# Project export: Chimera This document was generated by HackStack to give an AI agent context about a hackathon project. Sections are labeled with their provenance; content marked as truncated was cut to keep this document small. ## Project metadata - Hackathon: TreeHacks 2024 - Tagline: Have you ever "lost" a file in your messy file manager? Chimera, your semantic file searcher, will easily understand anything you're looking for and ensure any code, paper, or document will appear! - Devpost: https://devpost.com/software/semantic-file-searcher - GitHub: https://github.com/couplefire/chimera - Team: 4 GitHub contributor(s) — Nathan Wang (25 commits), Locke Cai (15 commits), Jeff (7 commits), Michael Huang (5 commits) ## Devpost submission (written by the team) ### Inspiration Examining our difficulties in our daily workflow, we realized that we often could not find files that we knew we had written before. Lots of us had saved countless papers and books, only to be unsure of where we saved them. Not knowing where our old code was led us to rewrite old libraries and codebases. Thus, we all wished for a better file searcher. One which, instead of solely relying on filenames or exact string matching in the file content, also took into account how ### What it does Chimera will first ask you to describe the file you are looking for. Then it will go through all the files in your directory and rank them according to which fit the prompt the best. ### How we built it We used the Tauri framework to build the app and wrote it in Rust. The app first embedded the prompts and the files into a vector space (using OpenAI's embedding models), then compared the similarity between the prompts and each file, then outputted them in order of likelihood. ### Challenges we ran into Building it in Rust was quite difficult, as some of us didn't know Rust and had to learn it on the fly, and there were a lot of issues working with some Rust libraries. ### Accomplishments we're proud of The end product, being built with Rust, is highly performant. Our embeddings framework is also easily parallelizable, so it can easily be extended to index every file in a user's computer relatively quickly. ### What we learned We learned how to work with Rust (especially the Tauri framework and Apache Arrow's data format that LanceDB uses) and how to use a vector DB. ### What's next While this hackathon used OpenAI's embeddings API, the end goal would be to ship a local embeddings model to the user's computer so their files won't have to leave their personal laptop. In addition, we hope to fine-tune the embeddings model to perform better at the specific task of searching for files. ## README (from the GitHub repository) # Chimera ## Inspiration Examining our difficulties in our daily workflow, we realized that we often could not find files that we knew we had written before. Lots of us had saved countless papers and books, only to be unsure of where we saved them. Not knowing where our old code was led us to rewrite old libraries and codebases. Thus, we all wished for a better file searcher. One which, instead of solely relying on filenames or exact string matching in the file content, also took into account how ## What it does Chimera will first ask you to describe the file you are looking for. Then it will go through all the files in your directory and rank them according to which fit the prompt the best. ## How we built it We used the Tauri framework to build the app and wrote it in Rust. The app first embedded the prompts and the files into a vector space (using OpenAI's embedding models), then compared the similarity between the prompts and each file, then outputted them in order of likelihood. ## Challenges we ran into Building it in Rust was quite difficult, as some of us didn't know Rust and had to learn it on the fly, and there were a lot of issues working with some Rust libraries. ## Accomplishments that we're proud of The end product, being built with Rust, is highly performant. Our embeddings framework is also easily parallelizable, so it can easily be extended to index every file in a user's computer relatively quickly. ## What we learned We learned how to work with Rust (especially the Tauri framework and Apache Arrow's data format that LanceDB uses) and how to use a vector DB. ## What's next for Chimera While this hackathon used OpenAI's embeddings API, the end goal would be to ship a local embeddings model to the user's computer so their files won't have to leave their personal laptop. In addition, we hope to fine-tune the embeddings model to perform better at the specific task of searching for files. ## Detected evidence (automated analysis) Indexed codebase: 36 recognized source files, 117 KB. - CSS (language) — detected in the code - HTML (language) — detected in the code - Python (language) — detected in the code - React (technology) — detected in the code - Rust (language) — detected in the code - TypeScript (language) — detected in the code ## Codebase structure (from repository index) ### Files (90 of 90) ``` .DS_Store .gitignore .vscode/extensions.json .vscode/settings.json data/my_table.lance/_latest.manifest data/my_table.lance/_transactions/0-d7b4903c-e874-488f-b235-8d35d0190fbf.txn data/my_table.lance/_transactions/1-53371309-3f9a-48f8-81ef-5285513f72a9.txn data/my_table.lance/_versions/1.manifest data/my_table.lance/_versions/2.manifest data/my_table.lance/data/b370ad91-6389-49d0-8353-5062a5da98a8.lance files-to-index/6110-sp24-main/.gitignore files-to-index/6110-sp24-main/.gitmodules files-to-index/6110-sp24-main/.pre-commit-config.yaml files-to-index/6110-sp24-main/build.sh files-to-index/6110-sp24-main/Cargo.lock files-to-index/6110-sp24-main/Cargo.toml files-to-index/6110-sp24-main/README.md files-to-index/6110-sp24-main/run.sh files-to-index/6110-sp24-main/src/main.rs files-to-index/6110-sp24-main/src/parser.rs files-to-index/6110-sp24-main/src/snapshots/decaf_rust__tokenizer__tests__tokenizer@aGVsbG8gd29ybGQ=.snap files-to-index/6110-sp24-main/src/snapshots/decaf_rust__tokenizer__tests__tokenizer@IT0gLS0gKysgLSAhID0gPiA+PSA8IDw9ICogLyAlICsgLSA7IFsgXSB7IH0gKCAp.snap files-to-index/6110-sp24-main/src/snapshots/decaf_rust__tokenizer__tests__tokenizer@J1xcJw==.snap files-to-index/6110-sp24-main/src/snapshots/decaf_rust__tokenizer__tests__tokenizer@J1xuJw==.snap files-to-index/6110-sp24-main/src/snapshots/decaf_rust__tokenizer__tests__tokenizer@J211bHQn.snap files-to-index/6110-sp24-main/src/snapshots/decaf_rust__tokenizer__tests__tokenizer@J2En.snap files-to-index/6110-sp24-main/src/snapshots/decaf_rust__tokenizer__tests__tokenizer@Ly8gdGhpcyBpcyBhIGNvbW1lbnQKaW50IG1haW4=.snap files-to-index/6110-sp24-main/src/snapshots/decaf_rust__tokenizer__tests__tokenizer@MHgxMjNoZWxsbw==.snap files-to-index/6110-sp24-main/src/tokenizer.rs files-to-index/6110-sp24-main/src/utils/cli.rs files-to-index/6110-sp24-main/src/utils/mod.rs files-to-index/6110-sp24-main/test.dcf files-to-index/6110-sp24-main/test.sh files-to-index/lecture notes/l-bft.txt files-to-index/lecture notes/l-bitcoin.txt files-to-index/lecture notes/l-ethereum.txt files-to-index/lecture notes/l-farm.txt files-to-index/lecture notes/l-gfs.txt files-to-index/lecture notes/l-grove.txt files-to-index/lecture notes/l-linearizability.txt files-to-index/lecture notes/l-memcached.txt files-to-index/lecture notes/l-raft-QA.txt files-to-index/lecture notes/l-raft.txt files-to-index/lecture notes/l-raft2.txt files-to-index/lecture notes/l-rpc.txt files-to-index/lecture notes/l-spanner.txt files-to-index/lecture notes/l-sundr.txt files-to-index/lecture notes/l-zookeeper.txt README.md tauri/.gitignore tauri/files-to-index/a.txt tauri/files-to-index/test_create_table.py tauri/files-to-index/test_data/my_table.lance/_latest.manifest tauri/files-to-index/test_data/my_table.lance/_transactions/0-2bba53c4-3f2b-4fe2-b8b3-706cc5b85cd7.txn tauri/files-to-index/test_data/my_table.lance/_transactions/1-2d4f8c33-7fff-423c-a3db-8c00d67e9c27.txn tauri/files-to-index/test_data/my_table.lance/_versions/1.manifest tauri/files-to-index/test_data/my_table.lance/_versions/2.manifest tauri/files-to-index/test_data/my_table.lance/data/0ac366be-01cd-4b36-bfb6-5828927324d2.lance tauri/index.html tauri/package.json tauri/README.md tauri/src-tauri/.gitignore tauri/src-tauri/build.rs tauri/src-tauri/Cargo.lock tauri/src-tauri/Cargo.toml tauri/src-tauri/icons/icon.icns tauri/src-tauri/src/db.rs tauri/src-tauri/src/embeddings/file_embedding.rs tauri/src-tauri/src/embeddings/mod.rs tauri/src-tauri/src/embeddings/prompt_embedding.rs tauri/src-tauri/src/indexer.rs tauri/src-tauri/src/main.rs tauri/src-tauri/src/parser.rs tauri/src-tauri/src/similarity_search.rs tauri/src-tauri/tauri.conf.json tauri/src/App.css tauri/src/App.tsx tauri/src/components/resultbox.tsx tauri/src/components/searchbox.tsx tauri/src/main.tsx tauri/src/styles.css tauri/src/styles/resultbox.module.css tauri/src/styles/searchbox.module.css tauri/src/types/types.ts tauri/src/utils/ext_to_icon.ts tauri/src/utils/logoutils.ts tauri/src/vite-env.d.ts tauri/tsconfig.json tauri/tsconfig.node.json tauri/vite.config.ts ``` ### Dependencies - files-to-index/6110-sp24-main/Cargo.toml: base64@0.21.7, clap@4.4.18, insta@1.34.0, rstest@0.18.2, serde@1.0.196 - tauri/package.json: @tauri-apps/api@^1.5.2, @tauri-apps/cli@^1.5.9, @types/lodash@^4.14.202, @types/react@^18.2.15, @types/react-dom@^18.2.7, @vitejs/plugin-react@^4.2.1, lodash@^4.17.21, react@^18.2.0, react-dom@^18.2.0, typescript@^5.0.2, vite@^5.0.0 - tauri/src-tauri/Cargo.toml: anyhow@1.0.79, arrow-array@50.0.0, arrow-schema@50.0.0, async-openai@0.18.3, futures@0.3.30, lopdf@0.32.0, open@5.0.1, pdf-extract@0.7.4, serde@1.0, serde_json@1.0, tauri@1.5, tauri-build@1.5, tokio@1.36.0, vectordb@0.4.10, walkdir@2.4.0 ### Recent commits (newest first) - updated readme - changed name - polishes - better icons - better icons - better icons - normalize - fix conflict - open files (rust) - add more files; update parser - add more files; rearrange into folders - added highlight - merge - concate name and file content - Batch indexing - integrated w backend - migrate to async openai - add random files - better front end - working ish ## Key source files (fetched from GitHub, selected and truncated for size) ### tauri/package.json ``` { "name": "semantic-files", "private": true, "version": "0.0.0", "type": "module", "scripts": { "dev": "vite", "build": "tsc && vite build", "preview": "vite preview", "tauri": "tauri" }, "dependencies": { "@tauri-apps/api": "^1.5.2", "lodash": "^4.17.21", "react": "^18.2.0", "react-dom": "^18.2.0" }, "devDependencies": { "@tauri-apps/cli": "^1.5.9", "@types/lodash": "^4.14.202", "@types/react": "^18.2.15", "@types/react-dom": "^18.2.7", "@vitejs/plugin-react": "^4.2.1", "typescript": "^5.0.2", "vite": "^5.0.0" } } ``` ### files-to-index/6110-sp24-main/Cargo.toml ``` [package] name = "decaf-rust" version = "0.1.0" edition = "2021" # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html [dependencies] clap = { version = "4.4.18", features = ["derive"] } serde = { version = "1.0.196", features = ["derive"] } [dev-dependencies] base64 = "0.21.7" insta = { version = "1.34.0", features = ["yaml"] } rstest = "0.18.2" [profile.dev.package] insta.opt-level = 3 ``` ### tauri/src-tauri/Cargo.toml ``` [package] name = "semantic-files" version = "0.0.0" description = "A Tauri App" authors = ["you"] edition = "2021" # See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html [build-dependencies] tauri-build = { version = "1.5", features = [] } [dependencies] tauri = { version = "1.5", features = [ "window-hide", "global-shortcut-all", "macos-private-api", "shell-open"] } serde = { version = "1.0", features = ["derive"] } serde_json = "1.0" vectordb = "0.4.10" walkdir = "2.4.0" anyhow = "1.0.79" pdf-extract = "0.7.4" tokio = "1.36.0" futures = "0.3.30" arrow-schema = "50.0.0" arrow-array = "50.0.0" lopdf = "0.32.0" async-openai = "0.18.3" open = "5.0.1" [features] # this feature is used for production builds or when `devPath` points to the filesystem # DO NOT REMOVE!! custom-protocol = ["tauri/custom-protocol"] ``` ### tauri/src/main.tsx ```typescript import React from "react"; import ReactDOM from "react-dom/client"; import App from "./App"; import "./styles.css"; ReactDOM.createRoot(document.getElementById("root") as HTMLElement).render( , ); ``` ### tauri/src/App.tsx ```typescript import { useEffect, useState } from "react"; import { invoke } from "@tauri-apps/api/tauri"; import { appWindow } from "@tauri-apps/api/window"; import SearchBox from "./components/searchbox"; import ResultBox from "./components/resultbox"; import { SearchResult } from "./types/types"; import { debounce } from 'lodash' import "./App.css"; function App() { const [searchResults, setSearchResults] = useState([]); useEffect(() => { window.addEventListener('click', (e: MouseEvent) => { if (!document.getElementById('app')?.contains(e.target as Node)) { appWindow.hide(); } }); return () => { window.removeEventListener('click', () => { appWindow.hide(); }); } }, []) const handleSearchChange = async (text: string) => { const res = await invoke("search", { searchText: text}); if (res) { const searchResults = res as SearchResult[]; setSearchResults(searchResults); }; }; const debouncedSearch = debounce(handleSearchChange, 100, { trailing: true, leading: true }) return (
{searchResults.map((result, index) => ( ))}
); } export default App; ``` ### files-to-index/6110-sp24-main/src/main.rs ```rust use std::{error, process::exit}; mod parser; mod tokenizer; mod utils; fn get_writer(output: &Option) -> Box { match output { Some(path) => Box::new(std::fs::File::create(path.as_path()).unwrap()), None => Box::new(std::io::stdout()), } } fn execute( input: &str, writer: &mut Box, target: utils::cli::CompilerAction, ) -> Result<(), Box> { match target { utils::cli::CompilerAction::Default => { panic!("Invalid target"); } utils::cli::CompilerAction::Scan => { let tokens = tokenizer::tokenize(input.to_string())?; for token in tokens { writeln!(writer, "{token}").unwrap(); } } utils::cli::CompilerAction::Parse => { let tokens = tokenizer::tokenize(input.to_string())?; let program = parser::parse_program(tokens)?; writeln!(writer, "{:?}", program).unwrap(); } utils::cli::CompilerAction::Inter => { todo!("inter"); } utils::cli::CompilerAction::Assembly => { todo!("assembly"); } } Ok(()) } fn main() { let args = utils::cli::parse(); let input = std::fs::read_to_string(&args.input).expect("Filename is incorrect."); if args.debug { eprintln!( "Filename: {:?}\nDebug: {:?}\nOptimizations: {:?}\nOutput File: {:?}\nTarget: {:?}", args.input, args.debug, args.opt, args.output, args.target ); } // Use writeln!(writer, "template string") to write to stdout ot file. let mut writer = get_writer(&args.output); match execute(&input, &mut writer, args.target) { Ok(()) => {} Err(e) => { eprintln!("{e}"); exit(1); } } } ``` ### tauri/src-tauri/src/main.rs ```rust // Prevents additional console window on Windows in release, DO NOT REMOVE!! #![cfg_attr(not(debug_assertions), windows_subsystem = "windows")] use std::sync::Arc; use serde::Serialize; use tauri::Manager; use tauri::GlobalShortcutManager; use db::{DbConnection, init_db}; use indexer::start_indexing; mod indexer; mod db; mod similarity_search; mod embeddings; mod parser; pub const EMBEDDING_DIM: u32 = 128; #[derive(Serialize)] struct SearchResult { fileName: String, directory: String, fileSize: u64, numPages: Option, } #[tauri::command] async fn search(search_text: &str, state: tauri::State<'_, DbConnection>) -> Result, ()> { if search_text.is_empty() { return Ok(Vec::new()); } let prompt_embed = embeddings::create_embedding_prompt(search_text).await.unwrap(); let result = similarity_search::search(state.inner().clone(), prompt_embed).await.unwrap(); Ok(result) } #[tauri::command] async fn open(path: &str, _state: tauri::State<'_, DbConnection>) -> Result<(), ()> { open::that(path).unwrap(); Ok(()) } fn main() { tauri::Builder::default() .setup(|app| { let initialize_db = true; let handle = tauri::async_runtime::spawn(async move { let db = init_db(initialize_db).await; db }); let db = tauri::async_runtime::block_on(handle).unwrap(); app.manage(db.clone()); if initialize_db { tauri::async_runtime::spawn(async move { println!("Starting indexing process..."); match start_indexing(db).await { Ok(_) => println!("Indexing process finished successfully!"), Err(e) => println!("Error while indexing: {}", e), } }); } let window = app.get_window("main").unwrap(); window.hide().unwrap(); let window_rc1 = Arc::new(window); let window_rc2 = Arc::clone(&window_rc1); { let _toggle = app.app_handle().global_shortcut_manager().register("Cmd+]", move || { if window_rc1.is_visible().unwrap() { window_rc1.hide().unwrap(); } else { window_rc1.show().unwrap(); } }); } { let _hide_esc = app.app_handle().global_shortcut_manager().register("esc", move || { window_rc2.hide().unwrap(); }); } Ok(()) }) .invoke_handler(tauri::generate_handler![search, open]) .run(tauri::generate_context!()) .expect("error while running tauri application"); } ``` ### files-to-index/6110-sp24-main/src/utils/cli.rs ```rust /** * A generic command-line interface for 6.035 compilers. This class * provides command-line parsing for student projects. It recognizes * the required -target, -debug, -opt, and * -o switches, and generates a name for input and output * files. * * @author 6.1100 Staff, last updated January 2024 */ use clap::Parser; #[derive(Clone, clap::ValueEnum, Debug)] pub enum CompilerAction { Default, Scan, Parse, Inter, Assembly, } #[derive(Clone, clap::ValueEnum, Debug, PartialEq, Eq, Hash)] pub enum Optimization {} #[derive(Parser, Debug)] pub struct Args { /// compile to the given stage #[clap(short, long, value_enum, default_value_t=CompilerAction::Default, value_name = "stage")] pub target: CompilerAction, /// write output to #[clap(short, long, value_name = "outname")] pub output: Option, /// Perform the listed optimizations #[clap( short = 'O', long, value_delimiter = ',', value_enum, value_name = "optimization,.." )] pub opt: Vec, /// Print debugging information #[arg(short, long, default_value_t = false)] pub debug: bool, /// Decaf file pub input: std::path::PathBuf, } pub fn parse() -> Args { Args::parse() } ``` ### tauri/index.html ```html Tauri + React + TS
``` ### tauri/vite.config.ts ```typescript import { defineConfig } from "vite"; import react from "@vitejs/plugin-react"; // https://vitejs.dev/config/ export default defineConfig(async () => ({ plugins: [react()], // Vite options tailored for Tauri development and only applied in `tauri dev` or `tauri build` // // 1. prevent vite from obscuring rust errors clearScreen: false, // 2. tauri expects a fixed port, fail if that port is not available server: { port: 1420, strictPort: true, watch: { // 3. tell vite to ignore watching `src-tauri` ignored: ["**/src-tauri/**"], }, }, })); ``` [25 more indexed source files omitted to keep this export small. The full file list is in the Codebase structure section above.]