@promptbook/wizard 0.100.0-25 → 0.100.0-28

package/esm/typings/src/execution/createPipelineExecutor/40-executeAttempts.d.ts

@@ -1,10 +1,11 @@
-import type { ReadonlyDeep, WritableDeep } from 'type-fest';
+import type { PartialDeep, Promisable, ReadonlyDeep, WritableDeep } from 'type-fest';
 import type { PipelineJson } from '../../pipeline/PipelineJson/PipelineJson';
 import type { TaskJson } from '../../pipeline/PipelineJson/TaskJson';
 import type { Parameters } from '../../types/typeAliases';
 import type { string_parameter_name } from '../../types/typeAliases';
 import type { TODO_string } from '../../utils/organization/TODO_string';
 import type { ExecutionReportJson } from '../execution-report/ExecutionReportJson';
+import type { PipelineExecutorResult } from '../PipelineExecutorResult';
 import type { CreatePipelineExecutorOptions } from './00-CreatePipelineExecutorOptions';
 /**
  * Options for executing attempts of a pipeline task, including configuration for jokers, priority,
@@ -46,6 +47,10 @@ export type ExecuteAttemptsOptions = Required<Omit<CreatePipelineExecutorOptions
      * The pipeline structure prepared for execution, as a deeply immutable PipelineJson object.
      */
     readonly preparedPipeline: ReadonlyDeep<PipelineJson>;
+    /**
+     * Callback invoked with partial results as the execution progresses.
+     */
+    onProgress(newOngoingResult: PartialDeep<PipelineExecutorResult>): Promisable<void>;
     /**
      * The execution report object, which is updated during execution.
      */

package/esm/typings/src/llm-providers/anthropic-claude/AnthropicClaudeExecutionTools.d.ts

@@ -2,7 +2,6 @@ import Anthropic from '@anthropic-ai/sdk';
 import type { AvailableModel } from '../../execution/AvailableModel';
 import type { LlmExecutionTools } from '../../execution/LlmExecutionTools';
 import type { ChatPromptResult } from '../../execution/PromptResult';
-import type { CompletionPromptResult } from '../../execution/PromptResult';
 import type { Prompt } from '../../types/Prompt';
 import type { string_markdown } from '../../types/typeAliases';
 import type { string_markdown_text } from '../../types/typeAliases';
@@ -42,10 +41,6 @@ export declare class AnthropicClaudeExecutionTools implements LlmExecutionTools
      * Calls Anthropic Claude API to use a chat model.
      */
     callChatModel(prompt: Pick<Prompt, 'content' | 'parameters' | 'modelRequirements'>): Promise<ChatPromptResult>;
-    /**
-     * Calls Anthropic Claude API to use a completion model.
-     */
-    callCompletionModel(prompt: Pick<Prompt, 'content' | 'parameters' | 'modelRequirements'>): Promise<CompletionPromptResult>;
     /**
      * Get the model that should be used as default
      */

package/esm/typings/src/types/ModelRequirements.d.ts

@@ -87,8 +87,6 @@ export type CommonModelRequirements = {
     readonly seed?: number_seed;
     /**
      * Maximum number of tokens that can be generated by the model
-     *
-     * Note: [🌾]
      */
     readonly maxTokens?: number;
 };

package/esm/typings/src/version.d.ts

@@ -15,7 +15,7 @@ export declare const BOOK_LANGUAGE_VERSION: string_semantic_version;
 export declare const PROMPTBOOK_ENGINE_VERSION: string_promptbook_version;
 /**
  * Represents the version string of the Promptbook engine.
- * It follows semantic versioning (e.g., `0.100.0-24`).
+ * It follows semantic versioning (e.g., `0.100.0-27`).
  *
  * @generated
  */

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@promptbook/wizard",
-    "version": "0.100.0-25",
+    "version": "0.100.0-28",
     "description": "Promptbook: Run AI apps in plain human language across multiple models and platforms",
     "private": false,
     "sideEffects": false,
@@ -95,7 +95,7 @@
     "module": "./esm/index.es.js",
     "typings": "./esm/typings/src/_packages/wizard.index.d.ts",
     "peerDependencies": {
-        "@promptbook/core": "0.100.0-25"
+        "@promptbook/core": "0.100.0-28"
     },
     "dependencies": {
         "@ai-sdk/deepseek": "0.1.6",

package/umd/index.umd.js

@@ -49,7 +49,7 @@
      * @generated
      * @see https://github.com/webgptorg/promptbook
      */
-    const PROMPTBOOK_ENGINE_VERSION = '0.100.0-25';
+    const PROMPTBOOK_ENGINE_VERSION = '0.100.0-28';
     /**
      * TODO: string_promptbook_version should be constrained to the all versions of Promptbook engine
      * Note: [💞] Ignore a discrepancy between file name and entity name
@@ -2434,8 +2434,7 @@
             const rawPromptContent = templateParameters(content, { ...parameters, modelName });
             const rawRequest = {
                 model: modelRequirements.modelName || this.getDefaultChatModel().modelName,
-                max_tokens: modelRequirements.maxTokens || 4096,
-                //                                            <- TODO: [🌾] Make some global max cap for maxTokens
+                max_tokens: modelRequirements.maxTokens || 8192,
                 temperature: modelRequirements.temperature,
                 system: modelRequirements.systemMessage,
                 messages: [
@@ -2494,59 +2493,6 @@
                 },
             });
         }
-        /**
-         * Calls Anthropic Claude API to use a completion model.
-         */
-        async callCompletionModel(prompt) {
-            if (this.options.isVerbose) {
-                console.info('🖋 Anthropic Claude callCompletionModel call');
-            }
-            const { content, parameters, modelRequirements } = prompt;
-            if (modelRequirements.modelVariant !== 'COMPLETION') {
-                throw new PipelineExecutionError('Use callCompletionModel only for COMPLETION variant');
-            }
-            const client = await this.getClient();
-            const modelName = modelRequirements.modelName || this.getDefaultChatModel().modelName;
-            const rawPromptContent = templateParameters(content, { ...parameters, modelName });
-            const rawRequest = {
-                model: modelName,
-                max_tokens_to_sample: modelRequirements.maxTokens || 2000,
-                temperature: modelRequirements.temperature,
-                prompt: rawPromptContent,
-            };
-            const start = $getCurrentDate();
-            const rawResponse = await this.limiter
-                .schedule(() => client.completions.create(rawRequest))
-                .catch((error) => {
-                if (this.options.isVerbose) {
-                    console.info(colors__default["default"].bgRed('error'), error);
-                }
-                throw error;
-            });
-            if (this.options.isVerbose) {
-                console.info(colors__default["default"].bgWhite('rawResponse'), JSON.stringify(rawResponse, null, 4));
-            }
-            if (!rawResponse.completion) {
-                throw new PipelineExecutionError('No completion from Anthropic Claude');
-            }
-            const resultContent = rawResponse.completion;
-            const complete = $getCurrentDate();
-            const usage = computeAnthropicClaudeUsage(rawPromptContent, resultContent, rawResponse);
-            return exportJson({
-                name: 'promptResult',
-                message: `Result of \`AnthropicClaudeExecutionTools.callCompletionModel\``,
-                order: [],
-                value: {
-                    content: resultContent,
-                    modelName: rawResponse.model || modelName,
-                    timing: { start, complete },
-                    usage,
-                    rawPromptContent,
-                    rawRequest,
-                    rawResponse,
-                },
-            });
-        }
         // <- Note: [🤖] callXxxModel
         /**
          * Get the model that should be used as default
@@ -3252,7 +3198,6 @@
                 const modelName = prompt.modelRequirements.modelName || this.options.deploymentName;
                 const modelSettings = {
                     maxTokens: modelRequirements.maxTokens,
-                    //                                      <- TODO: [🌾] Make some global max cap for maxTokens
                     temperature: modelRequirements.temperature,
                     user: (_a = this.options.userId) === null || _a === void 0 ? void 0 : _a.toString(),
                     // <- TODO: [🈁] Use `seed` here AND/OR use is `isDeterministic` for entire execution tools
@@ -3358,8 +3303,7 @@
             try {
                 const modelName = prompt.modelRequirements.modelName || this.options.deploymentName;
                 const modelSettings = {
-                    maxTokens: modelRequirements.maxTokens || 2000,
-                    //                                                  <- TODO: [🌾] Make some global max cap for maxTokens
+                    maxTokens: modelRequirements.maxTokens,
                     temperature: modelRequirements.temperature,
                     user: (_a = this.options.userId) === null || _a === void 0 ? void 0 : _a.toString(),
                     // <- TODO: [🈁] Use `seed` here AND/OR use is `isDeterministic` for entire execution tools
@@ -4448,7 +4392,6 @@
             const modelSettings = {
                 model: modelName,
                 max_tokens: modelRequirements.maxTokens,
-                //                                   <- TODO: [🌾] Make some global max cap for maxTokens
                 temperature: modelRequirements.temperature,
                 // <- TODO: [🈁] Use `seed` here AND/OR use is `isDeterministic` for entire execution tools
                 // <- Note: [🧆]
@@ -4544,8 +4487,7 @@
             const modelName = modelRequirements.modelName || this.getDefaultCompletionModel().modelName;
             const modelSettings = {
                 model: modelName,
-                max_tokens: modelRequirements.maxTokens || 2000,
-                //                                                  <- TODO: [🌾] Make some global max cap for maxTokens
+                max_tokens: modelRequirements.maxTokens,
                 temperature: modelRequirements.temperature,
                 // <- TODO: [🈁] Use `seed` here AND/OR use is `isDeterministic` for entire execution tools
                 // <- Note: [🧆]
@@ -5264,8 +5206,6 @@
             const modelName = modelRequirements.modelName || this.getDefaultChatModel().modelName;
             const modelSettings = {
                 model: modelName,
-                max_tokens: modelRequirements.maxTokens,
-                //                                   <- TODO: [🌾] Make some global max cap for maxTokens
 
                 temperature: modelRequirements.temperature,
 
@@ -8943,7 +8883,7 @@
      */
     async function executeAttempts(options) {
         const { jokerParameterNames, priority, maxAttempts, // <- Note: [💂]
-        preparedContent, parameters, task, preparedPipeline, tools, $executionReport, pipelineIdentification, maxExecutionAttempts, } = options;
+        preparedContent, parameters, task, preparedPipeline, tools, $executionReport, pipelineIdentification, maxExecutionAttempts, onProgress, } = options;
         const $ongoingTaskResult = {
             $result: null,
             $resultString: null,
@@ -9187,6 +9127,10 @@
                     result: $ongoingTaskResult.$resultString,
                     error: error,
                 });
+                // Report failed attempt
+                onProgress({
+                    errors: [error],
+                });
             }
             finally {
                 if (!isJokerAttempt &&