ChutesAI: add support

This includes prettifying the model name, assuming the interfaces
(images are a force) and auto-sizing the context window based on the
response.

src/modules/llms/server/openai/models/chutesai.models.ts

@@ -0,0 +1,123 @@
+import { z } from 'zod';
+
+import { DModelInterfaceV1, LLM_IF_OAI_Chat, LLM_IF_OAI_Fn, LLM_IF_OAI_Vision } from '~/common/stores/llms/llms.types';
+
+import { serverCapitalizeFirstLetter } from '~/server/wire';
+
+import type { ModelDescriptionSchema } from '../../llm.server.types';
+
+import { fromManualMapping, ManualMappings } from './models.data';
+
+
+export function chutesAIHeuristic(hostname: string) {
+  return hostname.includes('.chutes.ai');
+}
+
+
+const _wireChutesAIListOutputSchema = z.array(z.object({
+
+  id: z.string(),
+  object: z.literal('model'),
+  created: z.number(),
+  owned_by: z.string().optional().nullable(),
+  root: z.string().optional().nullable(),
+
+  // ChutesAI specific field for context length
+  max_model_len: z.number().optional().nullable(),
+
+  // Optional fields that may be present
+  parent: z.string().nullable().optional(),
+  // permission: z.array(z.object({
+  //   id: z.string(),
+  //   object: z.literal('model_permission'),
+  //   created: z.number(),
+  //   allow_create_engine: z.boolean(),
+  //   allow_sampling: z.boolean(),
+  //   allow_logprobs: z.boolean(),
+  //   allow_search_indices: z.boolean(),
+  //   allow_view: z.boolean(),
+  //   allow_fine_tuning: z.boolean(),
+  //   organization: z.string(),
+  //   group: z.string().nullable(),
+  //   is_blocking: z.boolean(),
+  // })).optional(),
+}));
+
+const _chutesKnownModels: ManualMappings = [
+  // NOTE: we don't need manual patching as we have enough info for now
+] as const;
+
+const _chutesDenyListContains: string[] = [
+  // nothing to deny for now
+] as const;
+
+
+function _prettyModelId(id: string): string {
+  // example: "chutesai/Llama-4-Scout-17B-16E-Instruct" => "ChutesAI · Llama 4 Scout 17B 16E Instruct"
+  // example: "deepseek-ai/DeepSeek-R1" => "Deepseek AI · DeepSeek R1"
+  // example: "unsloth/Llama-3.2-1B-Instruct" => "Unsloth · Llama 3.2 1B Instruct"
+
+  return id
+    .replaceAll(/[_-]/g, ' ') // replace underscores or dashes with spaces
+    .replace('/', ' · ') // turn the first "/" into " · "
+    .split(' ')
+    .map(piece => {
+      // Handle special cases like version numbers
+      if (piece.match(/^\d+(\.\d+)*$/)) return piece; // keep version numbers as-is
+      if (piece.toLowerCase() === 'ai') return 'AI';
+      if (piece.toLowerCase() === 'v1' || piece.toLowerCase() === 'v2' || piece.toLowerCase() === 'v3') return piece.toUpperCase();
+      return serverCapitalizeFirstLetter(piece);
+    })
+    .join(' ')
+    .replace('Deepseek AI · DeepSeek', 'Deepseek AI · ') // special case for Deepseek
+    .trim();
+}
+
+
+export function chutesAIModelsToModelDescriptions(wireModels: unknown): ModelDescriptionSchema[] {
+  return _wireChutesAIListOutputSchema.parse(wireModels)
+
+    .filter((model) => {
+      return !_chutesDenyListContains.some(contains => model.id.includes(contains));
+    })
+
+    .map((model): ModelDescriptionSchema => {
+
+      // heuristics
+      const label = _prettyModelId(model.id);
+      const description = model.owned_by ? `${serverCapitalizeFirstLetter(model.owned_by)} model via ChutesAI.` : 'Model via ChutesAI.';
+
+      // Use max_model_len if available, otherwise fallback to 8192
+      const contextWindow = model.max_model_len || 8192;
+
+      const interfaces: DModelInterfaceV1[] = [
+        LLM_IF_OAI_Chat, // Assume all are chat models
+        LLM_IF_OAI_Vision, // Assume we can send them
+        LLM_IF_OAI_Fn, // Most models support function calling
+      ];
+
+      // Check for vision capabilities based on model name patterns
+      // if (model.id.toLowerCase().includes('vision') || model.id.toLowerCase().includes('vl')) {
+      //   interfaces.push(LLM_IF_OAI_Vision);
+      // }
+
+      // Most modern models support function calling
+      // interfaces.push(LLM_IF_OAI_Fn);
+
+      return fromManualMapping(_chutesKnownModels, model.id, model.created, undefined, {
+        idPrefix: model.id,
+        label,
+        description,
+        contextWindow,
+        interfaces,
+        hidden: false,
+      });
+    })
+
+    .sort((a: ModelDescriptionSchema, b: ModelDescriptionSchema): number => {
+      // Sort by creation date (newer first), then by id
+      if (a.created !== b.created)
+        return (b.created || 0) - (a.created || 0);
+      return a.id.localeCompare(b.id);
+    });
+}

src/modules/llms/server/openai/openai.router.ts

@@ -16,6 +16,7 @@ import { OpenAIWire_API_Images_Generations, OpenAIWire_API_Models_List, OpenAIWi
 import { ListModelsResponse_schema, ModelDescriptionSchema } from '../llm.server.types';
 import { alibabaModelSort, alibabaModelToModelDescription } from './models/alibaba.models';
 import { azureDeploymentFilter, azureDeploymentToModelDescription, azureParseFromDeploymentsAPI } from './models/azure.models';
+import { chutesAIHeuristic, chutesAIModelsToModelDescriptions } from './models/chutesai.models';
 import { deepseekModelFilter, deepseekModelSort, deepseekModelToModelDescription } from './models/deepseek.models';
 import { fastAPIHeuristic, fastAPIModels } from './models/fastapi.models';
 import { fireworksAIHeuristic, fireworksAIModelsToModelDescriptions } from './models/fireworksai.models';
@@ -241,6 +242,10 @@ export const llmOpenAIRouter = createTRPCRouter({
         // [OpenAI]: chat-only models, custom sort, manual mapping
         case 'openai':
 
+          // [ChutesAI] special case for model enumeration
+          if (chutesAIHeuristic(access.oaiHost))
+            return { models: chutesAIModelsToModelDescriptions(openAIModels) };
+
           // [FireworksAI] special case for model enumeration
           if (fireworksAIHeuristic(access.oaiHost))
             return { models: fireworksAIModelsToModelDescriptions(openAIModels) };