LLMChain

This is an experimental port of langchain( currently v0.0.139 ) to android/JVM/Kotlin Multiplatform. Please note that this project is currently in the proof-of-concept stage, and is subject to change.

Maven repository

Only the SNAPSHOT version is published.

Add to repositories

repositories {
    mavenCentral()
    maven { url "https://s01.oss.sonatype.org/content/repositories/snapshots/" }
}

Add to dependencies

• Android/JVM developers are advised to use the android branch README dependency directions.
• For Kotlin Multiplatform developers, try to add the following

dependencies {
    implementation("io.github.wangmuy.llmchain.kmp:core:0.0.1-SNAPSHOT") { changing=true }
    implementation("io.github.wangmuy.llmchain.kmp:serviceprovider-openai:0.0.1-SNAPSHOT") { changing=true }
}

Be sure to checkout the support matrix.

Multiplatform support matrix

package	platform	isCompiled	isTested
core	jvm	true	true
core	ios	false	false
core	native	true	true
core	js	true	true
core	wasm	true	false
serviceprovider-openai	jvm	true	true
serviceprovider-openai	ios	false	false
serviceprovider-openai	native	true	true
serviceprovider-openai	js	true	false
serviceprovider-openai	wasm	false	false

Quickstart

Here's the almost one-to-one translation of langchain Quickstart Guide in Quickstart.kt, including all the modules/components.

LLMs

val llm = OpenAIChat(APIKEY, proxy = PROXY)
llm.invocationParams[OpenAIChat.REQ_TEMPERATURE] = 0.9
val text = "What would be a good company name for a company that makes colorful socks?"
val output = llm.invoke(text, null)
println("output=\n$output")

Prompt templates

val prompt = PromptTemplate(
    inputVariables = listOf("product"),
    template = "What is a good name for a company that makes {product}?")
val formatted = prompt.format(mapOf("product" to "colorful socks"))
assertEquals("What is a good name for a company that makes colorful socks?", formatted)

Chains

val llm = OpenAIChat(APIKEY)
llm.invocationParams[OpenAIChat.REQ_TEMPERATURE] = 0.9
val prompt = PromptTemplate(
    inputVariables = listOf("product"),
    template = "What is a good name for a company that makes {product}?")
val chain = LLMChain(llm = llm, prompt = prompt)
val output = chain.run(mapOf("product" to "colorful socks"))
println("output=\n$output")

Agents

val llm = OpenAIChat(APIKEY)
llm.invocationParams[OpenAIChat.REQ_TEMPERATURE] = 0.0
val fakeSerpApiTool = Tool(
    name = "Search",
    description = "A search engine. Useful for when you need to answer questions about current events. Input should be a search query.",
    func = {_, _ -> "San Francisco Temperature Yesterday. Maximum temperature yesterday: 57 °F (at 1:56 pm) Minimum temperature yesterday: 49 °F (at 1:56 am)"}
)
val llmMathTool = LLMMathChain.asTool(llm)
val agentExecutor = Factory.initializeAgent(listOf(fakeSerpApiTool, llmMathTool), llm,
    Factory.AGENT_TYPE_ZERO_SHOT_REACT_DESCRIPTION)
agentExecutor.maxIterations = 4
val output = agentExecutor.run(mapOf("input" to "What was the high temperature in SF yesterday in Fahrenheit? What is that number raised to the .023 power?"))
println("output=\n$output")

Memory

val logCallbackHandler = object: DefaultCallbackHandler() {
    override fun onText(text: String, verbose: Boolean) {
        println(text)
    }
}
val callbackManager = CallbackManager(mutableListOf(logCallbackHandler))
val llm  = OpenAIChat(APIKEY).apply {
    invocationParams[OpenAIChat.REQ_MAX_TOKENS] = 50
}
llm.callbackManager = callbackManager
val conversation = ConversationChain(llm, verbose = true, callbackManager = callbackManager)
var output: Map<String, Any> = emptyMap()
var outputStr: String = ""
output = conversation.invoke(mapOf("input" to "Hi there!"))
outputStr = output[conversation.outputKey]!!.toString()
println("output=\n$outputStr")
output = conversation.invoke(mapOf("input" to "I'm doing well! Just having a conversation with an AI."))

implementations

• Schema

  • nearly all basic schema interfaces

• LLM

  • BaseLLM
  • LLM

• Prompt template

  • BasePromptTemplate
  • StringPromptTemplate
  • StringPromptValue
  • PromptTemplate

• Chain

  • Chain
  • LLMChain
  • LLMMathChain: currently use LLM, not the actual calculator
  • ConversationChain
  • RouterChain/MultiRouteChain/LLMRouterChain/MultiPromptChain Note: Only available in jvm/native due to limited implementation of RegexOption.DOT_MATCHES_ALL
  • SequentialChain/SimpleSequentialChain

• Agent

  • Agent
  • AgentExecutor
  • ZeroShotAgent

• Tools

  • BaseTool
  • Tool
  • InvalidTool

• Memory

  • SimpleMemory
  • ChatMemory

• DocStore

  • DocStore
  • InMemoryDocStore

• VectorStore

  • VectorStore
  • SimpleVectorStore
  • VectorStoreRetriever

• Embedding

  • Embeddings

• LLM service provider

  • OpenAI
    • OpenAIChat
    • OpenAIEmbedding
    • Function Calling

License

Copyright 2023 wangmuy

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.