feat: update DeepSeek models to V4 Flash and V4 Pro (#103)

* feat: update DeepSeek models to V4 Flash and V4 Pro

- Replace deprecated deepseek-chat / deepseek-reasoner with deepseek-v4-flash / deepseek-v4-pro
- Update base URL from api.deepseek.com/v1 to api.deepseek.com
- Update context window (1M) and max output (384K) per official API docs
- Old models deprecated 2026-07-24

* fix: resolveGeneralLLMConfig missing baseUrl + Admiral timeout 180s

resolveGeneralLLMConfig constructed a new LLMConfig but omitted
baseUrl, causing the adapter to fall back to api.openai.com/v1
regardless of provider — DeepSeek keys hit OpenAI (401).
Also raised Admiral preview timeout to 180s for thinking-mode models.

* fix: createTempestCommandInstance missing baseUrl for operator exec

Operator LLM calls also lacked baseUrl, same 401-for-DeepSeek bug
as resolveGeneralLLMConfig. Reads provider-specific baseUrl from config.

* fix(deepseek): migrate retired persisted defaults

---------

Co-authored-by: Joseph Magly <1159087+jmagly@users.noreply.github.com>

docs/index.html

@@ -7037,9 +7037,9 @@ Correlating findings across agents, identifying patterns, producing actionable i
             { id: 'z-ai/glm-4.7', name: 'GLM 4.7', provider: 'Z.AI' },
             // Meta
             { id: 'meta-llama/llama-3.3-70b', name: 'Llama 3.3 70B', provider: 'Meta' },
-            // DeepSeek (Dec 2025)
+            // DeepSeek (Jul 2026 — V4)
-            { id: 'deepseek/deepseek-r1', name: 'DeepSeek R1', provider: 'DeepSeek' },
+            { id: 'deepseek/deepseek-v4-pro', name: 'DeepSeek V4 Pro', provider: 'DeepSeek' },
-            { id: 'deepseek/deepseek-chat', name: 'DeepSeek V3', provider: 'DeepSeek' },
+            { id: 'deepseek/deepseek-v4-flash', name: 'DeepSeek V4 Flash', provider: 'DeepSeek' },
             // Mistral
             { id: 'mistralai/mistral-large', name: 'Mistral Large', provider: 'Mistral' },
             // Local
@@ -25836,8 +25836,8 @@ Be specific and actionable. Format as structured JSON.`;
     const bb = _admBackbone();
     try {
       // Op Admiral planning is a slow LLM call — cap it so a first-time user never gets a
-      // permanent "…drafting…" spinner if the model stalls (120s = generous but finite).
+      // permanent "…drafting…" spinner if the model stalls (180s).
-      const r = await fetch('/api/admiral/launch', { method:'POST', headers:{'Content-Type':'application/json'}, body: JSON.stringify(Object.assign({ brief: _admBrief('dry_run'), confirmed: false }, bb)), signal: AbortSignal.timeout(120000) });
+      const r = await fetch('/api/admiral/launch', { method:'POST', headers:{'Content-Type':'application/json'}, body: JSON.stringify(Object.assign({ brief: _admBrief('dry_run'), confirmed: false }, bb)), signal: AbortSignal.timeout(180000) });
       const d = await r.json();
       if (d.error) throw new Error(d.error);
       const plan = d.plan || d.directive || d;

src/__tests__/deepseek-provider.test.ts

@@ -1,5 +1,9 @@
 import { describe, expect, it, afterEach } from 'vitest';
-import { config, AVAILABLE_MODELS } from '../config/index.js';
+import {
+  config,
+  AVAILABLE_MODELS,
+  migrateLegacyDeepSeekSettings,
+} from '../config/index.js';
 import { createDeepSeekBackbone } from '../llm/index.js';
 
 const KEY = 'deepseek-key-abcdef123456';
@@ -7,12 +11,12 @@ const KEY = 'deepseek-key-abcdef123456';
 describe('DeepSeek provider wiring (#62)', () => {
   afterEach(() => { delete process.env.DEEPSEEK_API_KEY; });
 
-  it('resolves DeepSeek base URL, default model, and key from DEEPSEEK_API_KEY', () => {
+  it('resolves DeepSeek base URL, supported migrated model, and key from DEEPSEEK_API_KEY', () => {
     process.env.DEEPSEEK_API_KEY = KEY;
     const cfg = config.getLLMConfig('deepseek');
     expect(cfg.provider).toBe('deepseek');
-    expect(cfg.baseUrl).toBe('https://api.deepseek.com/v1');
+    expect(cfg.baseUrl).toBe('https://api.deepseek.com');
-    expect(cfg.model).toBe('deepseek-chat');
+    expect(['deepseek-v4-flash', 'deepseek-v4-pro']).toContain(cfg.model);
     expect(cfg.apiKey).toBe(KEY);
   });
 
@@ -26,8 +30,29 @@ describe('DeepSeek provider wiring (#62)', () => {
   it('surfaces native DeepSeek models and configured provider state', () => {
     process.env.DEEPSEEK_API_KEY = KEY;
     expect(AVAILABLE_MODELS.deepseek?.map(m => m.id)).toEqual(
-      expect.arrayContaining(['deepseek-chat', 'deepseek-reasoner']),
+      expect.arrayContaining(['deepseek-v4-flash', 'deepseek-v4-pro']),
     );
     expect(config.getConfiguredProviders()).toContain('deepseek');
   });
+
+  it('migrates only retired DeepSeek defaults and preserves custom settings', () => {
+    expect(migrateLegacyDeepSeekSettings({
+      baseUrl: 'https://api.deepseek.com/v1',
+      defaultModel: 'deepseek-chat',
+    })).toEqual({
+      baseUrl: 'https://api.deepseek.com',
+      defaultModel: 'deepseek-v4-flash',
+    });
+    expect(migrateLegacyDeepSeekSettings({
+      baseUrl: 'https://api.deepseek.com/v1',
+      defaultModel: 'deepseek-reasoner',
+    }).defaultModel).toBe('deepseek-v4-pro');
+    expect(migrateLegacyDeepSeekSettings({
+      baseUrl: 'https://gateway.example/v1',
+      defaultModel: 'custom/deepseek',
+    })).toEqual({
+      baseUrl: 'https://gateway.example/v1',
+      defaultModel: 'custom/deepseek',
+    });
+  });
 });

src/__tests__/provider-base-url-routing-static.test.ts

@@ -0,0 +1,19 @@
+import { describe, expect, it } from 'vitest';
+import { readFileSync } from 'node:fs';
+import { join } from 'node:path';
+
+describe('provider base URL routing regression', () => {
+  const serverSource = readFileSync(join(process.cwd(), 'src/server.ts'), 'utf8');
+
+  it('propagates the selected provider base URL into general LLM configuration', () => {
+    expect(serverSource).toMatch(
+      /function resolveGeneralLLMConfig[\s\S]*?baseUrl:\s*baseConfig\.baseUrl,/,
+    );
+  });
+
+  it('propagates the selected provider base URL into operator execution', () => {
+    expect(serverSource).toMatch(
+      /function createTempestCommandInstance[\s\S]*?const baseConfig = config\.getLLMConfig\([\s\S]*?baseUrl:\s*baseConfig\.baseUrl,/,
+    );
+  });
+});

src/config/index.ts

@@ -162,8 +162,8 @@ const DEFAULT_SETTINGS: TempestSettings = {
   },
 
   deepseek: {
-    baseUrl: 'https://api.deepseek.com/v1',
+    baseUrl: 'https://api.deepseek.com',
-    defaultModel: 'deepseek-chat',
+    defaultModel: 'deepseek-v4-pro',
   },
 
   codex: {
@@ -188,6 +188,26 @@ const DEFAULT_SETTINGS: TempestSettings = {
   },
 };
 
+export function migrateLegacyDeepSeekSettings(
+  settings: TempestSettings['deepseek'],
+): TempestSettings['deepseek'] {
+  const migrated = { ...settings };
+
+  // DeepSeek retires these compatibility aliases on 2026-07-24. Only migrate
+  // the exact historical defaults; operator-selected model IDs and endpoints
+  // remain untouched.
+  if (migrated.defaultModel === 'deepseek-chat') {
+    migrated.defaultModel = 'deepseek-v4-flash';
+  } else if (migrated.defaultModel === 'deepseek-reasoner') {
+    migrated.defaultModel = 'deepseek-v4-pro';
+  }
+  if (migrated.baseUrl === 'https://api.deepseek.com/v1') {
+    migrated.baseUrl = 'https://api.deepseek.com';
+  }
+
+  return migrated;
+}
+
 // =============================================================================
 // MODEL REGISTRY
 // =============================================================================
@@ -344,22 +364,22 @@ export const AVAILABLE_MODELS: Record<LLMProvider, ModelInfo[]> = {
       maxOutput: 4096,
       capabilities: ['reasoning', 'code', 'analysis'],
     },
-    // DeepSeek (Dec 2025)
+    // DeepSeek (Jul 2026 — V4)
     {
-      id: 'deepseek/deepseek-r1',
+      id: 'deepseek/deepseek-v4-pro',
-      name: 'DeepSeek R1',
+      name: 'DeepSeek V4 Pro',
       provider: 'DeepSeek',
-      contextWindow: 64000,
+      contextWindow: 1000000,
-      maxOutput: 8192,
+      maxOutput: 384000,
-      capabilities: ['reasoning', 'code', 'analysis', 'complex-tasks'],
+      capabilities: ['reasoning', 'code', 'analysis', 'complex-tasks', 'agents', 'tools'],
     },
     {
-      id: 'deepseek/deepseek-chat',
+      id: 'deepseek/deepseek-v4-flash',
-      name: 'DeepSeek V3',
+      name: 'DeepSeek V4 Flash',
       provider: 'DeepSeek',
-      contextWindow: 64000,
+      contextWindow: 1000000,
-      maxOutput: 8192,
+      maxOutput: 384000,
-      capabilities: ['reasoning', 'code', 'analysis'],
+      capabilities: ['reasoning', 'code', 'analysis', 'tools'],
     },
     // Mistral
     {
@@ -504,20 +524,20 @@ export const AVAILABLE_MODELS: Record<LLMProvider, ModelInfo[]> = {
   ],
   deepseek: [
     {
-      id: 'deepseek-chat',
+      id: 'deepseek-v4-pro',
-      name: 'DeepSeek V3 (native)',
+      name: 'DeepSeek V4 Pro (native)',
       provider: 'DeepSeek',
-      contextWindow: 64000,
+      contextWindow: 1000000,
-      maxOutput: 8192,
+      maxOutput: 384000,
-      capabilities: ['reasoning', 'code', 'analysis', 'tools'],
+      capabilities: ['reasoning', 'code', 'analysis', 'complex-tasks', 'agents', 'tools'],
     },
     {
-      id: 'deepseek-reasoner',
+      id: 'deepseek-v4-flash',
-      name: 'DeepSeek R1 (native)',
+      name: 'DeepSeek V4 Flash (native)',
       provider: 'DeepSeek',
-      contextWindow: 64000,
+      contextWindow: 1000000,
-      maxOutput: 8192,
+      maxOutput: 384000,
-      capabilities: ['reasoning', 'code', 'analysis', 'complex-tasks'],
+      capabilities: ['reasoning', 'code', 'analysis', 'tools'],
     },
   ],
   local: [
@@ -574,6 +594,15 @@ class ConfigManager {
       defaults: DEFAULT_SETTINGS,
     });
 
+    const deepseek = this.config.get('deepseek');
+    const migratedDeepSeek = migrateLegacyDeepSeekSettings(deepseek);
+    if (
+      migratedDeepSeek.baseUrl !== deepseek.baseUrl ||
+      migratedDeepSeek.defaultModel !== deepseek.defaultModel
+    ) {
+      this.config.set('deepseek', migratedDeepSeek);
+    }
+
     this.loadEnvVariables();
   }
 

src/server.ts

@@ -323,12 +323,15 @@ function createTempestCommandInstance(missionName: string, apiKey: string | unde
     tempestCommand.stop();
   }
 
+  const baseConfig = config.getLLMConfig(provider as any, model);
+
   tempestCommand = new TempestCommand({
     name: missionName,
     llm: {
       provider: provider as any,
       model,
       apiKey,
+      baseUrl: baseConfig.baseUrl,
       maxTokens: 4096,
       temperature: 0.7,
     },
@@ -6741,6 +6744,7 @@ function resolveGeneralLLMConfig(provider: string | undefined, model: string | u
   provider: any;
   model: string;
   apiKey?: string;
+  baseUrl?: string;
   maxTokens: number;
   temperature: number;
   timeout: number;
@@ -6769,6 +6773,7 @@ function resolveGeneralLLMConfig(provider: string | undefined, model: string | u
     provider: selectedProvider as any,
     model: model || baseConfig.model,
     apiKey: effectiveKey,
+    baseUrl: baseConfig.baseUrl,
     maxTokens: 8192,
     temperature: 0.4,
     timeout: readGeneralTimeoutEnv() ?? 300000, // General planning needs room (was a hardcoded 60s); override via env

src/setup.ts

@@ -196,7 +196,7 @@ async function setupAnthropicKey(): Promise<boolean> {
 
 async function setupDeepSeekKey(): Promise<boolean> {
   console.log('');
-  showInfo('DeepSeek provides OpenAI-compatible chat and reasoning models.');
+    showInfo('DeepSeek provides OpenAI-compatible chat and reasoning models (V4 Flash / V4 Pro).');
   showInfo('Get your API key at: ' + chalk.underline('https://platform.deepseek.com/api_keys'));
   console.log('');
 
@@ -220,8 +220,8 @@ async function setupDeepSeekKey(): Promise<boolean> {
   try {
     const llm = new LLMBackbone({
       provider: 'deepseek',
-      model: 'deepseek-chat',
+      model: 'deepseek-v4-pro',
-      baseUrl: 'https://api.deepseek.com/v1',
+      baseUrl: 'https://api.deepseek.com',
       apiKey,
       maxTokens: 10,
       temperature: 0,