cencori 1.0.3 → 1.1.0

package/README.md

@@ -134,10 +134,12 @@ console.log(response.embeddings[0]); // [0.1, 0.2, ...]
 
 ### Image Generation
 
+Generate images from text prompts using multiple providers:
+
 ```typescript
 const response = await cencori.ai.generateImage({
   prompt: 'A futuristic city at sunset with flying cars',
-  model: 'dall-e-3',
+  model: 'gpt-image-1.5',  // Best text rendering, top ELO rating
   size: '1024x1024',
   quality: 'hd'
 });
@@ -145,6 +147,12 @@ const response = await cencori.ai.generateImage({
 console.log(response.images[0].url); // https://...
 ```
 
+**Supported Models:**
+| Provider | Models | Description |
+|----------|--------|-------------|
+| **OpenAI** | `gpt-image-1.5`, `gpt-image-1`, `dall-e-3`, `dall-e-2` | Text rendering, creative |
+| **Google** | `gemini-3-pro-image`, `imagen-3` | High photorealism |
+
 ## Framework Integrations
 
 ### Vercel AI SDK
@@ -205,15 +213,29 @@ await cencori.workflow.trigger('data-enrichment', {
 });
 ```
 
-### Storage
+### Memory (Context Store)
 
 ```typescript
-// 🚧 Coming Soon
-const results = await cencori.storage.vectors.search('query', { 
-  limit: 5 
+// Store a memory with auto-embedding
+await cencori.memory.store({
+  namespace: 'docs',
+  content: 'Refund policy allows returns within 30 days',
+  metadata: { category: 'policy' }
 });
 
-await cencori.storage.knowledge.query('What is our refund policy?');
+// Semantic search
+const results = await cencori.memory.search({
+  namespace: 'docs',
+  query: 'what is the refund policy?',
+  limit: 5
+});
+
+// RAG helper
+const response = await cencori.ai.rag({
+  model: 'gpt-4o',
+  messages: [{ role: 'user', content: 'What is our refund policy?' }],
+  namespace: 'docs'
+});
 ```
 
 ## Why Cencori?

package/dist/ai/index.d.mts

@@ -95,6 +95,75 @@ declare class AINamespace {
      * console.log(response.images[0].url);
      */
     generateImage(request: ImageGenerationRequest): Promise<ImageGenerationResponse>;
+    /**
+     * RAG (Retrieval-Augmented Generation) - Chat with automatic memory context
+     *
+     * Searches your memory namespace for relevant context and includes it
+     * in the prompt automatically. Returns the AI response along with sources.
+     *
+     * @example
+     * const response = await cencori.ai.rag({
+     *   model: 'gpt-4o',
+     *   messages: [{ role: 'user', content: 'What are our company policies?' }],
+     *   namespace: 'company-docs',
+     *   limit: 5, // number of memories to retrieve
+     * });
+     * console.log(response.message.content);
+     * console.log(response.sources); // retrieved context
+     */
+    rag(request: RagRequest): Promise<RagResponse>;
+    /**
+     * Stream RAG responses with automatic memory context
+     *
+     * @example
+     * for await (const chunk of cencori.ai.ragStream({ model: 'gpt-4o', messages, namespace: 'docs' })) {
+     *   if (chunk.type === 'sources') console.log('Sources:', chunk.sources);
+     *   if (chunk.type === 'content') process.stdout.write(chunk.delta);
+     * }
+     */
+    ragStream(request: RagRequest): AsyncGenerator<RagStreamChunk, void, unknown>;
+}
+interface RagRequest {
+    model: string;
+    messages: Array<{
+        role: 'system' | 'user' | 'assistant';
+        content: string;
+    }>;
+    namespace: string;
+    temperature?: number;
+    maxTokens?: number;
+    limit?: number;
+    threshold?: number;
+    includeSources?: boolean;
+}
+interface RagResponse {
+    message: {
+        role: string;
+        content: string;
+    };
+    model: string;
+    provider: string;
+    usage: {
+        promptTokens: number;
+        completionTokens: number;
+        totalTokens: number;
+    };
+    sources?: Array<{
+        content: string;
+        metadata: Record<string, unknown>;
+        similarity: number;
+    }>;
+    latencyMs: number;
+}
+interface RagStreamChunk {
+    type: 'sources' | 'content';
+    delta?: string;
+    finish_reason?: string;
+    sources?: Array<{
+        content: string;
+        metadata: Record<string, unknown>;
+        similarity: number;
+    }>;
 }
 
-export { AINamespace, type StreamChunk };
+export { AINamespace, type RagRequest, type RagResponse, type RagStreamChunk, type StreamChunk };

package/dist/ai/index.d.ts

@@ -95,6 +95,75 @@ declare class AINamespace {
      * console.log(response.images[0].url);
      */
     generateImage(request: ImageGenerationRequest): Promise<ImageGenerationResponse>;
+    /**
+     * RAG (Retrieval-Augmented Generation) - Chat with automatic memory context
+     *
+     * Searches your memory namespace for relevant context and includes it
+     * in the prompt automatically. Returns the AI response along with sources.
+     *
+     * @example
+     * const response = await cencori.ai.rag({
+     *   model: 'gpt-4o',
+     *   messages: [{ role: 'user', content: 'What are our company policies?' }],
+     *   namespace: 'company-docs',
+     *   limit: 5, // number of memories to retrieve
+     * });
+     * console.log(response.message.content);
+     * console.log(response.sources); // retrieved context
+     */
+    rag(request: RagRequest): Promise<RagResponse>;
+    /**
+     * Stream RAG responses with automatic memory context
+     *
+     * @example
+     * for await (const chunk of cencori.ai.ragStream({ model: 'gpt-4o', messages, namespace: 'docs' })) {
+     *   if (chunk.type === 'sources') console.log('Sources:', chunk.sources);
+     *   if (chunk.type === 'content') process.stdout.write(chunk.delta);
+     * }
+     */
+    ragStream(request: RagRequest): AsyncGenerator<RagStreamChunk, void, unknown>;
+}
+interface RagRequest {
+    model: string;
+    messages: Array<{
+        role: 'system' | 'user' | 'assistant';
+        content: string;
+    }>;
+    namespace: string;
+    temperature?: number;
+    maxTokens?: number;
+    limit?: number;
+    threshold?: number;
+    includeSources?: boolean;
+}
+interface RagResponse {
+    message: {
+        role: string;
+        content: string;
+    };
+    model: string;
+    provider: string;
+    usage: {
+        promptTokens: number;
+        completionTokens: number;
+        totalTokens: number;
+    };
+    sources?: Array<{
+        content: string;
+        metadata: Record<string, unknown>;
+        similarity: number;
+    }>;
+    latencyMs: number;
+}
+interface RagStreamChunk {
+    type: 'sources' | 'content';
+    delta?: string;
+    finish_reason?: string;
+    sources?: Array<{
+        content: string;
+        metadata: Record<string, unknown>;
+        similarity: number;
+    }>;
 }
 
-export { AINamespace, type StreamChunk };
+export { AINamespace, type RagRequest, type RagResponse, type RagStreamChunk, type StreamChunk };

package/dist/ai/index.js

@@ -312,6 +312,131 @@ var AINamespace = class {
       provider: data.provider
     };
   }
+  /**
+   * RAG (Retrieval-Augmented Generation) - Chat with automatic memory context
+   * 
+   * Searches your memory namespace for relevant context and includes it
+   * in the prompt automatically. Returns the AI response along with sources.
+   * 
+   * @example
+   * const response = await cencori.ai.rag({
+   *   model: 'gpt-4o',
+   *   messages: [{ role: 'user', content: 'What are our company policies?' }],
+   *   namespace: 'company-docs',
+   *   limit: 5, // number of memories to retrieve
+   * });
+   * console.log(response.message.content);
+   * console.log(response.sources); // retrieved context
+   */
+  async rag(request) {
+    const response = await fetch(`${this.config.baseUrl}/api/ai/rag`, {
+      method: "POST",
+      headers: {
+        "CENCORI_API_KEY": this.config.apiKey,
+        "Content-Type": "application/json",
+        ...this.config.headers
+      },
+      body: JSON.stringify({
+        model: request.model,
+        messages: request.messages,
+        namespace: request.namespace,
+        temperature: request.temperature,
+        maxTokens: request.maxTokens,
+        limit: request.limit ?? 5,
+        threshold: request.threshold ?? 0.5,
+        include_sources: request.includeSources ?? true,
+        stream: false
+      })
+    });
+    if (!response.ok) {
+      const errorData = await response.json().catch(() => ({ error: "Unknown error" }));
+      throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);
+    }
+    const data = await response.json();
+    return {
+      message: {
+        role: "assistant",
+        content: data.message.content
+      },
+      model: data.model,
+      provider: data.provider,
+      usage: {
+        promptTokens: data.usage.prompt_tokens,
+        completionTokens: data.usage.completion_tokens,
+        totalTokens: data.usage.total_tokens
+      },
+      sources: data.sources?.map((s) => ({
+        content: s.content,
+        metadata: s.metadata,
+        similarity: s.similarity
+      })),
+      latencyMs: data.latency_ms
+    };
+  }
+  /**
+   * Stream RAG responses with automatic memory context
+   * 
+   * @example
+   * for await (const chunk of cencori.ai.ragStream({ model: 'gpt-4o', messages, namespace: 'docs' })) {
+   *   if (chunk.type === 'sources') console.log('Sources:', chunk.sources);
+   *   if (chunk.type === 'content') process.stdout.write(chunk.delta);
+   * }
+   */
+  async *ragStream(request) {
+    const response = await fetch(`${this.config.baseUrl}/api/ai/rag`, {
+      method: "POST",
+      headers: {
+        "CENCORI_API_KEY": this.config.apiKey,
+        "Content-Type": "application/json",
+        ...this.config.headers
+      },
+      body: JSON.stringify({
+        model: request.model,
+        messages: request.messages,
+        namespace: request.namespace,
+        temperature: request.temperature,
+        maxTokens: request.maxTokens,
+        limit: request.limit ?? 5,
+        threshold: request.threshold ?? 0.5,
+        include_sources: request.includeSources ?? true,
+        stream: true
+      })
+    });
+    if (!response.ok) {
+      const errorData = await response.json().catch(() => ({ error: "Unknown error" }));
+      throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);
+    }
+    if (!response.body) {
+      throw new Error("Response body is null");
+    }
+    const reader = response.body.getReader();
+    const decoder = new TextDecoder();
+    let buffer = "";
+    try {
+      while (true) {
+        const { done, value } = await reader.read();
+        if (done) break;
+        buffer += decoder.decode(value, { stream: true });
+        const lines = buffer.split("\n");
+        buffer = lines.pop() || "";
+        for (const line of lines) {
+          if (line.trim() === "") continue;
+          if (!line.startsWith("data: ")) continue;
+          const data = line.slice(6);
+          if (data === "[DONE]") {
+            return;
+          }
+          try {
+            const chunk = JSON.parse(data);
+            yield chunk;
+          } catch {
+          }
+        }
+      }
+    } finally {
+      reader.releaseLock();
+    }
+  }
 };
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {

package/dist/ai/index.js.map

@@ -1 +1 @@
-{"version":3,"sources":["../../src/ai/index.ts"],"sourcesContent":["/**\n * AI Gateway - Chat, Completions, Embeddings, and Streaming\n * \n * @example\n * const response = await cencori.ai.chat({\n *   model: 'gpt-4o',\n *   messages: [{ role: 'user', content: 'Hello!' }]\n * });\n */\n\nimport type {\n    CencoriConfig,\n    ChatRequest,\n    ChatResponse,\n    CompletionRequest,\n    EmbeddingRequest,\n    EmbeddingResponse,\n    GenerateObjectRequest,\n    GenerateObjectResponse,\n    ImageGenerationRequest,\n    ImageGenerationResponse,\n    ToolCall\n} from '../types';\n\n// API Response types\ninterface OpenAIChatResponse {\n    id: string;\n    model: string;\n    choices?: Array<{\n        message?: {\n            content?: string;\n            tool_calls?: Array<{\n                id: string;\n                type: 'function';\n                function: {\n                    name: string;\n                    arguments: string;\n                };\n            }>;\n        };\n        finish_reason?: string;\n    }>;\n    usage?: {\n        prompt_tokens?: number;\n        completion_tokens?: number;\n        total_tokens?: number;\n    };\n}\n\ninterface OpenAIEmbeddingResponse {\n    model: string;\n    data?: Array<{\n        embedding: number[];\n    }>;\n    usage?: {\n        total_tokens?: number;\n    };\n}\n\n/**\n * Stream chunk from chat stream\n */\nexport interface StreamChunk {\n    delta: string;\n    finish_reason?: 'stop' | 'length' | 'content_filter' | 'tool_calls' | 'error';\n    /** Tool calls in progress during streaming */\n    toolCalls?: ToolCall[];\n    /** Error message if the stream encountered an error */\n    error?: string;\n}\n\nexport class AINamespace {\n    private config: Required<CencoriConfig>;\n\n    constructor(config: Required<CencoriConfig>) {\n        this.config = config;\n    }\n\n    /**\n     * Create a chat completion\n     * \n     * @example\n     * const response = await cencori.ai.chat({\n     *   model: 'gpt-4o',\n     *   messages: [{ role: 'user', content: 'Hello!' }]\n     * });\n     */\n    async chat(request: ChatRequest): Promise<ChatResponse> {\n        const response = await fetch(`${this.config.baseUrl}/api/ai/chat`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                model: request.model,\n                messages: request.messages,\n                temperature: request.temperature,\n                maxTokens: request.maxTokens,\n                stream: false,\n                tools: request.tools,\n                toolChoice: request.toolChoice,\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string };\n            throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);\n        }\n\n        const data = await response.json() as OpenAIChatResponse;\n\n        const choice = data.choices?.[0];\n        const toolCalls = choice?.message?.tool_calls?.map(tc => ({\n            id: tc.id,\n            type: tc.type as 'function',\n            function: {\n                name: tc.function.name,\n                arguments: tc.function.arguments,\n            },\n        }));\n\n        return {\n            id: data.id,\n            model: data.model,\n            content: choice?.message?.content ?? '',\n            toolCalls,\n            finishReason: choice?.finish_reason as ChatResponse['finishReason'],\n            usage: {\n                promptTokens: data.usage?.prompt_tokens ?? 0,\n                completionTokens: data.usage?.completion_tokens ?? 0,\n                totalTokens: data.usage?.total_tokens ?? 0,\n            },\n        };\n    }\n\n    /**\n     * Stream chat completions\n     * Returns an async generator that yields chunks as they arrive\n     * \n     * @example\n     * for await (const chunk of cencori.ai.chatStream({ model: 'gpt-4o', messages })) {\n     *   process.stdout.write(chunk.delta);\n     * }\n     */\n    async *chatStream(request: ChatRequest): AsyncGenerator<StreamChunk, void, unknown> {\n        const response = await fetch(`${this.config.baseUrl}/api/ai/chat`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                model: request.model,\n                messages: request.messages,\n                temperature: request.temperature,\n                maxTokens: request.maxTokens,\n                stream: true,\n                tools: request.tools,\n                toolChoice: request.toolChoice,\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string };\n            throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);\n        }\n\n        if (!response.body) {\n            throw new Error('Response body is null');\n        }\n\n        const reader = response.body.getReader();\n        const decoder = new TextDecoder();\n        let buffer = '';\n\n        try {\n            while (true) {\n                const { done, value } = await reader.read();\n\n                if (done) break;\n\n                buffer += decoder.decode(value, { stream: true });\n                const lines = buffer.split('\\n');\n\n                // Keep the last incomplete line in the buffer\n                buffer = lines.pop() || '';\n\n                for (const line of lines) {\n                    if (line.trim() === '') continue;\n                    if (!line.startsWith('data: ')) continue;\n\n                    const data = line.slice(6); // Remove 'data: ' prefix\n\n                    if (data === '[DONE]') {\n                        return;\n                    }\n\n                    try {\n                        const chunk = JSON.parse(data) as StreamChunk;\n                        yield chunk;\n                    } catch {\n                        // Skip malformed JSON\n                    }\n                }\n            }\n        } finally {\n            reader.releaseLock();\n        }\n    }\n\n    /**\n     * Create a text completion\n     * \n     * @example\n     * const response = await cencori.ai.completions({\n     *   model: 'gpt-4o',\n     *   prompt: 'Write a haiku about coding'\n     * });\n     */\n    async completions(request: CompletionRequest): Promise<ChatResponse> {\n        // Convert to chat format internally\n        return this.chat({\n            model: request.model,\n            messages: [{ role: 'user', content: request.prompt }],\n            temperature: request.temperature,\n            maxTokens: request.maxTokens,\n        });\n    }\n\n    /**\n     * Create embeddings\n     * \n     * @example\n     * const response = await cencori.ai.embeddings({\n     *   model: 'text-embedding-3-small',\n     *   input: 'Hello world'\n     * });\n     */\n    async embeddings(request: EmbeddingRequest): Promise<EmbeddingResponse> {\n        const response = await fetch(`${this.config.baseUrl}/api/ai/embeddings`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                model: request.model,\n                input: request.input,\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string };\n            throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);\n        }\n\n        const data = await response.json() as OpenAIEmbeddingResponse;\n\n        return {\n            model: data.model,\n            embeddings: data.data?.map((d) => d.embedding) ?? [],\n            usage: {\n                totalTokens: data.usage?.total_tokens ?? 0,\n            },\n        };\n    }\n\n    /**\n     * Generate structured output with JSON schema\n     * \n     * @example\n     * const response = await cencori.ai.generateObject({\n     *   model: 'gpt-4o',\n     *   prompt: 'Generate a user profile',\n     *   schema: {\n     *     type: 'object',\n     *     properties: {\n     *       name: { type: 'string' },\n     *       age: { type: 'number' }\n     *     },\n     *     required: ['name', 'age']\n     *   }\n     * });\n     * console.log(response.object); // { name: 'John', age: 30 }\n     */\n    async generateObject<T = unknown>(request: GenerateObjectRequest): Promise<GenerateObjectResponse<T>> {\n        // Build messages from prompt or use provided messages\n        const messages = request.messages ?? [\n            { role: 'user' as const, content: request.prompt ?? '' }\n        ];\n\n        // Use function calling to enforce JSON schema\n        const response = await fetch(`${this.config.baseUrl}/api/ai/chat`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                model: request.model,\n                messages,\n                temperature: request.temperature,\n                maxTokens: request.maxTokens,\n                stream: false,\n                tools: [{\n                    type: 'function',\n                    function: {\n                        name: request.schemaName ?? 'generate_object',\n                        description: request.schemaDescription ?? 'Generate a structured object matching the schema',\n                        parameters: request.schema,\n                    },\n                }],\n                toolChoice: {\n                    type: 'function',\n                    function: { name: request.schemaName ?? 'generate_object' },\n                },\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string };\n            throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);\n        }\n\n        const data = await response.json() as OpenAIChatResponse;\n        const toolCall = data.choices?.[0]?.message?.tool_calls?.[0];\n\n        if (!toolCall) {\n            throw new Error('Model did not return structured output');\n        }\n\n        let parsedObject: T;\n        try {\n            parsedObject = JSON.parse(toolCall.function.arguments) as T;\n        } catch {\n            throw new Error('Failed to parse structured output as JSON');\n        }\n\n        return {\n            object: parsedObject,\n            usage: {\n                promptTokens: data.usage?.prompt_tokens ?? 0,\n                completionTokens: data.usage?.completion_tokens ?? 0,\n                totalTokens: data.usage?.total_tokens ?? 0,\n            },\n        };\n    }\n\n    /**\n     * Generate images from a text prompt\n     * \n     * @example\n     * const response = await cencori.ai.generateImage({\n     *   prompt: 'A futuristic city at sunset',\n     *   model: 'dall-e-3',\n     *   size: '1024x1024'\n     * });\n     * console.log(response.images[0].url);\n     */\n    async generateImage(request: ImageGenerationRequest): Promise<ImageGenerationResponse> {\n        const response = await fetch(`${this.config.baseUrl}/api/ai/images/generate`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                prompt: request.prompt,\n                model: request.model ?? 'dall-e-3',\n                n: request.n,\n                size: request.size,\n                quality: request.quality,\n                style: request.style,\n                responseFormat: request.responseFormat,\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string; message?: string };\n            throw new Error(`Cencori API error: ${errorData.message || errorData.error || response.statusText}`);\n        }\n\n        const data = await response.json() as {\n            images: Array<{ url?: string; b64_json?: string; revisedPrompt?: string }>;\n            model: string;\n            provider: string;\n        };\n\n        return {\n            images: data.images.map(img => ({\n                url: img.url,\n                b64Json: img.b64_json,\n                revisedPrompt: img.revisedPrompt,\n            })),\n            model: data.model,\n            provider: data.provider,\n        };\n    }\n}\n\n"],"mappings":";;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAuEO,IAAM,cAAN,MAAkB;AAAA,EAGrB,YAAY,QAAiC;AACzC,SAAK,SAAS;AAAA,EAClB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,MAAM,KAAK,SAA6C;AACpD,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,gBAAgB;AAAA,MAC/D,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf,UAAU,QAAQ;AAAA,QAClB,aAAa,QAAQ;AAAA,QACrB,WAAW,QAAQ;AAAA,QACnB,QAAQ;AAAA,QACR,OAAO,QAAQ;AAAA,QACf,YAAY,QAAQ;AAAA,MACxB,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AAEjC,UAAM,SAAS,KAAK,UAAU,CAAC;AAC/B,UAAM,YAAY,QAAQ,SAAS,YAAY,IAAI,SAAO;AAAA,MACtD,IAAI,GAAG;AAAA,MACP,MAAM,GAAG;AAAA,MACT,UAAU;AAAA,QACN,MAAM,GAAG,SAAS;AAAA,QAClB,WAAW,GAAG,SAAS;AAAA,MAC3B;AAAA,IACJ,EAAE;AAEF,WAAO;AAAA,MACH,IAAI,KAAK;AAAA,MACT,OAAO,KAAK;AAAA,MACZ,SAAS,QAAQ,SAAS,WAAW;AAAA,MACrC;AAAA,MACA,cAAc,QAAQ;AAAA,MACtB,OAAO;AAAA,QACH,cAAc,KAAK,OAAO,iBAAiB;AAAA,QAC3C,kBAAkB,KAAK,OAAO,qBAAqB;AAAA,QACnD,aAAa,KAAK,OAAO,gBAAgB;AAAA,MAC7C;AAAA,IACJ;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,OAAO,WAAW,SAAkE;AAChF,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,gBAAgB;AAAA,MAC/D,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf,UAAU,QAAQ;AAAA,QAClB,aAAa,QAAQ;AAAA,QACrB,WAAW,QAAQ;AAAA,QACnB,QAAQ;AAAA,QACR,OAAO,QAAQ;AAAA,QACf,YAAY,QAAQ;AAAA,MACxB,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,QAAI,CAAC,SAAS,MAAM;AAChB,YAAM,IAAI,MAAM,uBAAuB;AAAA,IAC3C;AAEA,UAAM,SAAS,SAAS,KAAK,UAAU;AACvC,UAAM,UAAU,IAAI,YAAY;AAChC,QAAI,SAAS;AAEb,QAAI;AACA,aAAO,MAAM;AACT,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAE1C,YAAI,KAAM;AAEV,kBAAU,QAAQ,OAAO,OAAO,EAAE,QAAQ,KAAK,CAAC;AAChD,cAAM,QAAQ,OAAO,MAAM,IAAI;AAG/B,iBAAS,MAAM,IAAI,KAAK;AAExB,mBAAW,QAAQ,OAAO;AACtB,cAAI,KAAK,KAAK,MAAM,GAAI;AACxB,cAAI,CAAC,KAAK,WAAW,QAAQ,EAAG;AAEhC,gBAAM,OAAO,KAAK,MAAM,CAAC;AAEzB,cAAI,SAAS,UAAU;AACnB;AAAA,UACJ;AAEA,cAAI;AACA,kBAAM,QAAQ,KAAK,MAAM,IAAI;AAC7B,kBAAM;AAAA,UACV,QAAQ;AAAA,UAER;AAAA,QACJ;AAAA,MACJ;AAAA,IACJ,UAAE;AACE,aAAO,YAAY;AAAA,IACvB;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,MAAM,YAAY,SAAmD;AAEjE,WAAO,KAAK,KAAK;AAAA,MACb,OAAO,QAAQ;AAAA,MACf,UAAU,CAAC,EAAE,MAAM,QAAQ,SAAS,QAAQ,OAAO,CAAC;AAAA,MACpD,aAAa,QAAQ;AAAA,MACrB,WAAW,QAAQ;AAAA,IACvB,CAAC;AAAA,EACL;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,MAAM,WAAW,SAAuD;AACpE,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,sBAAsB;AAAA,MACrE,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf,OAAO,QAAQ;AAAA,MACnB,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AAEjC,WAAO;AAAA,MACH,OAAO,KAAK;AAAA,MACZ,YAAY,KAAK,MAAM,IAAI,CAAC,MAAM,EAAE,SAAS,KAAK,CAAC;AAAA,MACnD,OAAO;AAAA,QACH,aAAa,KAAK,OAAO,gBAAgB;AAAA,MAC7C;AAAA,IACJ;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBA,MAAM,eAA4B,SAAoE;AAElG,UAAM,WAAW,QAAQ,YAAY;AAAA,MACjC,EAAE,MAAM,QAAiB,SAAS,QAAQ,UAAU,GAAG;AAAA,IAC3D;AAGA,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,gBAAgB;AAAA,MAC/D,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf;AAAA,QACA,aAAa,QAAQ;AAAA,QACrB,WAAW,QAAQ;AAAA,QACnB,QAAQ;AAAA,QACR,OAAO,CAAC;AAAA,UACJ,MAAM;AAAA,UACN,UAAU;AAAA,YACN,MAAM,QAAQ,cAAc;AAAA,YAC5B,aAAa,QAAQ,qBAAqB;AAAA,YAC1C,YAAY,QAAQ;AAAA,UACxB;AAAA,QACJ,CAAC;AAAA,QACD,YAAY;AAAA,UACR,MAAM;AAAA,UACN,UAAU,EAAE,MAAM,QAAQ,cAAc,kBAAkB;AAAA,QAC9D;AAAA,MACJ,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AACjC,UAAM,WAAW,KAAK,UAAU,CAAC,GAAG,SAAS,aAAa,CAAC;AAE3D,QAAI,CAAC,UAAU;AACX,YAAM,IAAI,MAAM,wCAAwC;AAAA,IAC5D;AAEA,QAAI;AACJ,QAAI;AACA,qBAAe,KAAK,MAAM,SAAS,SAAS,SAAS;AAAA,IACzD,QAAQ;AACJ,YAAM,IAAI,MAAM,2CAA2C;AAAA,IAC/D;AAEA,WAAO;AAAA,MACH,QAAQ;AAAA,MACR,OAAO;AAAA,QACH,cAAc,KAAK,OAAO,iBAAiB;AAAA,QAC3C,kBAAkB,KAAK,OAAO,qBAAqB;AAAA,QACnD,aAAa,KAAK,OAAO,gBAAgB;AAAA,MAC7C;AAAA,IACJ;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAaA,MAAM,cAAc,SAAmE;AACnF,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,2BAA2B;AAAA,MAC1E,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,QAAQ,QAAQ;AAAA,QAChB,OAAO,QAAQ,SAAS;AAAA,QACxB,GAAG,QAAQ;AAAA,QACX,MAAM,QAAQ;AAAA,QACd,SAAS,QAAQ;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf,gBAAgB,QAAQ;AAAA,MAC5B,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,WAAW,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IACvG;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AAMjC,WAAO;AAAA,MACH,QAAQ,KAAK,OAAO,IAAI,UAAQ;AAAA,QAC5B,KAAK,IAAI;AAAA,QACT,SAAS,IAAI;AAAA,QACb,eAAe,IAAI;AAAA,MACvB,EAAE;AAAA,MACF,OAAO,KAAK;AAAA,MACZ,UAAU,KAAK;AAAA,IACnB;AAAA,EACJ;AACJ;","names":[]}
+{"version":3,"sources":["../../src/ai/index.ts"],"sourcesContent":["/**\n * AI Gateway - Chat, Completions, Embeddings, and Streaming\n * \n * @example\n * const response = await cencori.ai.chat({\n *   model: 'gpt-4o',\n *   messages: [{ role: 'user', content: 'Hello!' }]\n * });\n */\n\nimport type {\n    CencoriConfig,\n    ChatRequest,\n    ChatResponse,\n    CompletionRequest,\n    EmbeddingRequest,\n    EmbeddingResponse,\n    GenerateObjectRequest,\n    GenerateObjectResponse,\n    ImageGenerationRequest,\n    ImageGenerationResponse,\n    ToolCall\n} from '../types';\n\n// API Response types\ninterface OpenAIChatResponse {\n    id: string;\n    model: string;\n    choices?: Array<{\n        message?: {\n            content?: string;\n            tool_calls?: Array<{\n                id: string;\n                type: 'function';\n                function: {\n                    name: string;\n                    arguments: string;\n                };\n            }>;\n        };\n        finish_reason?: string;\n    }>;\n    usage?: {\n        prompt_tokens?: number;\n        completion_tokens?: number;\n        total_tokens?: number;\n    };\n}\n\ninterface OpenAIEmbeddingResponse {\n    model: string;\n    data?: Array<{\n        embedding: number[];\n    }>;\n    usage?: {\n        total_tokens?: number;\n    };\n}\n\n/**\n * Stream chunk from chat stream\n */\nexport interface StreamChunk {\n    delta: string;\n    finish_reason?: 'stop' | 'length' | 'content_filter' | 'tool_calls' | 'error';\n    /** Tool calls in progress during streaming */\n    toolCalls?: ToolCall[];\n    /** Error message if the stream encountered an error */\n    error?: string;\n}\n\nexport class AINamespace {\n    private config: Required<CencoriConfig>;\n\n    constructor(config: Required<CencoriConfig>) {\n        this.config = config;\n    }\n\n    /**\n     * Create a chat completion\n     * \n     * @example\n     * const response = await cencori.ai.chat({\n     *   model: 'gpt-4o',\n     *   messages: [{ role: 'user', content: 'Hello!' }]\n     * });\n     */\n    async chat(request: ChatRequest): Promise<ChatResponse> {\n        const response = await fetch(`${this.config.baseUrl}/api/ai/chat`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                model: request.model,\n                messages: request.messages,\n                temperature: request.temperature,\n                maxTokens: request.maxTokens,\n                stream: false,\n                tools: request.tools,\n                toolChoice: request.toolChoice,\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string };\n            throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);\n        }\n\n        const data = await response.json() as OpenAIChatResponse;\n\n        const choice = data.choices?.[0];\n        const toolCalls = choice?.message?.tool_calls?.map(tc => ({\n            id: tc.id,\n            type: tc.type as 'function',\n            function: {\n                name: tc.function.name,\n                arguments: tc.function.arguments,\n            },\n        }));\n\n        return {\n            id: data.id,\n            model: data.model,\n            content: choice?.message?.content ?? '',\n            toolCalls,\n            finishReason: choice?.finish_reason as ChatResponse['finishReason'],\n            usage: {\n                promptTokens: data.usage?.prompt_tokens ?? 0,\n                completionTokens: data.usage?.completion_tokens ?? 0,\n                totalTokens: data.usage?.total_tokens ?? 0,\n            },\n        };\n    }\n\n    /**\n     * Stream chat completions\n     * Returns an async generator that yields chunks as they arrive\n     * \n     * @example\n     * for await (const chunk of cencori.ai.chatStream({ model: 'gpt-4o', messages })) {\n     *   process.stdout.write(chunk.delta);\n     * }\n     */\n    async *chatStream(request: ChatRequest): AsyncGenerator<StreamChunk, void, unknown> {\n        const response = await fetch(`${this.config.baseUrl}/api/ai/chat`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                model: request.model,\n                messages: request.messages,\n                temperature: request.temperature,\n                maxTokens: request.maxTokens,\n                stream: true,\n                tools: request.tools,\n                toolChoice: request.toolChoice,\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string };\n            throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);\n        }\n\n        if (!response.body) {\n            throw new Error('Response body is null');\n        }\n\n        const reader = response.body.getReader();\n        const decoder = new TextDecoder();\n        let buffer = '';\n\n        try {\n            while (true) {\n                const { done, value } = await reader.read();\n\n                if (done) break;\n\n                buffer += decoder.decode(value, { stream: true });\n                const lines = buffer.split('\\n');\n\n                // Keep the last incomplete line in the buffer\n                buffer = lines.pop() || '';\n\n                for (const line of lines) {\n                    if (line.trim() === '') continue;\n                    if (!line.startsWith('data: ')) continue;\n\n                    const data = line.slice(6); // Remove 'data: ' prefix\n\n                    if (data === '[DONE]') {\n                        return;\n                    }\n\n                    try {\n                        const chunk = JSON.parse(data) as StreamChunk;\n                        yield chunk;\n                    } catch {\n                        // Skip malformed JSON\n                    }\n                }\n            }\n        } finally {\n            reader.releaseLock();\n        }\n    }\n\n    /**\n     * Create a text completion\n     * \n     * @example\n     * const response = await cencori.ai.completions({\n     *   model: 'gpt-4o',\n     *   prompt: 'Write a haiku about coding'\n     * });\n     */\n    async completions(request: CompletionRequest): Promise<ChatResponse> {\n        // Convert to chat format internally\n        return this.chat({\n            model: request.model,\n            messages: [{ role: 'user', content: request.prompt }],\n            temperature: request.temperature,\n            maxTokens: request.maxTokens,\n        });\n    }\n\n    /**\n     * Create embeddings\n     * \n     * @example\n     * const response = await cencori.ai.embeddings({\n     *   model: 'text-embedding-3-small',\n     *   input: 'Hello world'\n     * });\n     */\n    async embeddings(request: EmbeddingRequest): Promise<EmbeddingResponse> {\n        const response = await fetch(`${this.config.baseUrl}/api/ai/embeddings`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                model: request.model,\n                input: request.input,\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string };\n            throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);\n        }\n\n        const data = await response.json() as OpenAIEmbeddingResponse;\n\n        return {\n            model: data.model,\n            embeddings: data.data?.map((d) => d.embedding) ?? [],\n            usage: {\n                totalTokens: data.usage?.total_tokens ?? 0,\n            },\n        };\n    }\n\n    /**\n     * Generate structured output with JSON schema\n     * \n     * @example\n     * const response = await cencori.ai.generateObject({\n     *   model: 'gpt-4o',\n     *   prompt: 'Generate a user profile',\n     *   schema: {\n     *     type: 'object',\n     *     properties: {\n     *       name: { type: 'string' },\n     *       age: { type: 'number' }\n     *     },\n     *     required: ['name', 'age']\n     *   }\n     * });\n     * console.log(response.object); // { name: 'John', age: 30 }\n     */\n    async generateObject<T = unknown>(request: GenerateObjectRequest): Promise<GenerateObjectResponse<T>> {\n        // Build messages from prompt or use provided messages\n        const messages = request.messages ?? [\n            { role: 'user' as const, content: request.prompt ?? '' }\n        ];\n\n        // Use function calling to enforce JSON schema\n        const response = await fetch(`${this.config.baseUrl}/api/ai/chat`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                model: request.model,\n                messages,\n                temperature: request.temperature,\n                maxTokens: request.maxTokens,\n                stream: false,\n                tools: [{\n                    type: 'function',\n                    function: {\n                        name: request.schemaName ?? 'generate_object',\n                        description: request.schemaDescription ?? 'Generate a structured object matching the schema',\n                        parameters: request.schema,\n                    },\n                }],\n                toolChoice: {\n                    type: 'function',\n                    function: { name: request.schemaName ?? 'generate_object' },\n                },\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string };\n            throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);\n        }\n\n        const data = await response.json() as OpenAIChatResponse;\n        const toolCall = data.choices?.[0]?.message?.tool_calls?.[0];\n\n        if (!toolCall) {\n            throw new Error('Model did not return structured output');\n        }\n\n        let parsedObject: T;\n        try {\n            parsedObject = JSON.parse(toolCall.function.arguments) as T;\n        } catch {\n            throw new Error('Failed to parse structured output as JSON');\n        }\n\n        return {\n            object: parsedObject,\n            usage: {\n                promptTokens: data.usage?.prompt_tokens ?? 0,\n                completionTokens: data.usage?.completion_tokens ?? 0,\n                totalTokens: data.usage?.total_tokens ?? 0,\n            },\n        };\n    }\n\n    /**\n     * Generate images from a text prompt\n     * \n     * @example\n     * const response = await cencori.ai.generateImage({\n     *   prompt: 'A futuristic city at sunset',\n     *   model: 'dall-e-3',\n     *   size: '1024x1024'\n     * });\n     * console.log(response.images[0].url);\n     */\n    async generateImage(request: ImageGenerationRequest): Promise<ImageGenerationResponse> {\n        const response = await fetch(`${this.config.baseUrl}/api/ai/images/generate`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                prompt: request.prompt,\n                model: request.model ?? 'dall-e-3',\n                n: request.n,\n                size: request.size,\n                quality: request.quality,\n                style: request.style,\n                responseFormat: request.responseFormat,\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string; message?: string };\n            throw new Error(`Cencori API error: ${errorData.message || errorData.error || response.statusText}`);\n        }\n\n        const data = await response.json() as {\n            images: Array<{ url?: string; b64_json?: string; revisedPrompt?: string }>;\n            model: string;\n            provider: string;\n        };\n\n        return {\n            images: data.images.map(img => ({\n                url: img.url,\n                b64Json: img.b64_json,\n                revisedPrompt: img.revisedPrompt,\n            })),\n            model: data.model,\n            provider: data.provider,\n        };\n    }\n\n    /**\n     * RAG (Retrieval-Augmented Generation) - Chat with automatic memory context\n     * \n     * Searches your memory namespace for relevant context and includes it\n     * in the prompt automatically. Returns the AI response along with sources.\n     * \n     * @example\n     * const response = await cencori.ai.rag({\n     *   model: 'gpt-4o',\n     *   messages: [{ role: 'user', content: 'What are our company policies?' }],\n     *   namespace: 'company-docs',\n     *   limit: 5, // number of memories to retrieve\n     * });\n     * console.log(response.message.content);\n     * console.log(response.sources); // retrieved context\n     */\n    async rag(request: RagRequest): Promise<RagResponse> {\n        const response = await fetch(`${this.config.baseUrl}/api/ai/rag`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                model: request.model,\n                messages: request.messages,\n                namespace: request.namespace,\n                temperature: request.temperature,\n                maxTokens: request.maxTokens,\n                limit: request.limit ?? 5,\n                threshold: request.threshold ?? 0.5,\n                include_sources: request.includeSources ?? true,\n                stream: false,\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string };\n            throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);\n        }\n\n        const data = await response.json() as RagApiResponse;\n\n        return {\n            message: {\n                role: 'assistant',\n                content: data.message.content,\n            },\n            model: data.model,\n            provider: data.provider,\n            usage: {\n                promptTokens: data.usage.prompt_tokens,\n                completionTokens: data.usage.completion_tokens,\n                totalTokens: data.usage.total_tokens,\n            },\n            sources: data.sources?.map(s => ({\n                content: s.content,\n                metadata: s.metadata,\n                similarity: s.similarity,\n            })),\n            latencyMs: data.latency_ms,\n        };\n    }\n\n    /**\n     * Stream RAG responses with automatic memory context\n     * \n     * @example\n     * for await (const chunk of cencori.ai.ragStream({ model: 'gpt-4o', messages, namespace: 'docs' })) {\n     *   if (chunk.type === 'sources') console.log('Sources:', chunk.sources);\n     *   if (chunk.type === 'content') process.stdout.write(chunk.delta);\n     * }\n     */\n    async *ragStream(request: RagRequest): AsyncGenerator<RagStreamChunk, void, unknown> {\n        const response = await fetch(`${this.config.baseUrl}/api/ai/rag`, {\n            method: 'POST',\n            headers: {\n                'CENCORI_API_KEY': this.config.apiKey,\n                'Content-Type': 'application/json',\n                ...this.config.headers,\n            },\n            body: JSON.stringify({\n                model: request.model,\n                messages: request.messages,\n                namespace: request.namespace,\n                temperature: request.temperature,\n                maxTokens: request.maxTokens,\n                limit: request.limit ?? 5,\n                threshold: request.threshold ?? 0.5,\n                include_sources: request.includeSources ?? true,\n                stream: true,\n            }),\n        });\n\n        if (!response.ok) {\n            const errorData = await response.json().catch(() => ({ error: 'Unknown error' })) as { error?: string };\n            throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);\n        }\n\n        if (!response.body) {\n            throw new Error('Response body is null');\n        }\n\n        const reader = response.body.getReader();\n        const decoder = new TextDecoder();\n        let buffer = '';\n\n        try {\n            while (true) {\n                const { done, value } = await reader.read();\n\n                if (done) break;\n\n                buffer += decoder.decode(value, { stream: true });\n                const lines = buffer.split('\\n');\n                buffer = lines.pop() || '';\n\n                for (const line of lines) {\n                    if (line.trim() === '') continue;\n                    if (!line.startsWith('data: ')) continue;\n\n                    const data = line.slice(6);\n\n                    if (data === '[DONE]') {\n                        return;\n                    }\n\n                    try {\n                        const chunk = JSON.parse(data) as RagStreamChunk;\n                        yield chunk;\n                    } catch {\n                        // Skip malformed JSON\n                    }\n                }\n            }\n        } finally {\n            reader.releaseLock();\n        }\n    }\n}\n\n// RAG Types\nexport interface RagRequest {\n    model: string;\n    messages: Array<{ role: 'system' | 'user' | 'assistant'; content: string }>;\n    namespace: string;\n    temperature?: number;\n    maxTokens?: number;\n    limit?: number;\n    threshold?: number;\n    includeSources?: boolean;\n}\n\nexport interface RagResponse {\n    message: { role: string; content: string };\n    model: string;\n    provider: string;\n    usage: {\n        promptTokens: number;\n        completionTokens: number;\n        totalTokens: number;\n    };\n    sources?: Array<{\n        content: string;\n        metadata: Record<string, unknown>;\n        similarity: number;\n    }>;\n    latencyMs: number;\n}\n\nexport interface RagStreamChunk {\n    type: 'sources' | 'content';\n    delta?: string;\n    finish_reason?: string;\n    sources?: Array<{\n        content: string;\n        metadata: Record<string, unknown>;\n        similarity: number;\n    }>;\n}\n\ninterface RagApiResponse {\n    message: { role: string; content: string };\n    model: string;\n    provider: string;\n    usage: {\n        prompt_tokens: number;\n        completion_tokens: number;\n        total_tokens: number;\n    };\n    sources?: Array<{\n        content: string;\n        metadata: Record<string, unknown>;\n        similarity: number;\n    }>;\n    latency_ms: number;\n}\n"],"mappings":";;;;;;;;;;;;;;;;;;;;AAAA;AAAA;AAAA;AAAA;AAAA;AAuEO,IAAM,cAAN,MAAkB;AAAA,EAGrB,YAAY,QAAiC;AACzC,SAAK,SAAS;AAAA,EAClB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,MAAM,KAAK,SAA6C;AACpD,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,gBAAgB;AAAA,MAC/D,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf,UAAU,QAAQ;AAAA,QAClB,aAAa,QAAQ;AAAA,QACrB,WAAW,QAAQ;AAAA,QACnB,QAAQ;AAAA,QACR,OAAO,QAAQ;AAAA,QACf,YAAY,QAAQ;AAAA,MACxB,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AAEjC,UAAM,SAAS,KAAK,UAAU,CAAC;AAC/B,UAAM,YAAY,QAAQ,SAAS,YAAY,IAAI,SAAO;AAAA,MACtD,IAAI,GAAG;AAAA,MACP,MAAM,GAAG;AAAA,MACT,UAAU;AAAA,QACN,MAAM,GAAG,SAAS;AAAA,QAClB,WAAW,GAAG,SAAS;AAAA,MAC3B;AAAA,IACJ,EAAE;AAEF,WAAO;AAAA,MACH,IAAI,KAAK;AAAA,MACT,OAAO,KAAK;AAAA,MACZ,SAAS,QAAQ,SAAS,WAAW;AAAA,MACrC;AAAA,MACA,cAAc,QAAQ;AAAA,MACtB,OAAO;AAAA,QACH,cAAc,KAAK,OAAO,iBAAiB;AAAA,QAC3C,kBAAkB,KAAK,OAAO,qBAAqB;AAAA,QACnD,aAAa,KAAK,OAAO,gBAAgB;AAAA,MAC7C;AAAA,IACJ;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,OAAO,WAAW,SAAkE;AAChF,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,gBAAgB;AAAA,MAC/D,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf,UAAU,QAAQ;AAAA,QAClB,aAAa,QAAQ;AAAA,QACrB,WAAW,QAAQ;AAAA,QACnB,QAAQ;AAAA,QACR,OAAO,QAAQ;AAAA,QACf,YAAY,QAAQ;AAAA,MACxB,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,QAAI,CAAC,SAAS,MAAM;AAChB,YAAM,IAAI,MAAM,uBAAuB;AAAA,IAC3C;AAEA,UAAM,SAAS,SAAS,KAAK,UAAU;AACvC,UAAM,UAAU,IAAI,YAAY;AAChC,QAAI,SAAS;AAEb,QAAI;AACA,aAAO,MAAM;AACT,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAE1C,YAAI,KAAM;AAEV,kBAAU,QAAQ,OAAO,OAAO,EAAE,QAAQ,KAAK,CAAC;AAChD,cAAM,QAAQ,OAAO,MAAM,IAAI;AAG/B,iBAAS,MAAM,IAAI,KAAK;AAExB,mBAAW,QAAQ,OAAO;AACtB,cAAI,KAAK,KAAK,MAAM,GAAI;AACxB,cAAI,CAAC,KAAK,WAAW,QAAQ,EAAG;AAEhC,gBAAM,OAAO,KAAK,MAAM,CAAC;AAEzB,cAAI,SAAS,UAAU;AACnB;AAAA,UACJ;AAEA,cAAI;AACA,kBAAM,QAAQ,KAAK,MAAM,IAAI;AAC7B,kBAAM;AAAA,UACV,QAAQ;AAAA,UAER;AAAA,QACJ;AAAA,MACJ;AAAA,IACJ,UAAE;AACE,aAAO,YAAY;AAAA,IACvB;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,MAAM,YAAY,SAAmD;AAEjE,WAAO,KAAK,KAAK;AAAA,MACb,OAAO,QAAQ;AAAA,MACf,UAAU,CAAC,EAAE,MAAM,QAAQ,SAAS,QAAQ,OAAO,CAAC;AAAA,MACpD,aAAa,QAAQ;AAAA,MACrB,WAAW,QAAQ;AAAA,IACvB,CAAC;AAAA,EACL;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,MAAM,WAAW,SAAuD;AACpE,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,sBAAsB;AAAA,MACrE,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf,OAAO,QAAQ;AAAA,MACnB,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AAEjC,WAAO;AAAA,MACH,OAAO,KAAK;AAAA,MACZ,YAAY,KAAK,MAAM,IAAI,CAAC,MAAM,EAAE,SAAS,KAAK,CAAC;AAAA,MACnD,OAAO;AAAA,QACH,aAAa,KAAK,OAAO,gBAAgB;AAAA,MAC7C;AAAA,IACJ;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAoBA,MAAM,eAA4B,SAAoE;AAElG,UAAM,WAAW,QAAQ,YAAY;AAAA,MACjC,EAAE,MAAM,QAAiB,SAAS,QAAQ,UAAU,GAAG;AAAA,IAC3D;AAGA,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,gBAAgB;AAAA,MAC/D,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf;AAAA,QACA,aAAa,QAAQ;AAAA,QACrB,WAAW,QAAQ;AAAA,QACnB,QAAQ;AAAA,QACR,OAAO,CAAC;AAAA,UACJ,MAAM;AAAA,UACN,UAAU;AAAA,YACN,MAAM,QAAQ,cAAc;AAAA,YAC5B,aAAa,QAAQ,qBAAqB;AAAA,YAC1C,YAAY,QAAQ;AAAA,UACxB;AAAA,QACJ,CAAC;AAAA,QACD,YAAY;AAAA,UACR,MAAM;AAAA,UACN,UAAU,EAAE,MAAM,QAAQ,cAAc,kBAAkB;AAAA,QAC9D;AAAA,MACJ,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AACjC,UAAM,WAAW,KAAK,UAAU,CAAC,GAAG,SAAS,aAAa,CAAC;AAE3D,QAAI,CAAC,UAAU;AACX,YAAM,IAAI,MAAM,wCAAwC;AAAA,IAC5D;AAEA,QAAI;AACJ,QAAI;AACA,qBAAe,KAAK,MAAM,SAAS,SAAS,SAAS;AAAA,IACzD,QAAQ;AACJ,YAAM,IAAI,MAAM,2CAA2C;AAAA,IAC/D;AAEA,WAAO;AAAA,MACH,QAAQ;AAAA,MACR,OAAO;AAAA,QACH,cAAc,KAAK,OAAO,iBAAiB;AAAA,QAC3C,kBAAkB,KAAK,OAAO,qBAAqB;AAAA,QACnD,aAAa,KAAK,OAAO,gBAAgB;AAAA,MAC7C;AAAA,IACJ;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAaA,MAAM,cAAc,SAAmE;AACnF,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,2BAA2B;AAAA,MAC1E,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,QAAQ,QAAQ;AAAA,QAChB,OAAO,QAAQ,SAAS;AAAA,QACxB,GAAG,QAAQ;AAAA,QACX,MAAM,QAAQ;AAAA,QACd,SAAS,QAAQ;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf,gBAAgB,QAAQ;AAAA,MAC5B,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,WAAW,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IACvG;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AAMjC,WAAO;AAAA,MACH,QAAQ,KAAK,OAAO,IAAI,UAAQ;AAAA,QAC5B,KAAK,IAAI;AAAA,QACT,SAAS,IAAI;AAAA,QACb,eAAe,IAAI;AAAA,MACvB,EAAE;AAAA,MACF,OAAO,KAAK;AAAA,MACZ,UAAU,KAAK;AAAA,IACnB;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAkBA,MAAM,IAAI,SAA2C;AACjD,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,eAAe;AAAA,MAC9D,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf,UAAU,QAAQ;AAAA,QAClB,WAAW,QAAQ;AAAA,QACnB,aAAa,QAAQ;AAAA,QACrB,WAAW,QAAQ;AAAA,QACnB,OAAO,QAAQ,SAAS;AAAA,QACxB,WAAW,QAAQ,aAAa;AAAA,QAChC,iBAAiB,QAAQ,kBAAkB;AAAA,QAC3C,QAAQ;AAAA,MACZ,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,UAAM,OAAO,MAAM,SAAS,KAAK;AAEjC,WAAO;AAAA,MACH,SAAS;AAAA,QACL,MAAM;AAAA,QACN,SAAS,KAAK,QAAQ;AAAA,MAC1B;AAAA,MACA,OAAO,KAAK;AAAA,MACZ,UAAU,KAAK;AAAA,MACf,OAAO;AAAA,QACH,cAAc,KAAK,MAAM;AAAA,QACzB,kBAAkB,KAAK,MAAM;AAAA,QAC7B,aAAa,KAAK,MAAM;AAAA,MAC5B;AAAA,MACA,SAAS,KAAK,SAAS,IAAI,QAAM;AAAA,QAC7B,SAAS,EAAE;AAAA,QACX,UAAU,EAAE;AAAA,QACZ,YAAY,EAAE;AAAA,MAClB,EAAE;AAAA,MACF,WAAW,KAAK;AAAA,IACpB;AAAA,EACJ;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWA,OAAO,UAAU,SAAoE;AACjF,UAAM,WAAW,MAAM,MAAM,GAAG,KAAK,OAAO,OAAO,eAAe;AAAA,MAC9D,QAAQ;AAAA,MACR,SAAS;AAAA,QACL,mBAAmB,KAAK,OAAO;AAAA,QAC/B,gBAAgB;AAAA,QAChB,GAAG,KAAK,OAAO;AAAA,MACnB;AAAA,MACA,MAAM,KAAK,UAAU;AAAA,QACjB,OAAO,QAAQ;AAAA,QACf,UAAU,QAAQ;AAAA,QAClB,WAAW,QAAQ;AAAA,QACnB,aAAa,QAAQ;AAAA,QACrB,WAAW,QAAQ;AAAA,QACnB,OAAO,QAAQ,SAAS;AAAA,QACxB,WAAW,QAAQ,aAAa;AAAA,QAChC,iBAAiB,QAAQ,kBAAkB;AAAA,QAC3C,QAAQ;AAAA,MACZ,CAAC;AAAA,IACL,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AACd,YAAM,YAAY,MAAM,SAAS,KAAK,EAAE,MAAM,OAAO,EAAE,OAAO,gBAAgB,EAAE;AAChF,YAAM,IAAI,MAAM,sBAAsB,UAAU,SAAS,SAAS,UAAU,EAAE;AAAA,IAClF;AAEA,QAAI,CAAC,SAAS,MAAM;AAChB,YAAM,IAAI,MAAM,uBAAuB;AAAA,IAC3C;AAEA,UAAM,SAAS,SAAS,KAAK,UAAU;AACvC,UAAM,UAAU,IAAI,YAAY;AAChC,QAAI,SAAS;AAEb,QAAI;AACA,aAAO,MAAM;AACT,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAE1C,YAAI,KAAM;AAEV,kBAAU,QAAQ,OAAO,OAAO,EAAE,QAAQ,KAAK,CAAC;AAChD,cAAM,QAAQ,OAAO,MAAM,IAAI;AAC/B,iBAAS,MAAM,IAAI,KAAK;AAExB,mBAAW,QAAQ,OAAO;AACtB,cAAI,KAAK,KAAK,MAAM,GAAI;AACxB,cAAI,CAAC,KAAK,WAAW,QAAQ,EAAG;AAEhC,gBAAM,OAAO,KAAK,MAAM,CAAC;AAEzB,cAAI,SAAS,UAAU;AACnB;AAAA,UACJ;AAEA,cAAI;AACA,kBAAM,QAAQ,KAAK,MAAM,IAAI;AAC7B,kBAAM;AAAA,UACV,QAAQ;AAAA,UAER;AAAA,QACJ;AAAA,MACJ;AAAA,IACJ,UAAE;AACE,aAAO,YAAY;AAAA,IACvB;AAAA,EACJ;AACJ;","names":[]}

package/dist/ai/index.mjs

@@ -288,6 +288,131 @@ var AINamespace = class {
       provider: data.provider
     };
   }
+  /**
+   * RAG (Retrieval-Augmented Generation) - Chat with automatic memory context
+   * 
+   * Searches your memory namespace for relevant context and includes it
+   * in the prompt automatically. Returns the AI response along with sources.
+   * 
+   * @example
+   * const response = await cencori.ai.rag({
+   *   model: 'gpt-4o',
+   *   messages: [{ role: 'user', content: 'What are our company policies?' }],
+   *   namespace: 'company-docs',
+   *   limit: 5, // number of memories to retrieve
+   * });
+   * console.log(response.message.content);
+   * console.log(response.sources); // retrieved context
+   */
+  async rag(request) {
+    const response = await fetch(`${this.config.baseUrl}/api/ai/rag`, {
+      method: "POST",
+      headers: {
+        "CENCORI_API_KEY": this.config.apiKey,
+        "Content-Type": "application/json",
+        ...this.config.headers
+      },
+      body: JSON.stringify({
+        model: request.model,
+        messages: request.messages,
+        namespace: request.namespace,
+        temperature: request.temperature,
+        maxTokens: request.maxTokens,
+        limit: request.limit ?? 5,
+        threshold: request.threshold ?? 0.5,
+        include_sources: request.includeSources ?? true,
+        stream: false
+      })
+    });
+    if (!response.ok) {
+      const errorData = await response.json().catch(() => ({ error: "Unknown error" }));
+      throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);
+    }
+    const data = await response.json();
+    return {
+      message: {
+        role: "assistant",
+        content: data.message.content
+      },
+      model: data.model,
+      provider: data.provider,
+      usage: {
+        promptTokens: data.usage.prompt_tokens,
+        completionTokens: data.usage.completion_tokens,
+        totalTokens: data.usage.total_tokens
+      },
+      sources: data.sources?.map((s) => ({
+        content: s.content,
+        metadata: s.metadata,
+        similarity: s.similarity
+      })),
+      latencyMs: data.latency_ms
+    };
+  }
+  /**
+   * Stream RAG responses with automatic memory context
+   * 
+   * @example
+   * for await (const chunk of cencori.ai.ragStream({ model: 'gpt-4o', messages, namespace: 'docs' })) {
+   *   if (chunk.type === 'sources') console.log('Sources:', chunk.sources);
+   *   if (chunk.type === 'content') process.stdout.write(chunk.delta);
+   * }
+   */
+  async *ragStream(request) {
+    const response = await fetch(`${this.config.baseUrl}/api/ai/rag`, {
+      method: "POST",
+      headers: {
+        "CENCORI_API_KEY": this.config.apiKey,
+        "Content-Type": "application/json",
+        ...this.config.headers
+      },
+      body: JSON.stringify({
+        model: request.model,
+        messages: request.messages,
+        namespace: request.namespace,
+        temperature: request.temperature,
+        maxTokens: request.maxTokens,
+        limit: request.limit ?? 5,
+        threshold: request.threshold ?? 0.5,
+        include_sources: request.includeSources ?? true,
+        stream: true
+      })
+    });
+    if (!response.ok) {
+      const errorData = await response.json().catch(() => ({ error: "Unknown error" }));
+      throw new Error(`Cencori API error: ${errorData.error || response.statusText}`);
+    }
+    if (!response.body) {
+      throw new Error("Response body is null");
+    }
+    const reader = response.body.getReader();
+    const decoder = new TextDecoder();
+    let buffer = "";
+    try {
+      while (true) {
+        const { done, value } = await reader.read();
+        if (done) break;
+        buffer += decoder.decode(value, { stream: true });
+        const lines = buffer.split("\n");
+        buffer = lines.pop() || "";
+        for (const line of lines) {
+          if (line.trim() === "") continue;
+          if (!line.startsWith("data: ")) continue;
+          const data = line.slice(6);
+          if (data === "[DONE]") {
+            return;
+          }
+          try {
+            const chunk = JSON.parse(data);
+            yield chunk;
+          } catch {
+          }
+        }
+      }
+    } finally {
+      reader.releaseLock();
+    }
+  }
 };
 export {
   AINamespace