Add add-models skill (#2786)

.claude/skills/add-models/SKILL.md

+---
+name: dyad:add-models
+description: Add one or more AI models to the language model constants file, researching specs from official docs.
+---
+
+# Add Models
+
+Add one or more AI models to `src/ipc/shared/language_model_constants.ts`, researching correct specifications from official documentation.
+
+## Arguments
+
+- `$ARGUMENTS`: Comma-separated list of model names to add (e.g., "gemini 3.1 pro, glm 5, sonnet 4.6").
+
+## Instructions
+
+1. **Parse the model list:**
+
+   Split `$ARGUMENTS` by commas to get individual model names. Trim whitespace from each.
+
+2. **Read the current constants file:**
+
+   Read `src/ipc/shared/language_model_constants.ts` to understand:
+   - Which providers exist and their current model entries
+   - The naming conventions for each provider (e.g., `claude-sonnet-4-20250514` for Anthropic, `gemini-2.5-pro` for Google)
+   - The structure of `ModelOption` entries (name, displayName, description, maxOutputTokens, contextWindow, temperature, dollarSigns)
+
+3. **Identify the provider for each model:**
+
+   Map each model to its provider based on the model name:
+   - **Anthropic** (`anthropic`): Claude models (Opus, Sonnet, Haiku)
+   - **OpenAI** (`openai`): GPT models
+   - **Google** (`google`): Gemini models
+   - **xAI** (`xai`): Grok models
+   - **OpenRouter** (`openrouter`): Models from other providers (DeepSeek, Qwen, Moonshot/Kimi, Z-AI/GLM, etc.)
+   - **Azure** (`azure`): Azure-hosted OpenAI models
+   - **Bedrock** (`bedrock`): AWS Bedrock-hosted Anthropic models
+   - **Vertex** (`vertex`): Google Vertex AI-hosted models
+
+   If a model could belong to multiple providers (e.g., a new Anthropic model should go in `anthropic` AND potentially `bedrock`), add it to the primary provider. Ask the user if they also want it added to secondary providers.
+
+4. **Research each model's specifications:**
+
+   For each model, use WebSearch and WebFetch to look up the official documentation:
+   - **Anthropic models**: Search `docs.anthropic.com` for model specs
+   - **OpenAI models**: Search `platform.openai.com/docs/models` for model specs
+   - **Google Gemini models**: Search `ai.google.dev/gemini-api/docs/models` for model specs
+   - **xAI models**: Search `docs.x.ai/docs/models` for model specs
+   - **OpenRouter models**: Search `openrouter.ai/<provider>/<model-name>` for model specs and pricing
+
+   For each model, determine:
+   - **API model name**: The exact string used in API calls (e.g., `claude-sonnet-4-5-20250929`, `gemini-2.5-pro`)
+   - **Display name**: Human-readable name (e.g., "Claude Sonnet 4.5", "Gemini 2.5 Pro")
+   - **Description**: Short description following the style of existing entries
+   - **Max output tokens**: The model's maximum output token limit
+   - **Context window**: The model's total context window size
+   - **Temperature**: Default temperature (0 for most models, 1 for OpenAI, 1.0 for Gemini 3.x models)
+   - **Dollar signs**: Cost tier from 0-6 based on pricing relative to other models in the same provider
+
+   **Dollar signs guide** (approximate, based on per-million-token input pricing):
+   - 0: Free
+   - 1: Very cheap (<$0.50/M input tokens)
+   - 2: Cheap ($0.50-$2/M)
+   - 3: Moderate ($2-$8/M)
+   - 4: Expensive ($8-$15/M)
+   - 5: Very expensive ($15-$30/M)
+   - 6: Premium ($30+/M)
+
+5. **Follow provider-specific conventions:**
+
+   Match the patterns used by existing entries:
+   - **OpenAI**: `maxOutputTokens: undefined` (OpenAI errors with `max_tokens`), `temperature: 1`
+   - **Anthropic**: `maxOutputTokens: 32_000`, `temperature: 0`
+   - **Google**: `maxOutputTokens: 65_536 - 1` (exclusive upper bound for Vertex), `temperature` varies
+   - **OpenRouter**: `maxOutputTokens: 32_000`, prefix model name with provider (e.g., `deepseek/deepseek-chat-v3.1`)
+   - **Azure**: `maxOutputTokens` commented out, `temperature: 1`
+   - **Bedrock**: Model names use ARN format (e.g., `us.anthropic.claude-sonnet-4-5-20250929-v1:0`)
+   - **xAI**: `maxOutputTokens: 32_000`, `temperature: 0`
+
+6. **Add the models to the constants file:**
+
+   Insert each new model entry into the appropriate provider's array in `MODEL_OPTIONS`. Place new models:
+   - At the **top** of the provider's array if it's the newest/most capable model
+   - After existing models of the same family but before older generations
+   - Add a comment with a link to the model's documentation page
+
+   Also check if related arrays need updating:
+   - `TURBO_MODELS`: If the model has a turbo variant
+   - `PROVIDERS_THAT_SUPPORT_THINKING`: If adding a new provider that supports thinking
+
+7. **Check for named constant exports:**
+
+   If the new model is likely to be referenced elsewhere (like `SONNET_4_5` or `GPT_5_2_MODEL_NAME`), create a named constant export for it. Search the codebase for references to similar constants to determine if one is needed:
+
+   ```
+   grep -r "SONNET_4_5\|GPT_5_2_MODEL_NAME\|GEMINI_3_FLASH" src/
+   ```
+
+8. **Verify the changes compile:**
+
+   ```
+   npm run ts
+   ```
+
+   Fix any type errors if they occur.
+
+9. **Summarize what was added:**
+
+   Report to the user:
+   - Which models were added and to which providers
+   - The key specs for each (context window, max output, pricing tier)
+   - Any models that couldn't be found or had ambiguous specifications
+   - Any decisions that were made (e.g., choosing between model versions)

package-lock.json

 {
   "name": "dyad",
-  "version": "0.37.0-beta.2",
+  "version": "0.37.0",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "dyad",
-      "version": "0.37.0-beta.2",
+      "version": "0.37.0",
       "license": "MIT",
       "dependencies": {
         "@ai-sdk/amazon-bedrock": "^4.0.46",

src/ipc/shared/language_model_constants.ts

@@ -18,7 +18,7 @@ export interface ModelOption {
 }
 
 export const GPT_5_2_MODEL_NAME = "gpt-5.2";
-export const SONNET_4_5 = "claude-sonnet-4-5-20250929";
+export const SONNET_4_6 = "claude-sonnet-4-6";
 export const GEMINI_3_FLASH = "gemini-3-flash-preview";
 
 export const MODEL_OPTIONS: Record<string, ModelOption[]> = {
@@ -123,6 +123,18 @@ export const MODEL_OPTIONS: Record<string, ModelOption[]> = {
       temperature: 0,
       dollarSigns: 6,
     },
+    // https://docs.anthropic.com/en/docs/about-claude/models/overview
+    {
+      name: SONNET_4_6,
+      displayName: "Claude Sonnet 4.6",
+      description:
+        "Anthropic's fast and intelligent model (note: >200k tokens is very expensive!)",
+      // Set to 32k since context window is 1M tokens
+      maxOutputTokens: 32_000,
+      contextWindow: 1_000_000,
+      temperature: 0,
+      dollarSigns: 5,
+    },
     {
       name: "claude-opus-4-5",
       displayName: "Claude Opus 4.5",
@@ -135,7 +147,7 @@ export const MODEL_OPTIONS: Record<string, ModelOption[]> = {
       dollarSigns: 5,
     },
     {
-      name: SONNET_4_5,
+      name: "claude-sonnet-4-5-20250929",
       displayName: "Claude Sonnet 4.5",
       description:
         "Anthropic's best model for coding (note: >200k tokens is very expensive!)",
@@ -157,11 +169,11 @@ export const MODEL_OPTIONS: Record<string, ModelOption[]> = {
     },
   ],
   google: [
-    // https://ai.google.dev/gemini-api/docs/models#gemini-3-pro
+    // https://ai.google.dev/gemini-api/docs/models/gemini-3.1-pro-preview
     {
-      name: "gemini-3-pro-preview",
-      displayName: "Gemini 3 Pro (Preview)",
-      description: "Google's latest Gemini model",
+      name: "gemini-3.1-pro-preview",
+      displayName: "Gemini 3.1 Pro (Preview)",
+      description: "Google's most capable Gemini model",
       // See Flash 2.5 comment below (go 1 below just to be safe, even though it seems OK now).
       maxOutputTokens: 65_536 - 1,
       // Gemini context window = input token + output token
@@ -249,6 +261,16 @@ export const MODEL_OPTIONS: Record<string, ModelOption[]> = {
       temperature: 1.0,
       dollarSigns: 2,
     },
+    // https://openrouter.ai/minimax/minimax-m2.5
+    {
+      name: "minimax/minimax-m2.5",
+      displayName: "MiniMax M2.5",
+      description: "Strong cost-effective model for real-world productivity",
+      maxOutputTokens: 32_000,
+      contextWindow: 196_608,
+      temperature: 0,
+      dollarSigns: 1,
+    },
     {
       name: "z-ai/glm-5",
       displayName: "GLM 5",

src/ipc/utils/get_model_client.ts

@@ -19,7 +19,7 @@ import {
   FREE_OPENROUTER_MODEL_NAMES,
   GEMINI_3_FLASH,
   GPT_5_2_MODEL_NAME,
-  SONNET_4_5,
+  SONNET_4_6,
 } from "../shared/language_model_constants";
 import { getLanguageModelProviders } from "../shared/language_model_helpers";
 import { LanguageModelProvider } from "@/ipc/types";
@@ -42,7 +42,7 @@ const AUTO_MODELS = [
   },
   {
     provider: "anthropic",
-    name: SONNET_4_5,
+    name: SONNET_4_6,
   },
   {
     provider: "google",
@@ -214,7 +214,7 @@ function getProModelClient({
         models: [
           // openai requires no prefix.
           provider.responses(`${GPT_5_2_MODEL_NAME}`, { providerId: "openai" }),
-          provider(`anthropic/${SONNET_4_5}`, { providerId: "anthropic" }),
+          provider(`anthropic/${SONNET_4_6}`, { providerId: "anthropic" }),
           provider(`gemini/${GEMINI_3_FLASH}`, { providerId: "google" }),
         ],
       }),