fix: add deepseek to all Record<SupportedProvider> types

- Add deepseek entries to error parsers/mappers in errors.ts
- Add fetchDeepSeekModels function and registry entries in routes.models.ts
- Add deepseek API key fallback and model fetcher
- Add deepseek to displayNames and providerEnvVars in seed.ts
- Add deepseek to pricesByProvider and rulesByProvider in optimization-rule.ts
- Add deepseek to preferredSourcePrefixes in models-dev-client.ts

Author: blastoiseclawd-hash

platform/backend/src/clients/models-dev-client.ts

@@ -406,6 +406,7 @@ class ModelsDevClient {
       anthropic: ["anthropic/"],
       cohere: ["cohere/"],
       cerebras: ["cerebras/"],
+      deepseek: ["deepseek/"],
       mistral: ["mistral/"],
       bedrock: ["amazon-bedrock/"],
       ollama: ["ollama/"],

platform/backend/src/database/seed.ts

@@ -317,6 +317,7 @@ async function seedChatApiKeysFromEnv(): Promise<void> {
     gemini: config.chat.gemini.apiKey,
     cerebras: config.chat.cerebras.apiKey,
     cohere: config.chat.cohere.apiKey,
+    deepseek: config.chat.deepseek.apiKey,
     mistral: config.chat.mistral.apiKey,
     ollama: config.chat.ollama.apiKey,
     vllm: config.chat.vllm.apiKey,
@@ -405,6 +406,7 @@ function getProviderDisplayName(provider: SupportedProvider): string {
     gemini: "Google",
     cerebras: "Cerebras",
     cohere: "Cohere",
+    deepseek: "DeepSeek",
     mistral: "Mistral",
     ollama: "Ollama",
     vllm: "vLLM",

platform/backend/src/models/optimization-rule.ts

@@ -272,6 +272,7 @@ class OptimizationRuleModel {
       gemini: [],
       cohere: [],
       cerebras: [],
+      deepseek: [], // DeepSeek model pricing varies, so no defaults
       mistral: [],
       vllm: [], // vLLM model pricing varies by deployment, so no defaults
       ollama: [], // Ollama model pricing varies by deployment, so no defaults
@@ -306,6 +307,7 @@ class OptimizationRuleModel {
         gemini: [],
         cohere: [],
         cerebras: [],
+        deepseek: [], // DeepSeek optimization rules are deployment-specific, no defaults
         mistral: [],
         vllm: [], // vLLM optimization rules are deployment-specific, no defaults
         ollama: [], // Ollama optimization rules are deployment-specific, no defaults

platform/backend/src/routes/chat/errors.ts

@@ -1107,6 +1107,7 @@ const providerParsers: Record<SupportedProvider, ErrorParser> = {
   bedrock: parseBedrockError,
   cerebras: parseOpenAIError, // Cerebras uses OpenAI-compatible API
   cohere: parseCohereError,
+  deepseek: parseOpenAIError, // DeepSeek uses OpenAI-compatible API
   mistral: parseOpenAIError, // Mistral uses OpenAI-compatible API
   vllm: parseVllmError,
   ollama: parseOllamaError,
@@ -1125,6 +1126,7 @@ const providerMappers: Record<SupportedProvider, ErrorMapper> = {
   bedrock: mapBedrockErrorWrapper,
   cerebras: mapOpenAIErrorWrapper, // Cerebras uses OpenAI-compatible API
   cohere: mapCohereErrorWrapper,
+  deepseek: mapOpenAIErrorWrapper, // DeepSeek uses OpenAI-compatible API
   mistral: mapOpenAIErrorWrapper, // Mistral uses OpenAI-compatible API
   vllm: mapVllmErrorWrapper,
   ollama: mapOllamaErrorWrapper,

platform/backend/src/routes/chat/routes.models.ts

@@ -237,6 +237,44 @@ async function fetchCerebrasModels(apiKey: string): Promise<ModelInfo[]> {
     }));
 }
 
+/**
+ * Fetch models from DeepSeek API (OpenAI-compatible)
+ */
+async function fetchDeepSeekModels(apiKey: string): Promise<ModelInfo[]> {
+  const baseUrl = config.llm.deepseek.baseUrl;
+  const url = `${baseUrl}/models`;
+
+  const response = await fetch(url, {
+    headers: {
+      Authorization: `Bearer ${apiKey}`,
+    },
+  });
+
+  if (!response.ok) {
+    const errorText = await response.text();
+    logger.error(
+      { status: response.status, error: errorText },
+      "Failed to fetch DeepSeek models",
+    );
+    throw new Error(`Failed to fetch DeepSeek models: ${response.status}`);
+  }
+
+  const data = (await response.json()) as {
+    data: Array<{
+      id: string;
+      created: number;
+      owned_by: string;
+    }>;
+  };
+
+  return data.data.map((model) => ({
+    id: model.id,
+    displayName: model.id,
+    provider: "deepseek" as const,
+    createdAt: new Date(model.created * 1000).toISOString(),
+  }));
+}
+
 /**
  * Fetch models from Mistral API (OpenAI-compatible)
  */
@@ -726,6 +764,7 @@ async function getProviderApiKey({
     anthropic: () => config.chat.anthropic.apiKey || null,
     cerebras: () => config.chat.cerebras.apiKey || null,
     cohere: () => config.chat.cohere?.apiKey || null,
+    deepseek: () => config.chat.deepseek?.apiKey || null,
     gemini: () => config.chat.gemini.apiKey || null,
     mistral: () => config.chat.mistral.apiKey || null,
     ollama: () => config.chat.ollama.apiKey || "", // Ollama typically doesn't require API keys
@@ -746,6 +785,7 @@ const modelFetchers: Record<
   anthropic: fetchAnthropicModels,
   bedrock: fetchBedrockModels,
   cerebras: fetchCerebrasModels,
+  deepseek: fetchDeepSeekModels,
   gemini: fetchGeminiModels,
   mistral: fetchMistralModels,
   openai: fetchOpenAiModels,