Support combining multiple output parsers

examples/src/prompts/combining_parser.ts

@@ -0,0 +1,66 @@
+import { OpenAI, PromptTemplate } from "langchain";
+import { StructuredOutputParser, RegexParser, CombiningOutputParser } from "langchain/output_parsers";
+
+export const run = async () => {
+
+  const answerParser = StructuredOutputParser.fromNamesAndDescriptions({
+    answer: "answer to the user's question",
+    source: "source used to answer the user's question, should be a website.",
+  });
+
+  const confidenceParser = new RegexParser(/Confidence: (A|B|C), Explanation: (.*)/, ["confidence", "explanation"], "noConfidence");
+
+  const parser = new CombiningOutputParser(answerParser, confidenceParser);
+  const formatInstructions = parser.getFormatInstructions();
+
+  const prompt = new PromptTemplate({
+    template:
+      "Answer the users question as best as possible.\n{format_instructions}\n{question}",
+    inputVariables: ["question"],
+    partialVariables: { format_instructions: formatInstructions },
+  });
+
+  const model = new OpenAI({ temperature: 0 });
+
+  const input = await prompt.format({
+    question: "What is the capital of France?",
+  });
+  const response = await model.call(input);
+
+  console.log(input);
+  /*
+  Answer the users question as best as possible.
+  For your first output: The output should be a markdown code snippet formatted in the following schema:
+
+  ```json
+  {
+      "answer": string // answer to the user's question
+      "source": string // source used to answer the user's question, should be a website.
+  }
+  ```
+  Complete that output fully. Then produce another output: Your response should match the following regex: //Confidence: (A|B|C), Explanation: (.*)//
+
+  What is the capital of France?
+  */
+
+  console.log(response);
+  /*
+  ```json
+  {
+      "answer": "Paris",
+      "source": "https://en.wikipedia.org/wiki/France"
+  }
+  ```
+  //Confidence: A, Explanation: Paris is the capital of France according to Wikipedia.//
+  */
+
+  console.log(await parser.parse(response));
+  /*
+  {
+    answer: 'Paris',
+    source: 'https://en.wikipedia.org/wiki/France',
+    confidence: 'A',
+    explanation: 'Paris is the capital of France according to Wikipedia.//'
+    }
+  */
+};

langchain/src/output_parsers/combining.ts

@@ -0,0 +1,29 @@
+import { BaseOutputParser } from "../schema/index.js";
+
+/**
+ * Class to combine multiple output parsers
+ * @augments BaseOutputParser
+ */
+export class CombiningOutputParser extends BaseOutputParser {
+  parsers: BaseOutputParser[];
+
+  constructor(...parsers: BaseOutputParser[]) {
+    super();
+    this.parsers = parsers;
+  }
+
+  async parse(input: string): Promise<Record<string, any>> {
+    let ret = {};
+    for (const p of this.parsers) {
+      ret = { ...ret, ...((await p.parse(input)) as Record<string, any>) };
+    }
+    return ret;
+  }
+
+  getFormatInstructions(): string {
+    const initial = "For your first output: " + this?.parsers?.[0]?.getFormatInstructions();
+    const subsequent = this.parsers.slice(1).map((p) => "Complete that output fully. Then produce another output: " + p.getFormatInstructions()).join("\n");
+    return initial + "\n" + subsequent;
+  }
+}
+

langchain/src/output_parsers/index.ts

@@ -2,3 +2,4 @@ export { ListOutputParser, CommaSeparatedListOutputParser } from "./list.js";
 export { RegexParser } from "./regex.js";
 export { StructuredOutputParser } from "./structured.js";
 export { OutputFixingParser } from "./fix.js";
+export { CombiningOutputParser } from "./combining.js";