feature: improve base script (#18)

README.md

@@ -1,6 +1,6 @@
-<h1 align="center"><big>Fail1</big></h1>
+<h1 align="center"><big>fail1</big></h1>
 
-<p align="center"><img src="assets/logo.png" alt="logo"/></p>
+<p align="center"><img src="assets/logo.png" alt="logo" width="200"/></p>
 
 > This project is built to fail 
 > (until it doesn't)
@@ -81,7 +81,7 @@ OPENAI_API_KEY=your_api_key_here
 To start the code generation process, run the following command:
 
 ```shell
-node generation-000.js -G "<goal>" -g <generations> -p "<persona>" -t <temperature> -c -m "<model>"
+node generation-000.js -G "<goal>" -g <generations> -p "<persona>" -t <temperature> -c -m "<model>" -n "<negative_prompt>"
 ```
 
 ## Options
@@ -94,6 +94,7 @@ node generation-000.js -G "<goal>" -g <generations> -p "<persona>" -t <temperatu
 | `--temperature` | `-t` | `number` | `0.2` | Sets the temperature for the generated code |
 | `--clean` | `-c` | `boolean` | `false` | Set to `true` if you want to remove any previously generated code |
 | `--model` | `-m` | `string` | `"gpt-3.5-turbo"` | Sets the model to use for generating the code |
+| `--negative` | `-n` | `string` | | Sets the negative prompt for the generated code |
 
 ## Functionality
 

base.js

@@ -45,6 +45,10 @@ Examples
  alias: "p",
  default: "expert node.js developer, creative, code optimizer, interaction expert",
  },
+ neg: {
+ type: "string",
+ alias: "n",
+ },
  temperature: {
  type: "number",
  alias: "t",
@@ -73,51 +77,67 @@ const configuration = new Configuration({
 export const openai = new OpenAIApi(configuration);
 
 const instructions = `
-The code should ALWAYS be IMPROVED or EXTENDED or REFACTORED or FIXED
-be creative and add new features
-The GOAL must be completed
-
-GOAL: ${flags.goal}
+GOAL: ${flags.goal}${
+ flags.neg
+ ? `, ${flags.neg
+ .split(",")
+ .map(neg => `no ${neg.trim()}`)
+ .join(", ")}`
+ : ""
+}
 
+RULES:
+The code should ALWAYS be EXTENDED or REFACTORED or FIXED
+The GOAL must be completed
 increment the generation constant ONCE per generation
-Keep track of changes, EXTEND the CHANGELOG
+EXTEND the CHANGELOG
 NEVER use external apis with secret or key
-ONLY use es module syntax (with imports)
+EXCLUSIVELY use esm (imports)
 NEVER explain anything
-ALWAYS output ONLY JavaScript
+NEVER output markdown
+EXCLUSIVELY output JavaScript
+EVERYTHING happens in one file
+VERY IMPORTANT: the entire answer has to be valid JavaScript
 `;
 
 const history = [];
 
 export const generations = flags.generations;
-const maxSpawns = 3;
-let spawns = maxSpawns;
 let run = 0;
 
+/**
+ *
+ * @param {number} generation
+ * @returns {Promise<void>}
+ */
 export async function evolve(generation) {
  if (flags.help) {
  return;
  }
- if (spawns <= 0) {
- spinner.fail("Maximum retries reached");
- return;
- }
+
  const nextGeneration = generation + 1;
+ spinner.start(`Evolution ${generation} -> ${nextGeneration}`);
 
  try {
  const filename = buildFilename(generation);
  const code = await fs.readFile(filename, "utf-8");
- spinner.start(`Generation ${generation} | ${spawns} spawns left`);
+
+ // Reduce history length
+ history.shift();
+ history.shift();
 
  if (flags.clean) {
  // Remove all older generations
- const files = (await globby(["generation-*.js", "!generation-000.js"])).filter(
- file => file > buildFilename(generation)
- );
+ const files = (
+ await globby(["generation-*.js", "generation-*.md", "!generation-000.js"])
+ ).filter(file => file > buildFilename(generation));
  await Promise.all(files.map(async file => await fs.unlink(file)));
  }
 
  if (run === 0) {
+ const promptFilename = buildPromptFilename(generation);
+ await fs.writeFile(promptFilename, buildPrompt(flags.goal, flags.neg));
+
  history.push(
  {
  role: "user",
@@ -129,12 +149,14 @@ export async function evolve(generation) {
  }
  );
  }
+
  run++;
+
  history.push({
  role: "user",
  content: "continue the code",
  });
- spinner.start(`Evolution ${generation} -> ${generation + 1}`);
+
  const completion = await openai.createChatCompletion({
  // model: "gpt-4",
  model: flags.model,
@@ -148,47 +170,121 @@ export async function evolve(generation) {
  max_tokens: 2048,
  temperature: flags.temperature,
  });
- spinner.stop();
+
  const { content } = completion.data.choices[0].message;
- const cleanContent = content
+
+ // Clean GPT output (might return code block)
+ const cleanContent = minify(content)
  .replace("```javascript", "")
  .replace("```js", "")
- .replace("```", "");
+ .replace("```", "")
+ .trim();
+
+ // Code should start with a comment (changelog). If it doesn't it is often not JavaScrip but
+ // a human language response
  if (cleanContent.startsWith("/*")) {
  history.push({
  role: "assistant",
  content: cleanContent,
  });
+
  const nextFilename = buildFilename(nextGeneration);
  await fs.writeFile(nextFilename, prettify(cleanContent));
- spinner.succeed(`Evolution ${generation} -> ${generation + 1}`);
+
+ spinner.succeed(`Evolution ${generation} -> ${nextGeneration}`);
+
  await import(`./${nextFilename}`);
  } else {
- throw new Error("NOT_JAVASCRIPT");
+ spinner.fail(`Evolution ${generation} -> ${nextGeneration}`);
+ await handleError(new Error("NOT_JAVASCRIPT"), generation);
  }
  } catch (error) {
- spawns--;
- spinner.fail(`Evolution ${generation} -> ${generation + 1}`);
+ spinner.fail(`Evolution ${generation} -> ${nextGeneration}`);
  await handleError(error, generation);
  }
 }
 
+/**
+ * Pads the given number or string with zeros to a length of 3 characters.
+ *
+ * @param {number} n - The input number or string to be padded.
+ * @returns {string} - The padded string.
+ */
 export function pad(n) {
  return n.toString().padStart(3, "0");
 }
 
+/**
+ * Builds a filename string for the given generation number.
+ *
+ * @param {number} currentGeneration - The input generation number.
+ * @returns {string} - The generated filename string.
+ */
 export function buildFilename(currentGeneration) {
  return path.join(".", `generation-${pad(currentGeneration)}.js`);
 }
 
+/**
+ * Builds a prompt filename string for the given generation number.
+ *
+ * @param {number} currentGeneration - The input generation number.
+ * @returns {string} - The generated prompt filename string.
+ */
+export function buildPromptFilename(currentGeneration) {
+ return path.join(".", `generation-${pad(currentGeneration)}.md`);
+}
+
+/**
+ * Builds a formatted string combining the given prompt and optional negativePrompt.
+ *
+ * @param {string} prompt - The main prompt to be included in the output.
+ * @param {string} [negativePrompt] - The optional negative prompt to be included in the output.
+ * @returns {string} - The formatted string combining the prompts.
+ */
+export function buildPrompt(prompt, negativePrompt = "") {
+ return `# Configuration
+
+## Prompt
+
+\`\`\`shell
+${prompt}
+\`\`\`
+
+## Negative Prompt
+
+\`\`\`shell
+${negativePrompt}
+\`\`\`
+`;
+}
+
+/**
+ * Minifies the given code string by removing leading whitespace.
+ *
+ * @param {string} code - The input code to be minified.
+ * @returns {string} - The minified code.
+ */
 export function minify(code) {
- return code.replace(/^\s+/gim, "");
+ return code.replace(/^\s+/g, "");
 }
 
+/**
+ * Prettifies the given code string using Prettier.
+ *
+ * @param {string} code - The input code to be prettified.
+ * @returns {string} - The prettified code.
+ */
 export function prettify(code) {
  return prettier.format(code, { semi: false, parser: "babel" });
 }
 
+/**
+ * Handles errors that occur during the code generation process.
+ *
+ * @param {Error} error - The error object containing information about the error.
+ * @param {number} generation - The current generation number.
+ * @returns {Promise<void>} - A promise that resolves when the error is handled.
+ */
 export async function handleError(error, generation) {
  const message = (
  error.response?.data?.error.message ??