From aa509273e6e9843459f6eac93dfe90c066d3cf3e Mon Sep 17 00:00:00 2001 From: extremeheat Date: Sun, 25 Feb 2024 14:38:54 -0500 Subject: [PATCH] Add a collectGithubRepoFiles tool (#8) * start work on codebase stripping * tools: add github repo lookup method, updates to code stripper * Add a `collectGithubRepoFiles` tool * update codebase.js * remove debug logging --- .gitignore | 140 +-------------- README.md | 2 +- src/index.d.ts | 11 ++ src/tools.js | 114 +------------ src/tools/codebase.js | 64 +++++++ src/tools/stripping.js | 379 +++++++++++++++++++++++++++++++++++++++++ src/tools/viz.js | 111 ++++++++++++ test/tools.js | 21 +++ test/viz.js | 10 -- 9 files changed, 598 insertions(+), 254 deletions(-) create mode 100644 src/tools/codebase.js create mode 100644 src/tools/stripping.js create mode 100644 src/tools/viz.js create mode 100644 test/tools.js delete mode 100644 test/viz.js diff --git a/.gitignore b/.gitignore index 039847f..5e15658 100644 --- a/.gitignore +++ b/.gitignore @@ -1,135 +1,7 @@ -# Logs -logs -*.log -npm-debug.log* -yarn-debug.log* -yarn-error.log* -lerna-debug.log* -.pnpm-debug.log* - -# Diagnostic reports (https://nodejs.org/api/report.html) -report.[0-9]*.[0-9]*.[0-9]*.[0-9]*.json - -# Runtime data -pids -*.pid -*.seed -*.pid.lock - -# Directory for instrumented libs generated by jscoverage/JSCover -lib-cov - -# Coverage directory used by tools like istanbul -coverage -*.lcov - -# nyc test coverage -.nyc_output - -# Grunt intermediate storage (https://gruntjs.com/creating-plugins#storing-task-files) -.grunt - -# Bower dependency directory (https://bower.io/) -bower_components - -# node-waf configuration -.lock-wscript - -# Compiled binary addons (https://nodejs.org/api/addons.html) -build/Release - -# Dependency directories -node_modules/ -jspm_packages/ - -# Snowpack dependency directory (https://snowpack.dev/) -web_modules/ - -# TypeScript cache -*.tsbuildinfo - -# Optional npm cache directory -.npm - -# Optional eslint cache -.eslintcache - -# Optional stylelint cache -.stylelintcache - -# Microbundle cache -.rpt2_cache/ -.rts2_cache_cjs/ -.rts2_cache_es/ -.rts2_cache_umd/ - -# Optional REPL history -.node_repl_history - -# Output of 'npm pack' -*.tgz - -# Yarn Integrity file -.yarn-integrity - -# dotenv environment variable files -.env -.env.development.local -.env.test.local -.env.production.local -.env.local - -# parcel-bundler cache (https://parceljs.org/) -.cache -.parcel-cache - -# Next.js build output -.next -out - -# Nuxt.js build / generate output -.nuxt -dist - -# Gatsby files -.cache/ -# Comment in the public line in if your project uses Gatsby and not Next.js -# https://nextjs.org/blog/next-9-1#public-directory-support -# public - -# vuepress build output -.vuepress/dist - -# vuepress v2.x temp and cache directory -.temp -.cache - -# Docusaurus cache and generated files -.docusaurus - -# Serverless directories -.serverless/ - -# FuseBox cache -.fusebox/ - -# DynamoDB Local files -.dynamodb/ - -# TernJS port file -.tern-port - -# Stores VSCode versions used for testing VSCode extensions -.vscode-test - -# yarn v2 -.yarn/cache -.yarn/unplugged -.yarn/build-state.yml -.yarn/install-state.gz -.pnp.* -playground +node_modules package-lock.json -*.key - -test/*.html \ No newline at end of file +playground +test/*.html +src/tools/repos +__* +*.key \ No newline at end of file diff --git a/README.md b/README.md index a8f6363..7e49d77 100644 --- a/README.md +++ b/README.md @@ -3,7 +3,7 @@ [![Build Status](https://github.com/extremeheat/LXL/actions/workflows/ci.yml/badge.svg)](https://github.com/extremeheat/LXL/actions/workflows/) [![Gitpod ready-to-code](https://img.shields.io/badge/Gitpod-ready--to--code-blue?logo=gitpod)](https://gitpod.io/#https://github.com/extremeheat/LXL) -LangXLang (LXL), a simple wrapper for Node.js to use OpenAI's GPT models and Google's Gemini and PaLM 2 models, with function calling support. +LangXLang (LXL), a Node.js library to use OpenAI's GPT models and Google's Gemini and PaLM 2 models, with function calling support. Supported models are: * OpenAI: `gpt-3.5-turbo-16k`, `gpt-3.5-turbo`, `gpt-4`, `gpt-4-turbo-preview` diff --git a/src/index.d.ts b/src/index.d.ts index 8df849d..adcb56b 100644 --- a/src/index.d.ts +++ b/src/index.d.ts @@ -15,6 +15,7 @@ declare module 'langxlang' { } interface Func { + // If default is not provided, the argument is required. Arg(options: { type: string[], description: string, example?: string, default?: any, required?: boolean }): string Arg(options: { type: object, description: string, example?: string, default?: any, required?: boolean }): object Arg(options: { type: T, description: string, example?: string, default?: any, required?: boolean }): T @@ -38,6 +39,16 @@ declare module 'langxlang' { interface Tools { // Generate HTML that shows side-by-side outputs for the system/user prompt across different models. makeVizForPrompt(systemPrompt: string, userPrompt: string, models: Model[]): Promise + // Returns a JS object with a list of files in a GitHub repo + collectGithubRepoFiles(repo: string, options: { + // What extension of files in the repo to include + extension?: string, + // The branch to use + branch?: string, + // Either a function that returns true if the file should be included + // or an array of regexes of which one needs to match for inclusion + matching?: (fileName: string) => boolean | RegExp[] + }): Promise<[absolutePath: string, relativePath: string, contents: string][]> } const tools: Tools diff --git a/src/tools.js b/src/tools.js index 81e9be5..75e0f69 100644 --- a/src/tools.js +++ b/src/tools.js @@ -1,111 +1,7 @@ -const { CompletionService } = require('./CompletionService') +const codebase = require('./tools/codebase') +const viz = require('./tools/viz') -function makeVizHtml (data) { - return ` - - - - - - LLM Output Viz - - - - -
-
-

System Prompt

- 
SYS PROMPT
-
-
-

User Prompt

- 
USR PROMPT
-
- ${ - data.models.map(([modelName, modelId]) => - `

${modelName}

MODEL OUTPUT
`).join('\n') - } -
- - - - - ` +module.exports = { + makeVizForPrompt: viz.makeVizForPrompt, + collectGithubRepoFiles: codebase.collectGithubRepoFiles } - -async function makeVizForPrompt (system, user, models) { - const service = new CompletionService() - const data = { models: [], outputs: {} } - for (const model of models) { - const { text } = await service.requestCompletion(model, system, user) - switch (model) { - case 'gpt-3.5-turbo-16k': - data.models.push(['GPT-3.5 Turbo 16k', '3516turbo']) - data.outputs['3516turbo'] = text - break - case 'gpt-3.5-turbo': - data.models.push(['GPT-3.5 Turbo', '35turbo']) - data.outputs['35turbo'] = text - break - case 'gpt-4': - data.models.push(['GPT-4', 'gpt4']) - data.outputs.gpt4 = text - break - case 'gpt-4-turbo-preview': - data.models.push(['GPT-4 Turbo Preview', 'gpt4turbo']) - data.outputs.gpt4turbo = text - break - case 'gemini-1.0-pro': - data.models.push(['Gemini 1.0 Pro', 'gemini']) - data.outputs.gemini = text - break - default: - data.models.push([model, model]) - data.outputs[model] = text - } - } - data.system = system - data.user = user - return makeVizHtml(data) -} - -module.exports = { makeVizForPrompt } diff --git a/src/tools/codebase.js b/src/tools/codebase.js new file mode 100644 index 0000000..a678094 --- /dev/null +++ b/src/tools/codebase.js @@ -0,0 +1,64 @@ +const fs = require('fs') +const cp = require('child_process') +const { join } = require('path') + +function fixSeparator (path) { + return path.replace(/\\/g, '/') +} + +function getAllFilesIn (folder) { + const files = [] + const entries = fs.readdirSync(folder, { withFileTypes: true }) + for (const entry of entries) { + const fullPath = join(folder, entry.name) + if (entry.isDirectory()) { + files.push(...getAllFilesIn(fullPath)) + } else { + files.push(fullPath) + } + } + return files.map(fixSeparator) +} + +// This function will clone a github repo, review all the files and merge relevant files into a single file +function collectGithubRepoFiles (repo, options) { + const extension = options.extension + const branch = options.branch || 'master' + // First, try to clone the repo inside a "repos" folder in this directory + const safeName = repo.replace(/\//g, ',') + const reposDir = join(__dirname, 'repos') + const repoPath = join(reposDir, safeName) + fs.mkdirSync(reposDir, { recursive: true }) + if (!fs.existsSync(repoPath)) { + cp.execSync(`git clone https://github.com/${repo}.git ${safeName}`, { cwd: reposDir }) + } + // Git pull origin/$branch + cp.execSync(`git pull origin ${branch}`, { cwd: repoPath }) + // Check out the branch + cp.execSync(`git checkout ${branch}`, { cwd: repoPath }) + // Now collect all the files inside repoPath, like `tree` + const allFiles = getAllFilesIn(repoPath) + .map(f => [f, f.replace(fixSeparator(repoPath), '')]) + + // Now figure out the relevant files + const relevantFiles = [] + for (const [file, relFile] of allFiles) { + if (extension && !file.endsWith(extension)) { + continue + } + if (options.matching) { + if (typeof options.matching === 'function') { + if (!options.matching(relFile)) { + continue + } + } else if (!options.matching.some(m => relFile.match(m))) { + continue + } + } + relevantFiles.push([file, relFile]) + } + const fileContents = relevantFiles.map(([abs, rel]) => [abs, rel, fs.readFileSync(abs, 'utf8').trim()]) + return fileContents +} + +module.exports = { collectGithubRepoFiles } diff --git a/src/tools/stripping.js b/src/tools/stripping.js new file mode 100644 index 0000000..1c73039 --- /dev/null +++ b/src/tools/stripping.js @@ -0,0 +1,379 @@ +// Stripping here refers to removing unnecessary tokens, comments and white-space in a program to minimize the amount of tokens +// that are needed to represent the program. In languages like Java, there's lots of syntax tokens that are needed for the program +// to run, but not needed for the purpose of abstractly understanding program logic. Think things like public/private, final, etc. + +function stripJava (code, options) { + // First, we need to "tokenize" the code, by splitting it into 3 types of data: comments, strings, and code. + const tokens = [] + let tokenSoFar = '' + let currentTokenType = 'code' // 'code' or 'multi-line-comment', 'single-line-comment' or 'string' + for (let i = 0; i < code.length; i++) { + const lastChar = code[i - 1] + const currentChar = code[i] + const nextChar = code[i + 1] + if (currentTokenType === 'code') { + if (currentChar === '/' && nextChar === '*') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'multi-line-comment' + } else if (currentChar === '/' && nextChar === '/') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'single-line-comment' + } else if (currentChar === '"') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'string' + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'multi-line-comment') { + if (currentChar === '*' && nextChar === '/') { + tokens.push([tokenSoFar + '*/', currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + i++ + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'single-line-comment') { + if (currentChar === '\n') { + tokens.push([tokenSoFar + '\n', currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'string') { + if (currentChar === '"' && lastChar !== '\\') { + tokens.push([tokenSoFar + '"', currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + } else { + tokenSoFar += currentChar + } + } + } + // Now we have an array of tokens, where every other token is a comment or string, and the others are code. + tokens.push([tokenSoFar, currentTokenType]) + // Now we can remove the keyword tokens that we don't want to keep. The always have spaces around them, so nothing fancy is needed. + const syntaxTokensToRemove = options.tokensToRemove || + ['protected', 'private', 'public', 'final', 'abstract', 'synchronized', 'volatile', 'transient', 'native', 'strictfp'] + + for (const entry of tokens) { + if (entry[1] === 'code') { + for (const forRemoval of syntaxTokensToRemove) { + entry[0] = entry[0].replace(new RegExp('\\b' + forRemoval + ' ', 'g'), '') + } + } + } + // Now we can replace some user specified tokens with other tokens. Useful for renaming variables + if (options.replacements) { + for (const entry of tokens) { + if (entry[1] === 'code') { + for (const [old, now] of options.replacements) { + entry[0] = entry[0].replaceAll(old, now) + } + } + } + } + + // First, make a new set of tokens, removing comments if the user wants + let newTokens = [] + for (const [tokenStr, tokenType] of tokens) { + if (options.removeComments && (tokenType === 'multi-line-comment' || tokenType === 'single-line-comment')) { + continue + } + newTokens.push([tokenStr, tokenType]) + } + // update the newTokens to merge adjacent code tokens (needed for correct space handling) + for (let i = 0; i < newTokens.length - 1; i++) { + const [tokenStr, tokenType] = newTokens[i] + const [nextTokenStr, nextTokenType] = newTokens[i + 1] + if (tokenType === 'code' && nextTokenType === 'code') { + newTokens[i + 1][0] = tokenStr + nextTokenStr + newTokens[i][0] = '' + } + } + newTokens = newTokens.filter(([tokenStr, tokenType]) => tokenStr !== '') + + // Now iterate through the new tokens and remove code with empty space lines + let result = '' + for (let i = 0; i < newTokens.length; i++) { + const [tokenStr, tokenType] = newTokens[i] + if (tokenType === 'code') { + const newStrLines = [] + for (const line of tokenStr.split('\n')) { + if (line.trim() === '') continue + newStrLines.push(line) + } + const now = newStrLines.join('\n') + result += now + } else { + result += tokenStr + } + } + return result +} + +function stripPHP (code, options = {}) { + // First, we need to "tokenize" the code, by splitting it into 3 types of data: comments, strings, and code. + const tokens = [] + let tokenSoFar = '' + // 'code' or 'multi-line-comment', 'single-line-comment' or 'double-quote-string', 'single-quote-string', 'heredoc-string', 'nowdoc-string' + let currentTokenType = 'code' + let currentTokenData = '' + for (let i = 0; i < code.length; i++) { + const lastChar = code[i - 1] + const currentChar = code[i] + const nextChar = code[i + 1] + const nextNextChar = code[i + 2] + if (currentTokenType === 'code') { + if (currentChar === '/' && nextChar === '*') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'multi-line-comment' + } else if (currentChar === '/' && nextChar === '/') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'single-line-comment' + } else if (currentChar === '"') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'double-quote-string' + } else if (currentChar === "'") { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'single-quote-string' + } else if (currentChar === '<' && nextChar === '<' && nextNextChar === '<') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + nextChar + nextNextChar + i += 2 + const end = code.indexOf('\n', i) + currentTokenData = code.substring(i, end).trim() + tokenSoFar += currentTokenData + if (currentTokenData.startsWith("'")) { + currentTokenType = 'nowdoc-string' + currentTokenData = currentTokenData.slice(1, -1) + } else { + currentTokenType = 'heredoc-string' + } + i = end + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'multi-line-comment') { + if (currentChar === '*' && nextChar === '/') { + tokens.push([tokenSoFar + '*/', currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + i++ + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'single-line-comment') { + if (currentChar === '\n') { + tokens.push([tokenSoFar + '\n', currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'double-quote-string' || currentTokenType === 'single-quote-string') { + if (currentChar === (currentTokenType === 'double-quote-string' ? '"' : "'") && lastChar !== '\\') { + tokens.push([tokenSoFar + currentChar, currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'heredoc-string' || currentTokenType === 'nowdoc-string') { + if (code.startsWith(currentTokenData, i) && (code[i + currentTokenData.length] === '\n' || code[i + currentTokenData.length] === ';')) { + tokenSoFar += currentTokenData + i += currentTokenData.length - 1 + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + } else { + tokenSoFar += currentChar + } + } + } + tokens.push([tokenSoFar, currentTokenType]) + // Now we can remove the keyword tokens that are not important for abstractly understanding the program + const syntaxTokensToRemove = options.tokensToRemove || ['public', 'private', 'protected', 'final', 'readonly'] + for (const entry of tokens) { + if (entry[1] === 'code') { + for (const forRemoval of syntaxTokensToRemove) { + entry[0] = entry[0].replace(new RegExp('\\b' + forRemoval + ' ', 'g'), '') + } + } + } + // Now we can replace some user specified tokens with other tokens. Useful for renaming variables + if (options.replacements) { + for (const entry of tokens) { + if (entry[1] === 'code') { + for (const [old, now] of options.replacements) { + entry[0] = entry[0].replaceAll(old, now) + } + } + } + } + + // First, make a new set of tokens, removing comments if the user wants + let newTokens = [] + for (const [tokenStr, tokenType] of tokens) { + if (options.removeComments && (tokenType === 'multi-line-comment' || tokenType === 'single-line-comment')) { + continue + } + newTokens.push([tokenStr, tokenType]) + } + // update the newTokens to merge adjacent code tokens (needed for correct space handling) + for (let i = 0; i < newTokens.length - 1; i++) { + const [tokenStr, tokenType] = newTokens[i] + const [nextTokenStr, nextTokenType] = newTokens[i + 1] + if (tokenType === 'code' && nextTokenType === 'code') { + newTokens[i + 1][0] = tokenStr + nextTokenStr + newTokens[i][0] = '' + } + } + newTokens = newTokens.filter(([tokenStr, tokenType]) => tokenStr !== '') + + // Now iterate through the new tokens and remove code with empty space lines + let result = '' + for (let i = 0; i < newTokens.length; i++) { + const [tokenStr, tokenType] = newTokens[i] + if (tokenType === 'code') { + const newStrLines = [] + for (const line of tokenStr.split('\n')) { + if (line.trim() === '') continue + newStrLines.push(line) + } + const now = newStrLines.join('\n') + result += now + } else { + result += tokenStr + } + } + return result +} + +function stripGo (code, options) { + const tokens = [] + let tokenSoFar = '' + let currentTokenType = 'code' + for (let i = 0; i < code.length; i++) { + const lastChar = code[i - 1] + const currentChar = code[i] + const nextChar = code[i + 1] + if (currentTokenType === 'code') { + if (currentChar === '/' && nextChar === '*') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'multi-line-comment' + } else if (currentChar === '/' && nextChar === '/') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'single-line-comment' + } else if (currentChar === '"') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'double-quote-string' + } else if (currentChar === "'") { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'single-quote-string' + } else if (currentChar === '`') { + tokens.push([tokenSoFar, currentTokenType]) + tokenSoFar = currentChar + currentTokenType = 'raw-string' + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'multi-line-comment') { + if (currentChar === '*' && nextChar === '/') { + tokens.push([tokenSoFar + '*/', currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + i++ + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'single-line-comment') { + if (currentChar === '\n') { + tokens.push([tokenSoFar + '\n', currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'double-quote-string' || currentTokenType === 'single-quote-string') { + if (currentChar === (currentTokenType === 'double-quote-string' ? '"' : "'") && lastChar !== '\\') { + tokens.push([tokenSoFar + currentChar, currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + } else { + tokenSoFar += currentChar + } + } else if (currentTokenType === 'raw-string') { + if (currentChar === '`') { + tokens.push([tokenSoFar + currentChar, currentTokenType]) + tokenSoFar = '' + currentTokenType = 'code' + } else { + tokenSoFar += currentChar + } + } + } + // Go doesn't have a lot of syntax tokens that can be removed, so we'll just remove comments and whitespace + tokens.push([tokenSoFar, currentTokenType]) + // Now we can replace some user specified tokens with other tokens. Useful for renaming variables + if (options.replacements) { + for (const entry of tokens) { + if (entry[1] === 'code') { + for (const [old, now] of options.replacements) { + entry[0] = entry[0].replaceAll(old, now) + } + } + } + } + + // First, make a new set of tokens, removing comments if the user wants + let newTokens = [] + for (const [tokenStr, tokenType] of tokens) { + if (options.removeComments && (tokenType === 'multi-line-comment' || tokenType === 'single-line-comment')) { + continue + } + newTokens.push([tokenStr, tokenType]) + } + // update the newTokens to merge adjacent code tokens (needed for correct space handling) + for (let i = 0; i < newTokens.length - 1; i++) { + const [tokenStr, tokenType] = newTokens[i] + const [nextTokenStr, nextTokenType] = newTokens[i + 1] + if (tokenType === 'code' && nextTokenType === 'code') { + newTokens[i + 1][0] = tokenStr + nextTokenStr + newTokens[i][0] = '' + } + } + newTokens = newTokens.filter(([tokenStr, tokenType]) => tokenStr !== '') + + // Now iterate through the new tokens and remove code with empty space lines + let result = '' + for (let i = 0; i < newTokens.length; i++) { + const [tokenStr, tokenType] = newTokens[i] + if (tokenType === 'code') { + const newStrLines = [] + for (const line of tokenStr.split('\n')) { + if (line.trim() === '') continue + newStrLines.push(line) + } + const now = newStrLines.join('\n') + result += now + } else { + result += tokenStr + } + } + return result +} + +module.exports = { stripJava, stripPHP, stripGo } diff --git a/src/tools/viz.js b/src/tools/viz.js new file mode 100644 index 0000000..9e184ae --- /dev/null +++ b/src/tools/viz.js @@ -0,0 +1,111 @@ +const { CompletionService } = require('../CompletionService') + +function makeVizHtml (data) { + return ` + + + + + + LLM Output Viz + + + + +
+
+

System Prompt

+ 
SYS PROMPT
+
+
+

User Prompt

+ 
USR PROMPT
+
+ ${ + data.models.map(([modelName, modelId]) => + `

${modelName}

MODEL OUTPUT
`).join('\n') + } +
+ + + + + ` +} + +async function makeVizForPrompt (system, user, models) { + const service = new CompletionService() + const data = { models: [], outputs: {} } + for (const model of models) { + const { text } = await service.requestCompletion(model, system, user) + switch (model) { + case 'gpt-3.5-turbo-16k': + data.models.push(['GPT-3.5 Turbo 16k', '3516turbo']) + data.outputs['3516turbo'] = text + break + case 'gpt-3.5-turbo': + data.models.push(['GPT-3.5 Turbo', '35turbo']) + data.outputs['35turbo'] = text + break + case 'gpt-4': + data.models.push(['GPT-4', 'gpt4']) + data.outputs.gpt4 = text + break + case 'gpt-4-turbo-preview': + data.models.push(['GPT-4 Turbo Preview', 'gpt4turbo']) + data.outputs.gpt4turbo = text + break + case 'gemini-1.0-pro': + data.models.push(['Gemini 1.0 Pro', 'gemini']) + data.outputs.gemini = text + break + default: + data.models.push([model, model]) + data.outputs[model] = text + } + } + data.system = system + data.user = user + return makeVizHtml(data) +} + +module.exports = { makeVizForPrompt } diff --git a/test/tools.js b/test/tools.js new file mode 100644 index 0000000..26357d7 --- /dev/null +++ b/test/tools.js @@ -0,0 +1,21 @@ +const { tools } = require('langxlang') + +async function testViz () { + const viz = await tools.makeVizForPrompt('', 'Why is the sky blue?', ['gpt-3.5-turbo']) + console.log(viz) +} + +async function testCodebase () { + const files = await tools.collectGithubRepoFiles('extremeheat/node-basic-args', { + extension: '.js', + matching: [/examples/] + }) + console.log(files) +} + +async function main () { + await testViz() + await testCodebase() +} + +main() diff --git a/test/viz.js b/test/viz.js deleted file mode 100644 index 23bb006..0000000 --- a/test/viz.js +++ /dev/null @@ -1,10 +0,0 @@ -const { tools } = require('langxlang') -const fs = require('fs') -const path = require('path') - -async function main () { - // const all = ['gpt-3.5-turbo-16k', 'gpt-3.5-turbo', 'gpt-4', 'gpt-4-turbo-preview', 'gemini-1.0-pro'] - const viz = await tools.makeVizForPrompt('', 'Why is the sky blue?', ['gpt-3.5-turbo']) - fs.writeFileSync(path.join(__dirname, 'viz.html'), viz) -} -main()