choco builder

ChocoBuilder (aka Chocolate Factory)

Read the docs →

What is ChocoBuilder?

ChocoBuilder (origin Chocolate Factory) is an open source LLM application development framework designed to help you easily create a powerful software development SDLC + LLM generation assistant.

For native (Android/iOS/Embedded device) SDK: see in https://github.com/unit-mesh/edge-infer

QuickStart

Method 1: Integrate into JVM project

Module list: https://central.sonatype.com/namespace/cc.unitmesh

dependencies {
    // 核心模块
    implementation ' cc.unitmesh:cocoa-core:1.0.0 '
    // Pinecone
    implementation ' cc.unitmesh:store-pinecone:1.0.0 '
    // ElasticSearch
    implementation ' cc.unitmesh:store-elasticsearch:1.0.0 '
    // ...其它模块
}

See more examples: examples/

Method 2: Use RAGScript

@file:DependsOn( " cc.unitmesh:rag-script:1.0.0 " )

import cc.unitmesh.rag.*

rag {
    indexing {
        val chunks = document( " README.md " ).split()
        store.indexing(chunks)
    }

    querying {
        store.findRelevant( " workflow dsl design " )
            .lowInMiddle()
            . also {
                println (it)
            }
    }
}

Method 3: Local deployment example

git clone https://github.com/unit-mesh/chocolate-factory
# modify OPENAI_API_KEY and OPENAI_HOST in docker-compose.yml
docker-compose up

Use Cases

Desktop/IDE:

• https://github.com/unit-mesh/auto-dev - The AI-powered coding wizard with multilingual support , auto code generation , and a helpful bug-slaying assistant
• https://github.com/phodal/shire - AI Coding Agent Language

Server:

• https://github.com/unit-mesh/devops-genius/

Android:

• https://github.com/unit-mesh/android-semantic-search-kit

Development

See in https://framework.unitmesh.cc/ or see in documents

Design Philosophy: Domain Driven Problem-Solving

The key concepts of ChocoBuilder are:

(PS: Origin made by Michael Plöd at Aligning organization and architecture with strategic DDD)

A user's problem is processed by the following steps:

1. ProblemClarifier.kt
2. ProblemAnalyzer.kt
3. SolutionDesigner.kt
4. SolutionReviewer.kt
5. SolutionExecutor.kt

• Online Demo: https://framework.unitmesh.cc/ (TODO)
• Video Demo: https://www.bilibili.com/video/BV1T14y1C7p2
• Gif Demo:

Example 1: Frontend Screenshot

• Step 1: ProblemClarifier: Use responsive layout to write a chat page
  • Step 1.1: ProblemClarifier: On the left is a navigation, in the middle is a chat area, and at the bottom of the chat area is an input button.
• Step 2: SolutionDesigner: Please confirm whether the following design meets your requirements. If it meets, please reply "YES". If it does not meet, please make your request.
• Step 3: SolutionExecutor: Generate a chat page

Example 2: Semantic Code Search

• Step 1: ProblemAnalyzer analyzes user needs and converts them into multiple semantic queries
  • Chinese, English, HyDE modes
• Step 2: SolutionExecutor: Retrieve the most relevant code snippets from the database according to the user's needs, and summarized by ChatGPT

Example input: How is Semantic Workflow implemented?

Final output:

Example 3: Testcase Generator

• Step 1: ProblemAnalyzer analyzes the user's requirements and confirms whether it is a requirement generated by a test case.
  • Multi-Temperature mode: TemperatureMode.Default, TemperatureMode.Creative
• Step 2: SolutionDesigner designs the solution for the test case generation
• Step 3: SolutionReviewer confirms whether the solution meets the user's needs

Example input: User published article

Final output:

Examples 4: Code Interpreter

• Step 1: SolutionExecutor

Example 1: Write a multiplication table

Output example:

1    2    3    4    5    6    7    8    9
2    4    6    8    10    12    14    16    18
3    6    9    12    15    18    21    24    27
4    8    12    16    20    24    28    32    36
5    10    15    20    25    30    35    40    45
6    12    18    24    30    36    42    48    54
7    14    21    28    35    42    49    56    63
8    16    24    32    40    48    56    64    72
9    18    27    36    45    54    63    72    81    

Example 2: Generate charts based on requirements (TODO)

Generate an electricity bill chart for the first half of 2023, the information is as follows: ###1~June: 201.2,222,234.3,120.2,90,90.4###

Process code:

 %use lets-plot

import kotlin.math.PI
import kotlin.random.Random


val incomeData = mapOf(
    "x" to listOf("一月", "二月", "三月", "四月", "五月", "六月"),
    "y" to listOf(201.2, 222, 234.3, 120.2, 90, 94.4)
)
letsPlot(incomeData) { x = "x"; y = "y" } +
        geomBar(stat = Stat.identity) +
        geomText(labelFormat = "${.2f}") { label = "y"; } +
        ggtitle("2023 年上半年电费")

Final output:

License

RAG relevant modules were inspired by

• LangChain4j
• LangChain
• LlamaIndex
• Spring AI

Some RAG modules based on LangChain4j and Spring AI which is licensed under the Apache License 2.0.

This code is distributed under the MPL 2.0 license. See LICENSE in this directory.