codemaxxing 1.1.2 → 1.1.4

package/README.md

@@ -10,7 +10,7 @@
 
 Open-source terminal coding agent. Connect **any** LLM — local or remote — and start building. Like Claude Code, but you bring your own model.
 
-**🆕 v1.1.0:** Use GPT-5.4 with your ChatGPT Plus subscription — no API key needed. Just `/login` → OpenAI → OAuth. Same access as Codex CLI.
+**🆕 v1.1.3:** GPT-5.4 via ChatGPT Plus OAuth, Anthropic OAuth auto-refresh, better Windows terminal behavior, Escape-to-cancel, model picker fixes, and smoother first-run auth/model selection.
 
 ## Why?
 
@@ -71,7 +71,7 @@ If you already have a local server running, Codemaxxing auto-detects common defa
 
 For LM Studio:
 1. Download [LM Studio](https://lmstudio.ai)
-2. Load a coding model (for example **Qwen 2.5 Coder 7B** for a lightweight test)
+2. Load a coding model (ideally **Qwen 2.5 Coder 14B** or **DeepSeek Coder V2 16B** if your machine can handle it; use 7B only as a fallback)
 3. Start the local server
 4. Run:
 
@@ -89,10 +89,11 @@ codemaxxing
 
 If no LLM is available, Codemaxxing can guide you through:
 - detecting your hardware
-- recommending a model
+- recommending a model (with stronger coding defaults first)
 - installing Ollama
 - downloading the model
 - connecting automatically
+- opening the model picker automatically after cloud auth
 
 ### Option C — ChatGPT Plus (GPT-5.4, easiest cloud option)
 

package/dist/agent.d.ts

@@ -1,3 +1,4 @@
+import type { ChatCompletionTool } from "openai/resources/chat/completions";
 import { type ConnectedServer } from "./utils/mcp.js";
 import type { ProviderConfig } from "./config.js";
 export declare function getModelCost(model: string): {
@@ -28,6 +29,7 @@ export declare class CodingAgent {
     private providerType;
     private currentApiKey;
     private currentBaseUrl;
+    private aborted;
     private messages;
     private tools;
     private cwd;
@@ -97,6 +99,19 @@ export declare class CodingAgent {
     /**
      * Switch to a different model mid-session
      */
+    /**
+     * Get available tools (for UI hints, capabilities display, etc.)
+     */
+    getTools(): ChatCompletionTool[];
+    /**
+     * Abort the current generation. Safe to call from any thread.
+     */
+    abort(): void;
+    /**
+     * Check if generation was aborted, and reset the flag.
+     */
+    isAborted(): boolean;
+    private resetAbort;
     switchModel(model: string, baseUrl?: string, apiKey?: string, providerType?: "openai" | "anthropic"): void;
     /**
      * Attempt to refresh an expired Anthropic OAuth token.

package/dist/agent.js

@@ -153,6 +153,7 @@ export class CodingAgent {
     providerType;
     currentApiKey = null;
     currentBaseUrl = "";
+    aborted = false;
     messages = [];
     tools = FILE_TOOLS;
     cwd;
@@ -268,6 +269,7 @@ export class CodingAgent {
      * and loops until the model responds with text (no more tool calls).
      */
     async chat(userMessage) {
+        this.resetAbort();
         const userMsg = { role: "user", content: userMessage };
         this.messages.push(userMsg);
         saveMessage(this.sessionId, userMsg);
@@ -312,6 +314,14 @@ export class CodingAgent {
             let chunkPromptTokens = 0;
             let chunkCompletionTokens = 0;
             for await (const chunk of stream) {
+                // Check for abort
+                if (this.aborted) {
+                    try {
+                        stream.controller?.abort();
+                    }
+                    catch { }
+                    break;
+                }
                 // Capture usage from the final chunk
                 if (chunk.usage) {
                     chunkPromptTokens = chunk.usage.prompt_tokens ?? 0;
@@ -382,6 +392,11 @@ export class CodingAgent {
                     (this.totalCompletionTokens / 1_000_000) * costs.output;
                 updateSessionCost(this.sessionId, this.totalPromptTokens, this.totalCompletionTokens, this.totalCost);
             }
+            // If aborted, return what we have so far
+            if (this.aborted) {
+                updateTokenEstimate(this.sessionId, this.estimateTokens());
+                return contentText ? contentText + "\n\n_(cancelled)_" : "_(cancelled)_";
+            }
             // If no tool calls, we're done — return the text
             if (toolCalls.size === 0) {
                 updateTokenEstimate(this.sessionId, this.estimateTokens());
@@ -542,12 +557,33 @@ export class CodingAgent {
                 });
             }
         }
-        return msgs;
+        // Sanitize: remove tool_result messages that don't have a matching tool_use
+        const validToolUseIds = new Set();
+        for (const m of msgs) {
+            if (m.role === "assistant" && Array.isArray(m.content)) {
+                for (const block of m.content) {
+                    if (block.type === "tool_use") {
+                        validToolUseIds.add(block.id);
+                    }
+                }
+            }
+        }
+        return msgs.filter((m) => {
+            if (m.role === "user" && Array.isArray(m.content)) {
+                const toolResults = m.content.filter((b) => b.type === "tool_result");
+                if (toolResults.length > 0) {
+                    // Only keep if ALL tool_results have matching tool_use
+                    return toolResults.every((tr) => validToolUseIds.has(tr.tool_use_id));
+                }
+            }
+            return true;
+        });
     }
     /**
      * Anthropic-native streaming chat
      */
     async chatAnthropic(_userMessage) {
+        this.resetAbort();
         let iterations = 0;
         const MAX_ITERATIONS = 20;
         while (iterations < MAX_ITERATIONS) {
@@ -587,6 +623,10 @@ export class CodingAgent {
                     tools: anthropicTools,
                 });
                 stream.on("text", (text) => {
+                    if (this.aborted) {
+                        stream.abort();
+                        return;
+                    }
                     contentText += text;
                     this.options.onToken?.(text);
                 });
@@ -607,6 +647,11 @@ export class CodingAgent {
                 // Re-throw if we can't handle it
                 throw err;
             }
+            // If aborted, return what we have
+            if (this.aborted) {
+                updateTokenEstimate(this.sessionId, this.estimateTokens());
+                return contentText ? contentText + "\n\n_(cancelled)_" : "_(cancelled)_";
+            }
             // Track usage
             if (finalMessage.usage) {
                 const promptTokens = finalMessage.usage.input_tokens;
@@ -735,6 +780,7 @@ export class CodingAgent {
      * OpenAI Responses API chat (for Codex OAuth tokens + GPT-5.4)
      */
     async chatOpenAIResponses(userMessage) {
+        this.resetAbort();
         let iterations = 0;
         const MAX_ITERATIONS = 20;
         while (iterations < MAX_ITERATIONS) {
@@ -771,6 +817,11 @@ export class CodingAgent {
                 }
                 this.messages.push(assistantMessage);
                 saveMessage(this.sessionId, assistantMessage);
+                // If aborted, return what we have
+                if (this.aborted) {
+                    updateTokenEstimate(this.sessionId, this.estimateTokens());
+                    return contentText ? contentText + "\n\n_(cancelled)_" : "_(cancelled)_";
+                }
                 // If no tool calls, we're done
                 if (toolCalls.length === 0) {
                     updateTokenEstimate(this.sessionId, this.estimateTokens());
@@ -868,6 +919,27 @@ export class CodingAgent {
     /**
      * Switch to a different model mid-session
      */
+    /**
+     * Get available tools (for UI hints, capabilities display, etc.)
+     */
+    getTools() {
+        return this.tools;
+    }
+    /**
+     * Abort the current generation. Safe to call from any thread.
+     */
+    abort() {
+        this.aborted = true;
+    }
+    /**
+     * Check if generation was aborted, and reset the flag.
+     */
+    isAborted() {
+        return this.aborted;
+    }
+    resetAbort() {
+        this.aborted = false;
+    }
     switchModel(model, baseUrl, apiKey, providerType) {
         this.model = model;
         if (apiKey)

package/dist/index.js

@@ -229,6 +229,7 @@ function App() {
             setApproval,
             setWizardScreen,
             setWizardIndex,
+            openModelPicker,
         });
     }, []);
     // Initialize agent on mount
@@ -252,6 +253,125 @@ function App() {
     showSuggestionsRef.current = showSuggestions;
     const pastedChunksRef = React.useRef(pastedChunks);
     pastedChunksRef.current = pastedChunks;
+    const openModelPicker = useCallback(async () => {
+        addMsg("info", "Fetching available models...");
+        const groups = {};
+        const providerEntries = [];
+        let localFound = false;
+        const localEndpoints = [
+            { name: "LM Studio", port: 1234 },
+            { name: "Ollama", port: 11434 },
+            { name: "vLLM", port: 8000 },
+            { name: "LocalAI", port: 8080 },
+        ];
+        for (const endpoint of localEndpoints) {
+            if (localFound)
+                break;
+            try {
+                const url = `http://localhost:${endpoint.port}/v1`;
+                const models = await listModels(url, "local");
+                if (models.length > 0) {
+                    groups["Local LLM"] = models.map(m => ({
+                        name: m,
+                        baseUrl: url,
+                        apiKey: "local",
+                        providerType: "openai",
+                    }));
+                    localFound = true;
+                }
+            }
+            catch { /* not running */ }
+        }
+        if (!localFound) {
+            try {
+                const ollamaModels = await listInstalledModelsDetailed();
+                if (ollamaModels.length > 0) {
+                    groups["Local LLM"] = ollamaModels.map(m => ({
+                        name: m.name,
+                        baseUrl: "http://localhost:11434/v1",
+                        apiKey: "ollama",
+                        providerType: "openai",
+                    }));
+                    localFound = true;
+                }
+            }
+            catch { /* Ollama not running */ }
+        }
+        if (localFound) {
+            providerEntries.push({ name: "Local LLM", description: "No auth needed — auto-detected", authed: true });
+        }
+        const anthropicCred = getCredential("anthropic");
+        const claudeModels = ["claude-sonnet-4-6", "claude-opus-4-6", "claude-haiku-4-5-20251001"];
+        if (anthropicCred) {
+            groups["Anthropic (Claude)"] = claudeModels.map(m => ({
+                name: m,
+                baseUrl: "https://api.anthropic.com",
+                apiKey: anthropicCred.apiKey,
+                providerType: "anthropic",
+            }));
+        }
+        providerEntries.push({ name: "Anthropic (Claude)", description: "Claude Opus, Sonnet, Haiku — use your subscription or API key", authed: !!anthropicCred });
+        const openaiCred = getCredential("openai");
+        const openaiModels = ["gpt-5.4", "gpt-5.4-pro", "gpt-5", "gpt-5-mini", "gpt-4.1", "gpt-4.1-mini", "o3", "o4-mini", "gpt-4o"];
+        if (openaiCred) {
+            const isOAuthToken = openaiCred.method === "oauth" || openaiCred.method === "cached-token" ||
+                (!openaiCred.apiKey.startsWith("sk-") && !openaiCred.apiKey.startsWith("sess-"));
+            const baseUrl = isOAuthToken
+                ? "https://chatgpt.com/backend-api"
+                : (openaiCred.baseUrl || "https://api.openai.com/v1");
+            groups["OpenAI (ChatGPT)"] = openaiModels.map(m => ({
+                name: m,
+                baseUrl,
+                apiKey: openaiCred.apiKey,
+                providerType: "openai",
+            }));
+        }
+        providerEntries.push({ name: "OpenAI (ChatGPT)", description: "GPT-5, GPT-4.1, o3 — use your ChatGPT subscription or API key", authed: !!openaiCred });
+        const openrouterCred = getCredential("openrouter");
+        if (openrouterCred) {
+            try {
+                const orModels = await listModels(openrouterCred.baseUrl || "https://openrouter.ai/api/v1", openrouterCred.apiKey);
+                if (orModels.length > 0) {
+                    groups["OpenRouter"] = orModels.slice(0, 20).map(m => ({
+                        name: m,
+                        baseUrl: openrouterCred.baseUrl || "https://openrouter.ai/api/v1",
+                        apiKey: openrouterCred.apiKey,
+                        providerType: "openai",
+                    }));
+                }
+            }
+            catch { /* skip */ }
+        }
+        providerEntries.push({ name: "OpenRouter", description: "200+ models (Claude, GPT, Gemini, Llama, etc.) — one login", authed: !!openrouterCred });
+        const qwenCred = getCredential("qwen");
+        if (qwenCred) {
+            groups["Qwen"] = ["qwen-max", "qwen-plus", "qwen-turbo"].map(m => ({
+                name: m,
+                baseUrl: qwenCred.baseUrl || "https://dashscope.aliyuncs.com/compatible-mode/v1",
+                apiKey: qwenCred.apiKey,
+                providerType: "openai",
+            }));
+        }
+        providerEntries.push({ name: "Qwen", description: "Qwen 3.5, Qwen Coder — use your Qwen CLI login or API key", authed: !!qwenCred });
+        const copilotCred = getCredential("copilot");
+        if (copilotCred) {
+            groups["GitHub Copilot"] = ["gpt-4o", "claude-3.5-sonnet"].map(m => ({
+                name: m,
+                baseUrl: copilotCred.baseUrl || "https://api.githubcopilot.com",
+                apiKey: copilotCred.apiKey,
+                providerType: "openai",
+            }));
+        }
+        providerEntries.push({ name: "GitHub Copilot", description: "Use your GitHub Copilot subscription", authed: !!copilotCred });
+        if (providerEntries.length > 0) {
+            setModelPickerGroups(groups);
+            setProviderPicker(providerEntries);
+            setProviderPickerIndex(0);
+            setSelectedProvider(null);
+            return;
+        }
+        addMsg("error", "No models available. Download one with /ollama pull or configure a provider.");
+    }, [addMsg]);
     const handleSubmit = useCallback(async (value) => {
         value = sanitizeInputArtifacts(value);
         // Skip autocomplete if input exactly matches a command (e.g. /models vs /model)
@@ -478,133 +598,7 @@ function App() {
             return;
         }
         if (trimmed === "/models" || trimmed === "/model") {
-            addMsg("info", "Fetching available models...");
-            const groups = {};
-            const providerEntries = [];
-            // Local LLM (Ollama/LM Studio) — always show, auto-detect
-            let localFound = false;
-            // Check common local LLM endpoints
-            const localEndpoints = [
-                { name: "LM Studio", port: 1234 },
-                { name: "Ollama", port: 11434 },
-                { name: "vLLM", port: 8000 },
-                { name: "LocalAI", port: 8080 },
-            ];
-            for (const endpoint of localEndpoints) {
-                if (localFound)
-                    break;
-                try {
-                    const url = `http://localhost:${endpoint.port}/v1`;
-                    const models = await listModels(url, "local");
-                    if (models.length > 0) {
-                        groups["Local LLM"] = models.map(m => ({
-                            name: m,
-                            baseUrl: url,
-                            apiKey: "local",
-                            providerType: "openai",
-                        }));
-                        localFound = true;
-                    }
-                }
-                catch { /* not running */ }
-            }
-            // Also check Ollama native API
-            if (!localFound) {
-                try {
-                    const ollamaModels = await listInstalledModelsDetailed();
-                    if (ollamaModels.length > 0) {
-                        groups["Local LLM"] = ollamaModels.map(m => ({
-                            name: m.name,
-                            baseUrl: "http://localhost:11434/v1",
-                            apiKey: "ollama",
-                            providerType: "openai",
-                        }));
-                        localFound = true;
-                    }
-                }
-                catch { /* Ollama not running */ }
-            }
-            if (localFound) {
-                providerEntries.push({ name: "Local LLM", description: "No auth needed — auto-detected", authed: true });
-            }
-            // Anthropic
-            const anthropicCred = getCredential("anthropic");
-            const claudeModels = ["claude-sonnet-4-6", "claude-opus-4-6", "claude-haiku-4-5-20251001"];
-            if (anthropicCred) {
-                groups["Anthropic (Claude)"] = claudeModels.map(m => ({
-                    name: m,
-                    baseUrl: "https://api.anthropic.com",
-                    apiKey: anthropicCred.apiKey,
-                    providerType: "anthropic",
-                }));
-            }
-            providerEntries.push({ name: "Anthropic (Claude)", description: "Claude Opus, Sonnet, Haiku — use your subscription or API key", authed: !!anthropicCred });
-            // OpenAI
-            const openaiCred = getCredential("openai");
-            const openaiModels = ["gpt-5.4", "gpt-5.4-pro", "gpt-5", "gpt-5-mini", "gpt-4.1", "gpt-4.1-mini", "o3", "o4-mini", "gpt-4o"];
-            if (openaiCred) {
-                // OAuth tokens (non sk- keys) must use ChatGPT backend, not api.openai.com
-                const isOAuthToken = openaiCred.method === "oauth" || openaiCred.method === "cached-token" ||
-                    (!openaiCred.apiKey.startsWith("sk-") && !openaiCred.apiKey.startsWith("sess-"));
-                const baseUrl = isOAuthToken
-                    ? "https://chatgpt.com/backend-api"
-                    : (openaiCred.baseUrl || "https://api.openai.com/v1");
-                groups["OpenAI (ChatGPT)"] = openaiModels.map(m => ({
-                    name: m,
-                    baseUrl,
-                    apiKey: openaiCred.apiKey,
-                    providerType: "openai",
-                }));
-            }
-            providerEntries.push({ name: "OpenAI (ChatGPT)", description: "GPT-5, GPT-4.1, o3 — use your ChatGPT subscription or API key", authed: !!openaiCred });
-            // OpenRouter
-            const openrouterCred = getCredential("openrouter");
-            if (openrouterCred) {
-                try {
-                    const orModels = await listModels(openrouterCred.baseUrl || "https://openrouter.ai/api/v1", openrouterCred.apiKey);
-                    if (orModels.length > 0) {
-                        groups["OpenRouter"] = orModels.slice(0, 20).map(m => ({
-                            name: m,
-                            baseUrl: openrouterCred.baseUrl || "https://openrouter.ai/api/v1",
-                            apiKey: openrouterCred.apiKey,
-                            providerType: "openai",
-                        }));
-                    }
-                }
-                catch { /* skip */ }
-            }
-            providerEntries.push({ name: "OpenRouter", description: "200+ models (Claude, GPT, Gemini, Llama, etc.) — one login", authed: !!openrouterCred });
-            // Qwen
-            const qwenCred = getCredential("qwen");
-            if (qwenCred) {
-                groups["Qwen"] = ["qwen-max", "qwen-plus", "qwen-turbo"].map(m => ({
-                    name: m,
-                    baseUrl: qwenCred.baseUrl || "https://dashscope.aliyuncs.com/compatible-mode/v1",
-                    apiKey: qwenCred.apiKey,
-                    providerType: "openai",
-                }));
-            }
-            providerEntries.push({ name: "Qwen", description: "Qwen 3.5, Qwen Coder — use your Qwen CLI login or API key", authed: !!qwenCred });
-            // GitHub Copilot
-            const copilotCred = getCredential("copilot");
-            if (copilotCred) {
-                groups["GitHub Copilot"] = ["gpt-4o", "claude-3.5-sonnet"].map(m => ({
-                    name: m,
-                    baseUrl: copilotCred.baseUrl || "https://api.githubcopilot.com",
-                    apiKey: copilotCred.apiKey,
-                    providerType: "openai",
-                }));
-            }
-            providerEntries.push({ name: "GitHub Copilot", description: "Use your GitHub Copilot subscription", authed: !!copilotCred });
-            // Show provider picker (step 1)
-            if (providerEntries.length > 0) {
-                setModelPickerGroups(groups);
-                setProviderPicker(providerEntries);
-                setProviderPickerIndex(0);
-                setSelectedProvider(null);
-                return;
-            }
-            addMsg("error", "No models available. Download one with /ollama pull or configure a provider.");
+            await openModelPicker();
             return;
         }
         if (trimmed.startsWith("/model ")) {
@@ -825,11 +819,14 @@ function App() {
             setLoading,
             setSpinnerMsg,
             agent,
+            streaming,
+            loading,
             setModelName,
             addMsg,
             exit,
             refreshConnectionBanner,
             connectToProvider,
+            openModelPicker,
             handleSubmit,
             _require,
         });
@@ -854,7 +851,7 @@ function App() {
                     default:
                         return _jsx(Text, { children: msg.text }, msg.id);
                 }
-            }), loading && !approval && !streaming && _jsx(NeonSpinner, { message: spinnerMsg, colors: theme.colors }), streaming && !loading && _jsx(StreamingIndicator, { colors: theme.colors }), approval && (_jsx(ApprovalPrompt, { approval: approval, colors: theme.colors })), loginPicker && (_jsx(LoginPicker, { loginPickerIndex: loginPickerIndex, colors: theme.colors })), loginMethodPicker && (_jsx(LoginMethodPickerUI, { loginMethodPicker: loginMethodPicker, loginMethodIndex: loginMethodIndex, colors: theme.colors })), skillsPicker === "menu" && (_jsx(SkillsMenu, { skillsPickerIndex: skillsPickerIndex, colors: theme.colors })), skillsPicker === "browse" && (_jsx(SkillsBrowse, { skillsPickerIndex: skillsPickerIndex, colors: theme.colors })), skillsPicker === "installed" && (_jsx(SkillsInstalled, { skillsPickerIndex: skillsPickerIndex, sessionDisabledSkills: sessionDisabledSkills, colors: theme.colors })), skillsPicker === "remove" && (_jsx(SkillsRemove, { skillsPickerIndex: skillsPickerIndex, colors: theme.colors })), themePicker && (_jsx(ThemePickerUI, { themePickerIndex: themePickerIndex, theme: theme })), sessionPicker && (_jsx(SessionPicker, { sessions: sessionPicker, selectedIndex: sessionPickerIndex, colors: theme.colors })), deleteSessionPicker && (_jsx(DeleteSessionPicker, { sessions: deleteSessionPicker, selectedIndex: deleteSessionPickerIndex, colors: theme.colors })), deleteSessionConfirm && (_jsx(DeleteSessionConfirm, { session: deleteSessionConfirm, colors: theme.colors })), providerPicker && !selectedProvider && (_jsx(ProviderPicker, { providers: providerPicker, selectedIndex: providerPickerIndex, colors: theme.colors })), selectedProvider && modelPickerGroups && modelPickerGroups[selectedProvider] && (_jsx(ModelPicker, { providerName: selectedProvider, models: modelPickerGroups[selectedProvider], selectedIndex: modelPickerIndex, activeModel: modelName, colors: theme.colors })), ollamaDeletePicker && (_jsx(OllamaDeletePicker, { models: ollamaDeletePicker.models, selectedIndex: ollamaDeletePickerIndex, colors: theme.colors })), ollamaPullPicker && (_jsx(OllamaPullPicker, { selectedIndex: ollamaPullPickerIndex, colors: theme.colors })), ollamaDeleteConfirm && (_jsx(OllamaDeleteConfirm, { model: ollamaDeleteConfirm.model, size: ollamaDeleteConfirm.size, colors: theme.colors })), ollamaPulling && (_jsx(OllamaPullProgress, { model: ollamaPulling.model, progress: ollamaPulling.progress, colors: theme.colors })), ollamaExitPrompt && (_jsx(OllamaExitPrompt, { colors: theme.colors })), wizardScreen === "connection" && (_jsx(WizardConnection, { wizardIndex: wizardIndex, colors: theme.colors })), wizardScreen === "models" && wizardHardware && (_jsx(WizardModels, { wizardIndex: wizardIndex, wizardHardware: wizardHardware, wizardModels: wizardModels, colors: theme.colors })), wizardScreen === "install-ollama" && (_jsx(WizardInstallOllama, { wizardHardware: wizardHardware, colors: theme.colors })), wizardScreen === "pulling" && (wizardSelectedModel || wizardPullProgress) && (_jsx(WizardPulling, { wizardSelectedModel: wizardSelectedModel, wizardPullProgress: wizardPullProgress, wizardPullError: wizardPullError, colors: theme.colors })), showSuggestions && (_jsx(CommandSuggestions, { cmdMatches: cmdMatches, cmdIndex: cmdIndex, colors: theme.colors })), _jsxs(Box, { borderStyle: "single", borderColor: approval ? theme.colors.warning : theme.colors.border, paddingX: 1, children: [_jsx(Text, { color: theme.colors.secondary, bold: true, children: "> " }), approval ? (_jsx(Text, { color: theme.colors.warning, children: "waiting for approval..." })) : ready && !loading && !wizardScreen ? (_jsxs(Box, { children: [pastedChunks.map((p) => (_jsxs(Text, { color: theme.colors.muted, children: ["[Pasted text #", p.id, " +", p.lines, " lines]"] }, p.id))), _jsx(TextInput, { value: input, onChange: (v) => { setInput(sanitizeInputArtifacts(v)); setCmdIndex(0); }, onSubmit: handleSubmit }, inputKey)] })) : (_jsx(Text, { dimColor: true, children: loading ? "waiting for response..." : "initializing..." }))] }), agent && (_jsx(StatusBar, { agent: agent, modelName: modelName, sessionDisabledSkills: sessionDisabledSkills }))] }));
+            }), loading && !approval && !streaming && _jsx(NeonSpinner, { message: spinnerMsg, colors: theme.colors }), streaming && !loading && _jsx(StreamingIndicator, { colors: theme.colors }), approval && (_jsx(ApprovalPrompt, { approval: approval, colors: theme.colors })), loginPicker && (_jsx(LoginPicker, { loginPickerIndex: loginPickerIndex, colors: theme.colors })), loginMethodPicker && (_jsx(LoginMethodPickerUI, { loginMethodPicker: loginMethodPicker, loginMethodIndex: loginMethodIndex, colors: theme.colors })), skillsPicker === "menu" && (_jsx(SkillsMenu, { skillsPickerIndex: skillsPickerIndex, colors: theme.colors })), skillsPicker === "browse" && (_jsx(SkillsBrowse, { skillsPickerIndex: skillsPickerIndex, colors: theme.colors })), skillsPicker === "installed" && (_jsx(SkillsInstalled, { skillsPickerIndex: skillsPickerIndex, sessionDisabledSkills: sessionDisabledSkills, colors: theme.colors })), skillsPicker === "remove" && (_jsx(SkillsRemove, { skillsPickerIndex: skillsPickerIndex, colors: theme.colors })), themePicker && (_jsx(ThemePickerUI, { themePickerIndex: themePickerIndex, theme: theme })), sessionPicker && (_jsx(SessionPicker, { sessions: sessionPicker, selectedIndex: sessionPickerIndex, colors: theme.colors })), deleteSessionPicker && (_jsx(DeleteSessionPicker, { sessions: deleteSessionPicker, selectedIndex: deleteSessionPickerIndex, colors: theme.colors })), deleteSessionConfirm && (_jsx(DeleteSessionConfirm, { session: deleteSessionConfirm, colors: theme.colors })), providerPicker && !selectedProvider && (_jsx(ProviderPicker, { providers: providerPicker, selectedIndex: providerPickerIndex, colors: theme.colors })), selectedProvider && modelPickerGroups && modelPickerGroups[selectedProvider] && (_jsx(ModelPicker, { providerName: selectedProvider, models: modelPickerGroups[selectedProvider], selectedIndex: modelPickerIndex, activeModel: modelName, colors: theme.colors })), ollamaDeletePicker && (_jsx(OllamaDeletePicker, { models: ollamaDeletePicker.models, selectedIndex: ollamaDeletePickerIndex, colors: theme.colors })), ollamaPullPicker && (_jsx(OllamaPullPicker, { selectedIndex: ollamaPullPickerIndex, colors: theme.colors })), ollamaDeleteConfirm && (_jsx(OllamaDeleteConfirm, { model: ollamaDeleteConfirm.model, size: ollamaDeleteConfirm.size, colors: theme.colors })), ollamaPulling && (_jsx(OllamaPullProgress, { model: ollamaPulling.model, progress: ollamaPulling.progress, colors: theme.colors })), ollamaExitPrompt && (_jsx(OllamaExitPrompt, { colors: theme.colors })), wizardScreen === "connection" && (_jsx(WizardConnection, { wizardIndex: wizardIndex, colors: theme.colors })), wizardScreen === "models" && wizardHardware && (_jsx(WizardModels, { wizardIndex: wizardIndex, wizardHardware: wizardHardware, wizardModels: wizardModels, colors: theme.colors })), wizardScreen === "install-ollama" && (_jsx(WizardInstallOllama, { wizardHardware: wizardHardware, colors: theme.colors })), wizardScreen === "pulling" && (wizardSelectedModel || wizardPullProgress) && (_jsx(WizardPulling, { wizardSelectedModel: wizardSelectedModel, wizardPullProgress: wizardPullProgress, wizardPullError: wizardPullError, colors: theme.colors })), showSuggestions && (_jsx(CommandSuggestions, { cmdMatches: cmdMatches, cmdIndex: cmdIndex, colors: theme.colors })), _jsxs(Box, { borderStyle: process.platform === "win32" && !process.env.WT_SESSION ? "classic" : "single", borderColor: approval ? theme.colors.warning : theme.colors.border, paddingX: 1, children: [_jsx(Text, { color: theme.colors.secondary, bold: true, children: "> " }), approval ? (_jsx(Text, { color: theme.colors.warning, children: "waiting for approval..." })) : ready && !loading && !wizardScreen ? (_jsxs(Box, { children: [pastedChunks.map((p) => (_jsxs(Text, { color: theme.colors.muted, children: ["[Pasted text #", p.id, " +", p.lines, " lines]"] }, p.id))), _jsx(TextInput, { value: input, onChange: (v) => { setInput(sanitizeInputArtifacts(v)); setCmdIndex(0); }, onSubmit: handleSubmit }, inputKey)] })) : (_jsx(Text, { dimColor: true, children: loading ? "waiting for response..." : "initializing..." }))] }), agent && (_jsx(StatusBar, { agent: agent, modelName: modelName, sessionDisabledSkills: sessionDisabledSkills }))] }));
 }
 // Clear screen before render
 process.stdout.write("\x1B[2J\x1B[3J\x1B[H");

package/dist/ui/connection-types.d.ts

@@ -30,4 +30,5 @@ export interface ConnectionContext {
     } | null) => void;
     setWizardScreen: (val: WizardScreen) => void;
     setWizardIndex: (val: number) => void;
+    openModelPicker: () => Promise<void>;
 }

package/dist/ui/connection.js

@@ -75,11 +75,11 @@ export async function connectToProvider(isRetry, ctx) {
                 !!getCredential("openrouter") || !!getCredential("qwen") ||
                 !!getCredential("copilot");
             if (hasAnyCreds) {
-                // User has auth'd before — skip wizard, go straight to /models picker
-                info.push("✔ Found saved credentials. Use /models to pick a model and start coding.");
+                // User has auth'd before — skip wizard and go straight to the model picker
+                info.push("✔ Found saved credentials. Opening model picker...");
                 ctx.setConnectionInfo([...info]);
                 ctx.setReady(true);
-                // The user will run /models, which now works without an agent
+                await ctx.openModelPicker();
                 return;
             }
             // No creds found — show the setup wizard
@@ -192,4 +192,15 @@ export async function connectToProvider(isRetry, ctx) {
     if (isRetry) {
         ctx.addMsg("info", `✅ Connected to ${provider.model}`);
     }
+    else {
+        // First-time connection — show capabilities hint
+        const tools = a.getTools();
+        const toolCount = tools.length;
+        const toolNames = tools
+            .map((t) => t.function.name.replace(/_/g, " "))
+            .slice(0, 3)
+            .join(", ");
+        ctx.addMsg("info", `💡 You can: ${toolNames}${toolCount > 3 ? `, +${toolCount - 3} more` : ""}\n` +
+            `   Try: "list files in src/" or "read main.ts"`);
+    }
 }

package/dist/ui/input-router.d.ts

@@ -169,6 +169,8 @@ export interface InputRouterContext extends WizardContext {
     setCtrlCPressed: (val: boolean) => void;
     agent: CodingAgent | null;
     setModelName: (val: string) => void;
+    streaming: boolean;
+    loading: boolean;
     exit: () => void;
     refreshConnectionBanner: () => Promise<void>;
     handleSubmit: (value: string) => void;

package/dist/ui/input-router.js

@@ -43,6 +43,13 @@ export function routeKeyPress(inputChar, key, ctx) {
         return true;
     if (handleApprovalPrompts(inputChar, key, ctx))
         return true;
+    // Escape to abort generation (when loading or streaming)
+    if (key.escape && (ctx.streaming || ctx.loading) && ctx.agent) {
+        ctx.agent.abort();
+        ctx.setLoading(false);
+        ctx.addMsg("info", "⏹ Generation cancelled.");
+        return true;
+    }
     if (handleCtrlCExit(inputChar, key, ctx))
         return true;
     return false;
@@ -97,9 +104,10 @@ function handleLoginMethodPicker(inputChar, key, ctx) {
             ctx.setLoading(true);
             ctx.setSpinnerMsg("Waiting for authorization...");
             openRouterOAuth((msg) => ctx.addMsg("info", msg))
-                .then(() => {
-                ctx.addMsg("info", `✅ OpenRouter authenticated! Access to 200+ models.`);
+                .then(async () => {
+                ctx.addMsg("info", `✅ OpenRouter authenticated! Access to 200+ models.\n  Opening model picker...`);
                 ctx.setLoading(false);
+                await ctx.openModelPicker();
             })
                 .catch((err) => { ctx.addMsg("error", `OAuth failed: ${err.message}`); ctx.setLoading(false); });
         }
@@ -108,9 +116,10 @@ function handleLoginMethodPicker(inputChar, key, ctx) {
             ctx.setLoading(true);
             ctx.setSpinnerMsg("Waiting for Anthropic authorization...");
             loginAnthropicOAuth((msg) => ctx.addMsg("info", msg))
-                .then((cred) => {
-                ctx.addMsg("info", `✅ Anthropic authenticated! (${cred.label})\n  Next: type /models to pick a model`);
+                .then(async (cred) => {
+                ctx.addMsg("info", `✅ Anthropic authenticated! (${cred.label})\n  Opening model picker...`);
                 ctx.setLoading(false);
+                await ctx.openModelPicker();
             })
                 .catch((err) => {
                 ctx.addMsg("error", `OAuth failed: ${err.message}\n  Fallback: set your key via CLI:  codemaxxing auth api-key anthropic <your-key>\n  Or set ANTHROPIC_API_KEY env var and restart.\n  Get key at: console.anthropic.com/settings/keys`);
@@ -121,7 +130,8 @@ function handleLoginMethodPicker(inputChar, key, ctx) {
             // Try cached Codex token first as a quick path
             const imported = importCodexToken((msg) => ctx.addMsg("info", msg));
             if (imported) {
-                ctx.addMsg("info", `✅ Imported Codex credentials! (${imported.label})`);
+                ctx.addMsg("info", `✅ Imported Codex credentials! (${imported.label})\n  Opening model picker...`);
+                void ctx.openModelPicker();
             }
             else {
                 // Primary flow: browser OAuth
@@ -129,9 +139,10 @@ function handleLoginMethodPicker(inputChar, key, ctx) {
                 ctx.setLoading(true);
                 ctx.setSpinnerMsg("Waiting for OpenAI authorization...");
                 loginOpenAICodexOAuth((msg) => ctx.addMsg("info", msg))
-                    .then((cred) => {
-                    ctx.addMsg("info", `✅ OpenAI authenticated! (${cred.label})\n  Next: type /models to pick a model`);
+                    .then(async (cred) => {
+                    ctx.addMsg("info", `✅ OpenAI authenticated! (${cred.label})\n  Opening model picker...`);
                     ctx.setLoading(false);
+                    await ctx.openModelPicker();
                 })
                     .catch((err) => {
                     ctx.addMsg("error", `OAuth failed: ${err.message}\n  Fallback: set your key via CLI:  codemaxxing auth api-key openai <your-key>\n  Or set OPENAI_API_KEY env var and restart.\n  Get key at: platform.openai.com/api-keys`);
@@ -142,7 +153,8 @@ function handleLoginMethodPicker(inputChar, key, ctx) {
         else if (method === "cached-token" && providerId === "qwen") {
             const imported = importQwenToken((msg) => ctx.addMsg("info", msg));
             if (imported) {
-                ctx.addMsg("info", `✅ Imported Qwen credentials! (${imported.label})`);
+                ctx.addMsg("info", `✅ Imported Qwen credentials! (${imported.label})\n  Opening model picker...`);
+                void ctx.openModelPicker();
             }
             else {
                 ctx.addMsg("info", "No Qwen CLI found. Install Qwen CLI and sign in first.");
@@ -153,7 +165,7 @@ function handleLoginMethodPicker(inputChar, key, ctx) {
             ctx.setLoading(true);
             ctx.setSpinnerMsg("Waiting for GitHub authorization...");
             copilotDeviceFlow((msg) => ctx.addMsg("info", msg))
-                .then(() => { ctx.addMsg("info", `✅ GitHub Copilot authenticated!\n  Next: type /models to pick a model`); ctx.setLoading(false); })
+                .then(async () => { ctx.addMsg("info", `✅ GitHub Copilot authenticated!\n  Opening model picker...`); ctx.setLoading(false); await ctx.openModelPicker(); })
                 .catch((err) => { ctx.addMsg("error", `Copilot auth failed: ${err.message}`); ctx.setLoading(false); });
         }
         else if (method === "api-key") {
@@ -189,7 +201,7 @@ function handleLoginPicker(_inputChar, key, ctx) {
                 ctx.setLoading(true);
                 ctx.setSpinnerMsg("Waiting for authorization...");
                 openRouterOAuth((msg) => ctx.addMsg("info", msg))
-                    .then(() => { ctx.addMsg("info", `✅ OpenRouter authenticated! Access to 200+ models.\n  Next: type /models to pick a model`); ctx.setLoading(false); })
+                    .then(async () => { ctx.addMsg("info", `✅ OpenRouter authenticated! Access to 200+ models.\n  Opening model picker...`); ctx.setLoading(false); await ctx.openModelPicker(); })
                     .catch((err) => { ctx.addMsg("error", `OAuth failed: ${err.message}`); ctx.setLoading(false); });
             }
             else if (methods[0] === "device-flow") {
@@ -198,7 +210,7 @@ function handleLoginPicker(_inputChar, key, ctx) {
                 ctx.setLoading(true);
                 ctx.setSpinnerMsg("Waiting for GitHub authorization...");
                 copilotDeviceFlow((msg) => ctx.addMsg("info", msg))
-                    .then(() => { ctx.addMsg("info", `✅ GitHub Copilot authenticated!\n  Next: type /models to pick a model`); ctx.setLoading(false); })
+                    .then(async () => { ctx.addMsg("info", `✅ GitHub Copilot authenticated!\n  Opening model picker...`); ctx.setLoading(false); await ctx.openModelPicker(); })
                     .catch((err) => { ctx.addMsg("error", `Copilot auth failed: ${err.message}`); ctx.setLoading(false); });
             }
             else if (methods[0] === "api-key") {
@@ -380,6 +392,9 @@ function handleProviderPicker(_inputChar, key, ctx) {
     }
     if (key.escape) {
         ctx.setProviderPicker(null);
+        if (!ctx.agent) {
+            ctx.addMsg("info", "Model selection cancelled. Use /login for cloud providers or choose local setup from the startup menu.");
+        }
         return true;
     }
     if (key.return) {
@@ -475,13 +490,13 @@ function handleOllamaPullPicker(_inputChar, key, ctx) {
     if (!ctx.ollamaPullPicker)
         return false;
     const pullModels = [
-        { id: "qwen2.5-coder:7b", name: "Qwen 2.5 Coder 7B", size: "5 GB", desc: "Best balance of speed & quality" },
-        { id: "qwen2.5-coder:14b", name: "Qwen 2.5 Coder 14B", size: "9 GB", desc: "Higher quality, needs 16GB+ RAM" },
-        { id: "qwen2.5-coder:3b", name: "Qwen 2.5 Coder 3B", size: "2 GB", desc: "\u26A0\uFE0F Basic \u2014 may struggle with tool calls" },
-        { id: "qwen2.5-coder:32b", name: "Qwen 2.5 Coder 32B", size: "20 GB", desc: "Premium quality, needs 48GB+" },
-        { id: "deepseek-coder-v2:16b", name: "DeepSeek Coder V2", size: "9 GB", desc: "Strong alternative" },
-        { id: "codellama:7b", name: "CodeLlama 7B", size: "4 GB", desc: "Meta's coding model" },
-        { id: "starcoder2:7b", name: "StarCoder2 7B", size: "4 GB", desc: "Code completion focused" },
+        { id: "qwen2.5-coder:14b", name: "Qwen 2.5 Coder 14B", size: "9 GB", desc: "Recommended default for coding if your machine can handle it" },
+        { id: "deepseek-coder-v2:16b", name: "DeepSeek Coder V2 16B", size: "9 GB", desc: "Strong higher-quality alternative" },
+        { id: "qwen2.5-coder:7b", name: "Qwen 2.5 Coder 7B", size: "5 GB", desc: "Fallback for mid-range machines" },
+        { id: "qwen2.5-coder:32b", name: "Qwen 2.5 Coder 32B", size: "20 GB", desc: "Premium quality, needs lots of RAM" },
+        { id: "codellama:7b", name: "CodeLlama 7B", size: "4 GB", desc: "Older fallback coding model" },
+        { id: "starcoder2:7b", name: "StarCoder2 7B", size: "4 GB", desc: "Completion-focused fallback" },
+        { id: "qwen2.5-coder:3b", name: "Qwen 2.5 Coder 3B", size: "2 GB", desc: "⚠️ Last resort — may struggle with tool calls" },
     ];
     if (key.upArrow) {
         ctx.setOllamaPullPickerIndex((prev) => (prev - 1 + pullModels.length) % pullModels.length);

package/dist/ui/paste-interceptor.js

@@ -17,6 +17,15 @@ import { consumePendingPasteEndMarkerChunk, shouldSwallowPostPasteDebris } from
  */
 export function setupPasteInterceptor() {
     const pasteEvents = new EventEmitter();
+    // Detect Windows CMD/conhost — these don't support bracketed paste or ANSI sequences well.
+    // On these terminals, skip paste interception entirely to avoid eating keystrokes.
+    const isWindowsLegacyTerminal = process.platform === "win32" && (!process.env.WT_SESSION && // Not Windows Terminal
+        !process.env.TERM_PROGRAM // Not a modern terminal emulator
+    );
+    if (isWindowsLegacyTerminal) {
+        // Just return a dummy event bus — no interception, no burst buffering
+        return pasteEvents;
+    }
     // Enable bracketed paste mode — terminal wraps pastes in escape sequences
     process.stdout.write("\x1b[?2004h");
     // ── Internal state ──

package/dist/ui/pickers.js

@@ -74,13 +74,13 @@ export function OllamaDeletePicker({ models, selectedIndex, colors }) {
 }
 // ── Ollama Pull Picker ──
 const PULL_MODELS = [
-    { id: "qwen2.5-coder:7b", name: "Qwen 2.5 Coder 7B", size: "5 GB", desc: "Best balance of speed & quality" },
-    { id: "qwen2.5-coder:14b", name: "Qwen 2.5 Coder 14B", size: "9 GB", desc: "Higher quality, needs 16GB+ RAM" },
-    { id: "qwen2.5-coder:3b", name: "Qwen 2.5 Coder 3B", size: "2 GB", desc: "\u26A0\uFE0F Basic \u2014 may struggle with tool calls" },
-    { id: "qwen2.5-coder:32b", name: "Qwen 2.5 Coder 32B", size: "20 GB", desc: "Premium, needs 48GB+" },
-    { id: "deepseek-coder-v2:16b", name: "DeepSeek Coder V2", size: "9 GB", desc: "Strong alternative" },
-    { id: "codellama:7b", name: "CodeLlama 7B", size: "4 GB", desc: "Meta's coding model" },
-    { id: "starcoder2:7b", name: "StarCoder2 7B", size: "4 GB", desc: "Code completion focused" },
+    { id: "qwen2.5-coder:14b", name: "Qwen 2.5 Coder 14B", size: "9 GB", desc: "Recommended default for coding if your machine can handle it" },
+    { id: "deepseek-coder-v2:16b", name: "DeepSeek Coder V2 16B", size: "9 GB", desc: "Strong higher-quality alternative" },
+    { id: "qwen2.5-coder:7b", name: "Qwen 2.5 Coder 7B", size: "5 GB", desc: "Fallback for mid-range machines" },
+    { id: "qwen2.5-coder:32b", name: "Qwen 2.5 Coder 32B", size: "20 GB", desc: "Premium quality, needs lots of RAM" },
+    { id: "codellama:7b", name: "CodeLlama 7B", size: "4 GB", desc: "Older fallback coding model" },
+    { id: "starcoder2:7b", name: "StarCoder2 7B", size: "4 GB", desc: "Completion-focused fallback" },
+    { id: "qwen2.5-coder:3b", name: "Qwen 2.5 Coder 3B", size: "2 GB", desc: "⚠️ Last resort — may struggle with tool calls" },
 ];
 export function OllamaPullPicker({ selectedIndex, colors }) {
     return (_jsxs(Box, { flexDirection: "column", borderStyle: "single", borderColor: colors.border, paddingX: 1, marginBottom: 0, children: [_jsx(Text, { bold: true, color: colors.secondary, children: "Download which model?" }), _jsx(Text, { children: "" }), PULL_MODELS.map((m, i) => (_jsxs(Text, { children: ["  ", i === selectedIndex ? _jsx(Text, { color: colors.primary, bold: true, children: "▸ " }) : "  ", _jsx(Text, { color: i === selectedIndex ? colors.primary : undefined, bold: true, children: m.name }), _jsxs(Text, { color: colors.muted, children: [" · ", m.size, " · ", m.desc] })] }, m.id))), _jsx(Text, { children: "" }), _jsx(Text, { dimColor: true, children: "  ↑↓ navigate · Enter to download · Esc cancel" })] }));
@@ -113,7 +113,7 @@ export function WizardModels({ wizardIndex, wizardHardware, wizardModels, colors
     return (_jsxs(Box, { flexDirection: "column", borderStyle: "single", borderColor: colors.border, paddingX: 1, marginBottom: 0, children: [_jsx(Text, { bold: true, color: colors.secondary, children: "Your hardware:" }), _jsxs(Text, { color: colors.muted, children: ["  CPU: ", wizardHardware.cpu.name, " (", wizardHardware.cpu.cores, " cores)"] }), _jsxs(Text, { color: colors.muted, children: ["  RAM: ", formatBytes(wizardHardware.ram)] }), wizardHardware.gpu ? (_jsxs(Text, { color: colors.muted, children: ["  GPU: ", wizardHardware.gpu.name, wizardHardware.gpu.vram > 0 ? ` (${formatBytes(wizardHardware.gpu.vram)})` : ""] })) : (_jsx(Text, { color: colors.muted, children: "  GPU: none detected" })), !isLlmfitAvailable() && (_jsx(Text, { dimColor: true, children: "  Tip: Install llmfit for smarter recommendations: brew install llmfit" })), _jsx(Text, { children: "" }), _jsx(Text, { bold: true, color: colors.secondary, children: "Recommended models:" }), _jsx(Text, { children: "" }), wizardModels.map((m, i) => (_jsxs(Text, { children: [i === wizardIndex ? _jsx(Text, { color: colors.suggestion, bold: true, children: "  \u25B8 " }) : _jsx(Text, { children: "    " }), _jsxs(Text, { children: [getFitIcon(m.fit), " "] }), _jsx(Text, { color: i === wizardIndex ? colors.suggestion : colors.primary, bold: true, children: m.name }), _jsxs(Text, { color: colors.muted, children: ["     ~", m.size, " GB \u00B7 ", m.quality === "best" ? "Best" : m.quality === "great" ? "Great" : "Good", " quality \u00B7 ", m.speed] })] }, m.ollamaId))), wizardModels.length === 0 && (_jsx(Text, { color: colors.error, children: "  No suitable models found for your hardware." })), _jsx(Text, { children: "" }), _jsx(Text, { dimColor: true, children: "  \u2191\u2193 navigate \u00B7 Enter to install \u00B7 Esc back" })] }));
 }
 export function WizardInstallOllama({ wizardHardware, colors }) {
-    return (_jsxs(Box, { flexDirection: "column", borderStyle: "single", borderColor: colors.warning, paddingX: 1, marginBottom: 0, children: [_jsx(Text, { bold: true, color: colors.warning, children: "Ollama is required for local models." }), _jsx(Text, { children: "" }), _jsx(Text, { color: colors.primary, children: "  Press Enter to install Ollama automatically" }), _jsxs(Text, { dimColor: true, children: ["  Or install manually: ", _jsx(Text, { children: getOllamaInstallCommand(wizardHardware?.os ?? "linux") })] }), _jsx(Text, { children: "" }), _jsx(Text, { dimColor: true, children: "  Enter to install · Esc to go back" })] }));
+    return (_jsxs(Box, { flexDirection: "column", borderStyle: "single", borderColor: colors.warning, paddingX: 1, marginBottom: 0, children: [_jsx(Text, { bold: true, color: colors.warning, children: "Ollama is required for local models." }), _jsx(Text, { children: "" }), _jsx(Text, { color: colors.primary, children: "  Press Enter to install Ollama automatically" }), _jsxs(Text, { dimColor: true, children: ["  Or install manually: ", _jsx(Text, { children: getOllamaInstallCommand(wizardHardware?.os ?? "linux") })] }), _jsx(Text, { children: "" }), _jsx(Text, { dimColor: true, children: "  Enter to install · Esc back to model list" })] }));
 }
 export function WizardPulling({ wizardSelectedModel, wizardPullProgress, wizardPullError, colors }) {
     return (_jsx(Box, { flexDirection: "column", borderStyle: "single", borderColor: colors.border, paddingX: 1, marginBottom: 0, children: wizardPullError ? (_jsxs(_Fragment, { children: [_jsxs(Text, { color: colors.error, bold: true, children: ["  \u274C Error: ", wizardPullError] }), _jsx(Text, { children: "" }), _jsx(Text, { dimColor: true, children: "  Press Enter to retry \u00B7 Esc to go back" })] })) : wizardPullProgress ? (_jsxs(_Fragment, { children: [_jsxs(Text, { bold: true, color: colors.secondary, children: ["  ", wizardSelectedModel ? `Downloading ${wizardSelectedModel.name}...` : wizardPullProgress?.status || "Working..."] }), wizardPullProgress.status === "downloading" || wizardPullProgress.percent > 0 ? (_jsx(_Fragment, { children: _jsxs(Text, { children: ["  ", _jsxs(Text, { color: colors.primary, children: ["\u2588".repeat(Math.floor(wizardPullProgress.percent / 5)), "\u2591".repeat(20 - Math.floor(wizardPullProgress.percent / 5))] }), "  ", _jsxs(Text, { bold: true, children: [wizardPullProgress.percent, "%"] }), wizardPullProgress.completed != null && wizardPullProgress.total != null ? (_jsxs(Text, { color: colors.muted, children: [" \u00B7 ", formatBytes(wizardPullProgress.completed), " / ", formatBytes(wizardPullProgress.total)] })) : null] }) })) : (_jsxs(Text, { color: colors.muted, children: ["  ", wizardPullProgress.status, "..."] }))] })) : null }));

package/dist/ui/wizard-types.d.ts

@@ -23,5 +23,6 @@ export interface WizardContext {
     setSpinnerMsg: (val: string) => void;
     addMsg: (type: "user" | "response" | "tool" | "tool-result" | "error" | "info", text: string) => void;
     connectToProvider: (isRetry: boolean) => Promise<void>;
+    openModelPicker: () => Promise<void>;
     _require: NodeRequire;
 }

package/dist/ui/wizard.js

@@ -35,9 +35,10 @@ export function handleWizardScreen(_inputChar, key, ctx) {
                 ctx.setLoading(true);
                 ctx.setSpinnerMsg("Waiting for authorization...");
                 openRouterOAuth((msg) => ctx.addMsg("info", msg))
-                    .then(() => {
-                    ctx.addMsg("info", "✅ OpenRouter authenticated! Use /connect to connect.");
+                    .then(async () => {
+                    ctx.addMsg("info", "✅ OpenRouter authenticated! Opening model picker...");
                     ctx.setLoading(false);
+                    await ctx.openModelPicker();
                 })
                     .catch((err) => { ctx.addMsg("error", `OAuth failed: ${err.message}`); ctx.setLoading(false); });
             }
@@ -45,10 +46,12 @@ export function handleWizardScreen(_inputChar, key, ctx) {
                 ctx.setWizardScreen(null);
                 ctx.setLoginPicker(true);
                 ctx.setLoginPickerIndex(() => 0);
+                ctx.addMsg("info", "Choose a cloud provider to continue.");
             }
             else if (selected === "existing") {
                 ctx.setWizardScreen(null);
-                ctx.addMsg("info", "Start your LLM server, then type /connect to retry.");
+                ctx.addMsg("info", "Checking for your running server...");
+                void ctx.connectToProvider(true);
             }
             return true;
         }
@@ -196,6 +199,7 @@ function startPullFlow(ctx, selected) {
                     return;
                 }
             }
+            ctx.addMsg("info", `Downloading ${selected.name}. This can take a while on first run depending on your internet and disk speed.`);
             await pullModel(selected.ollamaId, (p) => {
                 ctx.setWizardPullProgress(p);
             });

package/dist/utils/auth.js

@@ -149,10 +149,12 @@ function generatePKCE() {
 export async function openRouterOAuth(onStatus) {
     const { verifier, challenge } = generatePKCE();
     return new Promise((resolve, reject) => {
+        let handled = false; // Guard against duplicate callbacks
         // Start local callback server
         const server = createServer(async (req, res) => {
             const url = new URL(req.url ?? "/", `http://localhost`);
-            if (url.pathname === "/callback") {
+            if (url.pathname === "/callback" && !handled) {
+                handled = true;
                 const code = url.searchParams.get("code");
                 if (!code) {
                     res.writeHead(400, { "Content-Type": "text/html" });
@@ -175,6 +177,9 @@ export async function openRouterOAuth(onStatus) {
                     });
                     if (!exchangeRes.ok) {
                         const errText = await exchangeRes.text();
+                        if (exchangeRes.status === 409) {
+                            throw new Error(`OpenRouter returned 409 (Conflict) — this usually means the auth code was already used. Please try /login again.`);
+                        }
                         throw new Error(`Exchange failed (${exchangeRes.status}): ${errText}`);
                     }
                     const data = (await exchangeRes.json());

package/dist/utils/models.js

@@ -1,14 +1,24 @@
 import { execSync } from "child_process";
 const MODELS = [
     {
-        name: "Qwen 2.5 Coder 3B",
-        ollamaId: "qwen2.5-coder:3b",
-        size: 2,
-        ramRequired: 8,
-        vramOptimal: 4,
-        description: "\u26A0\uFE0F May not support tool calling well",
-        speed: "~60 tok/s on M1",
-        quality: "limited",
+        name: "Qwen 2.5 Coder 14B",
+        ollamaId: "qwen2.5-coder:14b",
+        size: 9,
+        ramRequired: 24,
+        vramOptimal: 12,
+        description: "Recommended default for coding when your machine can handle it",
+        speed: "~25 tok/s on M1 Pro",
+        quality: "best",
+    },
+    {
+        name: "DeepSeek Coder V2 16B",
+        ollamaId: "deepseek-coder-v2:16b",
+        size: 9,
+        ramRequired: 24,
+        vramOptimal: 12,
+        description: "Strong higher-quality alternative for coding",
+        speed: "~30 tok/s on M1 Pro",
+        quality: "great",
     },
     {
         name: "Qwen 2.5 Coder 7B",
@@ -16,20 +26,10 @@ const MODELS = [
         size: 5,
         ramRequired: 16,
         vramOptimal: 8,
-        description: "Sweet spot for most machines",
+        description: "Fallback for mid-range machines when 14B is too heavy",
         speed: "~45 tok/s on M1",
         quality: "great",
     },
-    {
-        name: "Qwen 2.5 Coder 14B",
-        ollamaId: "qwen2.5-coder:14b",
-        size: 9,
-        ramRequired: 32,
-        vramOptimal: 16,
-        description: "High quality coding",
-        speed: "~25 tok/s on M1 Pro",
-        quality: "best",
-    },
     {
         name: "Qwen 2.5 Coder 32B",
         ollamaId: "qwen2.5-coder:32b",
@@ -40,23 +40,13 @@ const MODELS = [
         speed: "~12 tok/s on M1 Max",
         quality: "best",
     },
-    {
-        name: "DeepSeek Coder V2 16B",
-        ollamaId: "deepseek-coder-v2:16b",
-        size: 9,
-        ramRequired: 32,
-        vramOptimal: 16,
-        description: "Strong alternative for coding",
-        speed: "~30 tok/s on M1 Pro",
-        quality: "great",
-    },
     {
         name: "CodeLlama 7B",
         ollamaId: "codellama:7b",
         size: 4,
         ramRequired: 16,
         vramOptimal: 8,
-        description: "Meta's coding model",
+        description: "Older fallback coding model",
         speed: "~40 tok/s on M1",
         quality: "good",
     },
@@ -66,10 +56,20 @@ const MODELS = [
         size: 4,
         ramRequired: 16,
         vramOptimal: 8,
-        description: "Good for code completion",
+        description: "Completion-focused fallback",
         speed: "~40 tok/s on M1",
         quality: "good",
     },
+    {
+        name: "Qwen 2.5 Coder 3B",
+        ollamaId: "qwen2.5-coder:3b",
+        size: 2,
+        ramRequired: 8,
+        vramOptimal: 4,
+        description: "⚠️ Last-resort fallback — may struggle with tool calling",
+        speed: "~60 tok/s on M1",
+        quality: "limited",
+    },
 ];
 function scoreModel(model, ramGB, vramGB) {
     if (ramGB < model.ramRequired)
@@ -86,6 +86,23 @@ function scoreModel(model, ramGB, vramGB) {
 }
 const qualityOrder = { best: 3, great: 2, good: 1, limited: 0 };
 const fitOrder = { perfect: 4, good: 3, tight: 2, skip: 1 };
+function recommendationPriority(model) {
+    if (model.ollamaId === "qwen2.5-coder:14b")
+        return 100;
+    if (model.ollamaId === "deepseek-coder-v2:16b")
+        return 95;
+    if (model.ollamaId === "qwen2.5-coder:7b")
+        return 80;
+    if (model.ollamaId === "qwen2.5-coder:32b")
+        return 75;
+    if (model.ollamaId === "codellama:7b")
+        return 60;
+    if (model.ollamaId === "starcoder2:7b")
+        return 55;
+    if (model.ollamaId === "qwen2.5-coder:3b")
+        return 10;
+    return 0;
+}
 export function getRecommendations(hardware) {
     const ramGB = hardware.ram / (1024 * 1024 * 1024);
     const vramGB = hardware.gpu?.vram ? hardware.gpu.vram / (1024 * 1024 * 1024) : 0;
@@ -95,11 +112,14 @@ export function getRecommendations(hardware) {
         ...m,
         fit: scoreModel(m, ramGB, effectiveVRAM),
     }));
-    // Sort: perfect first, then by quality descending
+    // Sort: fit first, then strongly prefer better coding defaults over tiny fallback models
     scored.sort((a, b) => {
         const fitDiff = (fitOrder[b.fit] ?? 0) - (fitOrder[a.fit] ?? 0);
         if (fitDiff !== 0)
             return fitDiff;
+        const priorityDiff = recommendationPriority(b) - recommendationPriority(a);
+        if (priorityDiff !== 0)
+            return priorityDiff;
         return (qualityOrder[b.quality] ?? 0) - (qualityOrder[a.quality] ?? 0);
     });
     return scored;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "codemaxxing",
-  "version": "1.1.2",
+  "version": "1.1.4",
   "description": "Open-source terminal coding agent. Connect any LLM. Max your code.",
   "main": "dist/index.js",
   "bin": {