@jazzmind/busibox-app 3.0.18

package/dist/lib/agent/llm-client.js

@@ -0,0 +1,290 @@
+/**
+ * LiteLLM Client Wrapper
+ *
+ * Centralized liteLLM API client with error handling and logging.
+ * All LLM operations go through liteLLM proxy.
+ */
+import OpenAI from 'openai';
+// Lazy initialization of liteLLM client
+function getLiteLLMClient() {
+    return new OpenAI({
+        apiKey: process.env.LITELLM_API_KEY || 'sk-litellm-master-key-change-me',
+        baseURL: process.env.LITELLM_BASE_URL || 'http://localhost:4000/v1',
+    });
+}
+/**
+ * Check if liteLLM is properly configured
+ */
+export function isOpenAIConfigured() {
+    // For backward compatibility, check if liteLLM is configured
+    return !!process.env.LITELLM_BASE_URL;
+}
+// Legacy export for backward compatibility (deprecated - use getLiteLLMClient instead)
+export const openai = {
+    chat: {
+        completions: {
+            create: async (...args) => {
+                const client = getLiteLLMClient();
+                return client.chat.completions.create(...args);
+            },
+        },
+    },
+    embeddings: {
+        create: async (...args) => {
+            const client = getLiteLLMClient();
+            return client.embeddings.create(...args);
+        },
+    },
+    moderations: {
+        create: async (...args) => {
+            const client = getLiteLLMClient();
+            return client.moderations.create(...args);
+        },
+    },
+};
+/**
+ * Generate a chat completion with streaming support
+ *
+ * @param messages - Array of chat messages
+ * @param options - Generation options
+ * @returns Stream or complete response based on stream flag
+ */
+export async function generateChatCompletion(messages, options = {}) {
+    if (!isOpenAIConfigured()) {
+        throw new Error('LiteLLM not configured');
+    }
+    const { model, temperature = 0.7, maxTokens = 2000, stream = false, tools, } = options;
+    const litellm = getLiteLLMClient();
+    // Get default model if not specified
+    let modelName = model;
+    if (!modelName) {
+        const { getModelName } = await import('./models');
+        modelName = await getModelName();
+    }
+    return litellm.chat.completions.create({
+        model: modelName,
+        messages,
+        temperature,
+        max_tokens: maxTokens,
+        stream,
+        tools,
+    });
+}
+/**
+ * Generate a concise title for a conversation based on the first message
+ *
+ * @param firstUserMessage - The first message in the conversation
+ * @returns A short, descriptive title (max 50 chars)
+ */
+export async function generateConversationTitle(firstUserMessage) {
+    // Use liteLLM if available, otherwise fallback
+    const useLiteLLM = !!process.env.LITELLM_BASE_URL;
+    if (useLiteLLM) {
+        try {
+            const litellm = new OpenAI({
+                apiKey: process.env.LITELLM_API_KEY || 'sk-litellm-master-key-change-me',
+                baseURL: process.env.LITELLM_BASE_URL || 'http://localhost:4000/v1',
+            });
+            // Use "fast" model for title generation (as specified)
+            const modelName = 'fast';
+            const response = await litellm.chat.completions.create({
+                model: modelName,
+                messages: [
+                    {
+                        role: 'system',
+                        content: 'Generate a concise, descriptive title (max 50 characters) for a conversation that starts with the following user message. Return only the title, no quotes or extra text.',
+                    },
+                    {
+                        role: 'user',
+                        content: firstUserMessage.substring(0, 500), // Limit input length
+                    },
+                ],
+                temperature: 0.5,
+                max_tokens: 20,
+            });
+            const title = response.choices[0]?.message?.content?.trim() || firstUserMessage.substring(0, 50);
+            return title.substring(0, 50).trim();
+        }
+        catch (error) {
+            console.error('Failed to generate title with liteLLM:', error);
+            // Fallback to simple truncation
+            return firstUserMessage.substring(0, 50).trim() + (firstUserMessage.length > 50 ? '...' : '');
+        }
+    }
+    // Fallback: Use first 50 chars of message
+    return firstUserMessage.substring(0, 50).trim() + (firstUserMessage.length > 50 ? '...' : '');
+}
+/**
+ * NOTE: Embedding generation has been moved to @/lib/embeddings/data-client
+ * which uses the local FastEmbed service (bge-large-en-v1.5, 1024 dimensions)
+ * instead of OpenAI/liteLLM embeddings.
+ */
+/**
+ * Extract insights from text (conversation or document)
+ *
+ * This function analyzes text and extracts key insights,
+ * learnings, or knowledge that can be stored for future retrieval.
+ *
+ * @param text - Text to analyze
+ * @returns Array of insight strings
+ */
+export async function extractInsightsFromText(text) {
+    if (!isOpenAIConfigured()) {
+        return [];
+    }
+    if (!text || text.trim().length === 0) {
+        return [];
+    }
+    try {
+        const litellm = getLiteLLMClient();
+        const { getModelName } = await import('./models');
+        const modelName = await getModelName('analysis'); // Use analysis model for insight extraction
+        const response = await litellm.chat.completions.create({
+            model: modelName,
+            messages: [
+                {
+                    role: 'system',
+                    content: `You are an expert at extracting key insights from conversations.
+          Analyze the conversation and extract 3-5 important insights, learnings, or pieces of knowledge.
+          Each insight should be:
+          - A complete, standalone statement (max 500 characters)
+          - Useful for future conversations
+          - Specific and actionable
+          - Focused on user preferences, context, or important information
+
+          Return insights as a JSON object with an "insights" array of strings. Example:
+          {"insights": ["The user prefers Python for data analysis", "User is working on a machine learning project for image classification"]}`,
+                },
+                {
+                    role: 'user',
+                    content: `Extract insights from this conversation:\n\n${text}`,
+                },
+            ],
+            temperature: 0.3,
+            max_tokens: 1000,
+            response_format: { type: 'json_object' },
+        });
+        const content = response.choices[0]?.message?.content;
+        if (!content) {
+            return [];
+        }
+        const parsed = JSON.parse(content);
+        return Array.isArray(parsed.insights) ? parsed.insights : [];
+    }
+    catch (error) {
+        console.error('Failed to extract insights:', error);
+        return [];
+    }
+}
+/**
+ * Extract insights from conversation messages
+ *
+ * This function analyzes a conversation and extracts key insights,
+ * learnings, or knowledge that can be stored for future retrieval.
+ *
+ * @param messages - Array of conversation messages
+ * @returns Array of insight strings
+ */
+export async function extractInsightStringsFromMessages(messages) {
+    if (!isOpenAIConfigured()) {
+        return [];
+    }
+    if (messages.length < 2) {
+        return []; // Need at least a question and answer
+    }
+    try {
+        const conversationText = messages
+            .map(m => `${m.role}: ${m.content}`)
+            .join('\n\n');
+        const litellm = getLiteLLMClient();
+        const { getModelName } = await import('./models');
+        const modelName = await getModelName('analysis'); // Use analysis model for insight extraction
+        const response = await litellm.chat.completions.create({
+            model: modelName,
+            messages: [
+                {
+                    role: 'system',
+                    content: `You are an expert at extracting key insights from conversations.
+          Analyze the conversation and extract 3-5 important insights, learnings, or pieces of knowledge.
+          Each insight should be:
+          - A complete, standalone statement
+          - Useful for future conversations
+          - Specific and actionable
+
+          Return insights as a JSON array of strings. Example:
+          ["The user prefers Python for data analysis", "User is working on a machine learning project for image classification"]`,
+                },
+                {
+                    role: 'user',
+                    content: `Extract insights from this conversation:\n\n${conversationText}`,
+                },
+            ],
+            temperature: 0.3,
+            max_tokens: 500,
+            response_format: { type: 'json_object' },
+        });
+        const content = response.choices[0]?.message?.content;
+        if (!content) {
+            return [];
+        }
+        const parsed = JSON.parse(content);
+        return Array.isArray(parsed.insights) ? parsed.insights : [];
+    }
+    catch (error) {
+        console.error('Failed to extract insights:', error);
+        return [];
+    }
+}
+/**
+ * Moderate content for safety using OpenAI's moderation API
+ *
+ * @param text - Text to moderate
+ * @returns Moderation result
+ */
+export async function moderateContent(text) {
+    if (!isOpenAIConfigured()) {
+        return { flagged: false, categories: {} };
+    }
+    try {
+        const litellm = getLiteLLMClient();
+        const response = await litellm.moderations.create({
+            input: text,
+        });
+        return response.results[0];
+    }
+    catch (error) {
+        console.error('Content moderation failed:', error);
+        return { flagged: false, categories: {} };
+    }
+}
+/**
+ * Count tokens in text (approximate)
+ *
+ * Note: This is a rough estimate. For exact counts, use tiktoken library.
+ *
+ * @param text - Text to count tokens for
+ * @returns Approximate token count
+ */
+export function estimateTokenCount(text) {
+    // Rough estimate: ~4 characters per token
+    // This is approximate and varies by content
+    return Math.ceil(text.length / 4);
+}
+/**
+ * Truncate text to fit within token limit
+ *
+ * @param text - Text to truncate
+ * @param maxTokens - Maximum number of tokens
+ * @returns Truncated text
+ */
+export function truncateToTokenLimit(text, maxTokens) {
+    const estimatedTokens = estimateTokenCount(text);
+    if (estimatedTokens <= maxTokens) {
+        return text;
+    }
+    // Approximate character limit (4 chars per token)
+    const maxChars = maxTokens * 4;
+    return text.substring(0, maxChars) + '...';
+}
+export default openai;
+//# sourceMappingURL=llm-client.js.map

package/dist/lib/agent/llm-client.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"llm-client.js","sourceRoot":"","sources":["../../../src/lib/agent/llm-client.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAEH,OAAO,MAAM,MAAM,QAAQ,CAAC;AAE5B,wCAAwC;AACxC,SAAS,gBAAgB;IACvB,OAAO,IAAI,MAAM,CAAC;QAChB,MAAM,EAAE,OAAO,CAAC,GAAG,CAAC,eAAe,IAAI,iCAAiC;QACxE,OAAO,EAAE,OAAO,CAAC,GAAG,CAAC,gBAAgB,IAAI,0BAA0B;KACpE,CAAC,CAAC;AACL,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,kBAAkB;IAChC,6DAA6D;IAC7D,OAAO,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,gBAAgB,CAAC;AACxC,CAAC;AAED,uFAAuF;AACvF,MAAM,CAAC,MAAM,MAAM,GAAG;IACpB,IAAI,EAAE;QACJ,WAAW,EAAE;YACX,MAAM,EAAE,KAAK,EAAE,GAAG,IAAyD,EAAE,EAAE;gBAC7E,MAAM,MAAM,GAAG,gBAAgB,EAAE,CAAC;gBAClC,OAAO,MAAM,CAAC,IAAI,CAAC,WAAW,CAAC,MAAM,CAAC,GAAG,IAAI,CAAC,CAAC;YACjD,CAAC;SACF;KACF;IACD,UAAU,EAAE;QACV,MAAM,EAAE,KAAK,EAAE,GAAG,IAAgD,EAAE,EAAE;YACpE,MAAM,MAAM,GAAG,gBAAgB,EAAE,CAAC;YAClC,OAAO,MAAM,CAAC,UAAU,CAAC,MAAM,CAAC,GAAG,IAAI,CAAC,CAAC;QAC3C,CAAC;KACF;IACD,WAAW,EAAE;QACX,MAAM,EAAE,KAAK,EAAE,GAAG,IAAiD,EAAE,EAAE;YACrE,MAAM,MAAM,GAAG,gBAAgB,EAAE,CAAC;YAClC,OAAO,MAAM,CAAC,WAAW,CAAC,MAAM,CAAC,GAAG,IAAI,CAAC,CAAC;QAC5C,CAAC;KACF;CACQ,CAAC;AAEZ;;;;;;GAMG;AACH,MAAM,CAAC,KAAK,UAAU,sBAAsB,CAC1C,QAAkD,EAClD,UAMI,EAAE;IAEN,IAAI,CAAC,kBAAkB,EAAE,EAAE,CAAC;QAC1B,MAAM,IAAI,KAAK,CAAC,wBAAwB,CAAC,CAAC;IAC5C,CAAC;IAED,MAAM,EACJ,KAAK,EACL,WAAW,GAAG,GAAG,EACjB,SAAS,GAAG,IAAI,EAChB,MAAM,GAAG,KAAK,EACd,KAAK,GACN,GAAG,OAAO,CAAC;IAEZ,MAAM,OAAO,GAAG,gBAAgB,EAAE,CAAC;IAEnC,qCAAqC;IACrC,IAAI,SAAS,GAAG,KAAK,CAAC;IACtB,IAAI,CAAC,SAAS,EAAE,CAAC;QACf,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,MAAM,CAAC,UAAU,CAAC,CAAC;QAClD,SAAS,GAAG,MAAM,YAAY,EAAE,CAAC;IACnC,CAAC;IAED,OAAO,OAAO,CAAC,IAAI,CAAC,WAAW,CAAC,MAAM,CAAC;QACrC,KAAK,EAAE,SAAS;QAChB,QAAQ;QACR,WAAW;QACX,UAAU,EAAE,SAAS;QACrB,MAAM;QACN,KAAK;KACN,CAAC,CAAC;AACL,CAAC;AAED;;;;;GAKG;AACH,MAAM,CAAC,KAAK,UAAU,yBAAyB,CAC7C,gBAAwB;IAExB,+CAA+C;IAC/C,MAAM,UAAU,GAAG,CAAC,CAAC,OAAO,CAAC,GAAG,CAAC,gBAAgB,CAAC;IAElD,IAAI,UAAU,EAAE,CAAC;QACf,IAAI,CAAC;YACH,MAAM,OAAO,GAAG,IAAI,MAAM,CAAC;gBACzB,MAAM,EAAE,OAAO,CAAC,GAAG,CAAC,eAAe,IAAI,iCAAiC;gBACxE,OAAO,EAAE,OAAO,CAAC,GAAG,CAAC,gBAAgB,IAAI,0BAA0B;aACpE,CAAC,CAAC;YAEH,uDAAuD;YACvD,MAAM,SAAS,GAAG,MAAM,CAAC;YAEzB,MAAM,QAAQ,GAAG,MAAM,OAAO,CAAC,IAAI,CAAC,WAAW,CAAC,MAAM,CAAC;gBACrD,KAAK,EAAE,SAAS;gBAChB,QAAQ,EAAE;oBACR;wBACE,IAAI,EAAE,QAAQ;wBACd,OAAO,EAAE,2KAA2K;qBACrL;oBACD;wBACE,IAAI,EAAE,MAAM;wBACZ,OAAO,EAAE,gBAAgB,CAAC,SAAS,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE,qBAAqB;qBACnE;iBACF;gBACD,WAAW,EAAE,GAAG;gBAChB,UAAU,EAAE,EAAE;aACf,CAAC,CAAC;YAEH,MAAM,KAAK,GAAG,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,IAAI,gBAAgB,CAAC,SAAS,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC;YACjG,OAAO,KAAK,CAAC,SAAS,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,CAAC;QACvC,CAAC;QAAC,OAAO,KAAK,EAAE,CAAC;YACf,OAAO,CAAC,KAAK,CAAC,wCAAwC,EAAE,KAAK,CAAC,CAAC;YAC/D,gCAAgC;YAChC,OAAO,gBAAgB,CAAC,SAAS,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,GAAG,CAAC,gBAAgB,CAAC,MAAM,GAAG,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;QAChG,CAAC;IACH,CAAC;IAED,0CAA0C;IAC1C,OAAO,gBAAgB,CAAC,SAAS,CAAC,CAAC,EAAE,EAAE,CAAC,CAAC,IAAI,EAAE,GAAG,CAAC,gBAAgB,CAAC,MAAM,GAAG,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC;AAChG,CAAC;AAED;;;;GAIG;AAEH;;;;;;;;GAQG;AACH,MAAM,CAAC,KAAK,UAAU,uBAAuB,CAAC,IAAY;IACxD,IAAI,CAAC,kBAAkB,EAAE,EAAE,CAAC;QAC1B,OAAO,EAAE,CAAC;IACZ,CAAC;IAED,IAAI,CAAC,IAAI,IAAI,IAAI,CAAC,IAAI,EAAE,CAAC,MAAM,KAAK,CAAC,EAAE,CAAC;QACtC,OAAO,EAAE,CAAC;IACZ,CAAC;IAED,IAAI,CAAC;QACH,MAAM,OAAO,GAAG,gBAAgB,EAAE,CAAC;QACnC,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,MAAM,CAAC,UAAU,CAAC,CAAC;QAClD,MAAM,SAAS,GAAG,MAAM,YAAY,CAAC,UAAU,CAAC,CAAC,CAAC,4CAA4C;QAE9F,MAAM,QAAQ,GAAG,MAAM,OAAO,CAAC,IAAI,CAAC,WAAW,CAAC,MAAM,CAAC;YACrD,KAAK,EAAE,SAAS;YAChB,QAAQ,EAAE;gBACR;oBACE,IAAI,EAAE,QAAQ;oBACd,OAAO,EAAE;;;;;;;;;gJAS6H;iBACvI;gBACD;oBACE,IAAI,EAAE,MAAM;oBACZ,OAAO,EAAE,+CAA+C,IAAI,EAAE;iBAC/D;aACF;YACD,WAAW,EAAE,GAAG;YAChB,UAAU,EAAE,IAAI;YAChB,eAAe,EAAE,EAAE,IAAI,EAAE,aAAa,EAAE;SACzC,CAAC,CAAC;QAEH,MAAM,OAAO,GAAG,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,OAAO,CAAC;QACtD,IAAI,CAAC,OAAO,EAAE,CAAC;YACb,OAAO,EAAE,CAAC;QACZ,CAAC;QAED,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;QACnC,OAAO,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,CAAC;IAC/D,CAAC;IAAC,OAAO,KAAK,EAAE,CAAC;QACf,OAAO,CAAC,KAAK,CAAC,6BAA6B,EAAE,KAAK,CAAC,CAAC;QACpD,OAAO,EAAE,CAAC;IACZ,CAAC;AACH,CAAC;AAED;;;;;;;;GAQG;AACH,MAAM,CAAC,KAAK,UAAU,iCAAiC,CACrD,QAAkD;IAElD,IAAI,CAAC,kBAAkB,EAAE,EAAE,CAAC;QAC1B,OAAO,EAAE,CAAC;IACZ,CAAC;IAED,IAAI,QAAQ,CAAC,MAAM,GAAG,CAAC,EAAE,CAAC;QACxB,OAAO,EAAE,CAAC,CAAC,sCAAsC;IACnD,CAAC;IAED,IAAI,CAAC;QACH,MAAM,gBAAgB,GAAG,QAAQ;aAC9B,GAAG,CAAC,CAAC,CAAC,EAAE,CAAC,GAAG,CAAC,CAAC,IAAI,KAAK,CAAC,CAAC,OAAO,EAAE,CAAC;aACnC,IAAI,CAAC,MAAM,CAAC,CAAC;QAEhB,MAAM,OAAO,GAAG,gBAAgB,EAAE,CAAC;QACnC,MAAM,EAAE,YAAY,EAAE,GAAG,MAAM,MAAM,CAAC,UAAU,CAAC,CAAC;QAClD,MAAM,SAAS,GAAG,MAAM,YAAY,CAAC,UAAU,CAAC,CAAC,CAAC,4CAA4C;QAE9F,MAAM,QAAQ,GAAG,MAAM,OAAO,CAAC,IAAI,CAAC,WAAW,CAAC,MAAM,CAAC;YACrD,KAAK,EAAE,SAAS;YAChB,QAAQ,EAAE;gBACR;oBACE,IAAI,EAAE,QAAQ;oBACd,OAAO,EAAE;;;;;;;;kIAQ+G;iBACzH;gBACD;oBACE,IAAI,EAAE,MAAM;oBACZ,OAAO,EAAE,+CAA+C,gBAAgB,EAAE;iBAC3E;aACF;YACD,WAAW,EAAE,GAAG;YAChB,UAAU,EAAE,GAAG;YACf,eAAe,EAAE,EAAE,IAAI,EAAE,aAAa,EAAE;SACzC,CAAC,CAAC;QAEH,MAAM,OAAO,GAAG,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,EAAE,OAAO,EAAE,OAAO,CAAC;QACtD,IAAI,CAAC,OAAO,EAAE,CAAC;YACb,OAAO,EAAE,CAAC;QACZ,CAAC;QAED,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,CAAC;QACnC,OAAO,KAAK,CAAC,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,QAAQ,CAAC,CAAC,CAAC,EAAE,CAAC;IAC/D,CAAC;IAAC,OAAO,KAAK,EAAE,CAAC;QACf,OAAO,CAAC,KAAK,CAAC,6BAA6B,EAAE,KAAK,CAAC,CAAC;QACpD,OAAO,EAAE,CAAC;IACZ,CAAC;AACH,CAAC;AAED;;;;;GAKG;AACH,MAAM,CAAC,KAAK,UAAU,eAAe,CAAC,IAAY;IAChD,IAAI,CAAC,kBAAkB,EAAE,EAAE,CAAC;QAC1B,OAAO,EAAE,OAAO,EAAE,KAAK,EAAE,UAAU,EAAE,EAAE,EAAE,CAAC;IAC5C,CAAC;IAED,IAAI,CAAC;QACH,MAAM,OAAO,GAAG,gBAAgB,EAAE,CAAC;QACnC,MAAM,QAAQ,GAAG,MAAM,OAAO,CAAC,WAAW,CAAC,MAAM,CAAC;YAChD,KAAK,EAAE,IAAI;SACZ,CAAC,CAAC;QAEH,OAAO,QAAQ,CAAC,OAAO,CAAC,CAAC,CAAC,CAAC;IAC7B,CAAC;IAAC,OAAO,KAAK,EAAE,CAAC;QACf,OAAO,CAAC,KAAK,CAAC,4BAA4B,EAAE,KAAK,CAAC,CAAC;QACnD,OAAO,EAAE,OAAO,EAAE,KAAK,EAAE,UAAU,EAAE,EAAE,EAAE,CAAC;IAC5C,CAAC;AACH,CAAC;AAED;;;;;;;GAOG;AACH,MAAM,UAAU,kBAAkB,CAAC,IAAY;IAC7C,0CAA0C;IAC1C,4CAA4C;IAC5C,OAAO,IAAI,CAAC,IAAI,CAAC,IAAI,CAAC,MAAM,GAAG,CAAC,CAAC,CAAC;AACpC,CAAC;AAED;;;;;;GAMG;AACH,MAAM,UAAU,oBAAoB,CAAC,IAAY,EAAE,SAAiB;IAClE,MAAM,eAAe,GAAG,kBAAkB,CAAC,IAAI,CAAC,CAAC;IAEjD,IAAI,eAAe,IAAI,SAAS,EAAE,CAAC;QACjC,OAAO,IAAI,CAAC;IACd,CAAC;IAED,kDAAkD;IAClD,MAAM,QAAQ,GAAG,SAAS,GAAG,CAAC,CAAC;IAC/B,OAAO,IAAI,CAAC,SAAS,CAAC,CAAC,EAAE,QAAQ,CAAC,GAAG,KAAK,CAAC;AAC7C,CAAC;AAED,eAAe,MAAM,CAAC"}

package/dist/lib/agent/models.d.ts

@@ -0,0 +1,46 @@
+/**
+ * Model Registry Service
+ *
+ * Provides model selection utilities with all metadata sourced from liteLLM.
+ * No hardcoded model names, descriptions, or capabilities -- everything
+ * comes from liteLLM's /model/info endpoint, which is populated from
+ * model_registry.yml at config generation time.
+ */
+import { type ModelOption } from './litellm';
+export type { ModelOption };
+/**
+ * Get available model options for chat UI.
+ * All metadata comes from liteLLM's model_info.
+ */
+export declare function getAvailableModels(): Promise<ModelOption[]>;
+/**
+ * Get model name from model ID. Returns the ID as-is since
+ * liteLLM expects the purpose name directly.
+ */
+export declare function getModelName(modelId?: string): Promise<string>;
+/**
+ * Get model display name from liteLLM model_info.
+ */
+export declare function getModelDisplayName(modelId?: string): Promise<string>;
+/**
+ * Get native model capabilities from liteLLM model_info.
+ * These reflect what the underlying model actually supports.
+ */
+export declare function getModelCapabilities(modelId?: string): Promise<{
+    multimodal: boolean;
+    toolCalling: boolean;
+} | undefined>;
+/**
+ * Get native model capabilities (synchronous lookup from cache).
+ * Returns undefined if cache is not populated yet.
+ * Prefer the async getModelCapabilities() when possible.
+ */
+export declare function getNativeModelCapabilities(modelId?: string): {
+    multimodal: boolean;
+    toolCalling: boolean;
+} | undefined;
+/**
+ * Clear the model cache (useful for testing or manual refresh)
+ */
+export declare function clearModelCache(): void;
+//# sourceMappingURL=models.d.ts.map

package/dist/lib/agent/models.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"models.d.ts","sourceRoot":"","sources":["../../../src/lib/agent/models.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAEH,OAAO,EAA4D,KAAK,WAAW,EAAkB,MAAM,WAAW,CAAC;AAEvH,YAAY,EAAE,WAAW,EAAE,CAAC;AAwB5B;;;GAGG;AACH,wBAAsB,kBAAkB,IAAI,OAAO,CAAC,WAAW,EAAE,CAAC,CAGjE;AAED;;;GAGG;AACH,wBAAsB,YAAY,CAAC,OAAO,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC,CAEpE;AAED;;GAEG;AACH,wBAAsB,mBAAmB,CAAC,OAAO,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,CAAC,CAS3E;AAED;;;GAGG;AACH,wBAAsB,oBAAoB,CAAC,OAAO,CAAC,EAAE,MAAM,GAAG,OAAO,CAAC;IACpE,UAAU,EAAE,OAAO,CAAC;IACpB,WAAW,EAAE,OAAO,CAAC;CACtB,GAAG,SAAS,CAAC,CAWb;AAED;;;;GAIG;AACH,wBAAgB,0BAA0B,CAAC,OAAO,CAAC,EAAE,MAAM,GAAG;IAC5D,UAAU,EAAE,OAAO,CAAC;IACpB,WAAW,EAAE,OAAO,CAAC;CACtB,GAAG,SAAS,CAUZ;AAED;;GAEG;AACH,wBAAgB,eAAe,IAAI,IAAI,CAItC"}

package/dist/lib/agent/models.js

@@ -0,0 +1,96 @@
+/**
+ * Model Registry Service
+ *
+ * Provides model selection utilities with all metadata sourced from liteLLM.
+ * No hardcoded model names, descriptions, or capabilities -- everything
+ * comes from liteLLM's /model/info endpoint, which is populated from
+ * model_registry.yml at config generation time.
+ */
+import { fetchAvailableModels, fetchAllModelInfo, formatModelName } from './litellm';
+// Cache for models (refreshed every 60 seconds)
+let cachedModels = null;
+let cachedModelInfo = null;
+let cacheTimestamp = 0;
+const CACHE_TTL = 60000;
+async function ensureCache() {
+    const now = Date.now();
+    if (cachedModels && cachedModelInfo && (now - cacheTimestamp) < CACHE_TTL) {
+        return;
+    }
+    const [models, allInfo] = await Promise.all([
+        fetchAvailableModels(),
+        fetchAllModelInfo(),
+    ]);
+    cachedModels = models;
+    cachedModelInfo = allInfo;
+    cacheTimestamp = Date.now();
+}
+/**
+ * Get available model options for chat UI.
+ * All metadata comes from liteLLM's model_info.
+ */
+export async function getAvailableModels() {
+    await ensureCache();
+    return cachedModels || [];
+}
+/**
+ * Get model name from model ID. Returns the ID as-is since
+ * liteLLM expects the purpose name directly.
+ */
+export async function getModelName(modelId) {
+    return modelId || 'chat';
+}
+/**
+ * Get model display name from liteLLM model_info.
+ */
+export async function getModelDisplayName(modelId) {
+    if (!modelId)
+        return 'Chat';
+    await ensureCache();
+    const info = cachedModelInfo?.find(m => m.id === modelId);
+    if (info?.display_name)
+        return info.display_name;
+    const model = cachedModels?.find(m => m.id === modelId);
+    return model?.name || formatModelName(modelId);
+}
+/**
+ * Get native model capabilities from liteLLM model_info.
+ * These reflect what the underlying model actually supports.
+ */
+export async function getModelCapabilities(modelId) {
+    if (!modelId)
+        return undefined;
+    await ensureCache();
+    const info = cachedModelInfo?.find(m => m.id === modelId);
+    if (!info)
+        return undefined;
+    return {
+        multimodal: info.multimodal ?? false,
+        toolCalling: info.tool_calling ?? false,
+    };
+}
+/**
+ * Get native model capabilities (synchronous lookup from cache).
+ * Returns undefined if cache is not populated yet.
+ * Prefer the async getModelCapabilities() when possible.
+ */
+export function getNativeModelCapabilities(modelId) {
+    if (!modelId || !cachedModelInfo)
+        return undefined;
+    const info = cachedModelInfo.find(m => m.id === modelId);
+    if (!info)
+        return undefined;
+    return {
+        multimodal: info.multimodal ?? false,
+        toolCalling: info.tool_calling ?? false,
+    };
+}
+/**
+ * Clear the model cache (useful for testing or manual refresh)
+ */
+export function clearModelCache() {
+    cachedModels = null;
+    cachedModelInfo = null;
+    cacheTimestamp = 0;
+}
+//# sourceMappingURL=models.js.map

package/dist/lib/agent/models.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"models.js","sourceRoot":"","sources":["../../../src/lib/agent/models.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAEH,OAAO,EAAE,oBAAoB,EAAE,iBAAiB,EAAE,eAAe,EAAoC,MAAM,WAAW,CAAC;AAIvH,gDAAgD;AAChD,IAAI,YAAY,GAAyB,IAAI,CAAC;AAC9C,IAAI,eAAe,GAAuB,IAAI,CAAC;AAC/C,IAAI,cAAc,GAAW,CAAC,CAAC;AAC/B,MAAM,SAAS,GAAG,KAAK,CAAC;AAExB,KAAK,UAAU,WAAW;IACxB,MAAM,GAAG,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;IACvB,IAAI,YAAY,IAAI,eAAe,IAAI,CAAC,GAAG,GAAG,cAAc,CAAC,GAAG,SAAS,EAAE,CAAC;QAC1E,OAAO;IACT,CAAC;IAED,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,GAAG,MAAM,OAAO,CAAC,GAAG,CAAC;QAC1C,oBAAoB,EAAE;QACtB,iBAAiB,EAAE;KACpB,CAAC,CAAC;IAEH,YAAY,GAAG,MAAM,CAAC;IACtB,eAAe,GAAG,OAAO,CAAC;IAC1B,cAAc,GAAG,IAAI,CAAC,GAAG,EAAE,CAAC;AAC9B,CAAC;AAED;;;GAGG;AACH,MAAM,CAAC,KAAK,UAAU,kBAAkB;IACtC,MAAM,WAAW,EAAE,CAAC;IACpB,OAAO,YAAY,IAAI,EAAE,CAAC;AAC5B,CAAC;AAED;;;GAGG;AACH,MAAM,CAAC,KAAK,UAAU,YAAY,CAAC,OAAgB;IACjD,OAAO,OAAO,IAAI,MAAM,CAAC;AAC3B,CAAC;AAED;;GAEG;AACH,MAAM,CAAC,KAAK,UAAU,mBAAmB,CAAC,OAAgB;IACxD,IAAI,CAAC,OAAO;QAAE,OAAO,MAAM,CAAC;IAE5B,MAAM,WAAW,EAAE,CAAC;IACpB,MAAM,IAAI,GAAG,eAAe,EAAE,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,CAAC,CAAC;IAC1D,IAAI,IAAI,EAAE,YAAY;QAAE,OAAO,IAAI,CAAC,YAAY,CAAC;IAEjD,MAAM,KAAK,GAAG,YAAY,EAAE,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,CAAC,CAAC;IACxD,OAAO,KAAK,EAAE,IAAI,IAAI,eAAe,CAAC,OAAO,CAAC,CAAC;AACjD,CAAC;AAED;;;GAGG;AACH,MAAM,CAAC,KAAK,UAAU,oBAAoB,CAAC,OAAgB;IAIzD,IAAI,CAAC,OAAO;QAAE,OAAO,SAAS,CAAC;IAE/B,MAAM,WAAW,EAAE,CAAC;IACpB,MAAM,IAAI,GAAG,eAAe,EAAE,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,CAAC,CAAC;IAC1D,IAAI,CAAC,IAAI;QAAE,OAAO,SAAS,CAAC;IAE5B,OAAO;QACL,UAAU,EAAE,IAAI,CAAC,UAAU,IAAI,KAAK;QACpC,WAAW,EAAE,IAAI,CAAC,YAAY,IAAI,KAAK;KACxC,CAAC;AACJ,CAAC;AAED;;;;GAIG;AACH,MAAM,UAAU,0BAA0B,CAAC,OAAgB;IAIzD,IAAI,CAAC,OAAO,IAAI,CAAC,eAAe;QAAE,OAAO,SAAS,CAAC;IAEnD,MAAM,IAAI,GAAG,eAAe,CAAC,IAAI,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,KAAK,OAAO,CAAC,CAAC;IACzD,IAAI,CAAC,IAAI;QAAE,OAAO,SAAS,CAAC;IAE5B,OAAO;QACL,UAAU,EAAE,IAAI,CAAC,UAAU,IAAI,KAAK;QACpC,WAAW,EAAE,IAAI,CAAC,YAAY,IAAI,KAAK;KACxC,CAAC;AACJ,CAAC;AAED;;GAEG;AACH,MAAM,UAAU,eAAe;IAC7B,YAAY,GAAG,IAAI,CAAC;IACpB,eAAe,GAAG,IAAI,CAAC;IACvB,cAAc,GAAG,CAAC,CAAC;AACrB,CAAC"}

package/dist/lib/agent/openai.d.ts

@@ -0,0 +1,94 @@
+/**
+ * LiteLLM Client Wrapper
+ *
+ * Centralized liteLLM API client with error handling and logging.
+ * All LLM operations go through liteLLM proxy.
+ */
+import OpenAI from 'openai';
+/**
+ * Check if liteLLM is properly configured
+ */
+export declare function isOpenAIConfigured(): boolean;
+export declare const openai: OpenAI;
+/**
+ * Generate a chat completion with streaming support
+ *
+ * @param messages - Array of chat messages
+ * @param options - Generation options
+ * @returns Stream or complete response based on stream flag
+ */
+export declare function generateChatCompletion(messages: OpenAI.Chat.ChatCompletionMessageParam[], options?: {
+    model?: string;
+    temperature?: number;
+    maxTokens?: number;
+    stream?: boolean;
+    tools?: OpenAI.Chat.ChatCompletionTool[];
+}): Promise<(import("openai/core/streaming.mjs").Stream<OpenAI.Chat.Completions.ChatCompletionChunk> & {
+    _request_id?: string | null;
+}) | (OpenAI.Chat.Completions.ChatCompletion & {
+    _request_id?: string | null;
+})>;
+/**
+ * Generate a concise title for a conversation based on the first message
+ *
+ * @param firstUserMessage - The first message in the conversation
+ * @returns A short, descriptive title (max 50 chars)
+ */
+export declare function generateConversationTitle(firstUserMessage: string): Promise<string>;
+/**
+ * NOTE: Embedding generation has been moved to @/lib/embeddings/data-client
+ * which uses the local FastEmbed service (bge-large-en-v1.5, 1024 dimensions)
+ * instead of OpenAI/liteLLM embeddings.
+ */
+/**
+ * Extract insights from text (conversation or document)
+ *
+ * This function analyzes text and extracts key insights,
+ * learnings, or knowledge that can be stored for future retrieval.
+ *
+ * @param text - Text to analyze
+ * @returns Array of insight strings
+ */
+export declare function extractInsightsFromText(text: string): Promise<string[]>;
+/**
+ * Extract insights from conversation messages
+ *
+ * This function analyzes a conversation and extracts key insights,
+ * learnings, or knowledge that can be stored for future retrieval.
+ *
+ * @param messages - Array of conversation messages
+ * @returns Array of insight strings
+ */
+export declare function extractConversationInsights(messages: Array<{
+    role: string;
+    content: string;
+}>): Promise<string[]>;
+/**
+ * Moderate content for safety using OpenAI's moderation API
+ *
+ * @param text - Text to moderate
+ * @returns Moderation result
+ */
+export declare function moderateContent(text: string): Promise<OpenAI.Moderations.Moderation | {
+    flagged: boolean;
+    categories: {};
+}>;
+/**
+ * Count tokens in text (approximate)
+ *
+ * Note: This is a rough estimate. For exact counts, use tiktoken library.
+ *
+ * @param text - Text to count tokens for
+ * @returns Approximate token count
+ */
+export declare function estimateTokenCount(text: string): number;
+/**
+ * Truncate text to fit within token limit
+ *
+ * @param text - Text to truncate
+ * @param maxTokens - Maximum number of tokens
+ * @returns Truncated text
+ */
+export declare function truncateToTokenLimit(text: string, maxTokens: number): string;
+export default openai;
+//# sourceMappingURL=openai.d.ts.map

package/dist/lib/agent/openai.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"openai.d.ts","sourceRoot":"","sources":["../../../src/lib/agent/openai.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAEH,OAAO,MAAM,MAAM,QAAQ,CAAC;AAU5B;;GAEG;AACH,wBAAgB,kBAAkB,IAAI,OAAO,CAG5C;AAGD,eAAO,MAAM,MAAM,EAqBd,MAAM,CAAC;AAEZ;;;;;;GAMG;AACH,wBAAsB,sBAAsB,CAC1C,QAAQ,EAAE,MAAM,CAAC,IAAI,CAAC,0BAA0B,EAAE,EAClD,OAAO,GAAE;IACP,KAAK,CAAC,EAAE,MAAM,CAAC;IACf,WAAW,CAAC,EAAE,MAAM,CAAC;IACrB,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,MAAM,CAAC,EAAE,OAAO,CAAC;IACjB,KAAK,CAAC,EAAE,MAAM,CAAC,IAAI,CAAC,kBAAkB,EAAE,CAAC;CACrC;;;;IA+BP;AAED;;;;;GAKG;AACH,wBAAsB,yBAAyB,CAC7C,gBAAgB,EAAE,MAAM,GACvB,OAAO,CAAC,MAAM,CAAC,CAyCjB;AAED;;;;GAIG;AAEH;;;;;;;;GAQG;AACH,wBAAsB,uBAAuB,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,MAAM,EAAE,CAAC,CAmD7E;AAED;;;;;;;;GAQG;AACH,wBAAsB,2BAA2B,CAC/C,QAAQ,EAAE,KAAK,CAAC;IAAE,IAAI,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAA;CAAE,CAAC,GACjD,OAAO,CAAC,MAAM,EAAE,CAAC,CAsDnB;AAED;;;;;GAKG;AACH,wBAAsB,eAAe,CAAC,IAAI,EAAE,MAAM;;;GAgBjD;AAED;;;;;;;GAOG;AACH,wBAAgB,kBAAkB,CAAC,IAAI,EAAE,MAAM,GAAG,MAAM,CAIvD;AAED;;;;;;GAMG;AACH,wBAAgB,oBAAoB,CAAC,IAAI,EAAE,MAAM,EAAE,SAAS,EAAE,MAAM,GAAG,MAAM,CAU5E;AAED,eAAe,MAAM,CAAC"}