genaicode 0.0.34 → 0.0.37

package/README.md

@@ -1,3 +1,13 @@
+<p align="center">
+  <picture>
+    <source media="(prefers-color-scheme: dark)" srcset="media/logo-dark.png">
+    <source media="(prefers-color-scheme: light)" srcset="media/logo.png">
+    <img alt="GenAIcode Logo." src="media/logo.png" width="100%" height="auto">
+  </picture>
+</p>
+
+<div align="center">
+
 # Programming on steroids
 
   <a href="https://www.npmjs.com/package/genaicode">
@@ -69,6 +79,9 @@ GenAIcode supports various command-line options to customize its behavior:
 - `--imagen`: Enables image generation capabilities using AI models.
 - `--cheap`: Uses a cheaper, faster model for code generation, which may provide lower quality results but is more cost-effective for simpler tasks.
 - `--content-mask=<path>`: Applies a content mask to limit the initial source code files included in the request. The value should be a prefix of the path relative to rootDir.
+- `--ignore-pattern="glob/regex"`: Specify a pattern of files to ignore during the initial source code fetching. This saves initial token usage.
+- `--ask-question`: Allows the AI assistant to ask questions for clarification during the code generation process.
+- `--disable-cache`: Disables caching for the application, which can be useful if caching is causing issues or if you want to ensure fresh data is used for each operation.
 - `--help`: Displays the help message with all available options.
 
 ## Configuration (.genaicoderc)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "genaicode",
-  "version": "0.0.34",
+  "version": "0.0.37",
   "author": "Grzegorz Tańczyk",
   "repository": {
     "type": "git",
@@ -43,12 +43,13 @@
     "vitest": "^2.0.4"
   },
   "dependencies": {
-    "@anthropic-ai/sdk": "^0.24.3",
+    "@anthropic-ai/sdk": "^0.26.1",
     "@anthropic-ai/vertex-sdk": "^0.4.1",
     "@google-cloud/aiplatform": "^3.25.0",
     "@google-cloud/vertexai": "^1.3.0",
     "@imgly/background-removal-node": "^1.4.5",
     "diff": "^5.2.0",
+    "glob-regex": "^0.3.2",
     "image-size": "^1.1.1",
     "jsonschema": "^1.4.1",
     "mime-types": "^2.1.35",

package/src/ai-service/anthropic.js

@@ -1,5 +1,6 @@
 import Anthropic from '@anthropic-ai/sdk';
 import { printTokenUsageAndCost, processFunctionCalls } from './common.js';
+import { disableCache } from '../cli/cli-params.js';
 
 /**
  * This function generates content using the Anthropic Claude model.
@@ -7,7 +8,7 @@ import { printTokenUsageAndCost, processFunctionCalls } from './common.js';
 export async function generateContent(prompt, functionDefs, requiredFunctionName, temperature, cheap = false) {
   const anthropic = new Anthropic({
     defaultHeaders: {
-      'anthropic-beta': 'max-tokens-3-5-sonnet-2024-07-15',
+      'anthropic-beta': 'max-tokens-3-5-sonnet-2024-07-15' + (!disableCache ? ',prompt-caching-2024-07-31' : ''),
     },
   });
 
@@ -35,6 +36,7 @@ export async function generateContent(prompt, functionDefs, requiredFunctionName
             {
               type: 'text',
               text: item.text,
+              ...(item.cache && !disableCache ? { cache_control: { type: 'ephemeral' } } : {}),
             },
           ],
         };
@@ -57,22 +59,51 @@ export async function generateContent(prompt, functionDefs, requiredFunctionName
   const model = cheap ? 'claude-3-haiku-20240307' : 'claude-3-5-sonnet-20240620';
   console.log(`Using Anthropic model: ${model}`);
 
-  const response = await anthropic.messages.create({
-    model: model,
-    system: prompt.find((item) => item.type === 'systemPrompt').systemPrompt,
-    messages,
-    tools: functionDefs.map((fd) => ({
-      name: fd.name,
-      description: fd.description,
-      input_schema: fd.parameters,
-    })),
-    tool_choice: requiredFunctionName ? { type: 'tool', name: requiredFunctionName } : { type: 'any' },
-    max_tokens: cheap ? 4096 : 8192,
-    temperature: temperature,
-  });
+  let retryCount = 0;
+  let response;
+  while (retryCount < 3) {
+    try {
+      response = await anthropic.beta.promptCaching.messages.create({
+        model: model,
+        system: prompt.find((item) => item.type === 'systemPrompt').systemPrompt,
+        messages,
+        tools: functionDefs.map((fd) => ({
+          name: fd.name,
+          description: fd.description,
+          input_schema: fd.parameters,
+        })),
+        tool_choice: requiredFunctionName ? { type: 'tool', name: requiredFunctionName } : { type: 'any' },
+        max_tokens: cheap ? 4096 : 8192,
+        temperature: temperature,
+      });
+      break; // Exit loop if successful
+    } catch (error) {
+      if (error.headers?.['retry-after']) {
+        let retryAfter;
+        if (error.headers['retry-after'] === '0') {
+          retryAfter = (new Date(error.headers['anthropic-ratelimit-tokens-reset']).getTime() - Date.now()) / 1000;
+        } else {
+          retryAfter = Math.max(parseInt(error.headers['retry-after'], 10), 10);
+        }
+        console.log(`Rate limited. Retrying after ${retryAfter} seconds. Attempt ${retryCount + 1} of 3.`);
+        await new Promise((resolve) => setTimeout(resolve, retryAfter * 1000));
+        retryCount++;
+      } else {
+        console.error('An error occurred:', error);
+        throw error; // Re-throw the error if it's not a rate limit error
+      }
+    }
+  }
+
+  if (retryCount === 3) {
+    console.error('Failed to complete request after 3 attempts due to rate limiting.');
+    throw new Error('Rate limit exceeded. Operation aborted.');
+  }
 
   // Print token usage for Anthropic
   const usage = {
+    cacheCreateTokens: response.usage.cache_creation_input_tokens,
+    cacheReadTokens: response.usage.cache_read_input_tokens,
     inputTokens: response.usage.input_tokens,
     outputTokens: response.usage.output_tokens,
     totalTokens: response.usage.input_tokens + response.usage.output_tokens,

package/src/ai-service/chat-gpt.js

@@ -58,13 +58,40 @@ export async function generateContent(prompt, functionDefs, requiredFunctionName
   const model = cheap ? 'gpt-4o-mini' : 'gpt-4o-2024-08-06';
   console.log(`Using OpenAI model: ${model}`);
 
-  const response = await openai.chat.completions.create({
-    model: model,
-    messages,
-    tools: functionDefs.map((funDef) => ({ type: 'function', function: funDef })),
-    tool_choice: requiredFunctionName ? { type: 'function', function: { name: requiredFunctionName } } : 'required',
-    temperature: temperature,
-  });
+  let retryCount = 0;
+  let response;
+  while (retryCount < 3) {
+    try {
+      response = await openai.chat.completions.create({
+        model: model,
+        messages,
+        tools: functionDefs.map((funDef) => ({ type: 'function', function: funDef })),
+        tool_choice: requiredFunctionName ? { type: 'function', function: { name: requiredFunctionName } } : 'required',
+        temperature: temperature,
+      });
+      break; // Exit loop if successful
+    } catch (error) {
+      if (error.response?.headers?.['x-ratelimit-limit-tokens']) {
+        const rateLimitTokens = parseInt(error.response.headers['x-ratelimit-limit-tokens'], 10);
+        const retryAfter = error.response.headers['retry-after']
+          ? parseInt(error.response.headers['retry-after'], 10)
+          : 1;
+        console.log(
+          `Rate limited. Token limit: ${rateLimitTokens}. Retrying after ${retryAfter} seconds. Attempt ${retryCount + 1} of 3.`,
+        );
+        await new Promise((resolve) => setTimeout(resolve, retryAfter * 1000));
+        retryCount++;
+      } else {
+        console.error('An error occurred:', error);
+        throw error; // Re-throw the error if it's not a rate limit error
+      }
+    }
+  }
+
+  if (retryCount === 3) {
+    console.error('Failed to complete request after 3 attempts due to rate limiting.');
+    throw new Error('Rate limit exceeded. Operation aborted.');
+  }
 
   // Print token usage for chat gpt
   const usage = {

package/src/ai-service/common.js

@@ -12,8 +12,17 @@ export function printTokenUsageAndCost(usage, inputCostPerToken, outputCostPerTo
   console.log('  - Input tokens: ', usage.inputTokens);
   console.log('  - Output tokens: ', usage.outputTokens);
   console.log('  - Total tokens: ', usage.totalTokens);
+  if (usage.cacheCreateTokens) {
+    console.log('  - Cache create tokens: ', usage.cacheCreateTokens);
+  }
+  if (usage.cacheReadTokens) {
+    console.log('  - Cache read tokens: ', usage.cacheReadTokens);
+  }
 
-  const inputCost = usage.inputTokens * inputCostPerToken;
+  const inputCost =
+    usage.inputTokens * inputCostPerToken +
+    (usage.cacheCreateTokens ?? 0) * inputCostPerToken * 1.25 +
+    (usage.cacheReadTokens ?? 0) * inputCostPerToken * 0.2;
   const outputCost = usage.outputTokens * outputCostPerToken;
   const totalCost = inputCost + outputCost;
   console.log('  - Estimated cost: ', totalCost.toFixed(6), ' USD');

package/src/cli/cli-options.js

@@ -100,6 +100,18 @@ const cliOptions = [
     description:
       'Apply a content mask to limit the initial source code files included in the request. The value should be a prefix of the path relative to rootDir.',
   },
+  {
+    name: '--ignore-pattern=<pattern>',
+    description: 'Specify a pattern of files to ignore during the initial source code fetching.',
+  },
+  {
+    name: '--disable-cache',
+    description: 'Disable caching for the application.',
+  },
+  {
+    name: '--ask-question',
+    description: 'Allows the AI assistant to ask questions for clarification during the code generation process.',
+  },
 ];
 
 /**

package/src/cli/cli-params.js

@@ -17,6 +17,7 @@ export let vertexAi = params.includes('--vertex-ai');
 export let vertexAiClaude = params.includes('--vertex-ai-claude');
 export const dependencyTree = params.includes('--dependency-tree');
 export const verbosePrompt = params.includes('--verbose-prompt');
+export const disableCache = params.includes('--disable-cache');
 export let explicitPrompt = params.find((param) => param.startsWith('--explicit-prompt'))?.split('=')[1];
 export const disableContextOptimization = params.includes('--disable-context-optimization');
 export let taskFile = params.find((param) => param.startsWith('--task-file'))?.split('=')[1];
@@ -26,6 +27,7 @@ export const disableInitialLint = params.includes('--disable-initial-lint');
 export const vision = params.includes('--vision');
 export const imagen = params.find((param) => param.startsWith('--imagen'))?.split('=')[1];
 export const cheap = params.includes('--cheap');
+export const askQuestion = params.includes('--ask-question');
 
 // Add support for --help option
 export const helpRequested = params.includes('--help');
@@ -41,6 +43,11 @@ export const temperature = parseFloat(
 // New content mask parameter
 export const contentMask = params.find((param) => param.startsWith('--content-mask='))?.split('=')[1] || null;
 
+// New ignore pattern parameter
+export const ignorePatterns = params
+  .filter((param) => param.startsWith('--ignore-pattern='))
+  .map((param) => param.split('=')[1]);
+
 if (taskFile) {
   if (explicitPrompt) {
     throw new Error('The --task-file option is exclusive with the --explicit-prompt option');
@@ -97,3 +104,11 @@ if (cheap) {
 if (contentMask) {
   console.log(`Content mask: ${contentMask}`);
 }
+
+if (ignorePatterns) {
+  console.log(`Ignore pattern: ${ignorePatterns}`);
+}
+
+if (askQuestion) {
+  console.log('Assistant can ask questions to the user');
+}

package/src/cli/validate-cli-params.js

@@ -1,5 +1,6 @@
 import fs from 'fs';
 import path from 'path';
+import globRegex from 'glob-regex';
 import { rcConfig } from '../main/config.js';
 
 // List of allowed CLI parameters
@@ -26,8 +27,11 @@ const allowedParameters = [
   '--vision',
   '--imagen=',
   '--cheap',
+  '--ask-question',
   '--help',
   '--content-mask=',
+  '--disable-cache',
+  '--ignore-pattern=',
 ];
 
 /**
@@ -95,6 +99,22 @@ export function validateCliParams() {
       );
     }
   }
+
+  // Validate ignore pattern parameter
+  const ignorePatternParams = providedParameters.filter((param) => param.startsWith('--ignore-pattern='));
+  if (ignorePatternParams.length > 0) {
+    for (const ignorePatternParam of ignorePatternParams) {
+      const ignorePatternValue = ignorePatternParam.split('=')[1];
+      if (!ignorePatternValue) {
+        throw new Error('Invalid --ignore-pattern value. The pattern cannot be empty.');
+      }
+      try {
+        globRegex(ignorePatternValue);
+      } catch (e) {
+        throw new Error(`Invalid --ignore-pattern value. The pattern "${ignorePatternValue}" is not valid.`, e);
+      }
+    }
+  }
 }
 
 /**

package/src/files/read-files.js

@@ -2,11 +2,12 @@ import fs from 'fs';
 import mime from 'mime-types';
 import sizeOf from 'image-size';
 import path from 'path';
+import globRegex from 'glob-regex';
 
 import { getSourceFiles, getImageAssetFiles } from './find-files.js';
 import { rcConfig } from '../main/config.js';
 import { verifySourceCodeLimit } from '../prompt/limits.js';
-import { taskFile, contentMask } from '../cli/cli-params.js';
+import { taskFile, contentMask, ignorePatterns } from '../cli/cli-params.js';
 
 /**
  * Read contents of source files and create a map with file path as key and file content as value
@@ -23,8 +24,12 @@ function readSourceFiles(filterPaths) {
           continue;
         }
       }
-      const content = fs.readFileSync(file, 'utf-8');
-      sourceCode[file] = { content };
+      if (ignorePatterns.some((pattern) => globRegex(pattern).test(file))) {
+        sourceCode[file] = { content: null };
+      } else {
+        const content = fs.readFileSync(file, 'utf-8');
+        sourceCode[file] = { content };
+      }
     }
   }
   return sourceCode;

package/src/files/read-files.test.js

@@ -19,6 +19,7 @@ vi.mock('../prompt/limits.js');
 vi.mock('../cli/cli-params.js', () => ({
   taskFile: null,
   contentMask: null,
+  ignorePatterns: [],
 }));
 vi.mock('../main/config.js', () => ({
   rcConfig: { rootDir: '/home/project' },

package/src/main/codegen.test.js

@@ -37,6 +37,7 @@ vi.mock('../cli/cli-params.js', () => ({
   cheap: false,
   taskFile: undefined,
   disableInitialLint: undefined,
+  askQuestion: undefined,
 }));
 vi.mock('../files/find-files.js', () => ({
   getSourceFiles: () => [],