feat: Add legal document review use case (#467)

Author: leehuwuj

.changeset/stale-scissors-turn.md

@@ -0,0 +1,5 @@
+---
+"create-llama": patch
+---
+
+Add contract review use case (Python)

e2e/shared/extractor_template.spec.ts

@@ -0,0 +1,64 @@
+/* eslint-disable turbo/no-undeclared-env-vars */
+import { expect, test } from "@playwright/test";
+import { ChildProcess } from "child_process";
+import fs from "fs";
+import path from "path";
+import { TemplateAgents, TemplateFramework } from "../../helpers";
+import { createTestDir, runCreateLlama } from "../utils";
+
+const templateFramework: TemplateFramework = process.env.FRAMEWORK
+  ? (process.env.FRAMEWORK as TemplateFramework)
+  : "fastapi";
+const dataSource: string = process.env.DATASOURCE
+  ? process.env.DATASOURCE
+  : "--example-file";
+const templateAgents: TemplateAgents[] = ["extractor", "contract_review"];
+
+// The reflex template currently only works with FastAPI and files (and not on Windows)
+if (
+  process.platform !== "win32" &&
+  templateFramework === "fastapi" &&
+  dataSource === "--example-file"
+) {
+  for (const agents of templateAgents) {
+    test.describe(`Test reflex template ${agents} ${templateFramework} ${dataSource}`, async () => {
+      let appPort: number;
+      let name: string;
+      let appProcess: ChildProcess;
+      let cwd: string;
+
+      // Create reflex app
+      test.beforeAll(async () => {
+        cwd = await createTestDir();
+        appPort = Math.floor(Math.random() * 10000) + 10000;
+        const result = await runCreateLlama({
+          cwd,
+          templateType: "reflex",
+          templateFramework: "fastapi",
+          dataSource: "--example-file",
+          vectorDb: "none",
+          port: appPort,
+          postInstallAction: "runApp",
+          agents,
+        });
+        name = result.projectName;
+        appProcess = result.appProcess;
+      });
+
+      test.afterAll(async () => {
+        appProcess.kill();
+      });
+
+      test("App folder should exist", async () => {
+        const dirExists = fs.existsSync(path.join(cwd, name));
+        expect(dirExists).toBeTruthy();
+      });
+      test("Frontend should have a title", async ({ page }) => {
+        await page.goto(`http://localhost:${appPort}`);
+        await expect(page.getByText("Built by LlamaIndex")).toBeVisible({
+          timeout: 2000 * 60,
+        });
+      });
+    });
+  }
+}

e2e/shared/reflex_template.spec.ts

@@ -113,7 +113,7 @@ export async function runCreateLlama({
   if (observability) {
     commandArgs.push("--observability", observability);
   }
-  if (templateType === "multiagent" && agents) {
+  if ((templateType === "multiagent" || templateType === "reflex") && agents) {
     commandArgs.push("--agents", agents);
   }
 

e2e/utils.ts

@@ -18,6 +18,7 @@ export const EXAMPLE_10K_SEC_FILES: TemplateDataSource[] = [
       url: new URL(
         "https://s2.q4cdn.com/470004039/files/doc_earnings/2023/q4/filing/_10-K-Q4-2023-As-Filed.pdf",
       ),
+      filename: "apple_10k_report.pdf",
     },
   },
   {
@@ -26,10 +27,21 @@ export const EXAMPLE_10K_SEC_FILES: TemplateDataSource[] = [
       url: new URL(
         "https://ir.tesla.com/_flysystem/s3/sec/000162828024002390/tsla-20231231-gen.pdf",
       ),
+      filename: "tesla_10k_report.pdf",
     },
   },
 ];
 
+export const EXAMPLE_GDPR: TemplateDataSource = {
+  type: "file",
+  config: {
+    url: new URL(
+      "https://eur-lex.europa.eu/legal-content/EN/TXT/PDF/?uri=CELEX:32016R0679",
+    ),
+    filename: "gdpr.pdf",
+  },
+};
+
 export function getDataSources(
   files?: string,
   exampleFile?: boolean,

helpers/datasources.ts

@@ -118,7 +118,8 @@ const prepareContextData = async (
       const destPath = path.join(
         root,
         "data",
-        path.basename(dataSourceConfig.url.toString()),
+        dataSourceConfig.filename ??
+          path.basename(dataSourceConfig.url.toString()),
       );
       await downloadFile(dataSourceConfig.url.toString(), destPath);
     } else {
@@ -192,7 +193,7 @@ export const installTemplate = async (
     if (
       props.template === "streaming" ||
       props.template === "multiagent" ||
-      props.template === "extractor"
+      props.template === "reflex"
     ) {
       await createBackendEnvFile(props.root, props);
     }

helpers/index.ts

@@ -405,8 +405,8 @@ export const installPythonTemplate = async ({
 >) => {
   console.log("\nInitializing Python project with template:", template, "\n");
   let templatePath;
-  if (template === "extractor") {
-    templatePath = path.join(templatesDir, "types", "extractor", framework);
+  if (template === "reflex") {
+    templatePath = path.join(templatesDir, "types", "reflex");
   } else {
     templatePath = path.join(templatesDir, "types", "streaming", framework);
   }
@@ -472,24 +472,6 @@ export const installPythonTemplate = async ({
       cwd: path.join(compPath, "engines", "python", engine),
     });
 
-    // Copy agent code
-    if (template === "multiagent") {
-      if (agents) {
-        await copy("**", path.join(root), {
-          parents: true,
-          cwd: path.join(compPath, "agents", "python", agents),
-          rename: assetRelocator,
-        });
-      } else {
-        console.log(
-          red(
-            "There is no agent selected for multi-agent template. Please pick an agent to use via --agents flag.",
-          ),
-        );
-        process.exit(1);
-      }
-    }
-
     // Copy router code
     await copyRouterCode(root, tools ?? []);
   }
@@ -503,6 +485,28 @@ export const installPythonTemplate = async ({
     });
   }
 
+  if (template === "multiagent" || template === "reflex") {
+    if (agents) {
+      const sourcePath =
+        template === "multiagent"
+          ? path.join(compPath, "agents", "python", agents)
+          : path.join(compPath, "reflex", agents);
+
+      await copy("**", path.join(root), {
+        parents: true,
+        cwd: sourcePath,
+        rename: assetRelocator,
+      });
+    } else {
+      console.log(
+        red(
+          `There is no agent selected for ${template} template. Please pick an agent to use via --agents flag.`,
+        ),
+      );
+      process.exit(1);
+    }
+  }
+
   console.log("Adding additional dependencies");
 
   const addOnDependencies = getAdditionalDependencies(

helpers/python.ts

@@ -1,5 +1,5 @@
 import { SpawnOptions, spawn } from "child_process";
-import { TemplateFramework } from "./types";
+import { TemplateFramework, TemplateType } from "./types";
 
 const createProcess = (
   command: string,
@@ -58,17 +58,17 @@ export function runTSApp(appPath: string, port: number) {
 
 export async function runApp(
   appPath: string,
-  template: string,
+  template: TemplateType,
   framework: TemplateFramework,
   port?: number,
 ): Promise<void> {
   try {
     // Start the app
     const defaultPort =
-      framework === "nextjs" || template === "extractor" ? 3000 : 8000;
+      framework === "nextjs" || template === "reflex" ? 3000 : 8000;
 
     const appRunner =
-      template === "extractor"
+      template === "reflex"
         ? runReflexApp
         : framework === "fastapi"
           ? runFastAPIApp

helpers/run-app.ts

@@ -20,11 +20,11 @@ export type ModelConfig = {
   isConfigured(): boolean;
 };
 export type TemplateType =
-  | "extractor"
   | "streaming"
   | "community"
   | "llamapack"
-  | "multiagent";
+  | "multiagent"
+  | "reflex";
 export type TemplateFramework = "nextjs" | "express" | "fastapi";
 export type TemplateUI = "html" | "shadcn";
 export type TemplateVectorDB =
@@ -49,14 +49,21 @@ export type TemplateDataSource = {
 };
 export type TemplateDataSourceType = "file" | "web" | "db";
 export type TemplateObservability = "none" | "traceloop" | "llamatrace";
-export type TemplateAgents = "financial_report" | "blog" | "form_filling";
+export type TemplateAgents =
+  | "financial_report"
+  | "blog"
+  | "form_filling"
+  | "extractor"
+  | "contract_review";
 // Config for both file and folder
 export type FileSourceConfig =
   | {
       path: string;
+      filename?: string;
     }
   | {
       url: URL;
+      filename?: string;
     };
 export type WebSourceConfig = {
   baseUrl?: string;

helpers/types.ts

@@ -153,7 +153,7 @@ export const installTSTemplate = async ({
     } else {
       console.log(
         red(
-          "There is no agent selected for multi-agent template. Please pick an agent to use via --agents flag.",
+          `There is no agent selected for ${template} template. Please pick an agent to use via --agents flag.`,
         ),
       );
       process.exit(1);

helpers/typescript.ts

@@ -215,7 +215,7 @@ const options = program.opts();
 
 if (
   process.argv.includes("--no-llama-parse") ||
-  options.template === "extractor"
+  options.template === "reflex"
 ) {
   options.useLlamaParse = false;
 }

index.ts

@@ -49,7 +49,7 @@ export const getDataSourceChoices = (
     );
   }
 
-  if (framework === "fastapi" && template !== "extractor") {
+  if (framework === "fastapi" && template !== "reflex") {
     choices.push({
       title: "Use website content (requires Chrome)",
       value: "web",

questions/datasources.ts

@@ -95,8 +95,8 @@ export const askProQuestions = async (program: QuestionArgs) => {
     return; // early return - no further questions needed for llamapack projects
   }
 
-  if (program.template === "extractor") {
-    // Extractor template only supports FastAPI, empty data sources, and llamacloud
+  if (program.template === "reflex") {
+    // Reflex template only supports FastAPI, empty data sources, and llamacloud
     // So we just use example file for extractor template, this allows user to choose vector database later
     program.dataSources = [EXAMPLE_FILE];
     program.framework = "fastapi";
@@ -354,11 +354,8 @@ export const askProQuestions = async (program: QuestionArgs) => {
     // default to use LlamaParse if using LlamaCloud
     program.useLlamaParse = true;
   } else {
-    // Extractor template doesn't support LlamaParse and LlamaCloud right now (cannot use asyncio loop in Reflex)
-    if (
-      program.useLlamaParse === undefined &&
-      program.template !== "extractor"
-    ) {
+    // Reflex template doesn't support LlamaParse and LlamaCloud right now (cannot use asyncio loop in Reflex)
+    if (program.useLlamaParse === undefined && program.template !== "reflex") {
       // if already set useLlamaParse, don't ask again
       if (program.dataSources.some((ds) => ds.type === "file")) {
         const { useLlamaParse } = await prompts(