@promptbook/browser 0.104.0-9 → 0.104.0

package/README.md

@@ -27,10 +27,6 @@ Turn your company's scattered knowledge into AI ready Books
 
 
 
-<blockquote style="color: #ff8811">
-    <b>⚠ Warning:</b> This is a pre-release version of the library. It is not yet ready for production use. Please look at <a href="https://www.npmjs.com/package/@promptbook/core?activeTab=versions">latest stable release</a>.
-</blockquote>
-
 ## 📦 Package `@promptbook/browser`
 
 - Promptbooks are [divided into several](#-packages) packages, all are published from [single monorepo](https://github.com/webgptorg/promptbook).

package/esm/index.es.js

@@ -19,7 +19,7 @@ const BOOK_LANGUAGE_VERSION = '2.0.0';
  * @generated
  * @see https://github.com/webgptorg/promptbook
  */
-const PROMPTBOOK_ENGINE_VERSION = '0.104.0-9';
+const PROMPTBOOK_ENGINE_VERSION = '0.104.0';
 /**
  * TODO: string_promptbook_version should be constrained to the all versions of Promptbook engine
  * Note: [💞] Ignore a discrepancy between file name and entity name
@@ -2208,9 +2208,10 @@ function isValidFilePath(filename) {
  *
  * Note: [🔂] This function is idempotent.
  * Note: Dataurl are considered perfectly valid.
- * Note: There are two similar functions:
- * - `isValidUrl` which tests any URL
- * - `isValidPipelineUrl` *(this one)* which tests just promptbook URL
+ * Note: There are few similar functions:
+ * - `isValidUrl` *(this one)* which tests any URL
+ * - `isValidAgentUrl` which tests just agent URL
+ * - `isValidPipelineUrl` which tests just pipeline URL
  *
  * @public exported from `@promptbook/utils`
  */
@@ -2488,8 +2489,8 @@ for (let i = 0; i < defaultDiacriticsRemovalMap.length; i++) {
  */
 function removeDiacritics(input) {
     /*eslint no-control-regex: "off"*/
-    return input.replace(/[^\u0000-\u007E]/g, (a) => {
-        return DIACRITIC_VARIANTS_LETTERS[a] || a;
+    return input.replace(/[^\u0000-\u007E]/g, (character) => {
+        return DIACRITIC_VARIANTS_LETTERS[character] || character;
     });
 }
 /**
@@ -2620,6 +2621,39 @@ function titleToName(value) {
  * Note: [💞] Ignore a discrepancy between file name and entity name
  */
 
+/**
+ * Tests if given string is valid agent URL
+ *
+ * Note: There are few similar functions:
+ * - `isValidUrl` which tests any URL
+ * - `isValidAgentUrl` *(this one)* which tests just agent URL
+ * - `isValidPipelineUrl` which tests just pipeline URL
+ *
+ * @public exported from `@promptbook/utils`
+ */
+function isValidAgentUrl(url) {
+    if (!isValidUrl(url)) {
+        return false;
+    }
+    if (!url.startsWith('https://') && !url.startsWith('http://') /* <- Note: [👣] */) {
+        return false;
+    }
+    if (url.includes('#')) {
+        // TODO: [🐠]
+        return false;
+    }
+    /*
+    Note: [👣][🧠] Is it secure to allow pipeline URLs on private and unsecured networks?
+    if (isUrlOnPrivateNetwork(url)) {
+        return false;
+    }
+    */
+    return true;
+}
+/**
+ * TODO: [🐠] Maybe more info why the URL is invalid
+ */
+
 /**
  * Normalizes agent name from arbitrary string to valid agent name
  *
@@ -2651,8 +2685,8 @@ function createDefaultAgentName(agentSource) {
  */
 function createCommitmentRegex(commitment, aliases = [], requiresContent = true) {
     const allCommitments = [commitment, ...aliases];
-    const patterns = allCommitments.map((c) => {
-        const escapedCommitment = c.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+    const patterns = allCommitments.map((commitment) => {
+        const escapedCommitment = commitment.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
         return escapedCommitment.split(/\s+/).join('\\s+');
     });
     const keywordPattern = patterns.join('|');
@@ -2674,8 +2708,8 @@ function createCommitmentRegex(commitment, aliases = [], requiresContent = true)
  */
 function createCommitmentTypeRegex(commitment, aliases = []) {
     const allCommitments = [commitment, ...aliases];
-    const patterns = allCommitments.map((c) => {
-        const escapedCommitment = c.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
+    const patterns = allCommitments.map((commitment) => {
+        const escapedCommitment = commitment.replace(/[.*+?^${}()|[\]\\]/g, '\\$&');
         return escapedCommitment.split(/\s+/).join('\\s+');
     });
     const keywordPattern = patterns.join('|');
@@ -3185,9 +3219,7 @@ class DictionaryCommitmentDefinition extends BaseCommitmentDefinition {
         // Get existing dictionary entries from metadata
         const existingDictionary = ((_a = requirements.metadata) === null || _a === void 0 ? void 0 : _a.DICTIONARY) || '';
         // Merge the new dictionary entry with existing entries
-        const mergedDictionary = existingDictionary
-            ? `${existingDictionary}\n${trimmedContent}`
-            : trimmedContent;
+        const mergedDictionary = existingDictionary ? `${existingDictionary}\n${trimmedContent}` : trimmedContent;
         // Store the merged dictionary in metadata for debugging and inspection
         const updatedMetadata = {
             ...requirements.metadata,
@@ -3338,21 +3370,35 @@ class FromCommitmentDefinition extends BaseCommitmentDefinition {
     applyToAgentModelRequirements(requirements, content) {
         const trimmedContent = content.trim();
         if (!trimmedContent) {
-            return requirements;
+            return {
+                ...requirements,
+                parentAgentUrl: undefined,
+            };
         }
-        // Validate URL
-        try {
-            // eslint-disable-next-line @typescript-eslint/no-unused-vars
-            const url = new URL(trimmedContent);
-            // TODO: Add more validation if needed (e.g. check for valid protocol)
+        if (trimmedContent.toUpperCase() === 'VOID' ||
+            trimmedContent.toUpperCase() === 'NULL' ||
+            trimmedContent.toUpperCase() === 'NONE' ||
+            trimmedContent.toUpperCase() === 'NIL') {
+            return {
+                ...requirements,
+                parentAgentUrl: null,
+            };
         }
-        catch (error) {
-            console.warn(`Invalid URL in FROM commitment: ${trimmedContent}`);
-            return requirements;
+        if (!isValidAgentUrl(trimmedContent)) {
+            throw new Error(spaceTrim$1((block) => `
+                        Invalid agent URL in FROM commitment: "${trimmedContent}"
+
+                        \`\`\`book
+                        ${block(content)}
+                        \`\`\`
+
+
+                `));
         }
+        const parentAgentUrl = trimmedContent;
         return {
             ...requirements,
-            parentAgentUrl: trimmedContent,
+            parentAgentUrl,
         };
     }
 }
@@ -5716,19 +5762,38 @@ class UseBrowserCommitmentDefinition extends BaseCommitmentDefinition {
         `);
     }
     applyToAgentModelRequirements(requirements, content) {
-        // We simply mark that browser capability is enabled in metadata
-        // Get existing metadata
-        const existingMetadata = requirements.metadata || {};
         // Get existing tools array or create new one
-        const existingTools = existingMetadata.tools || [];
-        // Add 'browser' to tools if not already present
-        const updatedTools = existingTools.includes('browser') ? existingTools : [...existingTools, 'browser'];
-        // Return requirements with updated metadata
+        const existingTools = requirements.tools || [];
+        // Add 'web_browser' to tools if not already present
+        const updatedTools = existingTools.some((tool) => tool.name === 'web_browser')
+            ? existingTools
+            : [
+                // TODO: [🔰] Use through proper MCP server
+                ...existingTools,
+                {
+                    name: 'web_browser',
+                    description: spaceTrim$1(`
+                        A tool that can browse the web.
+                        Use this tool when you need to access specific websites or browse the internet.
+                    `),
+                    parameters: {
+                        type: 'object',
+                        properties: {
+                            url: {
+                                type: 'string',
+                                description: 'The URL to browse',
+                            },
+                        },
+                        required: ['url'],
+                    },
+                },
+            ];
+        // Return requirements with updated tools and metadata
         return {
             ...requirements,
+            tools: updatedTools,
             metadata: {
-                ...existingMetadata,
-                tools: updatedTools,
+                ...requirements.metadata,
                 useBrowser: true,
             },
         };
@@ -5821,13 +5886,13 @@ class UseMcpCommitmentDefinition extends BaseCommitmentDefinition {
  * The `USE SEARCH ENGINE` commitment indicates that the agent should utilize a search engine tool
  * to access and retrieve up-to-date information from the internet when necessary.
  *
- * The content following `USE SEARCH ENGINE` is ignored (similar to NOTE).
+ * The content following `USE SEARCH ENGINE` is an arbitrary text that the agent should know (e.g. search scope or instructions).
  *
  * Example usage in agent source:
  *
  * ```book
  * USE SEARCH ENGINE
- * USE SEARCH ENGINE This will be ignored
+ * USE SEARCH ENGINE Hledej informace o Přemyslovcích
  * ```
  *
  * @private [🪔] Maybe export the commitments through some package
@@ -5859,7 +5924,7 @@ class UseSearchEngineCommitmentDefinition extends BaseCommitmentDefinition {
 
             ## Key aspects
 
-            - The content following \`USE SEARCH ENGINE\` is ignored (similar to NOTE)
+            - The content following \`USE SEARCH ENGINE\` is an arbitrary text that the agent should know (e.g. search scope or instructions).
             - The actual search engine tool usage is handled by the agent runtime
             - Allows the agent to search for current information from the web
             - Useful for research tasks, finding facts, and accessing dynamic content
@@ -5884,20 +5949,45 @@ class UseSearchEngineCommitmentDefinition extends BaseCommitmentDefinition {
         `);
     }
     applyToAgentModelRequirements(requirements, content) {
-        // We simply mark that search engine capability is enabled in metadata
-        // Get existing metadata
-        const existingMetadata = requirements.metadata || {};
         // Get existing tools array or create new one
-        const existingTools = existingMetadata.tools || [];
-        // Add 'search-engine' to tools if not already present
-        const updatedTools = existingTools.includes('search-engine') ? existingTools : [...existingTools, 'search-engine'];
-        // Return requirements with updated metadata
+        const existingTools = requirements.tools || [];
+        // Add 'web_search' to tools if not already present
+        const updatedTools = existingTools.some((tool) => tool.name === 'web_search')
+            ? existingTools
+            : [
+                ...existingTools,
+                { type: 'web_search' },
+                // <- Note: [🔰] This is just using simple native search tool by OpenAI @see https://platform.openai.com/docs/guides/tools-web-search
+                //          In future we will use proper MCP search tool:
+                /*
+ 
+                {
+                    name: 'web_search',
+                    description: spaceTrim(`
+                      Search the internet for information.
+                      Use this tool when you need to find up-to-date information or facts that you don't know.
+                      ${!content ? '' : `Search scope / instructions: ${content}`}
+                  `),
+                    parameters: {
+                        type: 'object',
+                        properties: {
+                            query: {
+                                type: 'string',
+                                description: 'The search query',
+                            },
+                        },
+                        required: ['query'],
+                    },
+                },
+                */
+            ];
+        // Return requirements with updated tools and metadata
         return {
             ...requirements,
+            tools: updatedTools,
             metadata: {
-                ...existingMetadata,
-                tools: updatedTools,
-                useSearchEngine: true,
+                ...requirements.metadata,
+                useSearchEngine: content || true,
             },
         };
     }
@@ -6258,8 +6348,8 @@ function parseParameters(text) {
         return match;
     });
     // Remove duplicates based on name (keep the first occurrence)
-    const uniqueParameters = parameters.filter((param, index, array) => {
-        return array.findIndex((p) => p.name === param.name) === index;
+    const uniqueParameters = parameters.filter((parameter, index, array) => {
+        return array.findIndex((parameterItem) => parameterItem.name === parameter.name) === index;
     });
     return uniqueParameters;
 }
@@ -6301,7 +6391,57 @@ function parseAgentSource(agentSource) {
     }
     const meta = {};
     const links = [];
+    const capabilities = [];
     for (const commitment of parseResult.commitments) {
+        if (commitment.type === 'USE BROWSER') {
+            capabilities.push({
+                type: 'browser',
+                label: 'Browser',
+                iconName: 'Globe',
+            });
+            continue;
+        }
+        if (commitment.type === 'USE SEARCH ENGINE') {
+            capabilities.push({
+                type: 'search-engine',
+                label: 'Search Internet',
+                iconName: 'Search',
+            });
+            continue;
+        }
+        if (commitment.type === 'KNOWLEDGE') {
+            const content = spaceTrim$2(commitment.content).split('\n')[0] || '';
+            let label = content;
+            let iconName = 'Book';
+            if (content.startsWith('http://') || content.startsWith('https://')) {
+                try {
+                    const url = new URL(content);
+                    if (url.pathname.endsWith('.pdf')) {
+                        label = url.pathname.split('/').pop() || 'Document.pdf';
+                        iconName = 'FileText';
+                    }
+                    else {
+                        label = url.hostname.replace(/^www\./, '');
+                    }
+                }
+                catch (e) {
+                    // Invalid URL, treat as text
+                }
+            }
+            else {
+                // Text content - take first few words
+                const words = content.split(/\s+/);
+                if (words.length > 4) {
+                    label = words.slice(0, 4).join(' ') + '...';
+                }
+            }
+            capabilities.push({
+                type: 'knowledge',
+                label,
+                iconName,
+            });
+            continue;
+        }
         if (commitment.type === 'META LINK') {
             const linkValue = spaceTrim$2(commitment.content);
             links.push(linkValue);
@@ -6312,6 +6452,10 @@ function parseAgentSource(agentSource) {
             meta.image = spaceTrim$2(commitment.content);
             continue;
         }
+        if (commitment.type === 'META DESCRIPTION') {
+            meta.description = spaceTrim$2(commitment.content);
+            continue;
+        }
         if (commitment.type === 'META COLOR') {
             meta.color = normalizeSeparator(commitment.content);
             continue;
@@ -6348,6 +6492,7 @@ function parseAgentSource(agentSource) {
         meta,
         links,
         parameters,
+        capabilities,
     };
 }
 /**