Semantic-code-search

search code semantically for function search, logic multiplexing, can save develop time, support docker deploy.

Usage

1. install dependencies

   pnpm install
   

2. set OPENAI_API_KEY in .example.env
3. parse your code by run npm run parse, this command will generate repo code embeddings to data/{repoName}-{parserName}.csv

   npm run parse:example
   

4. search code

   npm run search:example
   

   also you can deploy the search server by run

   docker compose build
   docker compose up

   then call http://localhost:3060/search?keyword=testServer&target=example, you can set DEFAULT_TARGET in docker-compose.yml that you don't have to pass target

Command Line Parameters

target

when you run "npm run parse -- target=example", this program will read .example.env file and load environment variables

Environment Variables

OPENAI_API_KEY

get from https://platform.openai.com/account/api-keys

PARSER

put your parser function file to src/parser and should export default parser function in the file

EMBEDDING_MODEL

embedding model name

CHAT_COMPLETION_MODEL

compeletion model name

BATCH_GEN_COUNT

call openai createEmbedding api input array max length, default = 12

REPOPATH

the repo path to be parsed, absolute path

EXCLUDE_DIRS

ignore parse dirs and files, relative path, type to be string[]

EXT

parse file ext, type to be string[]

RELATIVE_ROOT_URL

searchResultItem.filePath will concat RELATIVE_ROOT_URL

Search API CALL Parameters

keyword

query text

target

the env file, if set target 'example', will load .example.env

n

top n results

lines

max code lines in searchResult item

Welcome to contribute😀