diff --git a/.prettierrc.js b/.prettierrc.js
index d2eb489..44838e8 100644
--- a/.prettierrc.js
+++ b/.prettierrc.js
@@ -2,5 +2,21 @@ module.exports = {
   trailingComma: "es5",
   tabWidth: 2,
   semi: true,
-  singleQuote: false
-};
+  singleQuote: false,
+  arrowParens: "always",
+  bracketSameLine: false,
+  bracketSpacing: true,
+  experimentalTernaries: false,
+  jsxSingleQuote: false,
+  quoteProps: "as-needed",
+  singleAttributePerLine: false,
+  htmlWhitespaceSensitivity: "css",
+  vueIndentScriptAndStyle: false,
+  proseWrap: "preserve",
+  insertPragma: false,
+  printWidth: 80,
+  requirePragma: false,
+  useTabs: false,
+  embeddedLanguageFormatting: "auto",
+  spaceBeforeFunctionParen: false
+};
\ No newline at end of file
diff --git a/README.md b/README.md
index 5685d77..3b2d81f 100644
--- a/README.md
+++ b/README.md
@@ -22,6 +22,7 @@ Some of the popular LLMs that we recommend are:
 
 - [Mistral](https://mistral.ai/)
 - [CodeLLama](https://github.com/facebookresearch/codellama)
+- [DeepSeek Coder](https://github.com/deepseek-ai/DeepSeek-Coder)
 
 ## Quick Install
 
diff --git a/lib/extension/package.json b/lib/extension/package.json
index 5d49775..1c63c53 100644
--- a/lib/extension/package.json
+++ b/lib/extension/package.json
@@ -15,7 +15,7 @@
     "@privy/common": "*",
     "handlebars": "4.7.8",
     "marked": "4.2.12",
-    "modelfusion": "0.113.0",
+    "modelfusion": "0.122.0",
     "secure-json-parse": "2.7.0",
     "simple-git": "3.21.0",
     "zod": "3.22.4"
diff --git a/lib/extension/src/ai/AIClient.ts b/lib/extension/src/ai/AIClient.ts
index 0a2d278..9833dea 100644
--- a/lib/extension/src/ai/AIClient.ts
+++ b/lib/extension/src/ai/AIClient.ts
@@ -1,5 +1,4 @@
 import {
-  Llama2Prompt,
   OpenAITextEmbeddingResponse,
   InstructionPrompt,
   TextStreamingModel,
@@ -9,7 +8,6 @@ import {
   ollama,
   openai,
   streamText,
-  MistralInstructPrompt,
 } from "modelfusion";
 import * as vscode from "vscode";
 import { z } from "zod";
@@ -30,14 +28,14 @@ function getProviderBaseUrl(): string {
   );
 }
 
-function getModel() {
-  return z
+function getModel(): string {
+  let model = z
     .enum(["mistral:instruct", "codellama:instruct", "custom"])
     .parse(vscode.workspace.getConfiguration("privy").get("model"));
-}
-
-function getCustomModel(): string {
-  return vscode.workspace.getConfiguration("privy").get("customModel", "");
+  if (model === "custom") {
+    return vscode.workspace.getConfiguration("privy").get("customModel", "");
+  }
+  return model;
 }
 
 function getProvider() {
@@ -46,6 +44,17 @@ function getProvider() {
     .parse(vscode.workspace.getConfiguration("privy").get("provider"));
 }
 
+function getPromptTemplate() {
+  const model = getModel();
+  if (model.startsWith("mistral")) {
+    return ollama.prompt.Mistral;
+  } else if (model.startsWith("deepseek")) {
+    return ollama.prompt.Text;
+  }
+
+  return ollama.prompt.Llama2;
+}
+
 export class AIClient {
   private readonly apiKeyManager: ApiKeyManager;
   private readonly logger: Logger;
@@ -78,24 +87,29 @@ export class AIClient {
     stop?: string[] | undefined;
     temperature?: number | undefined;
   }): Promise<TextStreamingModel<InstructionPrompt>> {
-    const modelConfiguration = getModel();
     const provider = getProvider();
 
     if (provider.startsWith("llama")) {
       return llamacpp
-        .TextGenerator({
+        .CompletionTextGenerator({
           api: await this.getProviderApiConfiguration(),
+          // TODO the prompt format needs to be configurable for non-Llama2 models
+          promptTemplate: llamacpp.prompt.Llama2,
           maxGenerationTokens: maxTokens,
           stopSequences: stop,
           temperature,
         })
-        .withTextPromptTemplate(Llama2Prompt.instruction());
+        .withInstructionPrompt();
     }
 
     return ollama
-      .ChatTextGenerator({
+      .CompletionTextGenerator({
         api: await this.getProviderApiConfiguration(),
-        model: getModel() === "custom" ? getCustomModel() : getModel(),
+        promptTemplate: getPromptTemplate(),
+        model: getModel(),
+        maxGenerationTokens: maxTokens,
+        stopSequences: stop,
+        temperature,
       })
       .withInstructionPrompt();
   }
@@ -112,28 +126,29 @@ export class AIClient {
     temperature?: number | undefined;
   }) {
     this.logger.log(["--- Start prompt ---", prompt, "--- End prompt ---"]);
-
-    return streamText(
-      await this.getTextStreamingModel({ maxTokens, stop, temperature }),
-      { instruction: prompt }
-    );
+    return streamText({
+      model: await this.getTextStreamingModel({ maxTokens, stop, temperature }),
+      prompt: {
+        instruction: prompt,
+      },
+    });
   }
 
   async generateEmbedding({ input }: { input: string }) {
     try {
-      const { embedding, response } = await embed(
-        openai.TextEmbedder({
+      const { embedding, rawResponse } = await embed({
+        model: openai.TextEmbedder({
           api: await this.getProviderApiConfiguration(),
           model: "text-embedding-ada-002",
         }),
-        input,
-        { fullResponse: true }
-      );
+        value: input,
+        fullResponse: true,
+      });
 
       return {
         type: "success" as const,
         embedding,
-        totalTokenCount: (response as OpenAITextEmbeddingResponse).usage
+        totalTokenCount: (rawResponse as OpenAITextEmbeddingResponse).usage
           .total_tokens,
       };
     } catch (error: any) {
diff --git a/package-lock.json b/package-lock.json
index a6d5e55..ff87dde 100644
--- a/package-lock.json
+++ b/package-lock.json
@@ -13,6 +13,7 @@
         "lib/*"
       ],
       "dependencies": {
+        "modelfusion": "^0.122.0",
         "pnpm": "^8.13.1"
       },
       "devDependencies": {
@@ -56,7 +57,7 @@
         "@privy/common": "*",
         "handlebars": "4.7.8",
         "marked": "4.2.12",
-        "modelfusion": "0.113.0",
+        "modelfusion": "0.122.0",
         "secure-json-parse": "2.7.0",
         "simple-git": "3.21.0",
         "zod": "3.22.4"
@@ -5553,9 +5554,9 @@
       }
     },
     "node_modules/modelfusion": {
-      "version": "0.113.0",
-      "resolved": "https://registry.npmjs.org/modelfusion/-/modelfusion-0.113.0.tgz",
-      "integrity": "sha512-yYv6xEZj5jjR/vIJh3M/orgEdFy5zzmTcfAkUCnnjxTm6Xngiu3aMG8XA0Xcqd4dQKvvs/ujpnt5azwV8Ddd6g==",
+      "version": "0.122.0",
+      "resolved": "https://registry.npmjs.org/modelfusion/-/modelfusion-0.122.0.tgz",
+      "integrity": "sha512-FMltMozwTStdjGW3r24pMLj3+Rxjsh9eceMh8b38Ava4HdHR4dbF2xOJOVeC/bmqJYaZHbIu2BWpu45rF+mOxA==",
       "dependencies": {
         "eventsource-parser": "1.1.1",
         "js-tiktoken": "1.0.7",
diff --git a/package.json b/package.json
index e8ac564..df997d2 100644
--- a/package.json
+++ b/package.json
@@ -34,6 +34,7 @@
     "lib/*"
   ],
   "dependencies": {
+    "modelfusion": "^0.122.0",
     "pnpm": "^8.13.1"
   }
 }
diff --git a/template/chat/chat-en.rdt.md b/template/chat/chat-en.rdt.md
index 69709df..24d7261 100644
--- a/template/chat/chat-en.rdt.md
+++ b/template/chat/chat-en.rdt.md
@@ -67,13 +67,6 @@ Developer: {{content}}
 {{/if}}
 {{/each}}
 
-## Task
-Write a response that continues the conversation.
-Stay focused on current developer request.
-Consider the possibility that there might not be a solution.
-Ask for clarification if the message does not make sense or more input is needed.
-Omit any links.
-Include code snippets (using Markdown) and examples where appropriate.
 
 ## Response
 Bot: