@promptbook/browser 0.104.0-8 → 0.104.0

package/umd/index.umd.js

@@ -23,7 +23,7 @@
      * @generated
      * @see https://github.com/webgptorg/promptbook
      */
-    const PROMPTBOOK_ENGINE_VERSION = '0.104.0-8';
+    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
@@ -2212,9 +2212,10 @@
      *
      * 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`
      */
@@ -2492,8 +2493,8 @@
      */
     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;
         });
     }
     /**
@@ -2624,6 +2625,39 @@
      * 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
      *
@@ -2655,8 +2689,8 @@
      */
     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('|');
@@ -2678,8 +2712,8 @@
      */
     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('|');
@@ -3189,9 +3223,7 @@
             // 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,
@@ -3342,21 +3374,35 @@
         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$1.spaceTrim((block) => `
+                        Invalid agent URL in FROM commitment: "${trimmedContent}"
+
+                        \`\`\`book
+                        ${block(content)}
+                        \`\`\`
+
+
+                `));
             }
+            const parentAgentUrl = trimmedContent;
             return {
                 ...requirements,
-                parentAgentUrl: trimmedContent,
+                parentAgentUrl,
             };
         }
     }
@@ -5720,19 +5766,38 @@
         `);
         }
         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.spaceTrim(`
+                        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,
                 },
             };
@@ -5825,13 +5890,13 @@
      * 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
@@ -5863,7 +5928,7 @@
 
             ## 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
@@ -5888,20 +5953,45 @@
         `);
         }
         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,
                 },
             };
         }
@@ -6262,8 +6352,8 @@
             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;
     }
@@ -6305,7 +6395,57 @@
         }
         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__default["default"](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__default["default"](commitment.content);
                 links.push(linkValue);
@@ -6316,6 +6456,10 @@
                 meta.image = spaceTrim__default["default"](commitment.content);
                 continue;
             }
+            if (commitment.type === 'META DESCRIPTION') {
+                meta.description = spaceTrim__default["default"](commitment.content);
+                continue;
+            }
             if (commitment.type === 'META COLOR') {
                 meta.color = normalizeSeparator(commitment.content);
                 continue;
@@ -6352,6 +6496,7 @@
             meta,
             links,
             parameters,
+            capabilities,
         };
     }
     /**