- Fix the embeddings ranking - the wrong query was used

- Removed the free DeepSeek model from open router as not working now
- Setting endpoint_tools is now enought to start agent
- Agent View is now not hidden if the tools model is deselected
- Refactoring

.vscodeignore

@@ -1,15 +1,50 @@
+# Exclude VCS and workspace config
+.git/**
+.github/**
 .vscode/**
-.idea/**
-.vscode-test/**
+
+# Do not ship TypeScript sources or source maps
 src/**
-.gitignore
-.yarnrc
-vsc-extension-quickstart.md
-**/tsconfig.json
-**/eslint.config.mjs
 **/*.map
 **/*.ts
-**/.vscode-test.*
-ui/node_modules/**
-ui/src/**
-**/test
+**/*.tsx
+**/*.test.*
+**/*.spec.*
+
+# VS Code testing/coverage artifacts
+.vscode-test/**
+coverage/**
+
+# Build artifacts and local packages
+*.vsix
+.vsix/**
+
+# UI workspace: keep only built assets
+ui/**
+!ui/dist/**
+
+# Keep built extension output and resources
+!dist/**
+!resources/**
+!llama.png
+!README.md
+!LICENSE
+!package.json
+
+# Node modules: keep runtime code but drop heavy non-runtime extras
+node_modules/**/test/**
+node_modules/**/tests/**
+node_modules/**/__tests__/**
+node_modules/**/benchmark/**
+node_modules/**/benchmarks/**
+node_modules/**/example/**
+node_modules/**/examples/**
+node_modules/**/doc/**
+node_modules/**/docs/**
+node_modules/**/man/**
+node_modules/**/.github/**
+node_modules/**/.vscode/**
+
+# Logs and misc
+npm-debug.log*
+yarn-error.log*

package.json

@@ -1946,7 +1946,6 @@
     "format": "prettier --write --ignore-path .gitignore '**/*'"
   },
   "dependencies": {
-    "@vscode/test-cli": "^0.0.11",
     "axios": "^1.1.2",
     "globby": "^14.1.0",
     "ignore": "^7.0.4",
@@ -1956,6 +1955,7 @@
     "simple-git": "^3.28.0"
   },
   "devDependencies": {
+    "@vscode/test-cli": "^0.0.11",
     "@babel/types": "^7.28.4",
     "@types/micromatch": "^4.0.9",
     "@types/mocha": "^10.0.10",

src/architect.ts

@@ -8,7 +8,7 @@ import {LlamaWebviewProvider} from './llama-webview-provider'
 import { Utils } from './utils';
 import { Env, LlmModel } from './types';
 import { env } from 'process';
-import { PERSISTENCE_KEYS, SETTING_NAME_FOR_LIST } from './constants';
+import { PERSISTENCE_KEYS, SETTING_NAME_FOR_LIST, UiView } from './constants';
 
 export class Architect {
     private app: Application
@@ -398,7 +398,8 @@ export class Architect {
             'extension.showLlamaWebview',
             async () => {
                 vscode.commands.executeCommand('llama-vscode.webview.focus');
-                this.app.llamaWebviewProvider.setView("agent")
+                if (this.app.isToolsModelSelected() || this.app.configuration.endpoint_tools) this.app.llamaWebviewProvider.setView(UiView.Agent)
+                else this.app.llamaWebviewProvider.setView(UiView.Environment)
                 const editor = vscode.window.activeTextEditor;
                 if (editor && editor.selection) {
                     let fileLongName = editor.document.fileName;

src/chat-context.ts

@@ -56,7 +56,7 @@ export class ChatContext {
         if ((this.app.getEmbeddingsModel().endpoint && this.app.getEmbeddingsModel().endpoint?.trim() != "") 
             || this.app.configuration.endpoint_embeddings.trim() != ""){
             this.app.statusbar.showTextInfo(this.app.configuration.getUiText("Filtering chunks with embeddings..."))
-            topContextChunks = await this.cosineSimilarityRank(query, topChunksBm25, this.app.configuration.rag_max_embedding_filter_chunks);
+            topContextChunks = await this.cosineSimilarityRank(prompt, topChunksBm25, this.app.configuration.rag_max_embedding_filter_chunks);
         } else {
             vscode.window.showInformationMessage('No embeddings server. Filtering chunks with embeddings will be skipped.');
             topContextChunks = topChunksBm25.slice(0, 5);

src/chat-with-ai.ts

@@ -56,21 +56,19 @@ export class ChatWithAi {
                         : this.app.configuration.endpoint_tools ? this.app.configuration.endpoint_tools + "/" : "";
 
         let chatModel = this.app.getChatModel();
-        if (!this.app.isChatModelSelected()) chatModel = this.app.getToolsModel();    
+        if (!this.app.isChatModelSelected() && !this.app.configuration.endpoint_chat) chatModel = this.app.getToolsModel();    
         if (chatModel.endpoint) {
             const chatEndpoint = Utils.trimTrailingSlash(chatModel.endpoint)
             targetUrl = chatEndpoint ? chatEndpoint + "/" : "";
         }
         if (!targetUrl) {
-            const shouldSelectModel = await Utils.showUserChoiceDialog("Select a chat or tools model run by llama-server or an env with chat or tools model run on llama-server to chat with AI.","Select")
-            if (shouldSelectModel){
-                this.app.llamaWebviewProvider.showEnvView();
-                vscode.window.showInformationMessage("After the chat/tools model is loaded, try again opening Chat with AI.")
-                return;
-            } else {
-                vscode.window.showErrorMessage("No endpoint for the chat or tools model. Select a chat or tools model run on llama-server or an env with chat or tools model or enter the endpoint of a running llama.cpp server with chat model in setting endpoint_chat. ")
-                return
-            }
+            await Utils.suggestModelSelection(
+                "Select a chat or tools model run by llama-server or an env with chat or tools model run on llama-server to chat with AI.",
+                "After the chat/tools model is loaded, try again opening Chat with AI.",
+                "No endpoint for the chat or tools model. Select a chat or tools model run on llama-server or an env with chat or tools model or enter the endpoint of a running llama.cpp server with chat model in setting endpoint_chat. ",
+                this.app
+            );
+            return
         }
 
         if (withContext){

src/configuration.ts

@@ -346,7 +346,7 @@ export class Configuration {
             };
         }
 
-        this.axiosRequestConfigTools = {};
+        this.axiosRequestConfigTools = {};  
         if (this.api_key_tools != undefined && this.api_key_tools.trim() != "") {
             this.axiosRequestConfigTools = {
                 headers: {

src/constants.ts

@@ -6,6 +6,13 @@ export enum ModelType {
   Tools = 'tools'
 }
 
+export enum UiView {
+  Agent = 'agent',
+  Environment = 'addenv',
+  Embeddings = 'embeddings',
+  AiRunner = 'airunner'
+}
+
 export enum Action {
   Select = 'select',
   AddLocal = 'addLocal',

src/git.ts

@@ -14,22 +14,20 @@ export class Git {
         let chatUrl = this.app.configuration.endpoint_chat
         if (!chatUrl) chatUrl = this.app.configuration.endpoint_tools;
         let chatModel = this.app.getChatModel();    
-        if (!this.app.isChatModelSelected()) chatModel = this.app.getToolsModel();
+        if (!this.app.isChatModelSelected() && !this.app.configuration.endpoint_chat) chatModel = this.app.getToolsModel();
         if (chatModel.endpoint) {
             const chatEndpoint = Utils.trimTrailingSlash(chatModel.endpoint)
             chatUrl = chatEndpoint ? chatEndpoint + "/" : "";
         }
         if (!chatUrl) {
-            const shouldSelectModel = await Utils.showUserChoiceDialog("Select a chat or tools model or an env with chat or tools model to generate a commit message.","Select")
-            if (shouldSelectModel){
-                this.app.llamaWebviewProvider.showEnvView();
-                vscode.window.showInformationMessage("After the chat/tools model is loaded, try again generating commit message.")
-                return;
-            } 
-            else {
-                vscode.window.showErrorMessage("No endpoint for the chat model. Select a chat or tools model or an env with chat or tools model or enter the endpoint of a running llama.cpp server with chat model in setting endpoint_chat. ")
-                return;
-            }
+            await Utils.suggestModelSelection(
+                "Select a chat or tools model or an env with chat or tools model to generate a commit message.",
+                "After the chat/tools model is loaded, try again generating commit message.",
+                "No endpoint for the chat model. Select a chat or tools model or an env with chat or tools model or enter the endpoint of a server with chat model in setting endpoint_chat.",
+                this.app
+            );
+
+            return
         }
 
         const gitExtension = vscode.extensions.getExtension('vscode.git')?.exports;

src/lists.ts

@@ -82,13 +82,6 @@ export const PREDEFINED_LISTS = new Map<string, any>([
               "aiModel": "",
               "isKeyRequired": false
             },
-            {
-            "name": "DeepSeek V3.1 (free) 163,800 context (OpenRouter)",
-            "localStartCommand": "",
-            "endpoint": "https://openrouter.ai/api",
-            "aiModel": "deepseek/deepseek-chat-v3.1:free",
-            "isKeyRequired": true
-            },
             {
               "name": "Z.AI: GLM 4.5 Air (free): GLM 4.5 Air - 128.000 context (OpenRouter)",
               "endpoint": "https://openrouter.ai/api",

src/llama-agent.ts

@@ -145,8 +145,8 @@ export class LlamaAgent {
             this.logText += "***" + query.replace("\n", "  \n") + "***" + "\n\n"; // Make sure markdown shows new lines correctly
 
             
-            if (!this.app.isToolsModelSelected()) {
-                vscode.window.showErrorMessage("Error: Tools model is not selected! Select tools model (or orchestra with tools model) if you want to to use Llama Agent.")
+            if (!this.app.isToolsModelSelected() && !this.app.configuration.endpoint_tools) {
+                vscode.window.showErrorMessage("Error: Tools model is not selected! Select tools model (or env with tools model) or set and endpoint in setting endpoint_tools if you want to to use Llama Agent View.")
                 this.app.llamaWebviewProvider.setState("AI is stopped")
                 return "Tools model is not selected"
             }

src/llama-server.ts

@@ -743,27 +743,31 @@ export class LlamaServer {
 
     private getChatModelProperties() {
         let selectedModel: LlmModel = this.app.getChatModel();
-        if (!this.app.isChatModelSelected()) selectedModel = this.app.getToolsModel();
-        
-        let model = this.app.configuration.ai_model;
-        if (selectedModel?.aiModel !== undefined && selectedModel.aiModel) model = selectedModel.aiModel;
+        if (!this.app.isChatModelSelected() && !this.app.configuration.endpoint_chat) selectedModel = this.app.getToolsModel();
 
         let endpoint = this.app.configuration.endpoint_chat;
-        if (!endpoint) endpoint = this.app.configuration.endpoint_tools;
-        if (selectedModel?.endpoint !== undefined && selectedModel.endpoint) endpoint = selectedModel.endpoint;
-
+        let model = this.app.configuration.ai_model;
         let requestConfig = this.app.configuration.axiosRequestConfigChat;
-        if (selectedModel?.isKeyRequired !== undefined && selectedModel.isKeyRequired) {
-            const apiKey = this.app.persistence.getApiKey(selectedModel.endpoint??"");
-            if (apiKey) {
-                requestConfig = {
-                    headers: {
-                        Authorization: `Bearer ${apiKey}`,
-                        "Content-Type": "application/json",
-                    },
-                };
+        if (!endpoint) { 
+            endpoint = this.app.configuration.endpoint_tools;
+            requestConfig = this.app.configuration.axiosRequestConfigTools;
+        }
+        if (selectedModel?.endpoint !== undefined && selectedModel.endpoint) {
+            endpoint = selectedModel.endpoint;
+            if (selectedModel?.aiModel !== undefined && selectedModel.aiModel) model = selectedModel.aiModel;
+            if (selectedModel?.isKeyRequired !== undefined && selectedModel.isKeyRequired) {
+                const apiKey = this.app.persistence.getApiKey(selectedModel.endpoint??"");
+                if (apiKey) {
+                    requestConfig = {
+                        headers: {
+                            Authorization: `Bearer ${apiKey}`,
+                            "Content-Type": "application/json",
+                        },
+                    };
+                }
             }
         }
+
         return { endpoint, model, requestConfig };
     }
 

src/menu.ts

@@ -1,7 +1,7 @@
 import {Application} from "./application";
 import vscode, { QuickPickItem } from "vscode";
 import { Utils } from "./utils";
-import { ModelType, AGENT_NAME, UI_TEXT_KEYS } from "./constants";
+import { ModelType, AGENT_NAME, UI_TEXT_KEYS, UiView } from "./constants";
 
 export class Menu {
     private app: Application
@@ -228,7 +228,7 @@ export class Menu {
                 break;
             case this.app.configuration.getUiText(UI_TEXT_KEYS.useAsLocalAIRunner):
                 vscode.commands.executeCommand('extension.showLlamaWebview');
-                this.app.llamaWebviewProvider.setView("airunner");
+                this.app.llamaWebviewProvider.setView(UiView.AiRunner);
                 break;
             default:
                 isHandled = false;

src/services/model-service.ts

@@ -408,14 +408,13 @@ export class ModelService {
             targetUrl = toolsEndpoint ? toolsEndpoint + "/" : "";
         }
         if (!targetUrl) {
-            const shouldSelectEnv = await Utils.showUserChoiceDialog("Select a tools model or an env with tools model to use Llama Agent.", "Select");
-            if (shouldSelectEnv) {
-                // await this.app.menu.selectEnvFromList(this.app.configuration.envs_list.filter(item => item.tools != undefined && item.tools.name));
-                this.app.llamaWebviewProvider.showEnvView()
-                vscode.window.showInformationMessage("After the tools model is loaded, try again opening llama agent.");
-            } else {
-                vscode.window.showErrorMessage("No endpoint for the tools model. Select an env with tools model or enter the endpoint of a running llama.cpp server with tools model in setting endpoint_tools. ");
-            }
+            await Utils.suggestModelSelection(
+                "Select a tools model or an env with tools model to use Llama Agent.",
+                "After the tools model is loaded, try again opening llama agent.",
+                "No endpoint for the tools model. Select an env with tools model or enter the endpoint of a running llama.cpp server with tools model in setting endpoint_tools.",
+                this.app
+            );
+
             return false;
         }
         else return true;

src/text-editor.ts

@@ -36,15 +36,13 @@ export class TextEditor {
             chatUrl = chatEndpoint ? chatEndpoint + "/" : "";
         }
         if (!chatUrl) { 
-            const shouldSelectModel = await Utils.showUserChoiceDialog("Select a chat or tools model or an env with chat or tools model to edit code with AI.","Select")
-            if (shouldSelectModel){
-                this.app.llamaWebviewProvider.showEnvView();
-                vscode.window.showInformationMessage("After the chat model is loaded, try again using Edit with AI.")
-                return;
-            } else {
-                vscode.window.showErrorMessage("No endpoint for the chat model. Select an env with chat model or enter the endpoint of a running llama.cpp server with chat model in setting endpoint_chat. ")
-                return
-            }
+            await Utils.suggestModelSelection(
+                "Select a chat or tools model or an env with chat or tools model to edit code with AI.",
+                "After the chat model is loaded, try again using Edit with AI.",
+                "No endpoint for the chat model. Select an env with chat model or enter the endpoint of a running llama.cpp server with chat model in setting endpoint_chat.",
+                this.app
+            );
+            return
         }
 
         if (editor.selection.isEmpty) {

src/utils.ts

@@ -7,6 +7,7 @@ import pm from 'picomatch'
 import * as https from 'https';
 import * as http from 'http';
 import { URL } from 'url';
+import { Application } from "./application";
 
 
 interface BM25Stats {
@@ -865,4 +866,14 @@ export class Utils {
         }
         return items;
     } 
+
+    static suggestModelSelection = async (choiceMsg: string, yesMsg: string, noMsg: string, app: Application) => {
+        const shouldSelectModel = await Utils.showUserChoiceDialog(choiceMsg, "Select");
+        if (shouldSelectModel) {
+            app.llamaWebviewProvider.showEnvView();
+            vscode.window.showInformationMessage(yesMsg);
+        } else {
+            vscode.window.showErrorMessage(noMsg);
+        }
+    }
 }

ui/src/components/AgentView.tsx

@@ -321,7 +321,8 @@ const AgentView: React.FC<AgentViewProps> = ({
       {/* Modern Header */}
       <div className="header">
         <div className="header-content">
-          {!currentToolsModel.includes('No model selected') && (
+          // TODO Remove this check once you are sure it works fine without it
+          {!currentToolsModel.includes('No model selected...') && (
             <div className="header-left">
               <button
                 onClick={handleClearText}
@@ -365,7 +366,7 @@ const AgentView: React.FC<AgentViewProps> = ({
       </div>
 
               {/* Main Content */}
-        {!currentToolsModel.includes('No model selected') && (
+        {!currentToolsModel.includes('No model selected...') && (
           <div className="content" style={{ flex: 1, display: 'flex', flexDirection: 'column', minHeight: 0 }}>
            {/* Chat Display Area */}
            {/* Markdown Display Area */}