Open-source, secure environment with real-world tools for enterprise-grade agents.
Python agentaiai-agentai-agentscode-interpretercopilotdevelopmentdevtoolsgptgpt-4javascriptllmnextjsopenaipythonreactsoftwaretypescript
fix(docs): Consistent order for JS & PY examples in guide
Pavel 'Strajk' Dolecek57401b922150
diff --git a/apps/docs/src/app/guide/simple-gpt4-code-interpreter/page.mdx b/apps/docs/src/app/guide/simple-gpt4-code-interpreter/page.mdxindex 3d221cf4a..c78a73dcf 100644--- a/apps/docs/src/app/guide/simple-gpt4-code-interpreter/page.mdx+++ b/apps/docs/src/app/guide/simple-gpt4-code-interpreter/page.mdx@@ -44,16 +44,15 @@ E2B allows you to execute the GPT-generated code in a sandboxed cloud environmenFirst, install the required packages. We need to install the E2B package and OpenAI package.<CodeGroup isRunnable={false}>-```bash {{ language: 'python' }}-pip install e2b-pip install openai-```-```bash {{ language: 'js' }}npm i @e2b/sdknpm i openai```+```bash {{ language: 'python' }}+pip install e2b+pip install openai+```</CodeGroup>## Get API keys@@ -68,6 +67,11 @@ Let's get our API keys for E2B and OpenAI. We'll need them to use the APIs.Now we need to import the packages we installed in the previous step.<CodeGroup isRunnable={false}>+```js+import e2b from '@e2b/sdk'+import OpenAI from 'openai'+```+```pythonimport e2bimport openai@@ -76,12 +80,6 @@ import openaiimport asyncioimport json```--```js-import e2b from '@e2b/sdk'-import OpenAI from 'openai'-```-</CodeGroup>## Make calls to GPT@@ -89,15 +87,6 @@ import OpenAI from 'openai'Here we set up the OpenAI chat completion endpoint that allows to call the GPT-4 (or GPT-3.5) models.<CodeGroup isRunnable={false}>-```python-response = openai.ChatCompletion.create(- model="gpt-4", # Or use gpt-3.5-turbo- messages=[- {"role": "system", "content": "You are a senior developer that can code in JavaScript. Always produce valid JSON."},- ]-)-```-```jsconst openai = new OpenAI()const chatCompletion = await openai.chat.completions.create({@@ -106,11 +95,22 @@ const chatCompletion = await openai.chat.completions.create({{role: 'system',content: 'You are a senior developer that can code in JavaScript. Always produce valid JSON.',- },+ }],})```+```python+response = openai.ChatCompletion.create(+ model="gpt-4", # Or use gpt-3.5-turbo+ messages=[+ {+ "role": "system",+ "content": "You are a senior developer that can code in JavaScript. Always produce valid JSON."+ },+ ]+)+```</CodeGroup>If you print the response, you should see something like this:@@ -144,25 +144,6 @@ If you print the response, you should see something like this:We're going to take an advantage of the [OpenAI functions](https://openai.com/blog/function-calling-and-other-api-updates). Defining these functions will make it easier instructing the model that it can write JavaScript code to complete request from the user. We'll execute this code later using E2B.<CodeGroup isRunnable={false}>-```python-functions = [- {- "name": "exec_code",- "description": "Executes the passed JavaScript code using Nodejs and returns the stdout and stderr.",- "parameters": {- "type": "object",- "properties": {- "code": {- "type": "string",- "description": "The JavaScript code to execute.",- },- },- "required": ["code"],- },- }-]-```-```jsconst functions = [{@@ -182,6 +163,24 @@ const functions = []```+```python+functions = [+ {+ "name": "exec_code",+ "description": "Executes the passed JavaScript code using Nodejs and returns the stdout and stderr.",+ "parameters": {+ "type": "object",+ "properties": {+ "code": {+ "type": "string",+ "description": "The JavaScript code to execute.",+ },+ },+ "required": ["code"],+ },+ }+]+```</CodeGroup>We created an OpenAI function `exec_code` that expects a single parameter `code`. The `code` parameter will be the Python code generated by GPT that we'll execute.@@ -190,19 +189,6 @@ Now we pass the `functions` variable to the GPT call we made earlier and also adThe new code is marked by the highlighted lines.<CodeGroup isRunnable={false}>-```python-response = openai.ChatCompletion.create(- model="gpt-4",- messages=[- {"role": "system", "content": "You are a senior developer that can code in JavaScript. Always produce valid JSON."},- {"role": "user", "content": "Write hello world"}, # $HighlightLine- {"role": "assistant", "content": "print(\"hello world\")", "name":"exec_code"}, # $HighlightLine- {"role": "user", "content": "Generate first 100 fibonacci numbers"}, # $HighlightLine- ],- functions=functions, # $HighlightLine-)-```-```jsconst chatCompletion = await openai.chat.completions.create({model: 'gpt-4',@@ -229,6 +215,18 @@ const chatCompletion = await openai.chat.completions.create({})```+```python+response = openai.ChatCompletion.create(+ model="gpt-4",+ messages=[+ {"role": "system", "content": "You are a senior developer that can code in JavaScript. Always produce valid JSON."},+ {"role": "user", "content": "Write hello world"}, # $HighlightLine+ {"role": "assistant", "content": "print(\"hello world\")", "name":"exec_code"}, # $HighlightLine+ {"role": "user", "content": "Generate first 100 fibonacci numbers"}, # $HighlightLine+ ],+ functions=functions, # $HighlightLine+)+```</CodeGroup>If you print the GPT response now, you'll most likely see the model is calling the `exec_code` function we defined earlier and is passing code for generating fibonacci numbers in the first element of the `choices` JSON array.@@ -259,29 +257,6 @@ If you print the GPT response now, you'll most likely see the model is calling tWe need to parse the GPT's response and detect if it contains a function call. If it does and it's actually the `exec_code` function, we need to parse the `arguments` field, get the `code` argument and actually run the code using E2B (we'll do it in the next step).<CodeGroup isRunnable={false}>-```python-# ... previous code--message = response["choices"][0]["message"]-if (func := message.get("function_call")):- func_name = func["name"]-- # Get rid of newlines and leading/trailing spaces in the raw function arguments JSON string.- # This sometimes help to avoid JSON parsing errors.- args = func["arguments"].strip().replace("\n", "")- # Parse the cleaned up JSON string.- func_args = json.loads(args)-- # If the model is calling the exec_code function we defined in the `functions` variable, we want to save the `code` argument to a variable.- if func_name == "exec_code":- code = func_args["code"]- # TODO: Execute the code using E2B.-else:- # The model didn't call a function, so we just print the message.- content = message["content"]- print(content)-```-```js// ... previous code@@ -309,7 +284,28 @@ if (func) {}```+```python+# ... previous code+message = response["choices"][0]["message"]+if (func := message.get("function_call")):+ func_name = func["name"]++ # Get rid of newlines and leading/trailing spaces in the raw function arguments JSON string.+ # This sometimes help to avoid JSON parsing errors.+ args = func["arguments"].strip().replace("\n", "")+ # Parse the cleaned up JSON string.+ func_args = json.loads(args)++ # If the model is calling the exec_code function we defined in the `functions` variable, we want to save the `code` argument to a variable.+ if func_name == "exec_code":+ code = func_args["code"]+ # TODO: Execute the code using E2B.+else:+ # The model didn't call a function, so we just print the message.+ content = message["content"]+ print(content)+```</CodeGroup>@@ -322,19 +318,6 @@ All we need to add is just a single line of code. I hgihtlighted it in the code<CodeGroup isRunnable={false}>-```python-# ... previous code--# If the model is calling the exec_code function we defined in the `functions` variable, we want to save the `code` argument to a variable.-if func_name == "exec_code":- code = func_args["code"]- stdout, stderr = await e2b.run_code("Node16", code) # $HighlightLine- print(stdout) # $HighlightLine- print(stderr) # $HighlightLine--# ... rest of the code-```-```js// ... previous code@@ -350,6 +333,18 @@ if (funcName === "exec_code") {// ... rest of the code```+```python+# ... previous code++# If the model is calling the exec_code function we defined in the `functions` variable, we want to save the `code` argument to a variable.+if func_name == "exec_code":+ code = func_args["code"]+ stdout, stderr = await e2b.run_code("Node16", code) # $HighlightLine+ print(stdout) # $HighlightLine+ print(stderr) # $HighlightLine++# ... rest of the code+```</CodeGroup>And this what the executed code for "Generate first 100 fibonacci numbers" prompt prints:@@ -400,39 +395,6 @@ We're going to solve this problem by letting GPT install missing libraries by adThis is how our new `functions` variable looks like. The highlighted lines are the new code.<CodeGroup isRunnable={false}>-```python-functions = [- {- "name": "exec_code",- "description": "Executes the passed JavaScript code using Nodejs and returns the stdout and stderr.",- "parameters": {- "type": "object",- "properties": {- "code": {- "type": "string",- "description": "The JavaScript code to execute.",- },- },- "required": ["code"],- },- },- { # $HighlightLine- "name": "install_package", # $HighlightLine- "description": "Installs the passed npm package.", # $HighlightLine- "parameters": { # $HighlightLine- "type": "object", # $HighlightLine- "properties": { # $HighlightLine- "name": { # $HighlightLine- "type": "string", # $HighlightLine- "description": "The name of an npm package to install.", # $HighlightLine- }, # $HighlightLine- }, # $HighlightLine- "required": ["name"], # $HighlightLine- }, # $HighlightLine- } # $HighlightLine-]-```-```jsconst functions = [{@@ -466,6 +428,38 @@ const functions = [}, // $HighlightLine];```+```python+functions = [+ {+ "name": "exec_code",+ "description": "Executes the passed JavaScript code using Nodejs and returns the stdout and stderr.",+ "parameters": {+ "type": "object",+ "properties": {+ "code": {+ "type": "string",+ "description": "The JavaScript code to execute.",+ },+ },+ "required": ["code"],+ },+ },+ { # $HighlightLine+ "name": "install_package", # $HighlightLine+ "description": "Installs the passed npm package.", # $HighlightLine+ "parameters": { # $HighlightLine+ "type": "object", # $HighlightLine+ "properties": { # $HighlightLine+ "name": { # $HighlightLine+ "type": "string", # $HighlightLine+ "description": "The name of an npm package to install.", # $HighlightLine+ }, # $HighlightLine+ }, # $HighlightLine+ "required": ["name"], # $HighlightLine+ }, # $HighlightLine+ } # $HighlightLine+]+```</CodeGroup>Now we need to add a new `if` statement to our code that will detect if the model is calling the `install_package` function and if it does, we need to install the package using E2B.@@ -482,4 +476,4 @@ TODO## Final code-You can find the final code for both Python and JavaScript in [this GitHub repository](https://github.com/e2b-dev/e2b-cookbook/tree/main). {{ className: 'lead' }}\ No newline at end of file+You can find the final code for both Python and JavaScript in [this GitHub repository](https://github.com/e2b-dev/e2b-cookbook/tree/main). {{ className: 'lead' }}