Implement dual-model filing pipeline with Ollama extraction

lib/server/ai.test.ts

@@ -2,6 +2,7 @@ import { beforeEach, describe, expect, it, mock } from 'bun:test';
 import {
   __resetAiWarningsForTests,
   getAiConfig,
+  getExtractionAiConfig,
   runAiAnalysis
 } from './ai';
 
@@ -154,4 +155,65 @@ describe('ai config and runtime', () => {
       })
     ).rejects.toThrow('AI SDK returned an empty response');
   });
+
+  it('uses ollama defaults for extraction workload config', () => {
+    const config = getExtractionAiConfig({
+      env: {},
+      warn: () => {}
+    });
+
+    expect(config.provider).toBe('ollama');
+    expect(config.baseUrl).toBe('http://127.0.0.1:11434');
+    expect(config.model).toBe('qwen3:8b');
+    expect(config.apiKey).toBe('ollama');
+    expect(config.temperature).toBe(0);
+  });
+
+  it('uses extraction workload and returns ollama provider on success', async () => {
+    const createModel = mock((config: {
+      provider: string;
+      apiKey?: string;
+      model: string;
+      baseUrl: string;
+      temperature: number;
+    }) => {
+      expect(config.provider).toBe('ollama');
+      expect(config.baseUrl).toBe('http://127.0.0.1:11434');
+      expect(config.model).toBe('qwen3:8b');
+      expect(config.temperature).toBe(0);
+      return { modelId: config.model };
+    });
+    const generate = mock(async () => ({ text: '{"summary":"ok","keyPoints":[],"redFlags":[],"followUpQuestions":[],"portfolioSignals":[],"confidence":0.6}' }));
+
+    const result = await runAiAnalysis('Extract this filing', 'Return JSON', {
+      env: {
+        OLLAMA_MODEL: 'qwen3:8b'
+      },
+      warn: () => {},
+      workload: 'extraction',
+      createModel,
+      generate
+    });
+
+    expect(createModel).toHaveBeenCalledTimes(1);
+    expect(generate).toHaveBeenCalledTimes(1);
+    expect(result.provider).toBe('ollama');
+    expect(result.model).toBe('qwen3:8b');
+  });
+
+  it('falls back to local text when extraction workload generation fails', async () => {
+    const result = await runAiAnalysis('Extract this filing', 'Return JSON', {
+      env: {},
+      warn: () => {},
+      workload: 'extraction',
+      createModel: () => ({}),
+      generate: async () => {
+        throw new Error('ollama unavailable');
+      }
+    });
+
+    expect(result.provider).toBe('local-fallback');
+    expect(result.model).toBe('qwen3:8b');
+    expect(result.text).toContain('AI SDK fallback mode is active');
+  });
 });