diff --git a/agent-code-review/src/providerDetector.ts b/agent-code-review/src/providerDetector.ts
index 7cd3672..eedbf8e 100644
--- a/agent-code-review/src/providerDetector.ts
+++ b/agent-code-review/src/providerDetector.ts
@@ -27,6 +27,14 @@ const DEFAULT_MODELS: Record<string, string> = {
 
 const OLLAMA_DEFAULT_URL = 'http://localhost:11434';
 
+const EMBEDDING_MODEL_PATTERN =
+  /embed|minilm|mxbai|bge[-_]|gte[-_]|nomic|snowflake-arctic|e5[-_]|clip$/i;
+
+export function pickOllamaChatModel(models: string[]): string | null {
+  const chatModel = models.find((m) => !EMBEDDING_MODEL_PATTERN.test(m));
+  return chatModel ?? null;
+}
+
 export async function detectProvider(options: DetectOptions = {}): Promise<ProviderConfig | null> {
   const env = options.env ?? process.env;
   const ollamaBaseUrl = options.ollamaBaseUrl ?? OLLAMA_DEFAULT_URL;
@@ -69,14 +77,17 @@ export async function detectProvider(options: DetectOptions = {}): Promise<Provi
 
   const ollamaModels = await detectOllama(ollamaBaseUrl);
   if (ollamaModels && ollamaModels.length > 0) {
-    const model = ollamaModels[0];
-    return {
-      provider: 'ollama',
-      model,
-      baseUrl: ollamaBaseUrl,
-      autoDetected: true,
-      detectedAt: new Date().toISOString(),
-    };
+    const model = pickOllamaChatModel(ollamaModels);
+    if (model) {
+      return {
+        provider: 'ollama',
+        model,
+        baseUrl: ollamaBaseUrl,
+        autoDetected: true,
+        detectedAt: new Date().toISOString(),
+      };
+    }
+    return null;
   }
 
   if (ollamaModels === null) {
diff --git a/agent-code-review/tests/providerDetector.test.ts b/agent-code-review/tests/providerDetector.test.ts
index 95ecf63..e9a8fe3 100644
--- a/agent-code-review/tests/providerDetector.test.ts
+++ b/agent-code-review/tests/providerDetector.test.ts
@@ -5,6 +5,7 @@ import os from 'os';
 import {
   detectProvider,
   detectOllama,
+  pickOllamaChatModel,
   loadConfig,
   saveConfig,
   getDefaultModel,
@@ -60,6 +61,37 @@ describe('providerDetector', () => {
       }
     });
 
+    it('skips embedding models and picks a chat-capable Ollama model', async () => {
+      const originalFetch = globalThis.fetch;
+      globalThis.fetch = vi.fn().mockResolvedValue({
+        ok: true,
+        json: async () => ({ models: [{ name: 'nomic-embed-text:latest' }, { name: 'llama3' }] }),
+      } as any);
+      try {
+        const result = await detectProvider({ env: {} });
+        expect(result).not.toBeNull();
+        expect(result!.provider).toBe('ollama');
+        expect(result!.model).toBe('llama3');
+        expect(result!.baseUrl).toBe('http://localhost:11434');
+      } finally {
+        globalThis.fetch = originalFetch;
+      }
+    });
+
+    it('returns null when Ollama only has embedding models', async () => {
+      const originalFetch = globalThis.fetch;
+      globalThis.fetch = vi.fn().mockResolvedValue({
+        ok: true,
+        json: async () => ({ models: [{ name: 'nomic-embed-text:latest' }] }),
+      } as any);
+      try {
+        const result = await detectProvider({ env: {} });
+        expect(result).toBeNull();
+      } finally {
+        globalThis.fetch = originalFetch;
+      }
+    });
+
     it('returns null when nothing is available', async () => {
       const originalFetch = globalThis.fetch;
       globalThis.fetch = vi.fn().mockRejectedValue(new Error('Connection refused'));
@@ -144,6 +176,24 @@ describe('providerDetector', () => {
     });
   });
 
+  describe('pickOllamaChatModel', () => {
+    it('returns null for an empty list', () => {
+      expect(pickOllamaChatModel([])).toBeNull();
+    });
+
+    it('skips embedding models and picks the first chat model', () => {
+      expect(pickOllamaChatModel(['nomic-embed-text:latest', 'llama3'])).toBe('llama3');
+    });
+
+    it('returns null when all models are embeddings', () => {
+      expect(pickOllamaChatModel(['nomic-embed-text:latest', 'mxbai-embed-large'])).toBeNull();
+    });
+
+    it('returns the first non-embedding model', () => {
+      expect(pickOllamaChatModel(['qwen2.5-coder:1.5b', 'llama3'])).toBe('qwen2.5-coder:1.5b');
+    });
+  });
+
   describe('loadConfig', () => {
     let tmpDir: string;
 
diff --git a/agent-doc-generator/src/providerDetector.ts b/agent-doc-generator/src/providerDetector.ts
index 7cd3672..eedbf8e 100644
--- a/agent-doc-generator/src/providerDetector.ts
+++ b/agent-doc-generator/src/providerDetector.ts
@@ -27,6 +27,14 @@ const DEFAULT_MODELS: Record<string, string> = {
 
 const OLLAMA_DEFAULT_URL = 'http://localhost:11434';
 
+const EMBEDDING_MODEL_PATTERN =
+  /embed|minilm|mxbai|bge[-_]|gte[-_]|nomic|snowflake-arctic|e5[-_]|clip$/i;
+
+export function pickOllamaChatModel(models: string[]): string | null {
+  const chatModel = models.find((m) => !EMBEDDING_MODEL_PATTERN.test(m));
+  return chatModel ?? null;
+}
+
 export async function detectProvider(options: DetectOptions = {}): Promise<ProviderConfig | null> {
   const env = options.env ?? process.env;
   const ollamaBaseUrl = options.ollamaBaseUrl ?? OLLAMA_DEFAULT_URL;
@@ -69,14 +77,17 @@ export async function detectProvider(options: DetectOptions = {}): Promise<Provi
 
   const ollamaModels = await detectOllama(ollamaBaseUrl);
   if (ollamaModels && ollamaModels.length > 0) {
-    const model = ollamaModels[0];
-    return {
-      provider: 'ollama',
-      model,
-      baseUrl: ollamaBaseUrl,
-      autoDetected: true,
-      detectedAt: new Date().toISOString(),
-    };
+    const model = pickOllamaChatModel(ollamaModels);
+    if (model) {
+      return {
+        provider: 'ollama',
+        model,
+        baseUrl: ollamaBaseUrl,
+        autoDetected: true,
+        detectedAt: new Date().toISOString(),
+      };
+    }
+    return null;
   }
 
   if (ollamaModels === null) {
diff --git a/agent-doc-generator/tests/providerDetector.test.ts b/agent-doc-generator/tests/providerDetector.test.ts
index 95ecf63..e9a8fe3 100644
--- a/agent-doc-generator/tests/providerDetector.test.ts
+++ b/agent-doc-generator/tests/providerDetector.test.ts
@@ -5,6 +5,7 @@ import os from 'os';
 import {
   detectProvider,
   detectOllama,
+  pickOllamaChatModel,
   loadConfig,
   saveConfig,
   getDefaultModel,
@@ -60,6 +61,37 @@ describe('providerDetector', () => {
       }
     });
 
+    it('skips embedding models and picks a chat-capable Ollama model', async () => {
+      const originalFetch = globalThis.fetch;
+      globalThis.fetch = vi.fn().mockResolvedValue({
+        ok: true,
+        json: async () => ({ models: [{ name: 'nomic-embed-text:latest' }, { name: 'llama3' }] }),
+      } as any);
+      try {
+        const result = await detectProvider({ env: {} });
+        expect(result).not.toBeNull();
+        expect(result!.provider).toBe('ollama');
+        expect(result!.model).toBe('llama3');
+        expect(result!.baseUrl).toBe('http://localhost:11434');
+      } finally {
+        globalThis.fetch = originalFetch;
+      }
+    });
+
+    it('returns null when Ollama only has embedding models', async () => {
+      const originalFetch = globalThis.fetch;
+      globalThis.fetch = vi.fn().mockResolvedValue({
+        ok: true,
+        json: async () => ({ models: [{ name: 'nomic-embed-text:latest' }] }),
+      } as any);
+      try {
+        const result = await detectProvider({ env: {} });
+        expect(result).toBeNull();
+      } finally {
+        globalThis.fetch = originalFetch;
+      }
+    });
+
     it('returns null when nothing is available', async () => {
       const originalFetch = globalThis.fetch;
       globalThis.fetch = vi.fn().mockRejectedValue(new Error('Connection refused'));
@@ -144,6 +176,24 @@ describe('providerDetector', () => {
     });
   });
 
+  describe('pickOllamaChatModel', () => {
+    it('returns null for an empty list', () => {
+      expect(pickOllamaChatModel([])).toBeNull();
+    });
+
+    it('skips embedding models and picks the first chat model', () => {
+      expect(pickOllamaChatModel(['nomic-embed-text:latest', 'llama3'])).toBe('llama3');
+    });
+
+    it('returns null when all models are embeddings', () => {
+      expect(pickOllamaChatModel(['nomic-embed-text:latest', 'mxbai-embed-large'])).toBeNull();
+    });
+
+    it('returns the first non-embedding model', () => {
+      expect(pickOllamaChatModel(['qwen2.5-coder:1.5b', 'llama3'])).toBe('qwen2.5-coder:1.5b');
+    });
+  });
+
   describe('loadConfig', () => {
     let tmpDir: string;
 
diff --git a/agent-refactor/src/providerDetector.ts b/agent-refactor/src/providerDetector.ts
index 7cd3672..eedbf8e 100644
--- a/agent-refactor/src/providerDetector.ts
+++ b/agent-refactor/src/providerDetector.ts
@@ -27,6 +27,14 @@ const DEFAULT_MODELS: Record<string, string> = {
 
 const OLLAMA_DEFAULT_URL = 'http://localhost:11434';
 
+const EMBEDDING_MODEL_PATTERN =
+  /embed|minilm|mxbai|bge[-_]|gte[-_]|nomic|snowflake-arctic|e5[-_]|clip$/i;
+
+export function pickOllamaChatModel(models: string[]): string | null {
+  const chatModel = models.find((m) => !EMBEDDING_MODEL_PATTERN.test(m));
+  return chatModel ?? null;
+}
+
 export async function detectProvider(options: DetectOptions = {}): Promise<ProviderConfig | null> {
   const env = options.env ?? process.env;
   const ollamaBaseUrl = options.ollamaBaseUrl ?? OLLAMA_DEFAULT_URL;
@@ -69,14 +77,17 @@ export async function detectProvider(options: DetectOptions = {}): Promise<Provi
 
   const ollamaModels = await detectOllama(ollamaBaseUrl);
   if (ollamaModels && ollamaModels.length > 0) {
-    const model = ollamaModels[0];
-    return {
-      provider: 'ollama',
-      model,
-      baseUrl: ollamaBaseUrl,
-      autoDetected: true,
-      detectedAt: new Date().toISOString(),
-    };
+    const model = pickOllamaChatModel(ollamaModels);
+    if (model) {
+      return {
+        provider: 'ollama',
+        model,
+        baseUrl: ollamaBaseUrl,
+        autoDetected: true,
+        detectedAt: new Date().toISOString(),
+      };
+    }
+    return null;
   }
 
   if (ollamaModels === null) {
diff --git a/agent-refactor/tests/providerDetector.test.ts b/agent-refactor/tests/providerDetector.test.ts
index 95ecf63..e9a8fe3 100644
--- a/agent-refactor/tests/providerDetector.test.ts
+++ b/agent-refactor/tests/providerDetector.test.ts
@@ -5,6 +5,7 @@ import os from 'os';
 import {
   detectProvider,
   detectOllama,
+  pickOllamaChatModel,
   loadConfig,
   saveConfig,
   getDefaultModel,
@@ -60,6 +61,37 @@ describe('providerDetector', () => {
       }
     });
 
+    it('skips embedding models and picks a chat-capable Ollama model', async () => {
+      const originalFetch = globalThis.fetch;
+      globalThis.fetch = vi.fn().mockResolvedValue({
+        ok: true,
+        json: async () => ({ models: [{ name: 'nomic-embed-text:latest' }, { name: 'llama3' }] }),
+      } as any);
+      try {
+        const result = await detectProvider({ env: {} });
+        expect(result).not.toBeNull();
+        expect(result!.provider).toBe('ollama');
+        expect(result!.model).toBe('llama3');
+        expect(result!.baseUrl).toBe('http://localhost:11434');
+      } finally {
+        globalThis.fetch = originalFetch;
+      }
+    });
+
+    it('returns null when Ollama only has embedding models', async () => {
+      const originalFetch = globalThis.fetch;
+      globalThis.fetch = vi.fn().mockResolvedValue({
+        ok: true,
+        json: async () => ({ models: [{ name: 'nomic-embed-text:latest' }] }),
+      } as any);
+      try {
+        const result = await detectProvider({ env: {} });
+        expect(result).toBeNull();
+      } finally {
+        globalThis.fetch = originalFetch;
+      }
+    });
+
     it('returns null when nothing is available', async () => {
       const originalFetch = globalThis.fetch;
       globalThis.fetch = vi.fn().mockRejectedValue(new Error('Connection refused'));
@@ -144,6 +176,24 @@ describe('providerDetector', () => {
     });
   });
 
+  describe('pickOllamaChatModel', () => {
+    it('returns null for an empty list', () => {
+      expect(pickOllamaChatModel([])).toBeNull();
+    });
+
+    it('skips embedding models and picks the first chat model', () => {
+      expect(pickOllamaChatModel(['nomic-embed-text:latest', 'llama3'])).toBe('llama3');
+    });
+
+    it('returns null when all models are embeddings', () => {
+      expect(pickOllamaChatModel(['nomic-embed-text:latest', 'mxbai-embed-large'])).toBeNull();
+    });
+
+    it('returns the first non-embedding model', () => {
+      expect(pickOllamaChatModel(['qwen2.5-coder:1.5b', 'llama3'])).toBe('qwen2.5-coder:1.5b');
+    });
+  });
+
   describe('loadConfig', () => {
     let tmpDir: string;
 
diff --git a/agent-security-audit/src/providerDetector.ts b/agent-security-audit/src/providerDetector.ts
index 7cd3672..eedbf8e 100644
--- a/agent-security-audit/src/providerDetector.ts
+++ b/agent-security-audit/src/providerDetector.ts
@@ -27,6 +27,14 @@ const DEFAULT_MODELS: Record<string, string> = {
 
 const OLLAMA_DEFAULT_URL = 'http://localhost:11434';
 
+const EMBEDDING_MODEL_PATTERN =
+  /embed|minilm|mxbai|bge[-_]|gte[-_]|nomic|snowflake-arctic|e5[-_]|clip$/i;
+
+export function pickOllamaChatModel(models: string[]): string | null {
+  const chatModel = models.find((m) => !EMBEDDING_MODEL_PATTERN.test(m));
+  return chatModel ?? null;
+}
+
 export async function detectProvider(options: DetectOptions = {}): Promise<ProviderConfig | null> {
   const env = options.env ?? process.env;
   const ollamaBaseUrl = options.ollamaBaseUrl ?? OLLAMA_DEFAULT_URL;
@@ -69,14 +77,17 @@ export async function detectProvider(options: DetectOptions = {}): Promise<Provi
 
   const ollamaModels = await detectOllama(ollamaBaseUrl);
   if (ollamaModels && ollamaModels.length > 0) {
-    const model = ollamaModels[0];
-    return {
-      provider: 'ollama',
-      model,
-      baseUrl: ollamaBaseUrl,
-      autoDetected: true,
-      detectedAt: new Date().toISOString(),
-    };
+    const model = pickOllamaChatModel(ollamaModels);
+    if (model) {
+      return {
+        provider: 'ollama',
+        model,
+        baseUrl: ollamaBaseUrl,
+        autoDetected: true,
+        detectedAt: new Date().toISOString(),
+      };
+    }
+    return null;
   }
 
   if (ollamaModels === null) {
diff --git a/agent-security-audit/tests/providerDetector.test.ts b/agent-security-audit/tests/providerDetector.test.ts
index 95ecf63..e9a8fe3 100644
--- a/agent-security-audit/tests/providerDetector.test.ts
+++ b/agent-security-audit/tests/providerDetector.test.ts
@@ -5,6 +5,7 @@ import os from 'os';
 import {
   detectProvider,
   detectOllama,
+  pickOllamaChatModel,
   loadConfig,
   saveConfig,
   getDefaultModel,
@@ -60,6 +61,37 @@ describe('providerDetector', () => {
       }
     });
 
+    it('skips embedding models and picks a chat-capable Ollama model', async () => {
+      const originalFetch = globalThis.fetch;
+      globalThis.fetch = vi.fn().mockResolvedValue({
+        ok: true,
+        json: async () => ({ models: [{ name: 'nomic-embed-text:latest' }, { name: 'llama3' }] }),
+      } as any);
+      try {
+        const result = await detectProvider({ env: {} });
+        expect(result).not.toBeNull();
+        expect(result!.provider).toBe('ollama');
+        expect(result!.model).toBe('llama3');
+        expect(result!.baseUrl).toBe('http://localhost:11434');
+      } finally {
+        globalThis.fetch = originalFetch;
+      }
+    });
+
+    it('returns null when Ollama only has embedding models', async () => {
+      const originalFetch = globalThis.fetch;
+      globalThis.fetch = vi.fn().mockResolvedValue({
+        ok: true,
+        json: async () => ({ models: [{ name: 'nomic-embed-text:latest' }] }),
+      } as any);
+      try {
+        const result = await detectProvider({ env: {} });
+        expect(result).toBeNull();
+      } finally {
+        globalThis.fetch = originalFetch;
+      }
+    });
+
     it('returns null when nothing is available', async () => {
       const originalFetch = globalThis.fetch;
       globalThis.fetch = vi.fn().mockRejectedValue(new Error('Connection refused'));
@@ -144,6 +176,24 @@ describe('providerDetector', () => {
     });
   });
 
+  describe('pickOllamaChatModel', () => {
+    it('returns null for an empty list', () => {
+      expect(pickOllamaChatModel([])).toBeNull();
+    });
+
+    it('skips embedding models and picks the first chat model', () => {
+      expect(pickOllamaChatModel(['nomic-embed-text:latest', 'llama3'])).toBe('llama3');
+    });
+
+    it('returns null when all models are embeddings', () => {
+      expect(pickOllamaChatModel(['nomic-embed-text:latest', 'mxbai-embed-large'])).toBeNull();
+    });
+
+    it('returns the first non-embedding model', () => {
+      expect(pickOllamaChatModel(['qwen2.5-coder:1.5b', 'llama3'])).toBe('qwen2.5-coder:1.5b');
+    });
+  });
+
   describe('loadConfig', () => {
     let tmpDir: string;
 
diff --git a/agent-test-generator/src/providerDetector.ts b/agent-test-generator/src/providerDetector.ts
index 7cd3672..eedbf8e 100644
--- a/agent-test-generator/src/providerDetector.ts
+++ b/agent-test-generator/src/providerDetector.ts
@@ -27,6 +27,14 @@ const DEFAULT_MODELS: Record<string, string> = {
 
 const OLLAMA_DEFAULT_URL = 'http://localhost:11434';
 
+const EMBEDDING_MODEL_PATTERN =
+  /embed|minilm|mxbai|bge[-_]|gte[-_]|nomic|snowflake-arctic|e5[-_]|clip$/i;
+
+export function pickOllamaChatModel(models: string[]): string | null {
+  const chatModel = models.find((m) => !EMBEDDING_MODEL_PATTERN.test(m));
+  return chatModel ?? null;
+}
+
 export async function detectProvider(options: DetectOptions = {}): Promise<ProviderConfig | null> {
   const env = options.env ?? process.env;
   const ollamaBaseUrl = options.ollamaBaseUrl ?? OLLAMA_DEFAULT_URL;
@@ -69,14 +77,17 @@ export async function detectProvider(options: DetectOptions = {}): Promise<Provi
 
   const ollamaModels = await detectOllama(ollamaBaseUrl);
   if (ollamaModels && ollamaModels.length > 0) {
-    const model = ollamaModels[0];
-    return {
-      provider: 'ollama',
-      model,
-      baseUrl: ollamaBaseUrl,
-      autoDetected: true,
-      detectedAt: new Date().toISOString(),
-    };
+    const model = pickOllamaChatModel(ollamaModels);
+    if (model) {
+      return {
+        provider: 'ollama',
+        model,
+        baseUrl: ollamaBaseUrl,
+        autoDetected: true,
+        detectedAt: new Date().toISOString(),
+      };
+    }
+    return null;
   }
 
   if (ollamaModels === null) {
diff --git a/agent-test-generator/tests/providerDetector.test.ts b/agent-test-generator/tests/providerDetector.test.ts
index 95ecf63..e9a8fe3 100644
--- a/agent-test-generator/tests/providerDetector.test.ts
+++ b/agent-test-generator/tests/providerDetector.test.ts
@@ -5,6 +5,7 @@ import os from 'os';
 import {
   detectProvider,
   detectOllama,
+  pickOllamaChatModel,
   loadConfig,
   saveConfig,
   getDefaultModel,
@@ -60,6 +61,37 @@ describe('providerDetector', () => {
       }
     });
 
+    it('skips embedding models and picks a chat-capable Ollama model', async () => {
+      const originalFetch = globalThis.fetch;
+      globalThis.fetch = vi.fn().mockResolvedValue({
+        ok: true,
+        json: async () => ({ models: [{ name: 'nomic-embed-text:latest' }, { name: 'llama3' }] }),
+      } as any);
+      try {
+        const result = await detectProvider({ env: {} });
+        expect(result).not.toBeNull();
+        expect(result!.provider).toBe('ollama');
+        expect(result!.model).toBe('llama3');
+        expect(result!.baseUrl).toBe('http://localhost:11434');
+      } finally {
+        globalThis.fetch = originalFetch;
+      }
+    });
+
+    it('returns null when Ollama only has embedding models', async () => {
+      const originalFetch = globalThis.fetch;
+      globalThis.fetch = vi.fn().mockResolvedValue({
+        ok: true,
+        json: async () => ({ models: [{ name: 'nomic-embed-text:latest' }] }),
+      } as any);
+      try {
+        const result = await detectProvider({ env: {} });
+        expect(result).toBeNull();
+      } finally {
+        globalThis.fetch = originalFetch;
+      }
+    });
+
     it('returns null when nothing is available', async () => {
       const originalFetch = globalThis.fetch;
       globalThis.fetch = vi.fn().mockRejectedValue(new Error('Connection refused'));
@@ -144,6 +176,24 @@ describe('providerDetector', () => {
     });
   });
 
+  describe('pickOllamaChatModel', () => {
+    it('returns null for an empty list', () => {
+      expect(pickOllamaChatModel([])).toBeNull();
+    });
+
+    it('skips embedding models and picks the first chat model', () => {
+      expect(pickOllamaChatModel(['nomic-embed-text:latest', 'llama3'])).toBe('llama3');
+    });
+
+    it('returns null when all models are embeddings', () => {
+      expect(pickOllamaChatModel(['nomic-embed-text:latest', 'mxbai-embed-large'])).toBeNull();
+    });
+
+    it('returns the first non-embedding model', () => {
+      expect(pickOllamaChatModel(['qwen2.5-coder:1.5b', 'llama3'])).toBe('qwen2.5-coder:1.5b');
+    });
+  });
+
   describe('loadConfig', () => {
     let tmpDir: string;