Merge branch 'develop' into feature/rag-knowledge-improvements-v2

Author: augchan42

packages/client-github/README.md

@@ -47,7 +47,6 @@ const client = await GitHubClientInterface.start(runtime);
 // Convert repository files to agent memories
 await client.createMemoriesFromFiles();
 
-typescript
 // Convert repository files to agent memories
 await client.createMemoriesFromFiles();
 ```
@@ -67,8 +66,6 @@ await client.createPullRequest(
     "Implements new functionality with tests"
 );
 
-
-typescript
 await client.createPullRequest(
 "Feature: Add new functionality",
 "feature/new-feature",
@@ -94,8 +91,6 @@ await client.createCommit(
         }
     ]
 );
-
-
 ```
 
 ## API Reference

packages/core/src/generation.ts

@@ -575,20 +575,23 @@ export async function generateText({
                 const openai = createOpenAI({
                     apiKey,
                     baseURL: endpoint,
-                    fetch: async (url: string, options: any) => {
+                    fetch: async (input: RequestInfo | URL, init?: RequestInit): Promise<Response> => {
+                        const url = typeof input === 'string' ? input : input.toString();
                         const chain_id =
                             runtime.getSetting("ETERNALAI_CHAIN_ID") || "45762";
+                        
+                        const options: RequestInit = { ...init };
                         if (options?.body) {
-                            const body = JSON.parse(options.body);
+                            const body = JSON.parse(options.body as string);
                             body.chain_id = chain_id;
                             options.body = JSON.stringify(body);
                         }
+
                         const fetching = await runtime.fetch(url, options);
-                        if (
-                            parseBooleanFromText(
-                                runtime.getSetting("ETERNALAI_LOG")
-                            )
-                        ) {
+                        
+                        if (parseBooleanFromText(
+                            runtime.getSetting("ETERNALAI_LOG")
+                        )) {
                             elizaLogger.info(
                                 "Request data: ",
                                 JSON.stringify(options, null, 2)
@@ -1272,7 +1275,6 @@ export async function splitChunks(
  * @param opts.presence_penalty The presence penalty to apply (0.0 to 2.0)
  * @param opts.temperature The temperature to control randomness (0.0 to 2.0)
  * @param opts.serverUrl The URL of the API server
- * @param opts.token The API token for authentication
  * @param opts.max_context_length Maximum allowed context length in tokens
  * @param opts.max_response_length Maximum allowed response length in tokens
  * @returns Promise resolving to a boolean value parsed from the model's response

packages/core/src/runtime.ts

@@ -380,9 +380,9 @@ export class AgentRuntime implements IAgentRuntime {
         this.imageVisionModelProvider =
             this.character.imageVisionModelProvider ?? this.modelProvider;
 
-        elizaLogger.info("Selected model provider:", this.modelProvider);
+        // elizaLogger.info("Selected model provider:", this.modelProvider); duplicated log ln: 343
         elizaLogger.info(
-            "Selected image model provider:",
+            "Selected image vision model provider:",
             this.imageVisionModelProvider
         );
 

packages/plugin-node/src/services/image.ts

@@ -189,6 +189,51 @@ class OpenAIImageProvider implements ImageProvider {
     }
 }
 
+
+class GroqImageProvider implements ImageProvider {
+    constructor(private runtime: IAgentRuntime) {}
+
+    async initialize(): Promise<void> {}
+
+    async describeImage(
+        imageData: Buffer,
+        mimeType: string
+    ): Promise<{ title: string; description: string }> {
+        const imageUrl = convertToBase64DataUrl(imageData, mimeType);
+
+        const content = [
+            { type: "text", text: IMAGE_DESCRIPTION_PROMPT },
+            { type: "image_url", image_url: { url: imageUrl } },
+        ];
+
+        const endpoint =
+            this.runtime.imageVisionModelProvider === ModelProviderName.GROQ
+                ? getEndpoint(this.runtime.imageVisionModelProvider)
+                : "https://api.groq.com/openai/v1/";
+
+        const response = await fetch(endpoint + "/chat/completions", {
+            method: "POST",
+            headers: {
+                "Content-Type": "application/json",
+                Authorization: `Bearer ${this.runtime.getSetting("GROQ_API_KEY")}`,
+            },
+            body: JSON.stringify({
+                model: /*this.runtime.imageVisionModelName ||*/ "llama-3.2-90b-vision-preview",
+                messages: [{ role: "user", content }],
+                max_tokens: 1024,
+            }),
+        });
+
+        if (!response.ok) {
+            await handleApiError(response, "Groq");
+        }
+
+        const data = await response.json();
+        return parseImageResponse(data.choices[0].message.content);
+    }
+}
+
+
 class GoogleImageProvider implements ImageProvider {
     constructor(private runtime: IAgentRuntime) {}
 
@@ -280,6 +325,12 @@ export class ImageDescriptionService
             ) {
                 this.provider = new OpenAIImageProvider(this.runtime);
                 elizaLogger.debug("Using openai for vision model");
+            } else if (
+                this.runtime.imageVisionModelProvider ===
+                ModelProviderName.GROQ
+            ) {
+                this.provider = new GroqImageProvider(this.runtime);
+                elizaLogger.debug("Using Groq for vision model");
             } else {
                 elizaLogger.error(
                     `Unsupported image vision model provider: ${this.runtime.imageVisionModelProvider}`
@@ -291,6 +342,9 @@ export class ImageDescriptionService
         } else if (model === models[ModelProviderName.GOOGLE]) {
             this.provider = new GoogleImageProvider(this.runtime);
             elizaLogger.debug("Using google for vision model");
+        } else if (model === models[ModelProviderName.GROQ]) {
+            this.provider = new GroqImageProvider(this.runtime);
+            elizaLogger.debug("Using groq for vision model");
         } else {
             elizaLogger.debug("Using default openai for vision model");
             this.provider = new OpenAIImageProvider(this.runtime);