diff --git a/e2e-tests/helpers/test_helper.ts b/e2e-tests/helpers/test_helper.ts
index 30e3045..4fd0438 100644
--- a/e2e-tests/helpers/test_helper.ts
+++ b/e2e-tests/helpers/test_helper.ts
@@ -73,6 +73,13 @@ class PageObject {
     await this.page.getByText("test-model").click();
   }
 
+  async selectTestOllamaModel() {
+    await this.page.getByRole("button", { name: "Model: Auto" }).click();
+    await this.page.getByText("Local models").click();
+    await this.page.getByText("Ollama", { exact: true }).click();
+    await this.page.getByText("Testollama", { exact: true }).click();
+  }
+
   async setUpTestProvider() {
     await this.page.getByText("Add custom providerConnect to").click();
     // Fill out provider dialog
@@ -201,6 +208,7 @@ export const test = base.extend<{
       const latestBuild = findLatestBuild();
       // parse the directory and find paths and other info
       const appInfo = parseElectronApp(latestBuild);
+      process.env.OLLAMA_HOST = "http://localhost:3500/ollama";
       process.env.E2E_TEST_BUILD = "true";
       // This is just a hack to avoid the AI setup screen.
       process.env.OPENAI_API_KEY = "sk-test";
diff --git a/e2e-tests/ollama.spec.ts b/e2e-tests/ollama.spec.ts
new file mode 100644
index 0000000..7a4712e
--- /dev/null
+++ b/e2e-tests/ollama.spec.ts
@@ -0,0 +1,7 @@
+import { test } from "./helpers/test_helper";
+
+test("send message to ollama", async ({ po }) => {
+  await po.selectTestOllamaModel();
+  await po.sendPrompt("hi");
+  await po.snapshotMessages();
+});
diff --git a/e2e-tests/snapshots/ollama.spec.ts_send-message-to-ollama-1.aria.yml b/e2e-tests/snapshots/ollama.spec.ts_send-message-to-ollama-1.aria.yml
new file mode 100644
index 0000000..c26e957
--- /dev/null
+++ b/e2e-tests/snapshots/ollama.spec.ts_send-message-to-ollama-1.aria.yml
@@ -0,0 +1,4 @@
+- paragraph: hi
+- paragraph: ollamachunkollamachunk
+- button "Retry":
+  - img
\ No newline at end of file
diff --git a/src/ipc/handlers/local_model_ollama_handler.ts b/src/ipc/handlers/local_model_ollama_handler.ts
index 8b9c9cf..c231ce6 100644
--- a/src/ipc/handlers/local_model_ollama_handler.ts
+++ b/src/ipc/handlers/local_model_ollama_handler.ts
@@ -4,7 +4,7 @@ import { LocalModelListResponse, LocalModel } from "../ipc_types";
 
 const logger = log.scope("ollama_handler");
 
-const OLLAMA_API_URL = "http://localhost:11434";
+const OLLAMA_API_URL = process.env.OLLAMA_HOST || "http://localhost:11434";
 
 interface OllamaModel {
   name: string;
diff --git a/src/ipc/utils/get_model_client.ts b/src/ipc/utils/get_model_client.ts
index d04f1f8..0293bc4 100644
--- a/src/ipc/utils/get_model_client.ts
+++ b/src/ipc/utils/get_model_client.ts
@@ -246,7 +246,7 @@ function getRegularModelClient(
     case "ollama": {
       // Ollama typically runs locally and doesn't require an API key in the same way
       const provider = createOllama({
-        baseURL: providerConfig.apiBaseUrl,
+        baseURL: process.env.OLLAMA_HOST,
       });
       return {
         modelClient: {
diff --git a/testing/fake-llm-server/index.ts b/testing/fake-llm-server/index.ts
index 759d5ec..b8668d7 100644
--- a/testing/fake-llm-server/index.ts
+++ b/testing/fake-llm-server/index.ts
@@ -56,8 +56,89 @@ app.get("/health", (req, res) => {
   res.send("OK");
 });
 
+// Ollama-specific endpoints
+app.get("/ollama/api/tags", (req, res) => {
+  const ollamaModels = {
+    models: [
+      {
+        name: "testollama",
+        modified_at: "2024-05-01T10:00:00.000Z",
+        size: 4700000000,
+        digest: "abcdef123456",
+        details: {
+          format: "gguf",
+          family: "llama",
+          families: ["llama"],
+          parameter_size: "8B",
+          quantization_level: "Q4_0",
+        },
+      },
+      {
+        name: "codellama:7b",
+        modified_at: "2024-04-25T12:30:00.000Z",
+        size: 3800000000,
+        digest: "fedcba654321",
+        details: {
+          format: "gguf",
+          family: "llama",
+          families: ["llama", "codellama"],
+          parameter_size: "7B",
+          quantization_level: "Q5_K_M",
+        },
+      },
+    ],
+  };
+  console.log("* Sending fake Ollama models");
+  res.json(ollamaModels);
+});
+
 let globalCounter = 0;
 
+app.post("/ollama/chat", (req, res) => {
+  // Tell the client we’re going to stream NDJSON
+  res.setHeader("Content-Type", "application/x-ndjson");
+  res.setHeader("Cache-Control", "no-cache");
+
+  // Chunk #1 – partial answer
+  const firstChunk = {
+    model: "llama3.2",
+    created_at: "2023-08-04T08:52:19.385406455-07:00",
+    message: {
+      role: "assistant",
+      content: "ollamachunk",
+      images: null,
+    },
+    done: false,
+  };
+
+  // Chunk #2 – final answer + metrics
+  const secondChunk = {
+    model: "llama3.2",
+    created_at: "2023-08-04T19:22:45.499127Z",
+    message: {
+      role: "assistant",
+      content: "",
+    },
+    done: true,
+    total_duration: 4883583458,
+    load_duration: 1334875,
+    prompt_eval_count: 26,
+    prompt_eval_duration: 342546000,
+    eval_count: 282,
+    eval_duration: 4535599000,
+  };
+
+  // Send the first object right away
+  res.write(JSON.stringify(firstChunk) + "\n");
+  res.write(JSON.stringify(firstChunk) + "\n");
+
+  // …and the second one a moment later to mimic streaming
+  setTimeout(() => {
+    res.write(JSON.stringify(secondChunk) + "\n");
+    res.end(); // Close the HTTP stream
+  }, 300); // 300 ms delay – tweak as you like
+});
+
 // Handle POST requests to /v1/chat/completions
 app.post("/v1/chat/completions", (req, res) => {
   const { stream = false, messages = [] } = req.body;
@@ -188,7 +269,7 @@ app.post("/v1/chat/completions", (req, res) => {
       clearInterval(interval);
       res.end();
     }
-  }, 1);
+  }, 10);
 });
 
 // Start the server