@promptbook/core 0.104.0-2 → 0.104.0-4

package/esm/index.es.js

@@ -27,7 +27,7 @@ const BOOK_LANGUAGE_VERSION = '2.0.0';
  * @generated
  * @see https://github.com/webgptorg/promptbook
  */
-const PROMPTBOOK_ENGINE_VERSION = '0.104.0-2';
+const PROMPTBOOK_ENGINE_VERSION = '0.104.0-4';
 /**
  * TODO: string_promptbook_version should be constrained to the all versions of Promptbook engine
  * Note: [💞] Ignore a discrepancy between file name and entity name
@@ -3744,6 +3744,15 @@ function countUsage(llmTools) {
             return promptResult;
         };
     }
+    if (llmTools.callImageGenerationModel !== undefined) {
+        proxyTools.callImageGenerationModel = async (prompt) => {
+            // console.info('[🚕] callImageGenerationModel through countTotalUsage');
+            const promptResult = await llmTools.callImageGenerationModel(prompt);
+            totalUsage = addUsage(totalUsage, promptResult.usage);
+            spending.next(promptResult.usage);
+            return promptResult;
+        };
+    }
     // <- Note: [🤖]
     return proxyTools;
 }
@@ -3853,6 +3862,12 @@ class MultipleLlmExecutionTools {
     callEmbeddingModel(prompt) {
         return this.callCommonModel(prompt);
     }
+    /**
+     * Calls the best available embedding model
+     */
+    callImageGenerationModel(prompt) {
+        return this.callCommonModel(prompt);
+    }
     // <- Note: [🤖]
     /**
      * Calls the best available model
@@ -3879,6 +3894,11 @@ class MultipleLlmExecutionTools {
                             continue llm;
                         }
                         return await llmExecutionTools.callEmbeddingModel(prompt);
+                    case 'IMAGE_GENERATION':
+                        if (llmExecutionTools.callImageGenerationModel === undefined) {
+                            continue llm;
+                        }
+                        return await llmExecutionTools.callImageGenerationModel(prompt);
                     // <- case [🤖]:
                     default:
                         throw new UnexpectedError(`Unknown model variant "${prompt.modelRequirements.modelVariant}" in ${llmExecutionTools.title}`);
@@ -6304,8 +6324,9 @@ async function executeAttempts(options) {
                                     $ongoingTaskResult.$resultString = $ongoingTaskResult.$completionResult.content;
                                     break variant;
                                 case 'EMBEDDING':
+                                case 'IMAGE_GENERATION':
                                     throw new PipelineExecutionError(spaceTrim$1((block) => `
-                                                Embedding model can not be used in pipeline
+                                                ${modelRequirements.modelVariant} model can not be used in pipeline
 
                                                 This should be catched during parsing
 
@@ -11044,17 +11065,64 @@ function parseAgentSourceWithCommitments(agentSource) {
         };
     }
     const lines = agentSource.split('\n');
-    const agentName = (((_a = lines[0]) === null || _a === void 0 ? void 0 : _a.trim()) || null);
+    let agentName = null;
+    let agentNameLineIndex = -1;
+    // Find the agent name: first non-empty line that is not a commitment and not a horizontal line
+    for (let i = 0; i < lines.length; i++) {
+        const line = lines[i];
+        if (line === undefined) {
+            continue;
+        }
+        const trimmed = line.trim();
+        if (!trimmed) {
+            continue;
+        }
+        const isHorizontal = HORIZONTAL_LINE_PATTERN.test(line);
+        if (isHorizontal) {
+            continue;
+        }
+        let isCommitment = false;
+        for (const definition of COMMITMENT_REGISTRY) {
+            const typeRegex = definition.createTypeRegex();
+            const match = typeRegex.exec(trimmed);
+            if (match && ((_a = match.groups) === null || _a === void 0 ? void 0 : _a.type)) {
+                isCommitment = true;
+                break;
+            }
+        }
+        if (!isCommitment) {
+            agentName = trimmed;
+            agentNameLineIndex = i;
+            break;
+        }
+    }
     const commitments = [];
     const nonCommitmentLines = [];
-    // Always add the first line (agent name) to non-commitment lines
-    if (lines[0] !== undefined) {
-        nonCommitmentLines.push(lines[0]);
+    // Add lines before agentName that are horizontal lines (they are non-commitment)
+    for (let i = 0; i < agentNameLineIndex; i++) {
+        const line = lines[i];
+        if (line === undefined) {
+            continue;
+        }
+        const trimmed = line.trim();
+        if (!trimmed) {
+            continue;
+        }
+        const isHorizontal = HORIZONTAL_LINE_PATTERN.test(line);
+        if (isHorizontal) {
+            nonCommitmentLines.push(line);
+        }
+        // Note: Commitments before agentName are not added to nonCommitmentLines
+    }
+    // Add the agent name line to non-commitment lines
+    if (agentNameLineIndex >= 0) {
+        nonCommitmentLines.push(lines[agentNameLineIndex]);
     }
     // Parse commitments with multiline support
     let currentCommitment = null;
-    // Process lines starting from the second line (skip agent name)
-    for (let i = 1; i < lines.length; i++) {
+    // Process lines starting from after the agent name line
+    const startIndex = agentNameLineIndex >= 0 ? agentNameLineIndex + 1 : 0;
+    for (let i = startIndex; i < lines.length; i++) {
         const line = lines[i];
         if (line === undefined) {
             continue;
@@ -16871,6 +16939,9 @@ function cacheLlmTools(llmTools, options = {}) {
             case 'EMBEDDING':
                 promptResult = await llmTools.callEmbeddingModel(prompt);
                 break variant;
+            case 'IMAGE_GENERATION':
+                promptResult = await llmTools.callImageGenerationModel(prompt);
+                break variant;
             // <- case [🤖]:
             default:
                 throw new PipelineExecutionError(`Unknown model variant "${prompt.modelRequirements.modelVariant}"`);
@@ -16958,6 +17029,11 @@ function cacheLlmTools(llmTools, options = {}) {
             return /* not await */ callCommonModel(prompt);
         };
     }
+    if (llmTools.callImageGenerationModel !== undefined) {
+        proxyTools.callImageGenerationModel = async (prompt) => {
+            return /* not await */ callCommonModel(prompt);
+        };
+    }
     // <- Note: [🤖]
     return proxyTools;
 }
@@ -16996,6 +17072,11 @@ function limitTotalUsage(llmTools, options = {}) {
             throw new LimitReachedError('Cannot call `callEmbeddingModel` because you are not allowed to spend any cost');
         };
     }
+    if (proxyTools.callImageGenerationModel !== undefined) {
+        proxyTools.callImageGenerationModel = async (prompt) => {
+            throw new LimitReachedError('Cannot call `callImageGenerationModel` because you are not allowed to spend any cost');
+        };
+    }
     // <- Note: [🤖]
     return proxyTools;
 }
@@ -18067,7 +18148,7 @@ class OpenAiCompatibleExecutionTools {
         let threadMessages = [];
         if ('thread' in prompt && Array.isArray(prompt.thread)) {
             threadMessages = prompt.thread.map((msg) => ({
-                role: msg.role === 'assistant' ? 'assistant' : 'user',
+                role: msg.sender === 'assistant' ? 'assistant' : 'user',
                 content: msg.content,
             }));
         }