{"id":3632,"date":"2023-12-10T18:12:58","date_gmt":"2023-12-11T02:12:58","guid":{"rendered":"https:\/\/devblogs.microsoft.com\/surface-duo\/?p=3632"},"modified":"2024-01-03T16:04:35","modified_gmt":"2024-01-04T00:04:35","slug":"android-openai-chatgpt-28","status":"publish","type":"post","link":"https:\/\/devblogs.microsoft.com\/surface-duo\/android-openai-chatgpt-28\/","title":{"rendered":"OpenAI Assistant code interpreter on Android"},"content":{"rendered":"

\n Hello prompt engineers,\n<\/p>\n

\n Over the last few weeks, we\u2019ve looked at different aspects of the new OpenAI Assistant API, both prototyping in the playground<\/a> and using Kotlin in the JetchatAI sample<\/a>. In this post we\u2019re going to add the Code Interpreter<\/a> feature which allows the Assistants API to write and run Python code in a sandboxed execution environment. By using the code interpreter, chat interactions can solve complex math problems, code problems, read and parse data files, and output formatted data files and charts.\n<\/p>\n

\n To keep with the theme of the last few examples, we are going to test the code interpreter with a simple math problem related to the fictitious Contoso health plans used in earlier posts.\n<\/p>\n

Enabling the code interpreter<\/h2>\n

\n The code interpreter is just a setting to be enabled, either in code or via the playground (depending on where you have set up your assistant). Figure 1 shows the Kotlin for creating an assistant using the code interpreter \u2013 including setting a very basic system prompt\/meta prompt\/instructions:\n<\/p>\n

val assistant = openAI.assistant(\r\n    request = AssistantRequest(\r\n        name = \"doc chat\",\r\n        instructions = \"answer questions about health plans\",\r\n        tools = listOf(AssistantTool.CodeInterpreter), \/\/ enables the code interpreter\r\n        model = ModelId(\"gpt-4-1106-preview\")\r\n    )\r\n)<\/pre>\n
Figure 1: enabling the code interpreter in Kotlin when creating an assistant<\/code><\/em>\n<\/p>\n
\n  As discussed in the first Kotlin assistant post<\/a>, JetchatAI<\/em> loads an assistant definition that was configured in the OpenAI playground, so it\u2019s even easier to enable the Code interpreter<\/strong> by flipping this switch:\n<\/p>\n
\n  \n<\/p>\n
Figure 2: enabling the code interpreter in the OpenAI playground<\/em>\n<\/p>\n
Extending JetchatAI #assistant-chat<\/h2>\n
\n  With the code interpreter enabled, we can test it both interactively in the playground and in the JetchatAI<\/em> app. The test question is \u201cif the health plan costs $1000 a month, what is deducted weekly from my paycheck?\u201d<\/strong>.\n<\/p>\n
\n  The playground output includes the code that was generated and executed in the interpreter \u2013 in Figure 3 you can see it creates variables from the values mentioned in the query and then a calculation that returns a value to the model, which is incorporated into the response to the user.\n<\/p>\n
\n  \"ScreenshotFigure 3: Testing the code interpreter in the playground with a simple math question (using GPT-4)<\/em>\n<\/p>\n
\n  Notice that the code interpreter just returns a numeric answer, and the model decides to \u2018round up\u2019 to \u201c$230.77<\/strong>\u201d and format as a dollar amount.\n<\/p>\n
\n  When implementing the Assistant API in an app, the code interpreter step (or steps) would not be rendered (in the same way that you don\u2019t render function call responses), although they are available in the run\u2019s step_details<\/code> data structure so you could still retrieve and display them to the user, or use for logging\/telemetry or some other purpose in your app.\n<\/p>\n
\n  \"\"Figure 4: Testing the same query on Android (also using GPT-4)<\/em><\/p>\n
No code changes were made to the JetchatAI Android sample<\/a> for this testing, it just needs the Assistant’s configuration updated as shown.<\/p>\n<\/p>\n
Why add the code interpreter?<\/h2>\n
\n  One of the reasons that the code interpreter option exists is because LLMs \u201con their own\u201d can be terribly bad at math<\/a>. Here are some examples using the same prompt on different models without a code interpreter to help:\n<\/p>\n