isabella232 / watertower

serverless search engine

watertower

`

© Copyright Steve Daniels and licensed under CC BY-SA 2.0

Search Engine for Serverless environment.

• Search via words in documents
• Filter by tag

Architecture

This API uses standard inverted index to achieve search. When searching, calculate score by TFIDF algorithm and sort.

This tool is using gocloud.dev's docstore as an storage. So it can run completely managed environment (DynamoDB, Google Firestore, Azure CosmosDB) and MongoDB.

Elasticsearch can use flexible document, but this watertower can use only the following structure. "title" and "content" are processed (tokenize, stemming, remove stop words) by packages in github.com/future-architect/nlp. Now only English and Japanese are implemented.

{
  "unique_key": "100",
  "lang": "en",
  "title": "100 Continue",
  "tags": ["100", "no-error"],
  "content": "This interim response indicates that everything so far is OK and that the client should continue the request, or ignore the response if the request is already finished.",
  "metadata": {"this is not":  "searchable, but you can store any strings"}
}

Go API

Basically, this is an document storage. All API you should handle is WaterTower struct. To create instance, you can pass URL.

Before using this search engine, you should import several packages:

import (
    // Natural languages you want to use
	_ "github.com/future-architect/watertower/nlp/english"
	_ "github.com/future-architect/watertower/nlp/japanese"

    // The storage backends you want to use.
	_ "gocloud.dev/docstore/memdocstore"
	_ "gocloud.dev/docstore/awsdynamodb"
	_ "gocloud.dev/docstore/gcpfirestore"
	_ "gocloud.dev/docstore/mongodocstore"
)

Each storage should be initialized table with "id" as a primary key(partition key for DynamoDB). Index will be a table name.

wt, err := watertower.NewWaterTower(ctx, watertower.Option{
    DocumentUrl: "dynamodb://", // default is "mem://"
    Index:       "docs",        // default is "index"
})

To store document, call PostDocument() method.

docID, err := wt.PostDocument("unique-document-key", &watertower.Document{
    Language: "en",                               // required
    Title: "title",                               // document title
    Content: "content",                           // document body
    Tags: []string{"tag"},                        // tags
    Metadata: map[string]string{"extra": "data"}, // extra data
})

To get document, you can use by unique-key or document-id.

doc, err := wt.FindDocumentByKey(uniqueKey)
docs, err := wt.FindDocuments(id1, id2)

To search document, use Search() method. First parameter is natural language search word for title and content. Second parameter is tags to filter result. Third parameter is natural language name to process search word.

docs, err := wt.Search(searchWord, tags, lang)

To use local index file, open watertower instance without DocumentUrl option.

Then call ReadIndex(r io.Reader) for/and WriteIndex(w io.Writer) to read/store index:

f, err := os.Open("watertower.idx")
if err != nil {
	panic(err)
}

wt, err := watertower.NewWaterTower(ctx, watertower.Option{
    DefaultLanguage: *defaultLanguage,
})
wt.ReadIndex(f)

Sample Codes

httpstatus in /samples/httpstatus

CLI tool to search http status code. It reads from bundled documents and registeres them when booting. It uses memdocstore as a backend datastore.

HTTP interface

watertower-server in /cmd/watertower-server implements Elasticsearch inspired API.

./watertower-server --port=8888

# Register document
$ curl -X POST "http://127.0.0.1:8888/index/_doc/"
　　-H "content-type: application/json"
　　-d '{ "unique_key": "id1", "title": "hello watertower",
　　　　　"content": "watertower is a full text search engine with tag filtering", "lang": "en" }'
{"_id":"d1","_index":"index","_type":"_doc","result":"created"}

# Get document by unique-key
$ curl -X GET "http://127.0.0.1:8888/index/_search?q=unique_key%3Aid1"
    -H"content-type: application/json"
{"hits":{"hits":[{"_id":"d1","_index":"index","_source":{"content":"watertower is a full text search engine with tag filtering","lang":"en","metadata":{},"tags":null,"title":"hello watertower","unique_key":"id1"},"_type":"_doc","sort":null}],"total":{"total":1}}}

# Get document by document ID
$ curl -X GET "http://127.0.0.1:8888/index/_source/d1"
(...)

# Search
$ curl -X GET "http://127.0.0.1:8888/index/_search"
  -H "content-type: application/json"
  -d '{"query": {"bool": {"must": {"match_phrase": {"content": {"query": "stay", "analyzer": "en"}}}}}}'
(...)

CLI tool

CLI tool watertower-cli in /cmd/watertower-cli.

$ ./watertower-cli --help
usage: watertower-cli [<flags>] <command> [<args> ...]

Flags:
  --help  Show context-sensitive help (also try --help-long and --help-man).
  --default-language=DEFAULT-LANGUAGE  
          Default language

Commands:
  help [<command>...]
    Show help.

  create-index [<flags>] <INPUT>
    Generate Index

  search [<flags>] [<WORDS>...]
    Search

It can create single file index by using create-index sub command.

You can try searching the index file with search sub command.

License

Apache2