code-puppy 0.0.214__py3-none-any.whl → 0.0.366__py3-none-any.whl

code_puppy/command_line/add_model_menu.py

@@ -0,0 +1,1079 @@
+"""Interactive terminal UI for browsing and adding models from models_dev_api.json.
+
+Provides a beautiful split-panel interface for browsing providers and models
+with live preview of model details and one-click addition to extra_models.json.
+"""
+
+import json
+import os
+import sys
+import time
+from pathlib import Path
+from typing import List, Optional
+
+from prompt_toolkit.application import Application
+from prompt_toolkit.key_binding import KeyBindings
+from prompt_toolkit.layout import Dimension, Layout, VSplit, Window
+from prompt_toolkit.layout.controls import FormattedTextControl
+from prompt_toolkit.widgets import Frame
+
+from code_puppy.command_line.utils import safe_input
+from code_puppy.config import EXTRA_MODELS_FILE, set_config_value
+from code_puppy.messaging import emit_error, emit_info, emit_warning
+from code_puppy.models_dev_parser import ModelInfo, ModelsDevRegistry, ProviderInfo
+from code_puppy.tools.command_runner import set_awaiting_user_input
+
+PAGE_SIZE = 15  # Items per page
+
+# Hardcoded OpenAI-compatible endpoints for providers that have dedicated SDKs
+# but actually work fine with custom_openai. These are fallbacks when provider.api is not set.
+PROVIDER_ENDPOINTS = {
+    "xai": "https://api.x.ai/v1",
+    "cohere": "https://api.cohere.com/compatibility/v1",  # Cohere's OpenAI-compatible endpoint
+    "groq": "https://api.groq.com/openai/v1",
+    "mistral": "https://api.mistral.ai/v1",
+    "togetherai": "https://api.together.xyz/v1",
+    "perplexity": "https://api.perplexity.ai",
+    "deepinfra": "https://api.deepinfra.com/v1/openai",
+    "aihubmix": "https://aihubmix.com/v1",
+}
+
+# Providers that require custom SDK implementations we don't support yet.
+# These use non-OpenAI-compatible APIs or require special authentication (AWS SigV4, GCP, etc.)
+UNSUPPORTED_PROVIDERS = {
+    "amazon-bedrock": "Requires AWS SigV4 authentication",
+    "google-vertex": "Requires GCP service account authentication",
+    "google-vertex-anthropic": "Requires GCP service account authentication",
+    "cloudflare-workers-ai": "Requires account ID in URL path",
+    "vercel": "Vercel AI Gateway - not yet supported",
+    "v0": "Vercel v0 - not yet supported",
+    "ollama-cloud": "Requires user-specific Ollama instance URL",
+}
+
+
+class AddModelMenu:
+    """Interactive TUI for browsing and adding models."""
+
+    def __init__(self):
+        """Initialize the model browser menu."""
+        self.registry: Optional[ModelsDevRegistry] = None
+        self.providers: List[ProviderInfo] = []
+        self.current_provider: Optional[ProviderInfo] = None
+        self.current_models: List[ModelInfo] = []
+
+        # State management
+        self.view_mode = "providers"  # "providers" or "models"
+        self.selected_provider_idx = 0
+        self.selected_model_idx = 0
+        self.current_page = 0
+        self.result = None  # Track if user added a model
+
+        # Pending model for credential prompting
+        self.pending_model: Optional[ModelInfo] = None
+        self.pending_provider: Optional[ProviderInfo] = None
+
+        # Custom model support
+        self.is_custom_model_selected = False
+        self.custom_model_name: Optional[str] = None
+
+        # Initialize registry
+        self._initialize_registry()
+
+    def _initialize_registry(self):
+        """Initialize the ModelsDevRegistry with error handling.
+
+        Fetches from live models.dev API first, falls back to bundled JSON.
+        """
+        try:
+            self.registry = (
+                ModelsDevRegistry()
+            )  # Will try API first, then bundled fallback
+            self.providers = self.registry.get_providers()
+            if not self.providers:
+                emit_error("No providers found in models database")
+        except FileNotFoundError as e:
+            emit_error(f"Models database unavailable: {e}")
+        except Exception as e:
+            emit_error(f"Error loading models registry: {e}")
+
+    def _get_current_provider(self) -> Optional[ProviderInfo]:
+        """Get the currently selected provider."""
+        if 0 <= self.selected_provider_idx < len(self.providers):
+            return self.providers[self.selected_provider_idx]
+        return None
+
+    def _get_current_model(self) -> Optional[ModelInfo]:
+        """Get the currently selected model.
+
+        Returns None if "Custom model" option is selected (which is at index len(current_models)).
+        """
+        if self.view_mode == "models" and self.current_provider:
+            # Check if custom model option is selected (it's the last item)
+            if self.selected_model_idx == len(self.current_models):
+                return None  # Custom model selected
+            if 0 <= self.selected_model_idx < len(self.current_models):
+                return self.current_models[self.selected_model_idx]
+        return None
+
+    def _is_custom_model_selected(self) -> bool:
+        """Check if the custom model option is currently selected."""
+        if self.view_mode == "models" and self.current_provider:
+            return self.selected_model_idx == len(self.current_models)
+        return False
+
+    def _render_provider_list(self) -> List:
+        """Render the provider list panel."""
+        lines = []
+
+        lines.append(("", " Providers"))
+        lines.append(("", "\n\n"))
+
+        if not self.providers:
+            lines.append(("fg:yellow", "  No providers available."))
+            lines.append(("", "\n\n"))
+            self._render_navigation_hints(lines)
+            return lines
+
+        # Show providers for current page
+        total_pages = (len(self.providers) + PAGE_SIZE - 1) // PAGE_SIZE
+        start_idx = self.current_page * PAGE_SIZE
+        end_idx = min(start_idx + PAGE_SIZE, len(self.providers))
+
+        for i in range(start_idx, end_idx):
+            provider = self.providers[i]
+            is_selected = i == self.selected_provider_idx
+            is_unsupported = provider.id in UNSUPPORTED_PROVIDERS
+
+            # Format: "> Provider Name (X models)" or "  Provider Name (X models)"
+            prefix = " > " if is_selected else "   "
+            suffix = " ⚠️" if is_unsupported else ""
+            label = f"{prefix}{provider.name} ({provider.model_count} models){suffix}"
+
+            # Use dimmed color for unsupported providers
+            if is_unsupported:
+                lines.append(("fg:ansibrightblack dim", label))
+            elif is_selected:
+                lines.append(("fg:ansibrightblack", label))
+            else:
+                lines.append(("fg:ansibrightblack", label))
+
+            lines.append(("", "\n"))
+
+        lines.append(("", "\n"))
+        lines.append(
+            ("fg:ansibrightblack", f" Page {self.current_page + 1}/{total_pages}")
+        )
+        lines.append(("", "\n"))
+
+        self._render_navigation_hints(lines)
+        return lines
+
+    def _render_model_list(self) -> List:
+        """Render the model list panel."""
+        lines = []
+
+        if not self.current_provider:
+            lines.append(("fg:yellow", "  No provider selected."))
+            lines.append(("", "\n\n"))
+            self._render_navigation_hints(lines)
+            return lines
+
+        lines.append(("", f" {self.current_provider.name} Models"))
+        lines.append(("", "\n\n"))
+
+        # Total items = models + 1 for custom model option
+        total_items = len(self.current_models) + 1
+        total_pages = (total_items + PAGE_SIZE - 1) // PAGE_SIZE
+        start_idx = self.current_page * PAGE_SIZE
+        end_idx = min(start_idx + PAGE_SIZE, total_items)
+
+        # Render models from the current page
+        for i in range(start_idx, end_idx):
+            # Check if this is the custom model option (last item)
+            if i == len(self.current_models):
+                is_selected = i == self.selected_model_idx
+                if is_selected:
+                    lines.append(("fg:ansicyan bold", " > ✨ Custom model..."))
+                else:
+                    lines.append(("fg:ansicyan", "   ✨ Custom model..."))
+                lines.append(("", "\n"))
+                continue
+
+            model = self.current_models[i]
+            is_selected = i == self.selected_model_idx
+
+            # Create capability icons
+            icons = []
+            if model.has_vision:
+                icons.append("👁")
+            if model.tool_call:
+                icons.append("🔧")
+            if model.reasoning:
+                icons.append("🧠")
+
+            icon_str = " ".join(icons) + " " if icons else ""
+
+            if is_selected:
+                lines.append(("fg:ansibrightblack", f" > {icon_str}{model.name}"))
+            else:
+                lines.append(("fg:ansibrightblack", f"   {icon_str}{model.name}"))
+
+            lines.append(("", "\n"))
+
+        lines.append(("", "\n"))
+        lines.append(
+            ("fg:ansibrightblack", f" Page {self.current_page + 1}/{total_pages}")
+        )
+        lines.append(("", "\n"))
+
+        self._render_navigation_hints(lines)
+        return lines
+
+    def _render_navigation_hints(self, lines: List):
+        """Render navigation hints at the bottom of the list panel."""
+        lines.append(("", "\n"))
+        lines.append(("fg:ansibrightblack", "  ↑/↓ "))
+        lines.append(("", "Navigate  "))
+        lines.append(("fg:ansibrightblack", "←/→ "))
+        lines.append(("", "Page\n"))
+        if self.view_mode == "providers":
+            lines.append(("fg:green", "  Enter  "))
+            lines.append(("", "Select\n"))
+        else:
+            lines.append(("fg:green", "  Enter  "))
+            lines.append(("", "Add Model\n"))
+            lines.append(("fg:ansibrightblack", "  Esc/Back  "))
+            lines.append(("", "Back\n"))
+        lines.append(("fg:ansibrightred", "  Ctrl+C "))
+        lines.append(("", "Cancel"))
+
+    def _render_model_details(self) -> List:
+        """Render the model details panel."""
+        lines = []
+
+        lines.append(("dim cyan", " MODEL DETAILS"))
+        lines.append(("", "\n\n"))
+
+        if self.view_mode == "providers":
+            provider = self._get_current_provider()
+            if not provider:
+                lines.append(("fg:yellow", "  No provider selected."))
+                return lines
+
+            lines.append(("bold", f"  {provider.name}"))
+            lines.append(("", "\n"))
+            lines.append(("fg:ansibrightblack", f"  ID: {provider.id}"))
+            lines.append(("", "\n"))
+            lines.append(("fg:ansibrightblack", f"  Models: {provider.model_count}"))
+            lines.append(("", "\n"))
+            lines.append(("fg:ansibrightblack", f"  API: {provider.api}"))
+            lines.append(("", "\n"))
+
+            # Show unsupported warning if applicable
+            if provider.id in UNSUPPORTED_PROVIDERS:
+                lines.append(("", "\n"))
+                lines.append(("fg:ansired bold", "  ⚠️  UNSUPPORTED PROVIDER"))
+                lines.append(("", "\n"))
+                lines.append(("fg:ansired", f"  {UNSUPPORTED_PROVIDERS[provider.id]}"))
+                lines.append(("", "\n"))
+                lines.append(
+                    (
+                        "fg:ansibrightblack",
+                        "  Models from this provider cannot be added.",
+                    )
+                )
+                lines.append(("", "\n"))
+
+            if provider.env:
+                lines.append(("", "\n"))
+                lines.append(("bold", "  Environment Variables:"))
+                lines.append(("", "\n"))
+                for env_var in provider.env:
+                    lines.append(("fg:ansibrightblack", f"    • {env_var}"))
+                    lines.append(("", "\n"))
+
+            if provider.doc:
+                lines.append(("", "\n"))
+                lines.append(("bold", "  Documentation:"))
+                lines.append(("", "\n"))
+                lines.append(("fg:ansibrightblack", f"    {provider.doc}"))
+                lines.append(("", "\n"))
+
+        else:  # models view
+            model = self._get_current_model()
+            provider = self.current_provider
+
+            if not provider:
+                lines.append(("fg:yellow", "  No model selected."))
+                return lines
+
+            # Handle custom model option
+            if self._is_custom_model_selected():
+                lines.append(("bold", "  ✨ Custom Model"))
+                lines.append(("", "\n\n"))
+                lines.append(("fg:ansicyan", "  Add a model not listed in models.dev"))
+                lines.append(("", "\n\n"))
+                lines.append(("bold", "  How it works:"))
+                lines.append(("", "\n"))
+                lines.append(("fg:ansibrightblack", "  1. Press Enter to select"))
+                lines.append(("", "\n"))
+                lines.append(("fg:ansibrightblack", "  2. Enter the model ID/name"))
+                lines.append(("", "\n"))
+                lines.append(
+                    ("fg:ansibrightblack", f"  3. Uses {provider.name}'s API endpoint")
+                )
+                lines.append(("", "\n\n"))
+                lines.append(("bold", "  Use cases:"))
+                lines.append(("", "\n"))
+                lines.append(("fg:ansibrightblack", "  • Newly released models"))
+                lines.append(("", "\n"))
+                lines.append(("fg:ansibrightblack", "  • Fine-tuned models"))
+                lines.append(("", "\n"))
+                lines.append(("fg:ansibrightblack", "  • Preview/beta models"))
+                lines.append(("", "\n"))
+                lines.append(("fg:ansibrightblack", "  • Custom deployments"))
+                lines.append(("", "\n\n"))
+                if provider.env:
+                    lines.append(("bold", "  Required credentials:"))
+                    lines.append(("", "\n"))
+                    for env_var in provider.env:
+                        lines.append(("fg:ansibrightblack", f"    • {env_var}"))
+                        lines.append(("", "\n"))
+                return lines
+
+            if not model:
+                lines.append(("fg:yellow", "  No model selected."))
+                return lines
+
+            lines.append(("bold", f"  {provider.name} - {model.name}"))
+            lines.append(("", "\n\n"))
+
+            # BIG WARNING for models without tool calling
+            if not model.tool_call:
+                lines.append(("fg:ansiyellow bold", "  ⚠️  NO TOOL CALLING SUPPORT"))
+                lines.append(("", "\n"))
+                lines.append(
+                    ("fg:ansiyellow", "  This model cannot use tools (file ops,")
+                )
+                lines.append(("", "\n"))
+                lines.append(
+                    ("fg:ansiyellow", "  shell commands, etc). It will be very")
+                )
+                lines.append(("", "\n"))
+                lines.append(("fg:ansiyellow", "  limited for coding tasks!"))
+                lines.append(("", "\n\n"))
+
+            # Capabilities
+            lines.append(("bold", "  Capabilities:"))
+            lines.append(("", "\n"))
+
+            capabilities = [
+                ("Vision", model.has_vision),
+                ("Tool Calling", model.tool_call),
+                ("Reasoning", model.reasoning),
+                ("Temperature", model.temperature),
+                ("Structured Output", model.structured_output),
+                ("Attachments", model.attachment),
+            ]
+
+            for cap_name, has_cap in capabilities:
+                if has_cap:
+                    lines.append(("fg:green", f"    ✓ {cap_name}"))
+                else:
+                    lines.append(("fg:ansibrightblack", f"    ✗ {cap_name}"))
+                lines.append(("", "\n"))
+
+            # Pricing
+            lines.append(("", "\n"))
+            lines.append(("bold", "  Pricing:"))
+            lines.append(("", "\n"))
+
+            if model.cost_input is not None or model.cost_output is not None:
+                if model.cost_input is not None:
+                    lines.append(
+                        (
+                            "fg:ansibrightblack",
+                            f"    Input: ${model.cost_input:.6f}/token",
+                        )
+                    )
+                    lines.append(("", "\n"))
+                if model.cost_output is not None:
+                    lines.append(
+                        (
+                            "fg:ansibrightblack",
+                            f"    Output: ${model.cost_output:.6f}/token",
+                        )
+                    )
+                    lines.append(("", "\n"))
+                if model.cost_cache_read is not None:
+                    lines.append(
+                        (
+                            "fg:ansibrightblack",
+                            f"    Cache Read: ${model.cost_cache_read:.6f}/token",
+                        )
+                    )
+                    lines.append(("", "\n"))
+            else:
+                lines.append(("fg:ansibrightblack", "    Pricing not available"))
+                lines.append(("", "\n"))
+
+            # Limits
+            lines.append(("", "\n"))
+            lines.append(("bold", "  Limits:"))
+            lines.append(("", "\n"))
+
+            if model.context_length > 0:
+                lines.append(
+                    (
+                        "fg:ansibrightblack",
+                        f"    Context: {model.context_length:,} tokens",
+                    )
+                )
+                lines.append(("", "\n"))
+            if model.max_output > 0:
+                lines.append(
+                    (
+                        "fg:ansibrightblack",
+                        f"    Max Output: {model.max_output:,} tokens",
+                    )
+                )
+                lines.append(("", "\n"))
+
+            # Modalities
+            if model.input_modalities or model.output_modalities:
+                lines.append(("", "\n"))
+                lines.append(("bold", "  Modalities:"))
+                lines.append(("", "\n"))
+
+                if model.input_modalities:
+                    lines.append(
+                        (
+                            "fg:ansibrightblack",
+                            f"    Input: {', '.join(model.input_modalities)}",
+                        )
+                    )
+                    lines.append(("", "\n"))
+                if model.output_modalities:
+                    lines.append(
+                        (
+                            "fg:ansibrightblack",
+                            f"    Output: {', '.join(model.output_modalities)}",
+                        )
+                    )
+                    lines.append(("", "\n"))
+
+            # Metadata
+            lines.append(("", "\n"))
+            lines.append(("bold", "  Metadata:"))
+            lines.append(("", "\n"))
+
+            lines.append(("fg:ansibrightblack", f"    Model ID: {model.model_id}"))
+            lines.append(("", "\n"))
+            lines.append(("fg:ansibrightblack", f"    Full ID: {model.full_id}"))
+            lines.append(("", "\n"))
+
+            if model.knowledge:
+                lines.append(
+                    ("fg:ansibrightblack", f"    Knowledge: {model.knowledge}")
+                )
+                lines.append(("", "\n"))
+
+            if model.release_date:
+                lines.append(
+                    ("fg:ansibrightblack", f"    Released: {model.release_date}")
+                )
+                lines.append(("", "\n"))
+
+            lines.append(
+                ("fg:ansibrightblack", f"    Open Weights: {model.open_weights}")
+            )
+            lines.append(("", "\n"))
+
+        return lines
+
+    def _add_model_to_extra_config(
+        self, model: ModelInfo, provider: ProviderInfo
+    ) -> bool:
+        """Add a model to the extra_models.json configuration file.
+
+        The extra_models.json format is a dictionary where:
+        - Keys are user-friendly model names (e.g., "provider-model-name")
+        - Values contain type, name, custom_endpoint (if needed), and context_length
+        """
+        try:
+            # Load existing extra models (dictionary format)
+            extra_models_path = Path(EXTRA_MODELS_FILE)
+            extra_models: dict = {}
+
+            if extra_models_path.exists():
+                try:
+                    with open(extra_models_path, "r", encoding="utf-8") as f:
+                        extra_models = json.load(f)
+                        if not isinstance(extra_models, dict):
+                            emit_error(
+                                "extra_models.json must be a dictionary, not a list"
+                            )
+                            return False
+                except json.JSONDecodeError as e:
+                    emit_error(f"Error parsing extra_models.json: {e}")
+                    return False
+
+            # Create a unique key for this model (provider-modelname format)
+            model_key = f"{provider.id}-{model.model_id}".replace("/", "-").replace(
+                ":", "-"
+            )
+
+            # Check for duplicates
+            if model_key in extra_models:
+                emit_info(f"Model {model_key} is already in extra_models.json")
+                return True  # Not an error, just already exists
+
+            # Convert to Code Puppy config format (dictionary value)
+            config = self._build_model_config(model, provider)
+            extra_models[model_key] = config
+
+            # Ensure directory exists
+            extra_models_path.parent.mkdir(parents=True, exist_ok=True)
+
+            # Save updated configuration
+            with open(extra_models_path, "w", encoding="utf-8") as f:
+                json.dump(extra_models, f, indent=4, ensure_ascii=False)
+
+            emit_info(f"Added {model_key} to extra_models.json")
+            return True
+
+        except Exception as e:
+            emit_error(f"Error adding model to extra_models.json: {e}")
+            return False
+
+    def _build_model_config(self, model: ModelInfo, provider: ProviderInfo) -> dict:
+        """Build a Code Puppy compatible model configuration.
+
+        Format matches models.json structure:
+        {
+            "type": "openai" | "anthropic" | "gemini" | "custom_openai" | etc.,
+            "name": "actual-model-id",
+            "custom_endpoint": {"url": "...", "api_key": "$ENV_VAR"},  # if needed
+            "context_length": 200000
+        }
+        """
+        # Map provider IDs to Code Puppy types
+        type_mapping = {
+            "openai": "openai",
+            "anthropic": "anthropic",
+            "google": "gemini",
+            "google-vertex": "gemini",
+            "mistral": "custom_openai",
+            "groq": "custom_openai",
+            "together-ai": "custom_openai",
+            "fireworks": "custom_openai",
+            "deepseek": "custom_openai",
+            "openrouter": "custom_openai",
+            "cerebras": "cerebras",
+            "cohere": "custom_openai",
+            "perplexity": "custom_openai",
+            "minimax": "custom_anthropic",
+        }
+
+        # Determine the model type
+        model_type = type_mapping.get(provider.id, "custom_openai")
+
+        # Special case: kimi-for-coding provider uses "kimi-for-coding" as the model name
+        # instead of the model_id from models.dev (which is "kimi-k2-thinking")
+        if provider.id == "kimi-for-coding":
+            model_name = "kimi-for-coding"
+        else:
+            model_name = model.model_id
+
+        config: dict = {
+            "type": model_type,
+            "name": model_name,
+        }
+
+        # Add custom endpoint for non-standard providers
+        if model_type == "custom_openai":
+            # Get the API URL - prefer provider.api, fall back to hardcoded endpoints
+            api_url = provider.api
+            if not api_url or api_url == "N/A":
+                api_url = PROVIDER_ENDPOINTS.get(provider.id)
+
+            if api_url:
+                # Determine the API key environment variable
+                api_key_env = f"${provider.env[0]}" if provider.env else "$API_KEY"
+                config["custom_endpoint"] = {"url": api_url, "api_key": api_key_env}
+
+        # Special handling for minimax: uses custom_anthropic but needs custom_endpoint
+        # and the URL needs /v1 stripped (comes as https://api.minimax.io/anthropic/v1)
+        if provider.id == "minimax" and provider.api:
+            api_url = provider.api
+            # Strip /v1 suffix if present
+            if api_url.endswith("/v1"):
+                api_url = api_url[:-3]
+            api_key_env = f"${provider.env[0]}" if provider.env else "$API_KEY"
+            config["custom_endpoint"] = {"url": api_url, "api_key": api_key_env}
+
+        # Add context length if available
+        if model.context_length and model.context_length > 0:
+            config["context_length"] = model.context_length
+
+        # Add supported settings based on model type
+        if model_type == "anthropic":
+            config["supported_settings"] = [
+                "temperature",
+                "extended_thinking",
+                "budget_tokens",
+            ]
+        elif model_type == "openai" and "gpt-5" in model.model_id:
+            # GPT-5 models have special settings
+            if "codex" in model.model_id:
+                config["supported_settings"] = ["reasoning_effort"]
+            else:
+                config["supported_settings"] = ["reasoning_effort", "verbosity"]
+        else:
+            # Default settings for most models (no top_p)
+            config["supported_settings"] = ["temperature", "seed"]
+
+        return config
+
+    def update_display(self):
+        """Update the display based on current state."""
+        if self.view_mode == "providers":
+            self.menu_control.text = self._render_provider_list()
+        else:
+            self.menu_control.text = self._render_model_list()
+
+        self.preview_control.text = self._render_model_details()
+
+    def _enter_provider(self):
+        """Enter the selected provider to view its models."""
+        provider = self._get_current_provider()
+        if not provider or not self.registry:
+            return
+
+        self.current_provider = provider
+        self.current_models = self.registry.get_models(provider.id)
+        self.view_mode = "models"
+        self.selected_model_idx = 0
+        self.current_page = 0
+        self.update_display()
+
+    def _go_back_to_providers(self):
+        """Go back to providers view."""
+        self.view_mode = "providers"
+        self.current_provider = None
+        self.current_models = []
+        self.selected_model_idx = 0
+        self.current_page = 0
+        self.update_display()
+
+    def _add_current_model(self):
+        """Add the currently selected model to extra_models.json."""
+        provider = self.current_provider
+
+        if not provider:
+            return
+
+        # Block unsupported providers
+        if provider.id in UNSUPPORTED_PROVIDERS:
+            self.result = "unsupported"
+            return
+
+        # Check if custom model option is selected
+        if self._is_custom_model_selected():
+            self.is_custom_model_selected = True
+            self.pending_provider = provider
+            self.result = (
+                "pending_custom_model"  # Signal to prompt for custom model name
+            )
+            return
+
+        model = self._get_current_model()
+        if model:
+            # Store model/provider for credential prompting after TUI exits
+            self.pending_model = model
+            self.pending_provider = provider
+            self.result = "pending_credentials"  # Signal to prompt for credentials
+
+    def _get_missing_env_vars(self, provider: ProviderInfo) -> List[str]:
+        """Check which required env vars are missing for a provider."""
+        missing = []
+        for env_var in provider.env:
+            if not os.environ.get(env_var):
+                missing.append(env_var)
+        return missing
+
+    def _prompt_for_credentials(self, provider: ProviderInfo) -> bool:
+        """Prompt user for missing credentials and save them.
+
+        Returns:
+            True if all credentials were provided (or none needed), False if user cancelled
+        """
+        missing_vars = self._get_missing_env_vars(provider)
+
+        if not missing_vars:
+            emit_info(
+                f"✅ All required credentials for {provider.name} are already set!"
+            )
+            return True
+
+        emit_info(f"\n🔑 {provider.name} requires the following credentials:\n")
+
+        for env_var in missing_vars:
+            # Show helpful hints based on common env var patterns
+            hint = self._get_env_var_hint(env_var)
+            if hint:
+                emit_info(f"  {hint}")
+
+            try:
+                # Use safe_input for cross-platform compatibility (Windows fix)
+                value = safe_input(f"  Enter {env_var} (or press Enter to skip): ")
+
+                if not value:
+                    emit_warning(
+                        f"Skipped {env_var} - you can set it later with /set {env_var}=<value>"
+                    )
+                    continue
+
+                # Save to config
+                set_config_value(env_var, value)
+                # Also set in current environment so it's immediately available
+                os.environ[env_var] = value
+                emit_info(f"✅ Saved {env_var} to config")
+
+            except (KeyboardInterrupt, EOFError):
+                emit_info("")  # Clean newline
+                emit_warning("Credential input cancelled")
+                return False
+
+        return True
+
+    def _create_custom_model_info(
+        self, model_name: str, context_length: int = 128000
+    ) -> ModelInfo:
+        """Create a ModelInfo object for a custom model.
+
+        Since we don't know the model's capabilities, we assume reasonable defaults.
+        """
+        provider_id = self.pending_provider.id if self.pending_provider else "custom"
+        return ModelInfo(
+            provider_id=provider_id,
+            model_id=model_name,
+            name=model_name,
+            tool_call=True,  # Assume true for usability
+            temperature=True,
+            context_length=context_length,
+            max_output=min(
+                16384, context_length // 4
+            ),  # Reasonable default based on context
+            input_modalities=["text"],
+            output_modalities=["text"],
+        )
+
+    def _prompt_for_custom_model(self) -> Optional[tuple[str, int]]:
+        """Prompt user for custom model details.
+
+        Returns:
+            Tuple of (model_name, context_length) if provided, None if cancelled
+        """
+        provider = self.pending_provider
+        if not provider:
+            return None
+
+        emit_info(f"\n✨ Adding custom model for {provider.name}\n")
+        emit_info("  Enter the model ID exactly as the provider expects it.")
+        emit_info(
+            "  Examples: gpt-4-turbo-preview, claude-3-opus-20240229, gemini-1.5-pro-latest\n"
+        )
+
+        try:
+            model_name = safe_input("  Model ID: ")
+
+            if not model_name:
+                emit_warning("No model name provided, cancelled.")
+                return None
+
+            # Ask for context size
+            emit_info("\n  Enter the context window size (in tokens).")
+            emit_info("  Common sizes: 8192, 32768, 128000, 200000, 1000000\n")
+
+            context_input = safe_input("  Context size [128000]: ")
+
+            if not context_input:
+                context_length = 128000  # Default
+            else:
+                # Handle k/K suffix (e.g., "128k" -> 128000)
+                context_input_lower = context_input.lower().replace(",", "")
+                if context_input_lower.endswith("k"):
+                    try:
+                        context_length = int(float(context_input_lower[:-1]) * 1000)
+                    except ValueError:
+                        emit_warning("Invalid context size, using default 128000")
+                        context_length = 128000
+                elif context_input_lower.endswith("m"):
+                    try:
+                        context_length = int(float(context_input_lower[:-1]) * 1000000)
+                    except ValueError:
+                        emit_warning("Invalid context size, using default 128000")
+                        context_length = 128000
+                else:
+                    try:
+                        context_length = int(context_input)
+                    except ValueError:
+                        emit_warning("Invalid context size, using default 128000")
+                        context_length = 128000
+
+            return (model_name, context_length)
+
+        except (KeyboardInterrupt, EOFError):
+            emit_info("")  # Clean newline
+            emit_warning("Custom model input cancelled")
+            return None
+
+    def _get_env_var_hint(self, env_var: str) -> str:
+        """Get a helpful hint for common environment variables."""
+        hints = {
+            "OPENAI_API_KEY": "💡 Get your API key from https://platform.openai.com/api-keys",
+            "ANTHROPIC_API_KEY": "💡 Get your API key from https://console.anthropic.com/",
+            "GEMINI_API_KEY": "💡 Get your API key from https://aistudio.google.com/apikey",
+            "GOOGLE_API_KEY": "💡 Get your API key from https://aistudio.google.com/apikey",
+            "AZURE_API_KEY": "💡 Get your API key from Azure Portal > Your OpenAI Resource > Keys",
+            "AZURE_RESOURCE_NAME": "💡 Your Azure OpenAI resource name (not the full URL)",
+            "GROQ_API_KEY": "💡 Get your API key from https://console.groq.com/keys",
+            "MISTRAL_API_KEY": "💡 Get your API key from https://console.mistral.ai/",
+            "COHERE_API_KEY": "💡 Get your API key from https://dashboard.cohere.com/api-keys",
+            "DEEPSEEK_API_KEY": "💡 Get your API key from https://platform.deepseek.com/",
+            "TOGETHER_API_KEY": "💡 Get your API key from https://api.together.xyz/settings/api-keys",
+            "FIREWORKS_API_KEY": "💡 Get your API key from https://fireworks.ai/api-keys",
+            "OPENROUTER_API_KEY": "💡 Get your API key from https://openrouter.ai/keys",
+            "PERPLEXITY_API_KEY": "💡 Get your API key from https://www.perplexity.ai/settings/api",
+            "CEREBRAS_API_KEY": "💡 Get your API key from https://cloud.cerebras.ai/",
+            "HUGGINGFACE_API_KEY": "💡 Get your API key from https://huggingface.co/settings/tokens",
+            "XAI_API_KEY": "💡 Get your API key from https://console.x.ai/",
+        }
+        return hints.get(env_var, "")
+
+    def run(self) -> bool:
+        """Run the interactive model browser (synchronous).
+
+        Returns:
+            True if a model was added, False otherwise
+        """
+        if not self.registry or not self.providers:
+            emit_warning("No models data available.")
+            return False
+
+        # Build UI
+        self.menu_control = FormattedTextControl(text="")
+        self.preview_control = FormattedTextControl(text="")
+
+        menu_window = Window(
+            content=self.menu_control, wrap_lines=True, width=Dimension(weight=30)
+        )
+        preview_window = Window(
+            content=self.preview_control, wrap_lines=True, width=Dimension(weight=70)
+        )
+
+        menu_frame = Frame(menu_window, width=Dimension(weight=30), title="Browse")
+        preview_frame = Frame(
+            preview_window, width=Dimension(weight=70), title="Details"
+        )
+
+        root_container = VSplit([menu_frame, preview_frame])
+
+        # Key bindings
+        kb = KeyBindings()
+
+        @kb.add("up")
+        def _(event):
+            if self.view_mode == "providers":
+                if self.selected_provider_idx > 0:
+                    self.selected_provider_idx -= 1
+                    self.current_page = self.selected_provider_idx // PAGE_SIZE
+            else:  # models view
+                if self.selected_model_idx > 0:
+                    self.selected_model_idx -= 1
+                    self.current_page = self.selected_model_idx // PAGE_SIZE
+            self.update_display()
+
+        @kb.add("down")
+        def _(event):
+            if self.view_mode == "providers":
+                if self.selected_provider_idx < len(self.providers) - 1:
+                    self.selected_provider_idx += 1
+                    self.current_page = self.selected_provider_idx // PAGE_SIZE
+            else:  # models view - include custom model option at the end
+                # Max index is len(current_models) which is the "Custom model" option
+                if self.selected_model_idx < len(self.current_models):
+                    self.selected_model_idx += 1
+                    self.current_page = self.selected_model_idx // PAGE_SIZE
+            self.update_display()
+
+        @kb.add("left")
+        def _(event):
+            """Previous page."""
+            if self.current_page > 0:
+                self.current_page -= 1
+                # Update selected index to first item on new page
+                if self.view_mode == "providers":
+                    self.selected_provider_idx = self.current_page * PAGE_SIZE
+                else:
+                    self.selected_model_idx = self.current_page * PAGE_SIZE
+                self.update_display()
+
+        @kb.add("right")
+        def _(event):
+            """Next page."""
+            if self.view_mode == "providers":
+                total_items = len(self.providers)
+            else:
+                total_items = len(self.current_models) + 1  # +1 for custom model option
+
+            total_pages = (total_items + PAGE_SIZE - 1) // PAGE_SIZE
+            if self.current_page < total_pages - 1:
+                self.current_page += 1
+                # Update selected index to first item on new page
+                if self.view_mode == "providers":
+                    self.selected_provider_idx = self.current_page * PAGE_SIZE
+                else:
+                    self.selected_model_idx = self.current_page * PAGE_SIZE
+                self.update_display()
+
+        @kb.add("enter")
+        def _(event):
+            if self.view_mode == "providers":
+                self._enter_provider()
+            elif self.view_mode == "models":
+                # Enter adds the model when viewing models
+                self._add_current_model()
+                event.app.exit()
+
+        @kb.add("escape")
+        def _(event):
+            if self.view_mode == "models":
+                self._go_back_to_providers()
+
+        @kb.add("backspace")
+        def _(event):
+            if self.view_mode == "models":
+                self._go_back_to_providers()
+
+        @kb.add("c-c")
+        def _(event):
+            event.app.exit()
+
+        layout = Layout(root_container)
+        app = Application(
+            layout=layout,
+            key_bindings=kb,
+            full_screen=False,
+            mouse_support=False,
+        )
+
+        set_awaiting_user_input(True)
+
+        # Enter alternate screen buffer once for entire session
+        sys.stdout.write("\033[?1049h")  # Enter alternate buffer
+        sys.stdout.write("\033[2J\033[H")  # Clear and home
+        sys.stdout.flush()
+        time.sleep(0.05)
+
+        try:
+            # Initial display
+            self.update_display()
+
+            # Just clear the current buffer (don't switch buffers)
+            sys.stdout.write("\033[2J\033[H")  # Clear screen within current buffer
+            sys.stdout.flush()
+
+            # Run application in a background thread to avoid event loop conflicts
+            # This is needed because code_puppy runs in an async context
+            app.run(in_thread=True)
+
+        finally:
+            # Exit alternate screen buffer once at end
+            sys.stdout.write("\033[?1049l")  # Exit alternate buffer
+            sys.stdout.flush()
+            # Reset awaiting input flag
+            set_awaiting_user_input(False)
+
+        # Clear exit message (unless we're about to prompt for more input)
+        if self.result not in ("pending_credentials", "pending_custom_model"):
+            emit_info("✓ Exited model browser")
+
+        # Handle unsupported provider
+        if self.result == "unsupported" and self.current_provider:
+            reason = UNSUPPORTED_PROVIDERS.get(
+                self.current_provider.id, "Not supported"
+            )
+            emit_error(f"Cannot add model from {self.current_provider.name}: {reason}")
+            return False
+
+        # Handle custom model flow after TUI exits
+        if self.result == "pending_custom_model" and self.pending_provider:
+            # Prompt for custom model details (name and context size)
+            custom_model_result = self._prompt_for_custom_model()
+            if not custom_model_result:
+                return False
+
+            model_name, context_length = custom_model_result
+
+            # Create a ModelInfo for the custom model
+            self.pending_model = self._create_custom_model_info(
+                model_name, context_length
+            )
+
+            # Prompt for any missing credentials
+            if self._prompt_for_credentials(self.pending_provider):
+                # Now add the model to config
+                if self._add_model_to_extra_config(
+                    self.pending_model, self.pending_provider
+                ):
+                    self.result = "added"
+                    return True
+            return False
+
+        # Handle pending credential flow after TUI exits
+        if (
+            self.result == "pending_credentials"
+            and self.pending_model
+            and self.pending_provider
+        ):
+            # Warn about non-tool-calling models
+            if not self.pending_model.tool_call:
+                emit_warning(
+                    f"⚠️  {self.pending_model.name} does NOT support tool calling!\n"
+                    f"   This model won't be able to edit files, run commands, or use any tools.\n"
+                    f"   It will be very limited for coding tasks."
+                )
+                try:
+                    confirm = safe_input(
+                        "\n  Are you sure you want to add this model? (y/N): "
+                    ).lower()
+                    if confirm not in ("y", "yes"):
+                        emit_info("Model addition cancelled.")
+                        return False
+                except (KeyboardInterrupt, EOFError):
+                    emit_info("")
+                    return False
+
+            # Prompt for any missing credentials
+            if self._prompt_for_credentials(self.pending_provider):
+                # Now add the model to config
+                if self._add_model_to_extra_config(
+                    self.pending_model, self.pending_provider
+                ):
+                    self.result = "added"
+                    return True
+            return False
+
+        return self.result == "added"
+
+
+def interactive_model_picker() -> bool:
+    """Show interactive terminal UI to browse and add models.
+
+    Returns:
+        True if a model was added, False otherwise
+    """
+    menu = AddModelMenu()
+    return menu.run()