Package com.hw.langchain.chains.llm

Class LLMChain

java.lang.Object

com.hw.langchain.chains.base.Chain

com.hw.langchain.chains.llm.LLMChain

Direct Known Subclasses:

ConversationChain

public class LLMChain
extends Chain

Chain to run queries against LLMs

Author:

HamaWhite

Field Summary

Fields

Modifier and Type	Field	Description
protected BaseLanguageModel	llm
protected String	outputKey
protected BaseLLMOutputParser<String>	outputParser	Output parser to use.
protected BasePromptTemplate	prompt	Prompt object to use.
protected boolean	returnFinalOnly	Whether to return only the final parsed result.

Fields inherited from class com.hw.langchain.chains.base.Chain

memory

Constructor Summary

Constructors

Constructor	Description
LLMChain(BaseLanguageModel llm, BasePromptTemplate prompt)
LLMChain(BaseLanguageModel llm, BasePromptTemplate prompt, String outputKey)

Method Summary

Modifier and Type	Method	Description
protected reactor.core.publisher.Flux<Map<String,String>>	asyncInnerCall(Map<String,Object> inputs)	Runs the logic of this chain and returns the async output.
reactor.core.publisher.Flux<String>	asyncPredict(Map<String,Object> kwargs)	Format prompt with kwargs and pass to LLM async.
String	chainType()
protected Map<String,String>	innerCall(Map<String,Object> inputs)	Runs the logic of this chain and returns the output.
List<String>	inputKeys()	Will be whatever keys the prompt expects.
List<String>	outputKeys()	Will always return text key.
String	predict(Map<String,Object> kwargs)	Format prompt with kwargs and pass to LLM.
<T> T	predictAndParse(Map<String,Object> kwargs)	Call predict and then parse the results.

Methods inherited from class com.hw.langchain.chains.base.Chain

asyncCall, asyncCall, asyncRun, asyncRun, call, call, prepInputs, run, run

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Field Details

llm

protected BaseLanguageModel llm

prompt

protected BasePromptTemplate prompt

Prompt object to use.

outputKey

protected String outputKey

outputParser

protected BaseLLMOutputParser<String> outputParser

Output parser to use. Defaults to one that takes the most likely string but does not change it.

returnFinalOnly

protected boolean returnFinalOnly

Whether to return only the final parsed result. Defaults to true. If false, will return a bunch of extra information about the generation.

Constructor Details

LLMChain

public LLMChain(BaseLanguageModel llm,
 BasePromptTemplate prompt)

LLMChain

public LLMChain(BaseLanguageModel llm,
 BasePromptTemplate prompt,
 String outputKey)

Method Details

chainType

public String chainType()

Specified by:

chainType in class Chain

inputKeys

public List<String> inputKeys()

Will be whatever keys the prompt expects.

Specified by:

inputKeys in class Chain

Returns:

the list of input keys

outputKeys

public List<String> outputKeys()

Will always return text key.

Specified by:

outputKeys in class Chain

Returns:

the list of output keys

innerCall

protected Map<String,String> innerCall(Map<String,Object> inputs)

Description copied from class: Chain

Runs the logic of this chain and returns the output.

Specified by:

innerCall in class Chain

Parameters:

inputs - the inputs to be processed by the chain

Returns:

a map containing the output generated by the chain

asyncInnerCall

protected reactor.core.publisher.Flux<Map<String,String>> asyncInnerCall(Map<String,Object> inputs)

Description copied from class: Chain

Runs the logic of this chain and returns the async output.

Overrides:

asyncInnerCall in class Chain

Parameters:

inputs - the inputs to be processed by the chain

Returns:

a map flux containing the output generated event by the chain

predict

public String predict(Map<String,Object> kwargs)

Format prompt with kwargs and pass to LLM.

Parameters:

kwargs - Keys to pass to prompt template.

Returns:

Completion from LLM.

asyncPredict

public reactor.core.publisher.Flux<String> asyncPredict(Map<String,Object> kwargs)

Format prompt with kwargs and pass to LLM async.

Parameters:

kwargs - Keys to pass to prompt template.

Returns:

Completion from LLM.

predictAndParse

public <T> T predictAndParse(Map<String,Object> kwargs)

Call predict and then parse the results.