llm-exe 2.3.5 → 2.3.7

package/dist/index.mjs

@@ -494,6 +494,21 @@ function toNumber(value) {
   return NaN;
 }
 
+// src/utils/modules/errors.ts
+var LlmExeError = class extends Error {
+  constructor(message, code, context) {
+    super(message ?? "");
+    __publicField(this, "code");
+    __publicField(this, "context");
+    this.name = this.constructor.name;
+    this.code = code ?? "unknown";
+    this.context = context;
+    if (Error.captureStackTrace) {
+      Error.captureStackTrace(this, this.constructor);
+    }
+  }
+};
+
 // src/parser/parsers/NumberParser.ts
 var NumberParser = class extends BaseParser {
   constructor(options) {
@@ -501,7 +516,18 @@ var NumberParser = class extends BaseParser {
   }
   parse(text) {
     const match = text.match(/-?\d+(\.\d+)?/);
-    return match && isFinite(toNumber(match[0])) ? toNumber(match[0]) : -1;
+    if (match && isFinite(toNumber(match[0]))) {
+      return toNumber(match[0]);
+    }
+    throw new LlmExeError(
+      `No numeric value found in input.`,
+      "parser",
+      {
+        parser: "number",
+        output: text,
+        error: `No numeric value found in input.`
+      }
+    );
   }
 };
 
@@ -1398,7 +1424,7 @@ var asyncCallWithTimeout = async (asyncPromise, timeLimit = 1e4) => {
   const timeoutPromise = new Promise((_resolve, reject) => {
     timeoutHandle = setTimeout(() => {
       return reject(
-        new Error("Unable to perform action. Try again, or use another action.")
+        new Error(`LLM call timed out after ${timeLimit}ms`)
       );
     }, timeLimit);
   });
@@ -1486,21 +1512,6 @@ function validateParserSchema(schema, parsed) {
   return null;
 }
 
-// src/utils/modules/errors.ts
-var LlmExeError = class extends Error {
-  constructor(message, code, context) {
-    super(message ?? "");
-    __publicField(this, "code");
-    __publicField(this, "context");
-    this.name = this.constructor.name;
-    this.code = code ?? "unknown";
-    this.context = context;
-    if (Error.captureStackTrace) {
-      Error.captureStackTrace(this, this.constructor);
-    }
-  }
-};
-
 // src/parser/parsers/JsonParser.ts
 var JsonParser = class extends BaseParserWithJson {
   constructor(options = {}) {
@@ -1544,6 +1555,8 @@ function camelCase(input) {
 var ListToJsonParser = class extends BaseParserWithJson {
   constructor(options = {}) {
     super("listToJson", options);
+    __publicField(this, "keyTransform");
+    this.keyTransform = options.keyTransform ?? "camelCase";
   }
   parse(text) {
     const lines = text.split("\n");
@@ -1554,7 +1567,8 @@ var ListToJsonParser = class extends BaseParserWithJson {
         const key = line.slice(0, colonIndex);
         const value = line.slice(colonIndex + 1).trim();
         if (value) {
-          output[camelCase(key)] = value;
+          const transformedKey = this.keyTransform === "preserve" ? key.trim() : camelCase(key);
+          output[transformedKey] = value;
         }
       }
     });
@@ -1727,7 +1741,15 @@ var StringExtractParser = class extends BaseParser {
         return option;
       }
     }
-    return "";
+    throw new LlmExeError(
+      `No matching enum value found in input.`,
+      "parser",
+      {
+        parser: "stringExtract",
+        output: text,
+        error: `No matching enum value found in input. Expected one of: ${this.enum.join(", ")}`
+      }
+    );
   }
 };
 
@@ -2244,6 +2266,7 @@ function cleanJsonSchemaFor(schema = {}, provider) {
 // src/llm/config/openai/compatible.ts
 function createOpenAiCompatibleConfiguration(overrides) {
   const [apiKeyPropertyKey, apiKeyPropertyValue] = overrides.apiKeyMapping;
+  const isReasoningModel = overrides.isReasoningModel ?? (() => false);
   const config = {
     key: overrides.key,
     provider: overrides.provider,
@@ -2297,10 +2320,7 @@ function createOpenAiCompatibleConfiguration(overrides) {
       effort: {
         key: "reasoning_effort",
         transform: (v, _s) => {
-          if (
-            // only supported reasoning models
-            ["gpt-5"].includes(_s.model) && typeof v === "string" && ["minimal", "low", "medium", "high"].includes(v)
-          ) {
+          if (typeof _s.model === "string" && isReasoningModel(_s.model) && typeof v === "string" && ["minimal", "low", "medium", "high"].includes(v)) {
             return v;
           }
           return void 0;
@@ -2347,7 +2367,8 @@ var openAiChatV1 = createOpenAiCompatibleConfiguration({
   key: "openai.chat.v1",
   provider: "openai.chat",
   endpoint: `https://api.openai.com/v1/chat/completions`,
-  apiKeyMapping: ["openAiApiKey", "OPENAI_API_KEY"]
+  apiKeyMapping: ["openAiApiKey", "OPENAI_API_KEY"],
+  isReasoningModel: (model) => model.startsWith("gpt-5") || model.startsWith("o3") || model.startsWith("o4")
 });
 var openAiChatMockV1 = {
   key: "openai.chat-mock.v1",
@@ -2393,10 +2414,12 @@ var openai = {
   "openai.gpt-4.1-nano": withDefaultModel(openAiChatV1, "gpt-4.1-nano"),
   // Reasoning models
   "openai.o3": withDefaultModel(openAiChatV1, "o3"),
-  "openai.o4-mini": withDefaultModel(openAiChatV1, "o4-mini"),
   // GPT-4o family
+  "openai.gpt-4": withDefaultModel(openAiChatV1, "gpt-4"),
   "openai.gpt-4o": withDefaultModel(openAiChatV1, "gpt-4o"),
-  "openai.gpt-4o-mini": withDefaultModel(openAiChatV1, "gpt-4o-mini")
+  "openai.gpt-4o-mini": withDefaultModel(openAiChatV1, "gpt-4o-mini"),
+  // Deprecated
+  "openai.o4-mini": withDefaultModel(openAiChatV1, "o4-mini")
 };
 
 // src/llm/config/anthropic/promptSanitizeMessageCallback.ts
@@ -2656,6 +2679,14 @@ var bedrock = {
 
 // src/llm/config/anthropic/index.ts
 var ANTHROPIC_VERSION = "2023-06-01";
+var MODELS_REJECTING_SAMPLING_PARAMS = ["claude-opus-4-7"];
+var isClaude4x = (model) => /^claude-(opus|sonnet|haiku)-4-/.test(model);
+var dropIfModelRejectsSamplingParams = (v, body) => MODELS_REJECTING_SAMPLING_PARAMS.includes(body.model) ? void 0 : v;
+var topPTransform = (v, body) => {
+  if (MODELS_REJECTING_SAMPLING_PARAMS.includes(body.model)) return void 0;
+  if (isClaude4x(body.model) && body.temperature !== void 0) return void 0;
+  return v;
+};
 var anthropicChatV1 = {
   key: "anthropic.chat.v1",
   provider: "anthropic.chat",
@@ -2688,13 +2719,16 @@ var anthropicChatV1 = {
       transform: anthropicPromptSanitize
     },
     temperature: {
-      key: "temperature"
+      key: "temperature",
+      transform: dropIfModelRejectsSamplingParams
     },
     topP: {
-      key: "top_p"
+      key: "top_p",
+      transform: topPTransform
     },
     topK: {
-      key: "top_k"
+      key: "top_k",
+      transform: dropIfModelRejectsSamplingParams
     },
     stopSequences: {
       key: "stop_sequences"
@@ -2730,6 +2764,11 @@ var anthropicChatV1 = {
 };
 var anthropic = {
   "anthropic.chat.v1": anthropicChatV1,
+  // Claude 4.7 models
+  "anthropic.claude-opus-4-7": withDefaultModel(
+    anthropicChatV1,
+    "claude-opus-4-7"
+  ),
   // Claude 4.6 models
   "anthropic.claude-opus-4-6": withDefaultModel(
     anthropicChatV1,
@@ -2739,7 +2778,20 @@ var anthropic = {
     anthropicChatV1,
     "claude-sonnet-4-6"
   ),
-  // Claude 4 models
+  // Claude 4.5 models
+  "anthropic.claude-haiku-4-5": withDefaultModel(
+    anthropicChatV1,
+    "claude-haiku-4-5"
+  ),
+  "anthropic.claude-sonnet-4-5": withDefaultModel(
+    anthropicChatV1,
+    "claude-sonnet-4-5"
+  ),
+  // Deprecated
+  "anthropic.claude-opus-4-1": withDefaultModel(
+    anthropicChatV1,
+    "claude-opus-4-1-20250805"
+  ),
   "anthropic.claude-sonnet-4": withDefaultModel(
     anthropicChatV1,
     "claude-sonnet-4-0"
@@ -2748,7 +2800,6 @@ var anthropic = {
     anthropicChatV1,
     "claude-opus-4-0"
   ),
-  // Deprecated
   "anthropic.claude-3-7-sonnet": withDefaultModel(
     anthropicChatV1,
     "claude-3-7-sonnet-20250219"
@@ -2764,10 +2815,6 @@ var anthropic = {
   "anthropic.claude-3-opus": withDefaultModel(
     anthropicChatV1,
     "claude-3-opus-20240229"
-  ),
-  "anthropic.claude-3-haiku": withDefaultModel(
-    anthropicChatV1,
-    "claude-3-haiku-20240307"
   )
 };
 
@@ -2776,7 +2823,13 @@ var xaiChatV1 = createOpenAiCompatibleConfiguration({
   key: "xai.chat.v1",
   provider: "xai.chat",
   endpoint: `https://api.x.ai/v1/chat/completions`,
-  apiKeyMapping: ["xAiApiKey", "XAI_API_KEY"]
+  apiKeyMapping: ["xAiApiKey", "XAI_API_KEY"],
+  // Per xAI docs (checked 2026-05-13), grok-4 errors if `reasoning_effort` is
+  // sent; grok-3-mini and the fast-non-reasoning variants don't accept it
+  // either. Only grok-4.3 and grok-4.20-multi-agent support it, and neither
+  // has a shorthand here yet. Leave the predicate empty so effort is dropped
+  // for every currently-shipped xAI shorthand.
+  isReasoningModel: () => false
 });
 var xai = {
   "xai.chat.v1": xaiChatV1,
@@ -2784,7 +2837,8 @@ var xai = {
   "xai.grok-3": withDefaultModel(xaiChatV1, "grok-3"),
   "xai.grok-3-mini": withDefaultModel(xaiChatV1, "grok-3-mini"),
   "xai.grok-4": withDefaultModel(xaiChatV1, "grok-4"),
-  "xai.grok-4-fast": withDefaultModel(xaiChatV1, "grok-4-fast-non-reasoning")
+  "xai.grok-4-fast": withDefaultModel(xaiChatV1, "grok-4-fast-non-reasoning"),
+  "xai.grok-4-1-fast": withDefaultModel(xaiChatV1, "grok-4-1-fast-non-reasoning")
 };
 
 // src/llm/output/_utils/combineJsonl.ts
@@ -2891,7 +2945,11 @@ var ollama = {
   "ollama.llama3.3": withDefaultModel(ollamaChatV1, "llama3.3"),
   "ollama.llama3.2": withDefaultModel(ollamaChatV1, "llama3.2"),
   "ollama.llama3.1": withDefaultModel(ollamaChatV1, "llama3.1"),
-  "ollama.qwq": withDefaultModel(ollamaChatV1, "qwq")
+  "ollama.qwq": withDefaultModel(ollamaChatV1, "qwq"),
+  "ollama.gemma3": withDefaultModel(ollamaChatV1, "gemma3"),
+  "ollama.mistral": withDefaultModel(ollamaChatV1, "mistral"),
+  "ollama.qwen2.5": withDefaultModel(ollamaChatV1, "qwen2.5"),
+  "ollama.qwen3": withDefaultModel(ollamaChatV1, "qwen3")
 };
 
 // src/llm/config/google/promptSanitizeMessageCallback.ts
@@ -3108,14 +3166,6 @@ var googleGeminiChatV1 = {
 };
 var google = {
   "google.chat.v1": googleGeminiChatV1,
-  "google.gemini-2.0-flash": withDefaultModel(
-    googleGeminiChatV1,
-    "gemini-2.0-flash"
-  ),
-  "google.gemini-2.0-flash-lite": withDefaultModel(
-    googleGeminiChatV1,
-    "gemini-2.0-flash-lite"
-  ),
   "google.gemini-2.5-flash": withDefaultModel(
     googleGeminiChatV1,
     "gemini-2.5-flash"
@@ -3124,13 +3174,22 @@ var google = {
     googleGeminiChatV1,
     "gemini-2.5-flash-lite"
   ),
-  "google.gemini-1.5-pro": withDefaultModel(
-    googleGeminiChatV1,
-    "gemini-1.5-pro"
-  ),
   "google.gemini-2.5-pro": withDefaultModel(
     googleGeminiChatV1,
     "gemini-2.5-pro"
+  ),
+  // Deprecated
+  "google.gemini-2.0-flash": withDefaultModel(
+    googleGeminiChatV1,
+    "gemini-2.0-flash"
+  ),
+  "google.gemini-2.0-flash-lite": withDefaultModel(
+    googleGeminiChatV1,
+    "gemini-2.0-flash-lite"
+  ),
+  "google.gemini-1.5-pro": withDefaultModel(
+    googleGeminiChatV1,
+    "gemini-1.5-pro"
   )
 };
 
@@ -3143,7 +3202,9 @@ var deepseekChatV1 = createOpenAiCompatibleConfiguration({
 });
 var deepseek = {
   "deepseek.chat.v1": deepseekChatV1,
-  "deepseek.chat": withDefaultModel(deepseekChatV1, "deepseek-chat")
+  "deepseek.chat": withDefaultModel(deepseekChatV1, "deepseek-chat"),
+  "deepseek.v4-flash": withDefaultModel(deepseekChatV1, "deepseek-v4-flash"),
+  "deepseek.v4-pro": withDefaultModel(deepseekChatV1, "deepseek-v4-pro")
 };
 
 // src/llm/config.ts
@@ -3258,12 +3319,19 @@ async function apiRequest(url, options) {
     const response = await fetch(url, finalOptions);
     if (!response.ok) {
       let message = `HTTP error. Status: ${response.status}. Error Message: ${response?.statusText || "Unknown error."}`;
-      if (url.startsWith("https://api.openai.com/")) {
-        try {
-          const body = await response.json();
-          message = `HTTP error. Status Code: ${response.status}. Error Message: ${body?.error?.message || "No further details provided."}`;
-        } catch (error) {
+      try {
+        const text = await response.text();
+        if (text) {
+          let detail = text;
+          try {
+            const body = JSON.parse(text);
+            detail = body?.error?.message || body?.error || body?.message || text;
+            if (typeof detail !== "string") detail = JSON.stringify(detail);
+          } catch {
+          }
+          message = `HTTP error. Status Code: ${response.status}. Error Message: ${detail}`;
         }
+      } catch {
       }
       throw new Error(message);
     }
@@ -3843,6 +3911,26 @@ function AmazonTitanEmbedding(result, config) {
   });
 }
 
+// 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/config.ts
 var embeddingConfigs = {
   "openai.embedding.v1": {
@@ -3904,6 +3992,43 @@ var embeddingConfigs = {
       }
     },
     transformResponse: AmazonTitanEmbedding
+  },
+  "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: {}
+    },
+    mapBody: {
+      input: {
+        key: "texts",
+        transform: (value) => Array.isArray(value) ? value : [value]
+      },
+      inputType: {
+        key: "input_type"
+      },
+      truncate: {
+        key: "truncate"
+      },
+      dimensions: {
+        key: "output_dimension"
+      }
+    },
+    transformResponse: CohereBedrockEmbedding
   }
 };
 function getEmbeddingConfig(provider) {
@@ -3924,6 +4049,8 @@ 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");
   }
@@ -4095,6 +4222,11 @@ var BasePrompt = class {
     return promptValue;
   }
   getReplacements(values) {
+    if (values === void 0 || values === null) {
+      throw new Error(
+        "format() requires an input object. Did you forget to pass arguments?"
+      );
+    }
     const { input = "", ...restOfValues } = values;
     const replacements = Object.assign(
       {},
@@ -4107,11 +4239,11 @@ var BasePrompt = class {
     return replacements;
   }
   /**
-   * validate description
-   * @return {boolean} Returns false if the template is not valid.
+   * Validates the prompt structure.
+   * @return {boolean} Returns false if the prompt has no messages defined.
    */
   validate() {
-    return true;
+    return this.messages.length > 0;
   }
 };
 
@@ -4708,14 +4840,6 @@ var ChatPrompt = class extends BasePrompt {
     }
     return messagesOut;
   }
-  /**
-   * validate Ensures there are not unresolved tokens in prompt.
-   * @TODO Make this work!
-   * @return Returns false if the template is not valid.
-   */
-  validate() {
-    return true;
-  }
 };
 
 // src/prompt/_functions.ts
@@ -4742,10 +4866,12 @@ var BaseStateItem = class {
     this.initialValue = initialValue;
   }
   setValue(value) {
-    assert(
-      typeof value === typeof this.value,
-      `Invalid value type. Expected ${typeof this.value}, received ${typeof value}`
-    );
+    if (this.value !== void 0 && this.value !== null) {
+      assert(
+        typeof value === typeof this.value,
+        `Invalid value type. Expected ${typeof this.value}, received ${typeof value}`
+      );
+    }
     this.value = value;
   }
   getKey() {
@@ -4881,11 +5007,42 @@ var Dialogue = class extends BaseStateItem {
     return this;
   }
   setMessageTurn(userMessage, assistantMessage, systemMessage = "") {
+    this.setSystemMessage(systemMessage);
     this.setUserMessage(userMessage);
     this.setAssistantMessage(assistantMessage);
-    this.setSystemMessage(systemMessage);
     return this;
   }
+  /**
+   * Aliases using `add*` naming to match ChatPrompt's API.
+   * These delegate to the corresponding `set*` methods.
+   */
+  addUserMessage(content, name) {
+    return this.setUserMessage(content, name);
+  }
+  addAssistantMessage(content) {
+    return this.setAssistantMessage(content);
+  }
+  addSystemMessage(content) {
+    return this.setSystemMessage(content);
+  }
+  addToolMessage(content, name, id) {
+    return this.setToolMessage(content, name, id);
+  }
+  addToolCallMessage(input) {
+    return this.setToolCallMessage(input);
+  }
+  addFunctionMessage(content, name, id) {
+    return this.setFunctionMessage(content, name, id);
+  }
+  addFunctionCallMessage(input) {
+    return this.setFunctionCallMessage(input);
+  }
+  addMessageTurn(userMessage, assistantMessage, systemMessage = "") {
+    return this.setMessageTurn(userMessage, assistantMessage, systemMessage);
+  }
+  addHistory(messages) {
+    return this.setHistory(messages);
+  }
   setHistory(messages) {
     for (const message of messages) {
       switch (message?.role) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "llm-exe",
-  "version": "2.3.5",
+  "version": "2.3.7",
   "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",

package/readme.md

@@ -104,11 +104,34 @@ const parser = createCustomParser("MyUppercaseParser", (output, input) => {
 
 #### State
 
+Manage conversation history and structured data across LLM calls:
+
 ```ts
+import { createState, createDialogue, createStateItem } from "llm-exe";
+
+// Create a state container
+const state = createState();
+
+// Dialogues — store conversation history
+const chat = state.createDialogue("chat");
+chat.setUserMessage("Hi");
+chat.setAssistantMessage("Hello!");
+chat.getHistory(); // returns message array
+
+// Standalone dialogue (without state)
 const dialogue = createDialogue("chat");
 dialogue.setUserMessage("Hi");
-dialogue.setAssistantMessage("Hello!");
-dialogue.getHistory(); // returns chat array
+
+// Context items — typed values with get/set/reset
+const intent = createStateItem("userIntent", "unknown");
+state.createContextItem(intent);
+intent.setValue("booking");
+intent.getValue();    // "booking"
+intent.resetValue();  // resets to "unknown"
+
+// Attributes — simple key-value metadata
+state.setAttribute("userId", "abc-123");
+state.attributes["userId"]; // "abc-123"
 ```
 
 #### Hooks