npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2024 – Pkg Stats / Ryan Hefner

initiative

v0.2.8

Published

Let's turn your LLM into Action Models with Zod

Downloads

16

Readme

Initiative

a langchain extension for building simple actions models

Use your zod schema to create a chain of actions and use it with any llm models.

npm install initiative

Usage

npm install initiative zod langchain @langchain/community

Import

import { TogetherAI } from "@langchain/community/llms/togetherai";

const model = new TogetherAI({
  modelName: "mistralai/Mixtral-8x7B-Instruct-v0.1", // Recommended model 7B
  apiKey: process.env.API_KEY,
});

Schema Definition

import { z } from "zod";
import { Implement, Schema, defaultPrompt } from "initiative";

const userState = {
  userSelected: z
    .enum(["YES", "NO"])
    .transform((x) => `User selected ${x} on text permissions`),
  userDragged: z.string().transform((x) => `User dragged ${x} out of the box`),
} satisfies State;

const schema = {
  getUserData: z
    .function()
    .describe("When user want to get the data of any other users")
    .args(z.string())
    .returns(z.string())
    .optional(),
  setName: z
    .function()
    .describe("When user input wants to update his/her name of the user")
    .args(z.string())
    .returns(z.void())
    .optional(),
  getTime: z
    .function()
    .args(z.enum(["NOW"]))
    .describe("When user input wants to know the current time")
    .returns(z.date())
    .optional(),
} satisfies Schema;

type FuncParam = {
  ctx: {};
  extra: {};
};

const { actionZod, combinedZod, dataZod, stateZod } = getZodCombined(
  schema,
  userState
);

const init = implement(combinedZod, {
  state: userState,
  functions: (z: FuncParam, y) => ({
    getUserData: async () => `${y?.userDragged}`,
    setName: () => {},
    getTime: () => new Date(),
  }),
  examples: [
    {
      Input: "What the time?",
      Output: {
        getTime: "NOW",
      },
    },
    {
      Input: "Set my name to Rajat",
      Output: {
        setName: "Rajat",
      },
    },
    {
      Input: "Find the person name Keanu",
      Output: {
        getUserData: "Keanu",
      },
    },
  ],
});

Initiate Action Chain with langchain

const chain = await createExtraction(model, init, {
  combinedZod,
  stateZod,
});

const userStateData = {
  userSelected: "YES",
  userDragged: "Toy",
};

const response = await chain.invoke("Get time", { state: userStateData });

console.log(response);

Output

{
  "input": "Get time",
  "response": {
    "raw": "\n<json>{\"getTime\":\"NOW\"}</json>",
    "validated": {
      "data": { "getTime": "NOW" },
      "json": { "getTime": "NOW" },
      "success": true
    }
  },
  "state": {
    "raw": {
      "userSelected": "YES",
      "userDragged": "Toy"
    },
    "validated": {
      "data": {
        "userSelected": "User selected YES on text permissions",
        "userDragged": "User dragged Toy out of the box"
      },
      "success": true
    }
  }
}

Execute Actions

const recipt = await executeActions(init, response, actionZod, {
  permissions: { getTime: true, setName: true, getUserData: true },
  params: {
    ctx: {},
    extra: {},
  },
});

console.log(recipt);

Output

{
  getTime: {
    value: 2024-03-30T19:12:45.614Z,
    key: "getTime",
    permission: true,
  }
}

Chained Action Execution

import { TogetherAI } from "@langchain/community/llms/togetherai";
import { z } from "zod";
import { createExtraction } from "./";
import {
  AvailableActions,
  executeChainActions,
  getZodChainedCombined,
  implementChain,
} from "./chain";
import { chainedActionPrompt } from "./lib/prompt";
import { State } from "./state";

const model = new TogetherAI({
  modelName: "mistralai/Mixtral-8x7B-Instruct-v0.1",
  apiKey: process.env.API_KEY,
});

const Schema = {
  searchUserWithName: z
    .function()
    .describe("When action needs imformation of a user to continue in order. ")
    .args(z.object({ name: z.string() }))
    .returns(z.string()),
  sentEmailToUser: z
    .function()
    .describe("When action is requisting to sent an email to someone. Pass name of user as param.")
    .args(z.object({ name: z.string() }))
    .returns(z.string()),
} satisfies AvailableActions;

const userState = {
  userSelected: z
    .enum(["YES", "NO"])
    .transform((x) => `User selected ${x} on text permissions`),
  userDragged: z.string().transform((x) => `User dragged ${x} out of the box`),
} satisfies State;

type FuncParam = {
  ctx: {};
  extra: {};
};

const userStateData = {
  userSelected: "YES",
  userDragged: "Toy",
};

const materials = getZodChainedCombined(Schema, userState);

const init = implementChain(Schema, materials, {
  functions: (x: FuncParam, y) => ({
    searchUserWithName: async ({ name }) => `Found ${name}`,
    sentEmailToUser: async ({name}) => `Senting email to ${name}`
  }),
  examples: [
    {
      Input: "Find user Rajat",
      Output: [{ searchUserWithName: { name: "Rajat" } }],
    },
    {
      Input: "Sent email to guy named Alex",
      Output: [
        { searchUserWithName: { name: "Alex" } },
        { sentEmailToUser: { name: "Alex" } }
    ],
    },
  ],
});

const chain = await createExtraction(
  model,
  init,
  {
    combinedZod: materials.combinedZod,
    stateZod: materials.stateZod,
  },
  chainedActionPrompt
);

const res = await chain.invoke("Find user Jovit and sent email to him", {
  state: userStateData,
});

console.log(res.response.validated?.success ? res.response.validated.data : "");

const x = await executeChainActions(init, res, {
  permissions: {
    searchUserWithName: true,
    sentEmailToUser: true,
  },
  params: {
    ctx: {},
    extra: {},
  },
});

console.log(x);

Output

Before execution

[
  {
    searchUserWithName: {
      name: "Jovit",
    },
  }, {
    sentEmailToUser: {
      name: "Jovit",
    },
  }
]

After execution

{
  searchUserWithName: {
    value: "Found Jovit",
    key: "searchUserWithName",
    iteration: 0,
    permission: true,
  },
  sentEmailToUser: {
    value: "Senting email to Jovit",
    key: "sentEmailToUser",
    iteration: 1,
    permission: true,
  },
}

Coming Soooooooooooon

  • [x] Add support direct actions initiation
  • [x] Add support normal zod schema
  • [x] Add support async actions
  • [x] Add support common cookies and headers
  • [x] Add chained actions support for multiple actions
  • [x] Proper validation and error handling for revalidation through LLM
  • [x] Permission for actions
  • [ ] RSC support with vercel/ai SDK

Packages used under the hood