feat: support human in the loop for TS

.changeset/eight-moons-perform.md

@@ -0,0 +1,6 @@
+---
+"@llamaindex/server": patch
+"create-llama": patch
+---
+
+feat: support human in the loop for TS

packages/create-llama/e2e/python/resolve_dependencies.spec.ts

@@ -20,6 +20,7 @@ const useCases: TemplateUseCase[] = [
   "financial_report",
   "code_generator",
   "document_generator",
+  "hitl",
 ];
 const dataSource: string = process.env.DATASOURCE
   ? process.env.DATASOURCE

packages/create-llama/e2e/shared/llamaindexserver_template.spec.ts

@@ -27,6 +27,7 @@ const templateUseCases = [
   "financial_report",
   "deep_research",
   "code_generator",
+  // "hitl",
 ];
 const ejectDir = "next";
 

packages/create-llama/e2e/typescript/resolve_dependencies.spec.ts

@@ -3,6 +3,7 @@ import { exec } from "child_process";
 import fs from "fs";
 import path from "path";
 import util from "util";
+import { NO_DATA_USE_CASES } from "../../helpers/constant";
 import {
   TemplateFramework,
   TemplateType,
@@ -25,6 +26,7 @@ const useCases: TemplateUseCase[] = [
   "financial_report",
   "code_generator",
   "document_generator",
+  "hitl",
 ];
 const dataSource: string = process.env.DATASOURCE
   ? process.env.DATASOURCE
@@ -83,7 +85,7 @@ test.describe("Test resolve TS dependencies", () => {
           });
         });
         // Skipping llamacloud for the use case doesn't use index.
-        if (useCase !== "code_generator" && useCase !== "document_generator") {
+        if (!useCase || !NO_DATA_USE_CASES.includes(useCase)) {
           test(`llamaParse - ${optionDescription}`, async () => {
             await runTest({
               templateType: templateType,

packages/create-llama/helpers/constant.ts

@@ -1,6 +1,15 @@
+import { TemplateUseCase } from "./types";
+
 export const COMMUNITY_OWNER = "run-llama";
 export const COMMUNITY_REPO = "create_llama_projects";
 export const LLAMA_PACK_OWNER = "run-llama";
 export const LLAMA_PACK_REPO = "llama_index";
 export const LLAMA_PACK_FOLDER = "llama-index-packs";
 export const LLAMA_PACK_FOLDER_PATH = `${LLAMA_PACK_OWNER}/${LLAMA_PACK_REPO}/main/${LLAMA_PACK_FOLDER}`;
+
+// these use cases don't have data folder, so no need to run generate and no need to getIndex
+export const NO_DATA_USE_CASES: TemplateUseCase[] = [
+  "code_generator",
+  "document_generator",
+  "hitl",
+];

packages/create-llama/helpers/index.ts

@@ -4,6 +4,7 @@ import path from "path";
 import picocolors, { cyan } from "picocolors";
 
 import fsExtra from "fs-extra";
+import { NO_DATA_USE_CASES } from "./constant";
 import { writeLoadersConfig } from "./datasources";
 import { createBackendEnvFile, createFrontendEnvFile } from "./env-variables";
 import { PackageManager } from "./get-pkg-manager";
@@ -98,8 +99,9 @@ async function generateContextData(
         }
       } else {
         console.log(`Running ${runGenerate} to generate the context data.`);
+
         const shouldRunGenerate =
-          useCase !== "code_generator" && useCase !== "document_generator"; // Artifact use case doesn't use index.
+          !useCase || !NO_DATA_USE_CASES.includes(useCase);
 
         if (shouldRunGenerate) {
           await callPackageManager(packageManager, true, ["run", "generate"]);

packages/create-llama/helpers/types.ts

@@ -59,7 +59,8 @@ export type TemplateUseCase =
   | "contract_review"
   | "agentic_rag"
   | "code_generator"
-  | "document_generator";
+  | "document_generator"
+  | "hitl";
 // Config for both file and folder
 export type FileSourceConfig =
   | {

packages/create-llama/helpers/typescript.ts

@@ -4,6 +4,7 @@ import path from "path";
 import { bold, cyan, red, yellow } from "picocolors";
 import { assetRelocator, copy } from "../helpers/copy";
 import { callPackageManager } from "../helpers/install";
+import { NO_DATA_USE_CASES } from "./constant";
 import { templatesDir } from "./dir";
 import { PackageManager } from "./get-pkg-manager";
 import { InstallTemplateArgs, ModelProvider, TemplateVectorDB } from "./types";
@@ -83,7 +84,7 @@ const installLlamaIndexServerTemplate = async ({
   }
 
   // Simplify use case code
-  if (useCase === "code_generator" || useCase === "document_generator") {
+  if (useCase && NO_DATA_USE_CASES.includes(useCase)) {
     // Artifact use case doesn't use index.
     // We don't need data.ts, generate.ts
     await fs.rm(path.join(root, "src", "app", "data.ts"));

packages/create-llama/questions/simple.ts

@@ -1,4 +1,5 @@
 import prompts from "prompts";
+import { NO_DATA_USE_CASES } from "../helpers/constant";
 import { EXAMPLE_10K_SEC_FILES, EXAMPLE_FILE } from "../helpers/datasources";
 import { askModelConfig } from "../helpers/providers";
 import { getTools } from "../helpers/tools";
@@ -11,7 +12,8 @@ type AppType =
   | "financial_report"
   | "deep_research"
   | "code_generator"
-  | "document_generator";
+  | "document_generator"
+  | "hitl";
 
 type SimpleAnswers = {
   appType: AppType;
@@ -57,6 +59,12 @@ export const askSimpleQuestions = async (
           value: "document_generator",
           description: "Build a OpenAI canvas-styled document generator.",
         },
+        {
+          title: "Human in the Loop",
+          value: "hitl",
+          description:
+            "Build a CLI command workflow that is reviewed by a human before execution",
+        },
       ],
     },
     questionHandlers,
@@ -81,7 +89,8 @@ export const askSimpleQuestions = async (
   );
   language = newLanguage;
 
-  if (appType !== "code_generator" && appType !== "document_generator") {
+  const shouldAskLlamaCloud = !NO_DATA_USE_CASES.includes(appType);
+  if (shouldAskLlamaCloud) {
     const { useLlamaCloud: newUseLlamaCloud } = await prompts(
       {
         type: "toggle",
@@ -170,6 +179,12 @@ const convertAnswers = async (
       tools: [],
       modelConfig: MODEL_GPT41,
     },
+    hitl: {
+      template: "llamaindexserver",
+      dataSources: [],
+      tools: [],
+      modelConfig: MODEL_GPT41,
+    },
   };
 
   const results = lookup[answers.appType];

packages/create-llama/templates/components/ui/use-cases/hitl/cli_human_input.tsx

@@ -0,0 +1,95 @@
+import { Button } from "@/components/ui/button";
+import { Card, CardContent, CardFooter } from "@/components/ui/card";
+import { JSONValue, useChatUI } from "@llamaindex/chat-ui";
+import React, { FC, useState } from "react";
+import { z } from "zod";
+
+// This schema is equivalent to the CLICommand model defined in events.py
+const CLIInputEventSchema = z.object({
+  command: z.string(),
+});
+type CLIInputEvent = z.infer<typeof CLIInputEventSchema>;
+
+const CLIHumanInput: FC<{
+  events: JSONValue[];
+}> = ({ events }) => {
+  const inputEvent = (events || [])
+    .map((ev) => {
+      const parseResult = CLIInputEventSchema.safeParse(ev);
+      return parseResult.success ? parseResult.data : null;
+    })
+    .filter((ev): ev is CLIInputEvent => ev !== null)
+    .at(-1);
+
+  const { append } = useChatUI();
+  const [confirmedValue, setConfirmedValue] = useState<boolean | null>(null);
+  const [editableCommand, setEditableCommand] = useState<string | undefined>(
+    inputEvent?.command,
+  );
+
+  // Update editableCommand if inputEvent changes (e.g. new event comes in)
+  React.useEffect(() => {
+    setEditableCommand(inputEvent?.command);
+  }, [inputEvent?.command]);
+
+  const handleConfirm = () => {
+    append({
+      content: "Yes",
+      role: "user",
+      annotations: [
+        {
+          type: "human_response",
+          data: {
+            execute: true,
+            command: editableCommand, // Use editable command
+          },
+        },
+      ],
+    });
+    setConfirmedValue(true);
+  };
+
+  const handleCancel = () => {
+    append({
+      content: "No",
+      role: "user",
+      annotations: [
+        {
+          type: "human_response",
+          data: {
+            execute: false,
+            command: inputEvent?.command,
+          },
+        },
+      ],
+    });
+    setConfirmedValue(false);
+  };
+
+  return (
+    <Card className="my-4">
+      <CardContent className="pt-6">
+        <p className="text-sm text-gray-700">
+          Do you want to execute the following command?
+        </p>
+        <input
+          disabled
+          type="text"
+          value={editableCommand || ""}
+          onChange={(e) => setEditableCommand(e.target.value)}
+          className="my-2 w-full overflow-x-auto rounded border border-gray-300 bg-gray-100 p-3 font-mono text-xs text-gray-800"
+        />
+      </CardContent>
+      {confirmedValue === null ? (
+        <CardFooter className="flex justify-end gap-2">
+          <>
+            <Button onClick={handleConfirm}>Yes</Button>
+            <Button onClick={handleCancel}>No</Button>
+          </>
+        </CardFooter>
+      ) : null}
+    </Card>
+  );
+};
+
+export default CLIHumanInput;

packages/create-llama/templates/components/use-cases/python/hitl/README-template.md

@@ -0,0 +1,109 @@
+This is a [LlamaIndex](https://www.llamaindex.ai/) project using [Workflows](https://docs.llamaindex.ai/en/stable/understanding/workflows/).
+
+## Getting Started
+
+First, setup the environment with uv:
+
+> **_Note:_** This step is not needed if you are using the dev-container.
+
+```shell
+uv sync
+```
+
+Then check the parameters that have been pre-configured in the `.env` file in this directory.
+Make sure you have set the `OPENAI_API_KEY` for the LLM.
+
+Then, run the development server:
+
+```shell
+uv run fastapi dev
+```
+
+Then open [http://localhost:8000](http://localhost:8000) with your browser to start the chat UI.
+
+To start the app optimized for **production**, run:
+
+```
+uv run fastapi run
+```
+
+## Configure LLM and Embedding Model
+
+You can configure [LLM model](https://docs.llamaindex.ai/en/stable/module_guides/models/llms) and [embedding model](https://docs.llamaindex.ai/en/stable/module_guides/models/embeddings) in [settings.py](app/settings.py).
+
+## Use Case
+
+This example shows how to use the LlamaIndexServer with a human in the loop. It allows you to start CLI commands that are reviewed by a human before execution.
+
+To update the workflow, you can modify the code in [`workflow.py`](app/workflow.py).
+
+You can start by sending an request on the [chat UI](http://localhost:8000) or you can test the `/api/chat` endpoint with the following curl request:
+
+```
+curl --location 'localhost:8000/api/chat' \
+--header 'Content-Type: application/json' \
+--data '{ "messages": [{ "role": "user", "content": "Show me the files in the current directory" }] }'
+```
+
+## How does HITL work?
+
+### Events
+
+The human-in-the-loop approach used here is based on a simple idea: the workflow pauses and waits for a human response before proceeding to the next step.
+
+To do this, you will need to implement two custom events:
+
+- [HumanInputEvent](https://github.com/run-llama/create-llama/blob/main/packages/server/src/utils/hitl/events.ts): This event is used to request input from the user.
+- [HumanResponseEvent](https://github.com/run-llama/create-llama/blob/main/packages/server/src/utils/hitl/events.ts): This event is sent to the workflow to resume execution with input from the user.
+
+In this example, we have implemented these two custom events in [`events.ts`](src/app/events.ts):
+
+- `cliHumanInputEvent` – to request input from the user for CLI command execution.
+- `cliHumanResponseEvent` – to resume the workflow with the response from the user.
+
+```typescript
+export const cliHumanInputEvent = humanInputEvent<{
+  type: "cli_human_input";
+  data: { command: string };
+  response: typeof cliHumanResponseEvent;
+}>();
+
+export const cliHumanResponseEvent = humanResponseEvent<{
+  type: "human_response";
+  data: { execute: boolean; command: string };
+}>();
+```
+
+### UI Component
+
+HITL also needs a custom UI component, that is shown when the LlamaIndexServer receives the `cliHumanInputEvent`. The name of the component is defined in the `type` field of the `cliHumanInputEvent` - in our case, it is `cli_human_input`, which corresponds to the [cli_human_input.tsx](./components/cli_human_input.tsx) component.
+
+The custom component must use `append` to send a message with a `human_response` annotation. The data of the annotation must be in the format of the response event `cliHumanResponseEvent`, in our case, for sending to execute the command `ls -l`, we would send:
+
+```tsx
+append({
+  content: "Yes",
+  role: "user",
+  annotations: [
+    {
+      type: "human_response",
+      data: {
+        execute: true,
+        command: "ls -l", // The command to execute
+      },
+    },
+  ],
+});
+```
+
+This component displays the command to execute and the user can choose to execute or cancel the command execution.
+
+## Learn More
+
+To learn more about LlamaIndex, take a look at the following resources:
+
+- [LlamaIndex Documentation](https://docs.llamaindex.ai) - learn about LlamaIndex.
+- [Workflows Introduction](https://docs.llamaindex.ai/en/stable/understanding/workflows/) - learn about LlamaIndex workflows.
+- [LlamaIndex Server](https://pypi.org/project/llama-index-server/)
+
+You can check out [the LlamaIndex GitHub repository](https://github.com/run-llama/llama_index) - your feedback and contributions are welcome!