llm params changed

Author: rawatSingh2022

src/index.ts

@@ -26,7 +26,8 @@ import codeboltTools from './modules/tools';
 import cbagent from './modules/agent';
 import cbutils from './modules/utils';
 
-
+// Export types for clean imports
+export type { Message, ToolCall, Tool, LLMInferenceParams } from './modules/llm';
 
 /**
  * @class Codebolt

src/modules/llm.ts

@@ -1,31 +1,125 @@
 import cbws from '../core/websocket';
 import {LLMResponse } from '@codebolt/types';
+
+/**
+ * Represents a message in the conversation with roles and content.
+ */
+export interface Message {
+    /** The role of the message sender: user, assistant, tool, or system */
+    role: 'user' | 'assistant' | 'tool' | 'system';
+    /** The content of the message, can be an array of content blocks or a string */
+    content: any[] | string;
+    /** Optional ID for tool calls */
+    tool_call_id?: string;
+    /** Optional tool calls for assistant messages */
+    tool_calls?: ToolCall[];
+    /** Additional properties that might be present */
+    [key: string]: any;
+}
+
+/**
+ * Represents a tool call in OpenAI format
+ */
+export interface ToolCall {
+    /** Unique identifier for this tool call */
+    id: string;
+    /** The type of tool call */
+    type: 'function';
+    /** Function call details */
+    function: {
+        /** Name of the function to call */
+        name: string;
+        /** Arguments for the function call as JSON string */
+        arguments: string;
+    };
+}
+
+/**
+ * Represents a tool definition in OpenAI format
+ */
+export interface Tool {
+    /** The type of tool */
+    type: 'function';
+    /** Function definition */
+    function: {
+        /** Name of the function */
+        name: string;
+        /** Description of what the function does */
+        description?: string;
+        /** JSON schema for the function parameters */
+        parameters?: any;
+    };
+}
+
+/**
+ * LLM inference request parameters
+ */
+export interface LLMInferenceParams {
+    /** Array of messages in the conversation */
+    messages: Message[];
+    /** Available tools for the model to use */
+    tools?: Tool[];
+    /** How the model should use tools */
+    tool_choice?: 'auto' | 'none' | 'required' | { type: 'function'; function: { name: string } };
+    /** The LLM role to determine which model to use */
+    llmrole: string;
+    /** Maximum number of tokens to generate */
+    max_tokens?: number;
+    /** Temperature for response generation */
+    temperature?: number;
+    /** Whether to stream the response */
+    stream?: boolean;
+}
+
 /**
  * A module for interacting with language learning models (LLMs) via WebSocket.
  */
 const cbllm = {
     /**
-     * Sends an inference request to the LLM and returns the model's response.
+     * Sends an inference request to the LLM using OpenAI message format with tools support.
      * The model is selected based on the provided `llmrole`. If the specific model
      * for the role is not found, it falls back to the default model for the current agent,
      * and ultimately to the default application-wide LLM if necessary.
      * 
-     * @param {string} message - The input message or prompt to be sent to the LLM.
-     * @param {string} llmrole - The role of the LLM to determine which model to use.
+     * @param {LLMInferenceParams} message - The inference parameters including messages, tools, and options.
      * @returns {Promise<LLMResponse>} A promise that resolves with the LLM's response.
      */
-    inference: async (message: string, llmrole: string): Promise<LLMResponse> => {
+    inference: async (message: LLMInferenceParams): Promise<LLMResponse> => {
         return cbws.messageManager.sendAndWaitForResponse(
             {
                 "type": "inference",
                 "message": {
-                    prompt: message,
-                    llmrole
+                    messages: message.messages,
+                    tools: message.tools,
+                    tool_choice: message.tool_choice,
+                    llmrole: message.llmrole
                 },
             },
             "llmResponse"
         );
+    },
+
+    /**
+     * Legacy method for backward compatibility - converts simple string prompt to message format.
+     * @deprecated Use the new inference method with proper message format instead.
+     * 
+     * @param {string} message - The input message or prompt to be sent to the LLM.
+     * @param {string} llmrole - The role of the LLM to determine which model to use.
+     * @returns {Promise<LLMResponse>} A promise that resolves with the LLM's response.
+     */
+    legacyInference: async (message: string, llmrole: string): Promise<LLMResponse> => {
+        const messages: Message[] = [
+            {
+                role: 'user',
+                content: message
+            }
+        ];
+        
+        return cbllm.inference({
+            messages,
+            llmrole
+        });
     }
 };
 
-export default cbllm;
+export default cbllm;