Lite Embeddings Dart Server

English · 中文

LLM Embedding tool HTTP service

Feature

• Support Vector Database: Chroma
• Support file type: pure text, include Markdown, TXT
• HTTP API Wrapper of Dart List Embeddings
• Base on Lite Embeddings Dart的EmbeddingsService(DTO included), add Controller、Router, wrapper to HTTP/WS API.

Usage

1. Prepare

1. Docs file, according to /example/docs/*.md
2. Separator in the file
   • If markdown file, recommend to use <!--SEPARATOR--> as separator, for NOT show it in markdown after rendering
3. Add .env file in the example folder, and add below content in the .env file:

   baseUrl = https://xxx.xxx.com         # LLM API BaseURL
   apiKey = sk-xxxxxxxxxxxxxxxxxxxx      # LLM API ApiKey

2. Develop run server

1. debug or run mode run /bin/server.dart file main()

3. HTTP API

• HTTP API

3.1 HTTP API

• Docs CRUD API, include:
  • /version: get version number, to confirm server running
  • /init: post the llm config to initial the embeddings service
  • /docs/create-by-text: Create docs embeddings, post whole text and separator, service will split and write to vector database
  • /docs/create: Create docs embeddings, post the split docs, service will write to vector database
  • /docs/delete: Delete docs, post docsId
  • /docs/list: List all docs, return docsId and docsName Array
  • /docs/rename: Rename docsName
  • /docs/query: Text query, return N segment array with distance sort
  • /docs/batch-query: Text array query, query multi text at once, return N segment array in array
  • /docs/multi-query: Docs array query, query multi docs with one text, return N segment with docsId array
  • /segment/list: List all segments in the docs
  • /segment/insert: Insert segment by index. If not index, new segment will be inserted at last
  • /segment/update: Update segment
  • /segment/delete: Delete segment
  • /dispose: Dispose vector database connection

BaseURL

• http://127.0.0.1:9537/api

[GET] /version

• Feature: get version number, to confirm server running

• Request params: null

• Response body sample:

  {
      "version": "0.1.0"
  }

[POST] /init

• Feature: post the llm config to initial the embeddings service

• Request params:

  • LLM config: baseUrl, apiKey, model
  • Sample:

  {
    "baseUrl": "<LLM API baseUrl, e.g. https://api.openai.com/v1>>",
    "apiKey": "<LLM API apiKey, e.g. sk-xxxxxxxxxx>",
    "model": "<LLM API embeddings model name, e.g. : text-embedding-ada-002>"
  }

• Response body:

  • Echo init info, include vectorDatabase name, baseUrl, embeddingsModel
  • Response body sample

    {
      "vectorDatabase": "<Vector Database name, e.g. : chroma>",
      "baseUrl": "<LLM API baseUrl, e.g. https://api.openai.com/v1>>",
      "embeddingsModel": "<LLM API embeddings model name, e.g. : text-embedding-ada-002>"
    }

[POST] /docs/create-by-text

• Feature: Create docs embeddings, post whole text and separator, service will split and write to vector database
• Request params:
  • Docs info: docsName, text, separator, metadata
  • Sample

  {
    "docsName": "<Docs name, e.g. Moore's Law for Everything.md>",
    "text": "<Docs full text, with separetor>",
    "separator": "<Separator text>",
    "metadata": "<json map, value only int, float, string, bool, NOT support object and array. Each segment with same metadata>"
  }

• Response body:
  • Create successful docs info: docsId, docsName
  • Response body sample

  {
    "docsId": "<Docs Id>",
    "docsName": "<Docs Name>"
  }

[POST] /docs/create

• Feature: Create docs embeddings, post the split docs, service will write to vector database
• Request params:
  • Docs info: docsName, segment and metadata array
  • Sample

  {
    "docsName": "<Docs name, e.g. Moore's Law for Everything.md>",
    "segmentList": [
      { 
        "text": "<Segment text>",
        "metadata": "<json map, value only int, float, string, bool, NOT support object and array>"
      }
    ]
  }

• Response body:
  • Create successful docs info: docsId, docsName
  • Response body sample

  {
    "docsId": "<Docs Id>",
    "docsName": "<Docs Name>"
  }

[POST] /docs/delete

• Feature: Delete docs, post docsId
• Request params:
  • DocsId
  • Sample

  {
    "docsId": "xxxxxxxx"
  }

• Response body:
  • Docs be deleted info: docsId
  • Response body sample

  {
    "docsId": "xxxxxxxx"
  }

[GET] /docs/list

• Feature: List all docs, return docsId and docsName Array
• Request params: null
• Response body:
  • Docs info list: docsId and docsName
  • Response body sample

  [ 
    {
      "docsId": "<Docs Id>",
      "docsName": "<Docs Name>"
    }
  ]

[POST] /docs/rename

• Feature: Rename docsName
• Request params:
  • docsId, new docsName
  • Sample

  {
    "docsId": "<Docs Id>",
    "docsName": "<Docs Name>"
  }

• Response body:
  • Docs be modified info: docsId and docsName
  • Response body sample

  {
    "docsId": "<Docs Id>",
    "docsName": "<Docs Name>"
  }

[POST] /docs/query

• Feature: Text query, return N segment array with distance sort
• Request params:
  • docsId, queryText, return query result count
  • Sample

  {
    "docsId": "xxxxxxxx",
    "queryText": "<query text string>",
    "nResults": "<UInt, return query result by sort number>"
  }

• Response body:
  • docsId, segmentResultList
  • Response body sample

  {
    "docsId": "xxxxxxxx",
    "segmentResultList": [
      {
        "segmentId": "<Segment Id>",
        "text": "<Segment text>",
        "metadata": "<json map, segment with>",
        "distance": "<0.x float, segment match distance, smaller means closer>"
      }
    ]
  }

[POST] /docs/batch-query

• Feature: Text array query, query multi text at once, return N segment array in array

• Request params:

  • docsId, queryText array, return query result count
  • Sample

  {
    "docsId": "xxxxxxxx",
    "queryTextList": [
      "<Query Text 1>",
      "<Query Text 2>"
    ],
    "nResults": "<UInt, return query result by sort number>"
  }

• Response body:

  • Query result: docsId and segmentResultList array
  • Response body sample

  [
    {
      "docsId": "xxxxxxxx",
      "segmentResultList": [
        {
          "segmentId": "<Segment Id>",
          "text": "<Segment text>",
          "metadata": "<json map, segment with>",
          "distance": "<0.x float, segment match distance, smaller means closer>"
        }
      ]
    }
  ]

• /docs/multi-query:

[POST] /docs/multi-query

• Feature: Docs array query, query multi docs with one text, return N segment with docsId array
• Request params:
  • docsId array, queryText, return query result count
  • Sample

  {
    "docsIdList": ["xxxxxxxx", "yyyyyyyy"],
    "queryText": "<Query Text 1>",
    "nResults": "<UInt, return query result by sort number>",
    "removeDuplicates": "<(Optional)boolean, default:true, return segments will be removed if same text>"
  }

• Response body:
  • Query result: segment info array, include docsId, segmentId, segment text, metadata, distance
  • Response body sample

  [
    {
      "docsId": "xxxxxxxx",
      "segmentId": "<Segment Id>",
      "text": "<Segment text>",
      "metadata": "<json map, segment with>",
      "distance": "<0.x float, segment match distance, smaller means closer>"
    }
  ]

[POST] /segment/list

• Feature: List all segments in the docs
• Request params:
  • docsId
  • Sample

  {
    "docsId": "xxxxxxxx"
  }

• Response body:
  • docsId, docsName, segment info list
  • Response body sample

  {
    "docsId": "xxxxxxxx",
    "docsName": "<Docs Name>",
    "segmentInfoList": [
      {
        "SegmentId": "<Segment Id>",
        "text": "<Segment text>",
        "metadata": "<json map, segment with>"
      }
    ] 
  }

[POST] /segment/insert

• Feature: Insert segment by index. If not index, new segment will be inserted at last
• Request params:
  • docsId, new segment, index
  • Sample

  {
    "docsId": "xxxxxxxx",
    "segment": {
      "text": "<Segment text>",
      "metadata": "<json map, segment with>"
    },
    "index": "(Optional) UInt, if null or large than length, be inserted at last"
  }

• Response body:
  • new Segment ID
  • Response body sample

  {
    "segmentId": "xxxxxxxx"
  }

[POST] /segment/update

• Feature: Update segment
• Request params:
  • docsId, segment
  • Sample

  {
    "docsId": "xxxxxxxx",
    "segment": {
      "segmentId": "Segment Id",
      "text": "<Segment text>",
      "metadata": "<json map, segment with>"
    }
  }

• Response body:
  • Segment Id
  • Response body sample

  {
    "segmentId": "xxxxxxxx"
  }

[POST] /segment/delete

• Feature: Delete segment
• Request params:
  • Segment be deleted docsId and segmentId
  • Sample

  {
    "docsId": "xxxxxxxx",
    "segmentId": "xxxxxxxx"
  }

• Response body:
  • Segment Id
  • Response body sample

  {
    "segmentId": "xxxxxxxx"
  }

[POST] /dispose

• Feature: Dispose vector database connection
• Request params:
• Response body:
  • Dispose successfully text
  • Response body sample

  "Dispose successfully."
  

Build and Run

1. Build in shell script:

   dart compile exe bin/server.dart -o build/lite_embeddings_dart_server

2. Then the lite_embeddings_dart_server file will be in build folder
3. Copy config.json file to lite_embeddings_dart_server same folder
4. Run in shell script:

   ./lite_embeddings_dart_server

5. Terminal will show:

   INFO: 2024-06-24 14:48:05.862057: PID 34567: [HTTP] Start Server - http://0.0.0.0:9537/api
   

6. After server running, will create log folder and embeddings.log file in the folder, to record server running logs.