbeddingOutputParser` instead.
+     */
+    transformResponse?: (result: any, _config?: Config<any>) => OutputResult;
 }
 
 declare function enforceResultAttributes<O>(input: any): {

package/dist/index.js

@@ -3903,77 +3903,6 @@ function useLlmConfiguration(config) {
   return (options = {}) => apiRequestWrapper(config, options, useLlm_call);
 }
 
-// src/embedding/output/BaseEmbeddingOutput.ts
-function BaseEmbeddingOutput(result) {
-  const __result = Object.freeze({
-    id: result.id || (0, import_uuid.v4)(),
-    model: result.model,
-    usage: result.usage,
-    embedding: [...result?.embedding || []],
-    created: result?.created || (/* @__PURE__ */ new Date()).getTime()
-  });
-  function getResult() {
-    return {
-      id: __result.id,
-      model: __result.model,
-      created: __result.created,
-      usage: __result.usage,
-      embedding: __result.embedding
-    };
-  }
-  function getEmbedding(index) {
-    if (index && index > 0) {
-      const arr = __result?.embedding;
-      const val = arr[index];
-      return val ? val : [];
-    }
-    return __result.embedding[0];
-  }
-  return {
-    getEmbedding,
-    getResult
-  };
-}
-
-// src/embedding/output/OpenAiEmbedding.ts
-function OpenAiEmbedding(result, config) {
-  const __result = deepClone(result);
-  const model = __result.model || config.model || "openai.unknown";
-  const created = (/* @__PURE__ */ new Date()).getTime();
-  const results = result?.data || [];
-  const embedding = results.map((a) => a.embedding);
-  const usage = {
-    output_tokens: 0,
-    input_tokens: result?.usage?.prompt_tokens,
-    total_tokens: result?.usage?.total_tokens
-  };
-  return BaseEmbeddingOutput({
-    model,
-    created,
-    usage,
-    embedding
-  });
-}
-
-// src/embedding/output/AmazonTitan.ts
-function AmazonTitanEmbedding(result, config) {
-  const __result = deepClone(result);
-  const model = config.model || "amazon.unknown";
-  const created = (/* @__PURE__ */ new Date()).getTime();
-  const embedding = [__result.embedding];
-  const usage = {
-    output_tokens: 0,
-    input_tokens: __result.inputTextTokenCount,
-    total_tokens: __result.inputTextTokenCount
-  };
-  return BaseEmbeddingOutput({
-    model,
-    created,
-    usage,
-    embedding
-  });
-}
-
 // src/embedding/config.ts
 var embeddingConfigs = {
   "openai.embedding.v1": {
@@ -4005,8 +3934,7 @@ var embeddingConfigs = {
       encodingFormat: {
         key: "encoding_format"
       }
-    },
-    transformResponse: OpenAiEmbedding
+    }
   },
   "amazon.embedding.v1": {
     key: "amazon.embedding.v1",
@@ -4033,8 +3961,59 @@ var embeddingConfigs = {
       dimensions: {
         key: "dimensions"
       }
+    }
+  },
+  "amazon:cohere.embedding.v1": {
+    key: "amazon:cohere.embedding.v1",
+    provider: "amazon:cohere.embedding",
+    endpoint: `https://bedrock-runtime.{{awsRegion}}.amazonaws.com/model/{{model}}/invoke`,
+    method: "POST",
+    headers: `{"Content-Type": "application/json" }`,
+    options: {
+      input: {},
+      inputType: {
+        default: "search_document"
+      },
+      truncate: {},
+      dimensions: {},
+      awsRegion: {
+        default: getEnvironmentVariable("AWS_REGION"),
+        required: [true, "aws region is required"]
+      },
+      awsSecretKey: {},
+      awsAccessKey: {}
     },
-    transformResponse: AmazonTitanEmbedding
+    mapBody: {
+      input: {
+        key: "texts",
+        transform: (value) => Array.isArray(value) ? value : [value]
+      },
+      inputType: {
+        key: "input_type"
+      },
+      truncate: {
+        key: "truncate"
+      },
+      dimensions: {
+        key: "output_dimension",
+        // Embed v3 has a fixed 1024-dim output and rejects the output_dimension
+        // field entirely — even when the value matches. Drop the field for v3
+        // when the user asked for the natural 1024; throw for any other value
+        // so we don't silently mutate user intent.
+        transform: (value, state) => {
+          if (typeof value === "undefined") return void 0;
+          const model = state?.model || "";
+          const isV3 = /embed-(english|multilingual)-v3/.test(model);
+          if (isV3) {
+            if (value === 1024) return void 0;
+            throw new Error(
+              `Cohere Embed v3 only supports 1024-dimensional output (model: "${model}", requested: ${value}). Use cohere.embed-v4:0 for configurable dimensions.`
+            );
+          }
+          return value;
+        }
+      }
+    }
   }
 };
 function getEmbeddingConfig(provider) {
@@ -4048,6 +4027,97 @@ function getEmbeddingConfig(provider) {
   throw new Error(`Invalid provider: ${provider}`);
 }
 
+// src/embedding/output/BaseEmbeddingOutput.ts
+function BaseEmbeddingOutput(result) {
+  const __result = Object.freeze({
+    id: result.id || (0, import_uuid.v4)(),
+    model: result.model,
+    usage: result.usage,
+    embedding: [...result?.embedding || []],
+    created: result?.created || (/* @__PURE__ */ new Date()).getTime()
+  });
+  function getResult() {
+    return {
+      id: __result.id,
+      model: __result.model,
+      created: __result.created,
+      usage: __result.usage,
+      embedding: __result.embedding
+    };
+  }
+  function getEmbedding(index) {
+    if (index && index > 0) {
+      const arr = __result?.embedding;
+      const val = arr[index];
+      return val ? val : [];
+    }
+    return __result.embedding[0];
+  }
+  return {
+    getEmbedding,
+    getResult
+  };
+}
+
+// src/embedding/output/AmazonTitan.ts
+function AmazonTitanEmbedding(result, config) {
+  const __result = deepClone(result);
+  const model = config.model || "amazon.unknown";
+  const created = (/* @__PURE__ */ new Date()).getTime();
+  const embedding = [__result.embedding];
+  const usage = {
+    output_tokens: 0,
+    input_tokens: __result.inputTextTokenCount,
+    total_tokens: __result.inputTextTokenCount
+  };
+  return BaseEmbeddingOutput({
+    model,
+    created,
+    usage,
+    embedding
+  });
+}
+
+// src/embedding/output/CohereBedrockEmbedding.ts
+function CohereBedrockEmbedding(result, config) {
+  const __result = deepClone(result);
+  const model = config.model || "cohere.unknown";
+  const created = (/* @__PURE__ */ new Date()).getTime();
+  const embedding = Array.isArray(__result.embeddings) ? __result.embeddings : [];
+  const usage = {
+    output_tokens: 0,
+    input_tokens: 0,
+    total_tokens: 0
+  };
+  return BaseEmbeddingOutput({
+    id: __result.id,
+    model,
+    created,
+    usage,
+    embedding
+  });
+}
+
+// src/embedding/output/OpenAiEmbedding.ts
+function OpenAiEmbedding(result, config) {
+  const __result = deepClone(result);
+  const model = __result.model || config.model || "openai.unknown";
+  const created = (/* @__PURE__ */ new Date()).getTime();
+  const results = result?.data || [];
+  const embedding = results.map((a) => a.embedding);
+  const usage = {
+    output_tokens: 0,
+    input_tokens: result?.usage?.prompt_tokens,
+    total_tokens: result?.usage?.total_tokens
+  };
+  return BaseEmbeddingOutput({
+    model,
+    created,
+    usage,
+    embedding
+  });
+}
+
 // src/embedding/output/getEmbeddingOutputParser.ts
 function getEmbeddingOutputParser(config, response) {
   switch (config.key) {
@@ -4055,6 +4125,8 @@ function getEmbeddingOutputParser(config, response) {
       return OpenAiEmbedding(response, config);
     case "amazon.embedding.v1":
       return AmazonTitanEmbedding(response, config);
+    case "amazon:cohere.embedding.v1":
+      return CohereBedrockEmbedding(response, config);
     default:
       throw new Error("Unsupported provider");
   }

package/dist/index.mjs

@@ -3840,77 +3840,6 @@ function useLlmConfiguration(config) {
   return (options = {}) => apiRequestWrapper(config, options, useLlm_call);
 }
 
-// src/embedding/output/BaseEmbeddingOutput.ts
-function BaseEmbeddingOutput(result) {
-  const __result = Object.freeze({
-    id: result.id || uuidv4(),
-    model: result.model,
-    usage: result.usage,
-    embedding: [...result?.embedding || []],
-    created: result?.created || (/* @__PURE__ */ new Date()).getTime()
-  });
-  function getResult() {
-    return {
-      id: __result.id,
-      model: __result.model,
-      created: __result.created,
-      usage: __result.usage,
-      embedding: __result.embedding
-    };
-  }
-  function getEmbedding(index) {
-    if (index && index > 0) {
-      const arr = __result?.embedding;
-      const val = arr[index];
-      return val ? val : [];
-    }
-    return __result.embedding[0];
-  }
-  return {
-    getEmbedding,
-    getResult
-  };
-}
-
-// src/embedding/output/OpenAiEmbedding.ts
-function OpenAiEmbedding(result, config) {
-  const __result = deepClone(result);
-  const model = __result.model || config.model || "openai.unknown";
-  const created = (/* @__PURE__ */ new Date()).getTime();
-  const results = result?.data || [];
-  const embedding = results.map((a) => a.embedding);
-  const usage = {
-    output_tokens: 0,
-    input_tokens: result?.usage?.prompt_tokens,
-    total_tokens: result?.usage?.total_tokens
-  };
-  return BaseEmbeddingOutput({
-    model,
-    created,
-    usage,
-    embedding
-  });
-}
-
-// src/embedding/output/AmazonTitan.ts
-function AmazonTitanEmbedding(result, config) {
-  const __result = deepClone(result);
-  const model = config.model || "amazon.unknown";
-  const created = (/* @__PURE__ */ new Date()).getTime();
-  const embedding = [__result.embedding];
-  const usage = {
-    output_tokens: 0,
-    input_tokens: __result.inputTextTokenCount,
-    total_tokens: __result.inputTextTokenCount
-  };
-  return BaseEmbeddingOutput({
-    model,
-    created,
-    usage,
-    embedding
-  });
-}
-
 // src/embedding/config.ts
 var embeddingConfigs = {
   "openai.embedding.v1": {
@@ -3942,8 +3871,7 @@ var embeddingConfigs = {
       encodingFormat: {
         key: "encoding_format"
       }
-    },
-    transformResponse: OpenAiEmbedding
+    }
   },
   "amazon.embedding.v1": {
     key: "amazon.embedding.v1",
@@ -3970,8 +3898,59 @@ var embeddingConfigs = {
       dimensions: {
         key: "dimensions"
       }
+    }
+  },
+  "amazon:cohere.embedding.v1": {
+    key: "amazon:cohere.embedding.v1",
+    provider: "amazon:cohere.embedding",
+    endpoint: `https://bedrock-runtime.{{awsRegion}}.amazonaws.com/model/{{model}}/invoke`,
+    method: "POST",
+    headers: `{"Content-Type": "application/json" }`,
+    options: {
+      input: {},
+      inputType: {
+        default: "search_document"
+      },
+      truncate: {},
+      dimensions: {},
+      awsRegion: {
+        default: getEnvironmentVariable("AWS_REGION"),
+        required: [true, "aws region is required"]
+      },
+      awsSecretKey: {},
+      awsAccessKey: {}
     },
-    transformResponse: AmazonTitanEmbedding
+    mapBody: {
+      input: {
+        key: "texts",
+        transform: (value) => Array.isArray(value) ? value : [value]
+      },
+      inputType: {
+        key: "input_type"
+      },
+      truncate: {
+        key: "truncate"
+      },
+      dimensions: {
+        key: "output_dimension",
+        // Embed v3 has a fixed 1024-dim output and rejects the output_dimension
+        // field entirely — even when the value matches. Drop the field for v3
+        // when the user asked for the natural 1024; throw for any other value
+        // so we don't silently mutate user intent.
+        transform: (value, state) => {
+          if (typeof value === "undefined") return void 0;
+          const model = state?.model || "";
+          const isV3 = /embed-(english|multilingual)-v3/.test(model);
+          if (isV3) {
+            if (value === 1024) return void 0;
+            throw new Error(
+              `Cohere Embed v3 only supports 1024-dimensional output (model: "${model}", requested: ${value}). Use cohere.embed-v4:0 for configurable dimensions.`
+            );
+          }
+          return value;
+        }
+      }
+    }
   }
 };
 function getEmbeddingConfig(provider) {
@@ -3985,6 +3964,97 @@ function getEmbeddingConfig(provider) {
   throw new Error(`Invalid provider: ${provider}`);
 }
 
+// src/embedding/output/BaseEmbeddingOutput.ts
+function BaseEmbeddingOutput(result) {
+  const __result = Object.freeze({
+    id: result.id || uuidv4(),
+    model: result.model,
+    usage: result.usage,
+    embedding: [...result?.embedding || []],
+    created: result?.created || (/* @__PURE__ */ new Date()).getTime()
+  });
+  function getResult() {
+    return {
+      id: __result.id,
+      model: __result.model,
+      created: __result.created,
+      usage: __result.usage,
+      embedding: __result.embedding
+    };
+  }
+  function getEmbedding(index) {
+    if (index && index > 0) {
+      const arr = __result?.embedding;
+      const val = arr[index];
+      return val ? val : [];
+    }
+    return __result.embedding[0];
+  }
+  return {
+    getEmbedding,
+    getResult
+  };
+}
+
+// src/embedding/output/AmazonTitan.ts
+function AmazonTitanEmbedding(result, config) {
+  const __result = deepClone(result);
+  const model = config.model || "amazon.unknown";
+  const created = (/* @__PURE__ */ new Date()).getTime();
+  const embedding = [__result.embedding];
+  const usage = {
+    output_tokens: 0,
+    input_tokens: __result.inputTextTokenCount,
+    total_tokens: __result.inputTextTokenCount
+  };
+  return BaseEmbeddingOutput({
+    model,
+    created,
+    usage,
+    embedding
+  });
+}
+
+// src/embedding/output/CohereBedrockEmbedding.ts
+function CohereBedrockEmbedding(result, config) {
+  const __result = deepClone(result);
+  const model = config.model || "cohere.unknown";
+  const created = (/* @__PURE__ */ new Date()).getTime();
+  const embedding = Array.isArray(__result.embeddings) ? __result.embeddings : [];
+  const usage = {
+    output_tokens: 0,
+    input_tokens: 0,
+    total_tokens: 0
+  };
+  return BaseEmbeddingOutput({
+    id: __result.id,
+    model,
+    created,
+    usage,
+    embedding
+  });
+}
+
+// src/embedding/output/OpenAiEmbedding.ts
+function OpenAiEmbedding(result, config) {
+  const __result = deepClone(result);
+  const model = __result.model || config.model || "openai.unknown";
+  const created = (/* @__PURE__ */ new Date()).getTime();
+  const results = result?.data || [];
+  const embedding = results.map((a) => a.embedding);
+  const usage = {
+    output_tokens: 0,
+    input_tokens: result?.usage?.prompt_tokens,
+    total_tokens: result?.usage?.total_tokens
+  };
+  return BaseEmbeddingOutput({
+    model,
+    created,
+    usage,
+    embedding
+  });
+}
+
 // src/embedding/output/getEmbeddingOutputParser.ts
 function getEmbeddingOutputParser(config, response) {
   switch (config.key) {
@@ -3992,6 +4062,8 @@ function getEmbeddingOutputParser(config, response) {
       return OpenAiEmbedding(response, config);
     case "amazon.embedding.v1":
       return AmazonTitanEmbedding(response, config);
+    case "amazon:cohere.embedding.v1":
+      return CohereBedrockEmbedding(response, config);
     default:
       throw new Error("Unsupported provider");
   }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "llm-exe",
-  "version": "2.3.6",
+  "version": "2.3.8",
   "description": "Simplify building LLM-powered apps with easy-to-use base components, supporting text and chat-based prompts with handlebars template engine, output parsers, and flexible function calling capabilities.",
   "keywords": [
     "ai",