Skip to main content

How to add values to a chain's state

An alternate way of passing data through steps of a chain is to leave the current values of the chain state unchanged while assigning a new value under a given key. The RunnablePassthrough.assign() static method takes an input value and adds the extra arguments passed to the assign function.

This is useful in the common LangChain Expression Language pattern of additively creating a dictionary to use as input to a later step.

Hereโ€™s an example:

import {
} from "@langchain/core/runnables";

const runnable = RunnableParallel.from({
extra: RunnablePassthrough.assign({
mult: (input: { num: number }) => input.num * 3,
modified: (input: { num: number }) => input.num + 1,

await runnable.invoke({ num: 1 });
{ extra: { num: 1, mult: 3, modified: 2 } }

Letโ€™s break down whatโ€™s happening here.

  • The input to the chain is {"num": 1}. This is passed into a RunnableParallel, which invokes the runnables it is passed in parallel with that input.
  • The value under the extra key is invoked. RunnablePassthrough.assign() keeps the original keys in the input dict ({"num": 1}), and assigns a new key called mult. The value is lambda x: x["num"] * 3), which is 3. Thus, the result is {"num": 1, "mult": 3}.
  • {"num": 1, "mult": 3} is returned to the RunnableParallel call, and is set as the value to the key extra.
  • At the same time, the modified key is called. The result is 2, since the lambda extracts a key called "num" from its input and adds one.

Thus, the result is {'extra': {'num': 1, 'mult': 3}, 'modified': 2}.


One convenient feature of this method is that it allows values to pass through as soon as they are available. To show this off, weโ€™ll use RunnablePassthrough.assign() to immediately return source docs in a retrieval chain:

yarn add @langchain/openai
import { StringOutputParser } from "@langchain/core/output_parsers";
import { ChatPromptTemplate } from "@langchain/core/prompts";
import {
} from "@langchain/core/runnables";
import { ChatOpenAI, OpenAIEmbeddings } from "@langchain/openai";
import { MemoryVectorStore } from "langchain/vectorstores/memory";

const vectorstore = await MemoryVectorStore.fromDocuments(
[{ pageContent: "harrison worked at kensho", metadata: {} }],
new OpenAIEmbeddings()

const retriever = vectorstore.asRetriever();

const template = `Answer the question based only on the following context:

Question: {question}

const prompt = ChatPromptTemplate.fromTemplate(template);

const model = new ChatOpenAI({ model: "gpt-4o" });

const generationChain = prompt.pipe(model).pipe(new StringOutputParser());

const retrievalChain = RunnableSequence.from([
context: retriever.pipe((docs) => docs[0].pageContent),
question: new RunnablePassthrough(),
RunnablePassthrough.assign({ output: generationChain }),

const stream = await"where did harrison work?");

for await (const chunk of stream) {
{ question: "where did harrison work?" }
{ context: "harrison worked at kensho" }
{ output: "" }
{ output: "H" }
{ output: "arrison" }
{ output: " worked" }
{ output: " at" }
{ output: " Kens" }
{ output: "ho" }
{ output: "." }
{ output: "" }

We can see that the first chunk contains the original "question" since that is immediately available. The second chunk contains "context" since the retriever finishes second. Finally, the output from the generation_chain streams in chunks as soon as it is available.

Next stepsโ€‹

Now youโ€™ve learned how to pass data through your chains to help to help format the data flowing through your chains.

To learn more, see the other how-to guides on runnables in this section.

Was this page helpful?

You can also leave detailed feedback on GitHub.