Release 0.1.3 (#84)

Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>

.changeset/config.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://unpkg.com/@changesets/config@3.0.0/schema.json",
   "changelog": "@changesets/cli/changelog",
-  "commit": true,
+  "commit": false,
   "fixed": [],
   "linked": [],
   "access": "public",

.changeset/curly-trains-obey.md

@@ -1,5 +0,0 @@
----
-"create-llama": patch
----
-
-Remove UI question (use shadcn as default). Use `html` UI by calling create-llama with --ui html parameter

.changeset/curvy-candles-mix.md

@@ -1,5 +0,0 @@
----
-"create-llama": patch
----
-
-Update loaders and tools config to yaml format

.changeset/dry-seals-sniff.md

@@ -1,5 +0,0 @@
----
-"create-llama": patch
----
-
-Let user select multiple datasources (URLs, files and folders)

.changeset/healthy-insects-check.md

@@ -1,5 +0,0 @@
----
-"create-llama": patch
----
-
-Add Dockerfile template

.changeset/itchy-seahorses-move.md

@@ -1,5 +0,0 @@
----
-"create-llama": patch
----
-
-Merge non-streaming and streaming template to one

.changeset/quiet-actors-help.md

@@ -1,5 +0,0 @@
----
-"create-llama": patch
----
-
-Add support for agent generation for Typescript

.github/workflows/e2e.yml

@@ -24,36 +24,46 @@ jobs:
     runs-on: ${{ matrix.os }}
     steps:
       - uses: actions/checkout@v4
+
       - name: Set up python ${{ matrix.python-version }}
         uses: actions/setup-python@v4
         with:
           python-version: ${{ matrix.python-version }}
+
       - name: Install Poetry
         uses: snok/install-poetry@v1
         with:
           version: ${{ env.POETRY_VERSION }}
-      - uses: pnpm/action-setup@v2
+
+      - uses: pnpm/action-setup@v3
+
       - name: Setup Node.js ${{ matrix.node-version }}
         uses: actions/setup-node@v4
         with:
           node-version: ${{ matrix.node-version }}
           cache: "pnpm"
+
       - name: Install dependencies
         run: pnpm install
+
       - name: Install Playwright Browsers
         run: pnpm exec playwright install --with-deps
         working-directory: .
+
       - name: Build create-llama
         run: pnpm run build
         working-directory: .
+
       - name: Install
         run: pnpm run pack-install
         working-directory: .
+
       - name: Run Playwright tests
         run: pnpm run e2e
         env:
           OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
         working-directory: .
+
       - uses: actions/upload-artifact@v3
         if: always()
         with:

.github/workflows/lint_on_push_or_pull.yml

@@ -13,17 +13,20 @@ jobs:
     runs-on: ubuntu-latest
     steps:
       - uses: actions/checkout@v4
-      - uses: pnpm/action-setup@v2
-        with:
-          version: latest
+
+      - uses: pnpm/action-setup@v3
+
       - name: Setup Node.js
         uses: actions/setup-node@v4
         with:
           node-version-file: ".nvmrc"
           cache: "pnpm"
+
       - name: Install dependencies
         run: pnpm install
+
       - name: Run lint
         run: pnpm run lint
+
       - name: Run Prettier
         run: pnpm run format

.github/workflows/publish_github.yml

@@ -0,0 +1,36 @@
+name: Publish to GitHub Releases
+
+on:
+  push:
+    tags:
+      - "v*"
+
+jobs:
+  build-and-publish:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout Repo
+        uses: actions/checkout@v4
+
+      - uses: pnpm/action-setup@v3
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version-file: ".nvmrc"
+          cache: "pnpm"
+
+      - name: Install dependencies
+        run: pnpm install
+
+      - name: Build tarball
+        run: |
+          pnpm pack
+
+      - name: Create release
+        uses: ncipollo/release-action@v1
+        with:
+          artifacts: "create-llama-*.tgz"
+          name: Release ${{ github.ref }}
+          bodyFile: "CHANGELOG.md"
+          token: ${{ secrets.GITHUB_TOKEN }}

.github/workflows/release.yml

@@ -0,0 +1,55 @@
+name: Release
+
+on:
+  push:
+    branches:
+      - main
+
+concurrency: ${{ github.workflow }}-${{ github.ref }}
+
+jobs:
+  release:
+    name: Release
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout Repo
+        uses: actions/checkout@v4
+
+      - uses: pnpm/action-setup@v3
+
+      - name: Setup Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version-file: ".nvmrc"
+          cache: "pnpm"
+
+      - name: Install dependencies
+        run: pnpm install
+
+      - name: Add auth token to .npmrc file
+        run: |
+          cat << EOF >> ".npmrc"
+            //registry.npmjs.org/:_authToken=$NPM_TOKEN
+          EOF
+        env:
+          NPM_TOKEN: ${{ secrets.NPM_TOKEN }}
+
+      - name: Get changeset status
+        id: get-changeset-status
+        run: |
+          pnpm changeset status --output .changeset/status.json
+          new_version=$(jq -r '.releases[0].newVersion' < .changeset/status.json)
+          rm -v .changeset/status.json
+          echo "new-version=${new_version}" >> "$GITHUB_OUTPUT"
+
+      - name: Create Release Pull Request or Publish to npm
+        id: changesets
+        uses: changesets/action@v1
+        with:
+          commit: Release ${{ steps.get-changeset-status.outputs.new-version }}
+          title: Release ${{ steps.get-changeset-status.outputs.new-version }}
+          # build package and call changeset publish
+          publish: pnpm release
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+          NPM_TOKEN: ${{ secrets.NPM_TOKEN }}

CHANGELOG.md

@@ -1,5 +1,57 @@
 # create-llama
 
+## 0.1.3
+
+### Patch Changes
+
+- 416073d: Directly import vector stores to work with NextJS
+
+## 0.1.2
+
+### Patch Changes
+
+- 056e376: Add support for displaying tool outputs (including weather widget as example)
+
+## 0.1.1
+
+### Patch Changes
+
+- 7bd3ed5: Support Anthropic and Gemini as model providers
+- 7bd3ed5: Support new agents from LITS 0.3
+- cfb5257: Display events (e.g. retrieving nodes) per chat message
+
+## 0.1.0
+
+### Minor Changes
+
+- f1c3e8d: Add Llama3 and Phi3 support using Ollama
+
+### Patch Changes
+
+- a0dec80: Use `gpt-4-turbo` model as default. Upgrade Python llama-index to 0.10.28
+- 753229d: Remove asking for AI models and use defaults instead (OpenAIs GPT-4 Vision Preview and Embeddings v3). Use `--ask-models` CLI parameter to select models.
+- 1d78202: Add observability for Python
+- 6acccd2: Use poetry run generate to generate embeddings for FastAPI
+- 9efcffe: Use Settings object for LlamaIndex configuration
+- 418bf9b: refactor: use tsx instead of ts-node
+- 1be69a5: Add Qdrant support
+
+## 0.0.32
+
+### Patch Changes
+
+- 625ed4d: Support Astra VectorDB
+- 922e0ce: Remove UI question (use shadcn as default). Use `html` UI by calling create-llama with --ui html parameter
+- ce2f24d: Update loaders and tools config to yaml format (for Python)
+- e8db041: Let user select multiple datasources (URLs, files and folders)
+- c06d4af: Add nodes to the response (Python)
+- 29b17ee: Allow using agents without any data source
+- 665c26c: Add redirect to documentation page when accessing the base URL (FastAPI)
+- 78ded9e: Add Dockerfile templates for Typescript and Python
+- 99e758f: Merge non-streaming and streaming template to one
+- b3f2685: Add support for agent generation for Typescript
+- 2739714: Use a database (MySQL or PostgreSQL) as a data source
+
 ## 0.0.31
 
 ### Patch Changes

CONTRIBUTING.md

@@ -7,7 +7,7 @@ Install NodeJS. Preferably v18 using nvm or n.
 Inside the `create-llama` directory:
 
 ```
-npm i -g pnpm ts-node
+npm i -g pnpm
 pnpm install
 ```
 

README.md

@@ -18,25 +18,19 @@ to start the development server. You can then visit [http://localhost:3000](http
 
 ## What you'll get
 
-- A Next.js-powered front-end. The app is set up as a chat interface that can answer questions about your data (see below)
-  - You can style it with HTML and CSS, or you can optionally use components from [shadcn/ui](https://ui.shadcn.com/)
+- A Next.js-powered front-end using components from [shadcn/ui](https://ui.shadcn.com/). The app is set up as a chat interface that can answer questions about your data (see below)
 - Your choice of 3 back-ends:
-  - **Next.js**: if you select this option, you’ll have a full stack Next.js application that you can deploy to a host like [Vercel](https://vercel.com/) in just a few clicks. This uses [LlamaIndex.TS](https://www.npmjs.com/package/llamaindex), our TypeScript library.
+  - **Next.js**: if you select this option, you’ll have a full-stack Next.js application that you can deploy to a host like [Vercel](https://vercel.com/) in just a few clicks. This uses [LlamaIndex.TS](https://www.npmjs.com/package/llamaindex), our TypeScript library.
   - **Express**: if you want a more traditional Node.js application you can generate an Express backend. This also uses LlamaIndex.TS.
-  - **Python FastAPI**: if you select this option you’ll get a backend powered by the [llama-index python package](https://pypi.org/project/llama-index/), which you can deploy to a service like Render or fly.io.
-- The back-end has a single endpoint that allows you to send the state of your chat and receive additional responses
-- You can choose whether you want a streaming or non-streaming back-end (if you're not sure, we recommend streaming)
-- You can choose whether you want to use `ContextChatEngine` or `SimpleChatEngine`
-  - `SimpleChatEngine` will just talk to the LLM directly without using your data
-  - `ContextChatEngine` will use your data to answer questions (see below).
+  - **Python FastAPI**: if you select this option, you’ll get a backend powered by the [llama-index python package](https://pypi.org/project/llama-index/), which you can deploy to a service like Render or fly.io.
+- The back-end has two endpoints (one streaming, the other one non-streaming) that allow you to send the state of your chat and receive additional responses
+- You add arbitrary data sources to your chat, like local files, websites, or data retrieved from a database.
+- Turn your chat into an AI agent by adding tools (functions called by the LLM).
 - The app uses OpenAI by default, so you'll need an OpenAI API key, or you can customize it to use any of the dozens of LLMs we support.
 
 ## Using your data
 
-If you've enabled `ContextChatEngine`, you can supply your own data and the app will index it and answer questions. Your generated app will have a folder called `data`:
-
-- With the Next.js backend this is `./data`
-- With the Express or Python backend this is in `./backend/data`
+You can supply your own data; the app will index it and answer questions. Your generated app will have a folder called `data` (If you're using Express or Python and generate a frontend, it will be `./backend/data`).
 
 The app will ingest any supported files you put in this directory. Your Next.js and Express apps use LlamaIndex.TS so they will be able to ingest any PDF, text, CSV, Markdown, Word and HTML files. The Python backend can read even more types, including video and audio files.
 
@@ -46,22 +40,28 @@ Before you can use your data, you need to index it. If you're using the Next.js
 npm run generate
 ```
 
-Then re-start your app. Remember you'll need to re-run `generate` if you add new files to your `data` folder. If you're using the Python backend, you can trigger indexing of your data by deleting the `./storage` folder and re-starting the app.
+Then re-start your app. Remember you'll need to re-run `generate` if you add new files to your `data` folder.
 
-## Don't want a front-end?
+If you're using the Python backend, you can trigger indexing of your data by calling:
 
-It's optional! If you've selected the Python or Express back-ends, just delete the `frontend` folder and you'll get an API without any front-end code.
+```bash
+poetry run generate
+```
 
-## Customizing the LLM
+## Want a front-end?
 
-By default the app will use OpenAI's gpt-3.5-turbo model. If you want to use GPT-4, you can modify this by editing a file:
+Optionally generate a frontend if you've selected the Python or Express back-ends. If you do so, `create-llama` will generate two folders: `frontend`, for your Next.js-based frontend code, and `backend` containing your API.
 
-- In the Next.js backend, edit `./app/api/chat/route.ts` and replace `gpt-3.5-turbo` with `gpt-4`
-- In the Express backend, edit `./backend/src/controllers/chat.controller.ts` and likewise replace `gpt-3.5-turbo` with `gpt-4`
-- In the Python backend, edit `./backend/app/utils/index.py` and once again replace `gpt-3.5-turbo` with `gpt-4`
+## Customizing the AI models
+
+The app will default to OpenAI's `gpt-4-turbo` LLM and `text-embedding-3-large` embedding model.
+
+If you want to use different OpenAI models, add the `--ask-models` CLI parameter.
 
 You can also replace OpenAI with one of our [dozens of other supported LLMs](https://docs.llamaindex.ai/en/stable/module_guides/models/llms/modules.html).
 
+To do so, you have to manually change the generated code (edit the `settings.ts` file for Typescript projects or the `settings.py` file for Python projects)
+
 ## Example
 
 The simplest thing to do is run `create-llama` in interactive mode:
@@ -84,13 +84,19 @@ Need to install the following packages:
   create-llama@latest
 Ok to proceed? (y) y
 ✔ What is your project named? … my-app
-✔ Which template would you like to use? › Chat with streaming
+✔ Which template would you like to use? › Chat
 ✔ Which framework would you like to use? › NextJS
-✔ Which UI would you like to use? › Just HTML
-✔ Which chat engine would you like to use? › ContextChatEngine
+✔ Would you like to set up observability? › No
 ✔ Please provide your OpenAI API key (leave blank to skip): …
-✔ Would you like to use ESLint? … No / Yes
-Creating a new LlamaIndex app in /home/my-app.
+✔ Which data source would you like to use? › Use an example PDF
+✔ Would you like to add another data source? › No
+✔ Would you like to use LlamaParse (improved parser for RAG - requires API key)? … no / yes
+✔ Would you like to use a vector database? › No, just store the data in the file system
+? How would you like to proceed? › - Use arrow-keys. Return to submit.
+   Just generate code (~1 sec)
+❯  Start in VSCode (~1 sec)
+   Generate code and install dependencies (~2 min)
+   Generate code, install dependencies, and run the app (~2 min)
 ```
 
 ### Running non-interactively

create-app.ts

@@ -30,10 +30,8 @@ export async function createApp({
   appPath,
   packageManager,
   frontend,
-  openAiKey,
+  modelConfig,
   llamaCloudKey,
-  model,
-  embeddingModel,
   communityProjectConfig,
   llamapack,
   vectorDb,
@@ -77,10 +75,8 @@ export async function createApp({
     ui,
     packageManager,
     isOnline,
-    openAiKey,
+    modelConfig,
     llamaCloudKey,
-    model,
-    embeddingModel,
     communityProjectConfig,
     llamapack,
     vectorDb,

e2e/utils.ts

@@ -12,8 +12,7 @@ import {
 } from "../helpers";
 
 export type AppType = "--frontend" | "--no-frontend" | "";
-const MODEL = "gpt-3.5-turbo";
-const EMBEDDING_MODEL = "text-embedding-ada-002";
+
 export type CreateLlamaResult = {
   projectName: string;
   appProcess: ChildProcess;
@@ -96,10 +95,6 @@ export async function runCreateLlama(
     templateUI,
     "--vector-db",
     vectorDb,
-    "--model",
-    MODEL,
-    "--embedding-model",
-    EMBEDDING_MODEL,
     "--open-ai-key",
     process.env.OPENAI_API_KEY,
     appType,

helpers/copy.ts

@@ -48,3 +48,21 @@ export const copy = async (
     }),
   );
 };
+
+export const assetRelocator = (name: string) => {
+  switch (name) {
+    case "gitignore":
+    case "npmrc":
+    case "eslintrc.json": {
+      return `.${name}`;
+    }
+    // README.md is ignored by webpack-asset-relocator-loader used by ncc:
+    // https://github.com/vercel/webpack-asset-relocator-loader/blob/e9308683d47ff507253e37c9bcbb99474603192b/src/asset-relocator.js#L227
+    case "README-template.md": {
+      return "README.md";
+    }
+    default: {
+      return name;
+    }
+  }
+};

helpers/datasources.ts

@@ -1,6 +1,8 @@
+import fs from "fs/promises";
 import path from "path";
+import yaml, { Document } from "yaml";
 import { templatesDir } from "./dir";
-import { TemplateDataSource } from "./types";
+import { DbSourceConfig, TemplateDataSource, WebSourceConfig } from "./types";
 
 export const EXAMPLE_FILE: TemplateDataSource = {
   type: "file",
@@ -28,3 +30,80 @@ export function getDataSources(
   }
   return dataSources;
 }
+
+export async function writeLoadersConfig(
+  root: string,
+  dataSources: TemplateDataSource[],
+  useLlamaParse?: boolean,
+) {
+  if (dataSources.length === 0) return; // no datasources, no config needed
+  const loaderConfig = new Document({});
+  // Web loader config
+  if (dataSources.some((ds) => ds.type === "web")) {
+    const webLoaderConfig = new Document({});
+
+    // Create config for browser driver arguments
+    const driverArgNodeValue = webLoaderConfig.createNode([
+      "--no-sandbox",
+      "--disable-dev-shm-usage",
+    ]);
+    driverArgNodeValue.commentBefore =
+      " The arguments to pass to the webdriver. E.g.: add --headless to run in headless mode";
+    webLoaderConfig.set("driver_arguments", driverArgNodeValue);
+
+    // Create config for urls
+    const urlConfigs = dataSources
+      .filter((ds) => ds.type === "web")
+      .map((ds) => {
+        const dsConfig = ds.config as WebSourceConfig;
+        return {
+          base_url: dsConfig.baseUrl,
+          prefix: dsConfig.prefix,
+          depth: dsConfig.depth,
+        };
+      });
+    const urlConfigNode = webLoaderConfig.createNode(urlConfigs);
+    urlConfigNode.commentBefore = ` base_url: The URL to start crawling with
+ prefix: Only crawl URLs matching the specified prefix
+ depth: The maximum depth for BFS traversal
+ You can add more websites by adding more entries (don't forget the - prefix from YAML)`;
+    webLoaderConfig.set("urls", urlConfigNode);
+
+    // Add web config to the loaders config
+    loaderConfig.set("web", webLoaderConfig);
+  }
+
+  // File loader config
+  if (dataSources.some((ds) => ds.type === "file")) {
+    // Add documentation to web loader config
+    const node = loaderConfig.createNode({
+      use_llama_parse: useLlamaParse,
+    });
+    node.commentBefore = ` use_llama_parse: Use LlamaParse if \`true\`. Needs a \`LLAMA_CLOUD_API_KEY\` from https://cloud.llamaindex.ai set as environment variable`;
+    loaderConfig.set("file", node);
+  }
+
+  // DB loader config
+  const dbLoaders = dataSources.filter((ds) => ds.type === "db");
+  if (dbLoaders.length > 0) {
+    const dbLoaderConfig = new Document({});
+    const configEntries = dbLoaders.map((ds) => {
+      const dsConfig = ds.config as DbSourceConfig;
+      return {
+        uri: dsConfig.uri,
+        queries: [dsConfig.queries],
+      };
+    });
+
+    const node = dbLoaderConfig.createNode(configEntries);
+    node.commentBefore = ` The configuration for the database loader, only supports MySQL and PostgreSQL databases for now.
+ uri: The URI for the database. E.g.: mysql+pymysql://user:password@localhost:3306/db or postgresql+psycopg2://user:password@localhost:5432/db
+ query: The query to fetch data from the database. E.g.: SELECT * FROM table`;
+    loaderConfig.set("db", node);
+  }
+
+  // Write loaders config
+  const loaderConfigPath = path.join(root, "config", "loaders.yaml");
+  await fs.mkdir(path.join(root, "config"), { recursive: true });
+  await fs.writeFile(loaderConfigPath, yaml.stringify(loaderConfig));
+}

helpers/devcontainer.ts

@@ -46,13 +46,16 @@ export const writeDevcontainer = async (
   framework: TemplateFramework,
   frontend: boolean,
 ) => {
-  console.log("Adding .devcontainer");
+  const devcontainerDir = path.join(root, ".devcontainer");
+  if (fs.existsSync(devcontainerDir)) {
+    console.log("Template already has a .devcontainer. Using it.");
+    return;
+  }
   const devcontainerContent = renderDevcontainerContent(
     templatesDir,
     framework,
     frontend,
   );
-  const devcontainerDir = path.join(root, ".devcontainer");
   fs.mkdirSync(devcontainerDir);
   await fs.promises.writeFile(
     path.join(devcontainerDir, "devcontainer.json"),

helpers/env-variables.ts

@@ -1,6 +1,7 @@
 import fs from "fs/promises";
 import path from "path";
 import {
+  ModelConfig,
   TemplateDataSource,
   TemplateFramework,
   TemplateVectorDB,
@@ -28,7 +29,10 @@ const renderEnvVar = (envVars: EnvVar[]): string => {
   );
 };
 
-const getVectorDBEnvs = (vectorDb: TemplateVectorDB) => {
+const getVectorDBEnvs = (vectorDb?: TemplateVectorDB): EnvVar[] => {
+  if (!vectorDb) {
+    return [];
+  }
   switch (vectorDb) {
     case "mongo":
       return [
@@ -93,19 +97,148 @@ const getVectorDBEnvs = (vectorDb: TemplateVectorDB) => {
           description: "The password to access the Milvus server.",
         },
       ];
+    case "astra":
+      return [
+        {
+          name: "ASTRA_DB_APPLICATION_TOKEN",
+          description: "The generated app token for your Astra database",
+        },
+        {
+          name: "ASTRA_DB_ENDPOINT",
+          description: "The API endpoint for your Astra database",
+        },
+        {
+          name: "ASTRA_DB_COLLECTION",
+          description: "The name of the collection in your Astra database",
+        },
+      ];
+    case "qdrant":
+      return [
+        {
+          name: "QDRANT_URL",
+          description:
+            "The qualified REST URL of the Qdrant server. Eg: http://localhost:6333",
+        },
+        {
+          name: "QDRANT_COLLECTION",
+          description: "The name of Qdrant collection to use.",
+        },
+        {
+          name: "QDRANT_API_KEY",
+          description:
+            "Optional API key for authenticating requests to Qdrant.",
+        },
+      ];
     default:
       return [];
   }
 };
 
+const getModelEnvs = (modelConfig: ModelConfig): EnvVar[] => {
+  return [
+    {
+      name: "MODEL_PROVIDER",
+      description: "The provider for the AI models to use.",
+      value: modelConfig.provider,
+    },
+    {
+      name: "MODEL",
+      description: "The name of LLM model to use.",
+      value: modelConfig.model,
+    },
+    {
+      name: "EMBEDDING_MODEL",
+      description: "Name of the embedding model to use.",
+      value: modelConfig.embeddingModel,
+    },
+    {
+      name: "EMBEDDING_DIM",
+      description: "Dimension of the embedding model to use.",
+      value: modelConfig.dimensions.toString(),
+    },
+    ...(modelConfig.provider === "openai"
+      ? [
+          {
+            name: "OPENAI_API_KEY",
+            description: "The OpenAI API key to use.",
+            value: modelConfig.apiKey,
+          },
+          {
+            name: "LLM_TEMPERATURE",
+            description: "Temperature for sampling from the model.",
+          },
+          {
+            name: "LLM_MAX_TOKENS",
+            description: "Maximum number of tokens to generate.",
+          },
+        ]
+      : []),
+    ...(modelConfig.provider === "anthropic"
+      ? [
+          {
+            name: "ANTHROPIC_API_KEY",
+            description: "The Anthropic API key to use.",
+            value: modelConfig.apiKey,
+          },
+        ]
+      : []),
+    ...(modelConfig.provider === "gemini"
+      ? [
+          {
+            name: "GOOGLE_API_KEY",
+            description: "The Google API key to use.",
+            value: modelConfig.apiKey,
+          },
+        ]
+      : []),
+  ];
+};
+
+const getFrameworkEnvs = (
+  framework?: TemplateFramework,
+  port?: number,
+): EnvVar[] => {
+  if (framework !== "fastapi") {
+    return [];
+  }
+  return [
+    {
+      name: "APP_HOST",
+      description: "The address to start the backend app.",
+      value: "0.0.0.0",
+    },
+    {
+      name: "APP_PORT",
+      description: "The port to start the backend app.",
+      value: port?.toString() || "8000",
+    },
+    // TODO: Once LlamaIndexTS supports string templates, move this to `getEngineEnvs`
+    {
+      name: "SYSTEM_PROMPT",
+      description: `Custom system prompt.
+Example:
+SYSTEM_PROMPT="You are a helpful assistant who helps users with their questions."`,
+    },
+  ];
+};
+
+const getEngineEnvs = (): EnvVar[] => {
+  return [
+    {
+      name: "TOP_K",
+      description:
+        "The number of similar embeddings to return when retrieving documents.",
+      value: "3",
+    },
+  ];
+};
+
 export const createBackendEnvFile = async (
   root: string,
   opts: {
-    openAiKey?: string;
     llamaCloudKey?: string;
     vectorDb?: TemplateVectorDB;
-    model?: string;
-    embeddingModel?: string;
+    modelConfig: ModelConfig;
     framework?: TemplateFramework;
     dataSources?: TemplateDataSource[];
     port?: number;
@@ -113,94 +246,20 @@ export const createBackendEnvFile = async (
 ) => {
   // Init env values
   const envFileName = ".env";
-  const defaultEnvs = [
-    {
-      render: true,
-      name: "MODEL",
-      description: "The name of LLM model to use.",
-      value: opts.model || "gpt-3.5-turbo",
-    },
-    {
-      render: true,
-      name: "OPENAI_API_KEY",
-      description: "The OpenAI API key to use.",
-      value: opts.openAiKey,
-    },
+  const envVars: EnvVar[] = [
     {
       name: "LLAMA_CLOUD_API_KEY",
       description: `The Llama Cloud API key.`,
       value: opts.llamaCloudKey,
     },
+    // Add model environment variables
+    ...getModelEnvs(opts.modelConfig),
+    // Add engine environment variables
+    ...getEngineEnvs(),
     // Add vector database environment variables
-    ...(opts.vectorDb ? getVectorDBEnvs(opts.vectorDb) : []),
+    ...getVectorDBEnvs(opts.vectorDb),
+    ...getFrameworkEnvs(opts.framework, opts.port),
   ];
-  let envVars: EnvVar[] = [];
-  if (opts.framework === "fastapi") {
-    envVars = [
-      ...defaultEnvs,
-      ...[
-        {
-          name: "APP_HOST",
-          description: "The address to start the backend app.",
-          value: "0.0.0.0",
-        },
-        {
-          name: "APP_PORT",
-          description: "The port to start the backend app.",
-          value: opts.port?.toString() || "8000",
-        },
-        {
-          name: "EMBEDDING_MODEL",
-          description: "Name of the embedding model to use.",
-          value: opts.embeddingModel,
-        },
-        {
-          name: "EMBEDDING_DIM",
-          description: "Dimension of the embedding model to use.",
-        },
-        {
-          name: "LLM_TEMPERATURE",
-          description: "Temperature for sampling from the model.",
-        },
-        {
-          name: "LLM_MAX_TOKENS",
-          description: "Maximum number of tokens to generate.",
-        },
-        {
-          name: "TOP_K",
-          description:
-            "The number of similar embeddings to return when retrieving documents.",
-          value: "3",
-        },
-        {
-          name: "SYSTEM_PROMPT",
-          description: `Custom system prompt.
-Example:
-SYSTEM_PROMPT="
-We have provided context information below.
----------------------
-{context_str}
----------------------
-Given this information, please answer the question: {query_str}
-"`,
-        },
-      ],
-    ];
-  } else {
-    envVars = [
-      ...defaultEnvs,
-      ...[
-        opts.framework === "nextjs"
-          ? {
-              name: "NEXT_PUBLIC_MODEL",
-              description:
-                "The LLM model to use (hardcode to front-end artifact).",
-              value: opts.model || "gpt-3.5-turbo",
-            }
-          : {},
-      ],
-    ];
-  }
   // Render and write env file
   const content = renderEnvVar(envVars);
   await fs.writeFile(path.join(root, envFileName), content);
@@ -211,20 +270,9 @@ export const createFrontendEnvFile = async (
   root: string,
   opts: {
     customApiPath?: string;
-    model?: string;
   },
 ) => {
   const defaultFrontendEnvs = [
-    {
-      name: "MODEL",
-      description: "The OpenAI model to use.",
-      value: opts.model,
-    },
-    {
-      name: "NEXT_PUBLIC_MODEL",
-      description: "The OpenAI model to use (hardcode to front-end artifact).",
-      value: opts.model,
-    },
     {
       name: "NEXT_PUBLIC_CHAT_API",
       description: "The backend API for chat endpoint.",

helpers/index.ts

@@ -4,15 +4,18 @@ import path from "path";
 import { cyan } from "picocolors";
 
 import fsExtra from "fs-extra";
+import { writeLoadersConfig } from "./datasources";
 import { createBackendEnvFile, createFrontendEnvFile } from "./env-variables";
 import { PackageManager } from "./get-pkg-manager";
 import { installLlamapackProject } from "./llama-pack";
 import { isHavingPoetryLockFile, tryPoetryRun } from "./poetry";
 import { installPythonTemplate } from "./python";
 import { downloadAndExtractRepo } from "./repo";
+import { ConfigFileType, writeToolsConfig } from "./tools";
 import {
   FileSourceConfig,
   InstallTemplateArgs,
+  ModelConfig,
   TemplateDataSource,
   TemplateFramework,
   TemplateVectorDB,
@@ -22,8 +25,8 @@ import { installTSTemplate } from "./typescript";
 // eslint-disable-next-line max-params
 async function generateContextData(
   framework: TemplateFramework,
+  modelConfig: ModelConfig,
   packageManager?: PackageManager,
-  openAiKey?: string,
   vectorDb?: TemplateVectorDB,
   llamaCloudKey?: string,
   useLlamaParse?: boolean,
@@ -31,20 +34,20 @@ async function generateContextData(
   if (packageManager) {
     const runGenerate = `${cyan(
       framework === "fastapi"
-        ? "poetry run python app/engine/generate.py"
+        ? "poetry run generate"
         : `${packageManager} run generate`,
     )}`;
-    const openAiKeyConfigured = openAiKey || process.env["OPENAI_API_KEY"];
+    const modelConfigured = modelConfig.isConfigured();
     const llamaCloudKeyConfigured = useLlamaParse
       ? llamaCloudKey || process.env["LLAMA_CLOUD_API_KEY"]
       : true;
     const hasVectorDb = vectorDb && vectorDb !== "none";
-    if (openAiKeyConfigured && llamaCloudKeyConfigured && !hasVectorDb) {
+    if (modelConfigured && llamaCloudKeyConfigured && !hasVectorDb) {
       // If all the required environment variables are set, run the generate script
       if (framework === "fastapi") {
         if (isHavingPoetryLockFile()) {
           console.log(`Running ${runGenerate} to generate the context data.`);
-          const result = tryPoetryRun("python app/engine/generate.py");
+          const result = tryPoetryRun("poetry run generate");
           if (!result) {
             console.log(`Failed to run ${runGenerate}.`);
             process.exit(1);
@@ -61,7 +64,7 @@ async function generateContextData(
 
     // generate the message of what to do to run the generate script manually
     const settings = [];
-    if (!openAiKeyConfigured) settings.push("your OpenAI key");
+    if (!modelConfigured) settings.push("your model provider API key");
     if (!llamaCloudKeyConfigured) settings.push("your Llama Cloud key");
     if (hasVectorDb) settings.push("your Vector DB environment variables");
     const settingsMessage =
@@ -117,20 +120,31 @@ export const installTemplate = async (
 
   if (props.framework === "fastapi") {
     await installPythonTemplate(props);
+    // write loaders configuration (currently Python only)
+    await writeLoadersConfig(
+      props.root,
+      props.dataSources,
+      props.useLlamaParse,
+    );
   } else {
     await installTSTemplate(props);
   }
 
+  // write tools configuration
+  await writeToolsConfig(
+    props.root,
+    props.tools,
+    props.framework === "fastapi" ? ConfigFileType.YAML : ConfigFileType.JSON,
+  );
+
   if (props.backend) {
     // This is a backend, so we need to copy the test data and create the env file.
 
     // Copy the environment file to the target directory.
     await createBackendEnvFile(props.root, {
-      openAiKey: props.openAiKey,
+      modelConfig: props.modelConfig,
       llamaCloudKey: props.llamaCloudKey,
       vectorDb: props.vectorDb,
-      model: props.model,
-      embeddingModel: props.embeddingModel,
       framework: props.framework,
       dataSources: props.dataSources,
       port: props.externalPort,
@@ -148,8 +162,8 @@ export const installTemplate = async (
       ) {
         await generateContextData(
           props.framework,
+          props.modelConfig,
           props.packageManager,
-          props.openAiKey,
           props.vectorDb,
           props.llamaCloudKey,
           props.useLlamaParse,
@@ -159,7 +173,6 @@ export const installTemplate = async (
   } else {
     // this is a frontend for a full-stack app, create .env file with model information
     await createFrontendEnvFile(props.root, {
-      model: props.model,
       customApiPath: props.customApiPath,
     });
   }

helpers/providers/anthropic.ts

@@ -0,0 +1,106 @@
+import ciInfo from "ci-info";
+import prompts from "prompts";
+import { ModelConfigParams } from ".";
+import { questionHandlers, toChoice } from "../../questions";
+
+const MODELS = [
+  "claude-3-opus",
+  "claude-3-sonnet",
+  "claude-3-haiku",
+  "claude-2.1",
+  "claude-instant-1.2",
+];
+const DEFAULT_MODEL = MODELS[0];
+
+// TODO: get embedding vector dimensions from the anthropic sdk (currently not supported)
+// Use huggingface embedding models for now
+enum HuggingFaceEmbeddingModelType {
+  XENOVA_ALL_MINILM_L6_V2 = "all-MiniLM-L6-v2",
+  XENOVA_ALL_MPNET_BASE_V2 = "all-mpnet-base-v2",
+}
+type ModelData = {
+  dimensions: number;
+};
+const EMBEDDING_MODELS: Record<HuggingFaceEmbeddingModelType, ModelData> = {
+  [HuggingFaceEmbeddingModelType.XENOVA_ALL_MINILM_L6_V2]: {
+    dimensions: 384,
+  },
+  [HuggingFaceEmbeddingModelType.XENOVA_ALL_MPNET_BASE_V2]: {
+    dimensions: 768,
+  },
+};
+const DEFAULT_EMBEDDING_MODEL = Object.keys(EMBEDDING_MODELS)[0];
+const DEFAULT_DIMENSIONS = Object.values(EMBEDDING_MODELS)[0].dimensions;
+
+type AnthropicQuestionsParams = {
+  apiKey?: string;
+  askModels: boolean;
+};
+
+export async function askAnthropicQuestions({
+  askModels,
+  apiKey,
+}: AnthropicQuestionsParams): Promise<ModelConfigParams> {
+  const config: ModelConfigParams = {
+    apiKey,
+    model: DEFAULT_MODEL,
+    embeddingModel: DEFAULT_EMBEDDING_MODEL,
+    dimensions: DEFAULT_DIMENSIONS,
+    isConfigured(): boolean {
+      if (config.apiKey) {
+        return true;
+      }
+      if (process.env["ANTHROPIC_API_KEY"]) {
+        return true;
+      }
+      return false;
+    },
+  };
+
+  if (!config.apiKey) {
+    const { key } = await prompts(
+      {
+        type: "text",
+        name: "key",
+        message:
+          "Please provide your Anthropic API key (or leave blank to use ANTHROPIC_API_KEY env variable):",
+      },
+      questionHandlers,
+    );
+    config.apiKey = key || process.env.ANTHROPIC_API_KEY;
+  }
+
+  // use default model values in CI or if user should not be asked
+  const useDefaults = ciInfo.isCI || !askModels;
+  if (!useDefaults) {
+    const { model } = await prompts(
+      {
+        type: "select",
+        name: "model",
+        message: "Which LLM model would you like to use?",
+        choices: MODELS.map(toChoice),
+        initial: 0,
+      },
+      questionHandlers,
+    );
+    config.model = model;
+
+    const { embeddingModel } = await prompts(
+      {
+        type: "select",
+        name: "embeddingModel",
+        message: "Which embedding model would you like to use?",
+        choices: Object.keys(EMBEDDING_MODELS).map(toChoice),
+        initial: 0,
+      },
+      questionHandlers,
+    );
+    config.embeddingModel = embeddingModel;
+    config.dimensions =
+      EMBEDDING_MODELS[
+        embeddingModel as HuggingFaceEmbeddingModelType
+      ].dimensions;
+  }
+
+  return config;
+}

helpers/providers/gemini.ts

@@ -0,0 +1,87 @@
+import ciInfo from "ci-info";
+import prompts from "prompts";
+import { ModelConfigParams } from ".";
+import { questionHandlers, toChoice } from "../../questions";
+
+const MODELS = ["gemini-1.5-pro-latest", "gemini-pro", "gemini-pro-vision"];
+type ModelData = {
+  dimensions: number;
+};
+const EMBEDDING_MODELS: Record<string, ModelData> = {
+  "embedding-001": { dimensions: 768 },
+  "text-embedding-004": { dimensions: 768 },
+};
+
+const DEFAULT_MODEL = MODELS[0];
+const DEFAULT_EMBEDDING_MODEL = Object.keys(EMBEDDING_MODELS)[0];
+const DEFAULT_DIMENSIONS = Object.values(EMBEDDING_MODELS)[0].dimensions;
+
+type GeminiQuestionsParams = {
+  apiKey?: string;
+  askModels: boolean;
+};
+
+export async function askGeminiQuestions({
+  askModels,
+  apiKey,
+}: GeminiQuestionsParams): Promise<ModelConfigParams> {
+  const config: ModelConfigParams = {
+    apiKey,
+    model: DEFAULT_MODEL,
+    embeddingModel: DEFAULT_EMBEDDING_MODEL,
+    dimensions: DEFAULT_DIMENSIONS,
+    isConfigured(): boolean {
+      if (config.apiKey) {
+        return true;
+      }
+      if (process.env["GOOGLE_API_KEY"]) {
+        return true;
+      }
+      return false;
+    },
+  };
+
+  if (!config.apiKey) {
+    const { key } = await prompts(
+      {
+        type: "text",
+        name: "key",
+        message:
+          "Please provide your Google API key (or leave blank to use GOOGLE_API_KEY env variable):",
+      },
+      questionHandlers,
+    );
+    config.apiKey = key || process.env.GOOGLE_API_KEY;
+  }
+
+  // use default model values in CI or if user should not be asked
+  const useDefaults = ciInfo.isCI || !askModels;
+  if (!useDefaults) {
+    const { model } = await prompts(
+      {
+        type: "select",
+        name: "model",
+        message: "Which LLM model would you like to use?",
+        choices: MODELS.map(toChoice),
+        initial: 0,
+      },
+      questionHandlers,
+    );
+    config.model = model;
+
+    const { embeddingModel } = await prompts(
+      {
+        type: "select",
+        name: "embeddingModel",
+        message: "Which embedding model would you like to use?",
+        choices: Object.keys(EMBEDDING_MODELS).map(toChoice),
+        initial: 0,
+      },
+      questionHandlers,
+    );
+    config.embeddingModel = embeddingModel;
+    config.dimensions = EMBEDDING_MODELS[embeddingModel].dimensions;
+  }
+
+  return config;
+}

helpers/providers/index.ts

@@ -0,0 +1,67 @@
+import ciInfo from "ci-info";
+import prompts from "prompts";
+import { questionHandlers } from "../../questions";
+import { ModelConfig, ModelProvider } from "../types";
+import { askAnthropicQuestions } from "./anthropic";
+import { askGeminiQuestions } from "./gemini";
+import { askOllamaQuestions } from "./ollama";
+import { askOpenAIQuestions } from "./openai";
+
+const DEFAULT_MODEL_PROVIDER = "openai";
+
+export type ModelConfigQuestionsParams = {
+  openAiKey?: string;
+  askModels: boolean;
+};
+
+export type ModelConfigParams = Omit<ModelConfig, "provider">;
+
+export async function askModelConfig({
+  askModels,
+  openAiKey,
+}: ModelConfigQuestionsParams): Promise<ModelConfig> {
+  let modelProvider: ModelProvider = DEFAULT_MODEL_PROVIDER;
+  if (askModels && !ciInfo.isCI) {
+    const { provider } = await prompts(
+      {
+        type: "select",
+        name: "provider",
+        message: "Which model provider would you like to use",
+        choices: [
+          {
+            title: "OpenAI",
+            value: "openai",
+          },
+          { title: "Ollama", value: "ollama" },
+          { title: "Anthropic", value: "anthropic" },
+          { title: "Gemini", value: "gemini" },
+        ],
+        initial: 0,
+      },
+      questionHandlers,
+    );
+    modelProvider = provider;
+  }
+
+  let modelConfig: ModelConfigParams;
+  switch (modelProvider) {
+    case "ollama":
+      modelConfig = await askOllamaQuestions({ askModels });
+      break;
+    case "anthropic":
+      modelConfig = await askAnthropicQuestions({ askModels });
+      break;
+    case "gemini":
+      modelConfig = await askGeminiQuestions({ askModels });
+      break;
+    default:
+      modelConfig = await askOpenAIQuestions({
+        openAiKey,
+        askModels,
+      });
+  }
+  return {
+    ...modelConfig,
+    provider: modelProvider,
+  };
+}

helpers/providers/ollama.ts

@@ -0,0 +1,95 @@
+import ciInfo from "ci-info";
+import ollama, { type ModelResponse } from "ollama";
+import { red } from "picocolors";
+import prompts from "prompts";
+import { ModelConfigParams } from ".";
+import { questionHandlers, toChoice } from "../../questions";
+
+type ModelData = {
+  dimensions: number;
+};
+const MODELS = ["llama3:8b", "wizardlm2:7b", "gemma:7b", "phi3"];
+const DEFAULT_MODEL = MODELS[0];
+// TODO: get embedding vector dimensions from the ollama sdk (currently not supported)
+const EMBEDDING_MODELS: Record<string, ModelData> = {
+  "nomic-embed-text": { dimensions: 768 },
+  "mxbai-embed-large": { dimensions: 1024 },
+  "all-minilm": { dimensions: 384 },
+};
+const DEFAULT_EMBEDDING_MODEL: string = Object.keys(EMBEDDING_MODELS)[0];
+
+type OllamaQuestionsParams = {
+  askModels: boolean;
+};
+
+export async function askOllamaQuestions({
+  askModels,
+}: OllamaQuestionsParams): Promise<ModelConfigParams> {
+  const config: ModelConfigParams = {
+    model: DEFAULT_MODEL,
+    embeddingModel: DEFAULT_EMBEDDING_MODEL,
+    dimensions: EMBEDDING_MODELS[DEFAULT_EMBEDDING_MODEL].dimensions,
+    isConfigured(): boolean {
+      return true;
+    },
+  };
+
+  // use default model values in CI or if user should not be asked
+  const useDefaults = ciInfo.isCI || !askModels;
+  if (!useDefaults) {
+    const { model } = await prompts(
+      {
+        type: "select",
+        name: "model",
+        message: "Which LLM model would you like to use?",
+        choices: MODELS.map(toChoice),
+        initial: 0,
+      },
+      questionHandlers,
+    );
+    await ensureModel(model);
+    config.model = model;
+
+    const { embeddingModel } = await prompts(
+      {
+        type: "select",
+        name: "embeddingModel",
+        message: "Which embedding model would you like to use?",
+        choices: Object.keys(EMBEDDING_MODELS).map(toChoice),
+        initial: 0,
+      },
+      questionHandlers,
+    );
+    await ensureModel(embeddingModel);
+    config.embeddingModel = embeddingModel;
+    config.dimensions = EMBEDDING_MODELS[embeddingModel].dimensions;
+  }
+
+  return config;
+}
+
+async function ensureModel(modelName: string) {
+  try {
+    if (modelName.split(":").length === 1) {
+      // model doesn't have a version suffix, use latest
+      modelName = modelName + ":latest";
+    }
+    const { models } = await ollama.list();
+    const found =
+      models.find((model: ModelResponse) => model.name === modelName) !==
+      undefined;
+    if (!found) {
+      console.log(
+        red(
+          `Model ${modelName} was not pulled yet. Call 'ollama pull ${modelName}' and try again.`,
+        ),
+      );
+      process.exit(1);
+    }
+  } catch (error) {
+    console.log(
+      red("Listing Ollama models failed. Is 'ollama' running? " + error),
+    );
+    process.exit(1);
+  }
+}

helpers/providers/openai.ts

@@ -0,0 +1,144 @@
+import ciInfo from "ci-info";
+import got from "got";
+import ora from "ora";
+import { red } from "picocolors";
+import prompts from "prompts";
+import { ModelConfigParams, ModelConfigQuestionsParams } from ".";
+import { questionHandlers } from "../../questions";
+
+const OPENAI_API_URL = "https://api.openai.com/v1";
+
+const DEFAULT_MODEL = "gpt-3.5-turbo";
+const DEFAULT_EMBEDDING_MODEL = "text-embedding-3-large";
+
+export async function askOpenAIQuestions({
+  openAiKey,
+  askModels,
+}: ModelConfigQuestionsParams): Promise<ModelConfigParams> {
+  const config: ModelConfigParams = {
+    apiKey: openAiKey,
+    model: DEFAULT_MODEL,
+    embeddingModel: DEFAULT_EMBEDDING_MODEL,
+    dimensions: getDimensions(DEFAULT_EMBEDDING_MODEL),
+    isConfigured(): boolean {
+      if (config.apiKey) {
+        return true;
+      }
+      if (process.env["OPENAI_API_KEY"]) {
+        return true;
+      }
+      return false;
+    },
+  };
+
+  if (!config.apiKey) {
+    const { key } = await prompts(
+      {
+        type: "text",
+        name: "key",
+        message: askModels
+          ? "Please provide your OpenAI API key (or leave blank to use OPENAI_API_KEY env variable):"
+          : "Please provide your OpenAI API key (leave blank to skip):",
+        validate: (value: string) => {
+          if (askModels && !value) {
+            if (process.env.OPENAI_API_KEY) {
+              return true;
+            }
+            return "OPENAI_API_KEY env variable is not set - key is required";
+          }
+          return true;
+        },
+      },
+      questionHandlers,
+    );
+    config.apiKey = key || process.env.OPENAI_API_KEY;
+  }
+
+  // use default model values in CI or if user should not be asked
+  const useDefaults = ciInfo.isCI || !askModels;
+  if (!useDefaults) {
+    const { model } = await prompts(
+      {
+        type: "select",
+        name: "model",
+        message: "Which LLM model would you like to use?",
+        choices: await getAvailableModelChoices(false, config.apiKey),
+        initial: 0,
+      },
+      questionHandlers,
+    );
+    config.model = model;
+
+    const { embeddingModel } = await prompts(
+      {
+        type: "select",
+        name: "embeddingModel",
+        message: "Which embedding model would you like to use?",
+        choices: await getAvailableModelChoices(true, config.apiKey),
+        initial: 0,
+      },
+      questionHandlers,
+    );
+    config.embeddingModel = embeddingModel;
+    config.dimensions = getDimensions(embeddingModel);
+  }
+
+  return config;
+}
+
+async function getAvailableModelChoices(
+  selectEmbedding: boolean,
+  apiKey?: string,
+) {
+  if (!apiKey) {
+    throw new Error("need OpenAI key to retrieve model choices");
+  }
+  const isLLMModel = (modelId: string) => {
+    return modelId.startsWith("gpt");
+  };
+
+  const isEmbeddingModel = (modelId: string) => {
+    return modelId.includes("embedding");
+  };
+
+  const spinner = ora("Fetching available models").start();
+  try {
+    const response = await got(`${OPENAI_API_URL}/models`, {
+      headers: {
+        Authorization: "Bearer " + apiKey,
+      },
+      timeout: 5000,
+      responseType: "json",
+    });
+    const data: any = await response.body;
+    spinner.stop();
+    return data.data
+      .filter((model: any) =>
+        selectEmbedding ? isEmbeddingModel(model.id) : isLLMModel(model.id),
+      )
+      .map((el: any) => {
+        return {
+          title: el.id,
+          value: el.id,
+        };
+      });
+  } catch (error) {
+    spinner.stop();
+    if ((error as any).response?.statusCode === 401) {
+      console.log(
+        red(
+          "Invalid OpenAI API key provided! Please provide a valid key and try again!",
+        ),
+      );
+    } else {
+      console.log(red("Request failed: " + error));
+    }
+    process.exit(1);
+  }
+}
+
+function getDimensions(modelName: string) {
+  // at 2024-04-24 all OpenAI embedding models support 1536 dimensions except
+  // "text-embedding-3-large", see https://openai.com/blog/new-embedding-models-and-api-updates
+  return modelName === "text-embedding-3-large" ? 1024 : 1536;
+}

helpers/python.ts

@@ -3,16 +3,16 @@ import path from "path";
 import { cyan, red } from "picocolors";
 import { parse, stringify } from "smol-toml";
 import terminalLink from "terminal-link";
-import yaml, { Document } from "yaml";
-import { copy } from "./copy";
+
+import { assetRelocator, copy } from "./copy";
 import { templatesDir } from "./dir";
 import { isPoetryAvailable, tryPoetryInstall } from "./poetry";
 import { Tool } from "./tools";
 import {
   InstallTemplateArgs,
+  ModelConfig,
   TemplateDataSource,
   TemplateVectorDB,
-  WebSourceConfig,
 } from "./types";
 
 interface Dependency {
@@ -22,8 +22,9 @@ interface Dependency {
 }
 
 const getAdditionalDependencies = (
+  modelConfig: ModelConfig,
   vectorDb?: TemplateVectorDB,
-  dataSource?: TemplateDataSource,
+  dataSources?: TemplateDataSource[],
   tools?: Tool[],
 ) => {
   const dependencies: Dependency[] = [];
@@ -42,6 +43,7 @@ const getAdditionalDependencies = (
         name: "llama-index-vector-stores-postgres",
         version: "^0.1.1",
       });
+      break;
     }
     case "pinecone": {
       dependencies.push({
@@ -61,30 +63,98 @@ const getAdditionalDependencies = (
       });
       break;
     }
+    case "astra": {
+      dependencies.push({
+        name: "llama-index-vector-stores-astra-db",
+        version: "^0.1.5",
+      });
+      break;
+    }
   }
 
   // Add data source dependencies
-  const dataSourceType = dataSource?.type;
-  if (dataSourceType === "file") {
-    // llama-index-readers-file (pdf, excel, csv) is already included in llama_index package
-    dependencies.push({
-      name: "docx2txt",
-      version: "^0.8",
-    });
-  } else if (dataSourceType === "web") {
-    dependencies.push({
-      name: "llama-index-readers-web",
-      version: "^0.1.6",
-    });
+  if (dataSources) {
+    for (const ds of dataSources) {
+      const dsType = ds?.type;
+      switch (dsType) {
+        case "file":
+          dependencies.push({
+            name: "docx2txt",
+            version: "^0.8",
+          });
+          break;
+        case "web":
+          dependencies.push({
+            name: "llama-index-readers-web",
+            version: "^0.1.6",
+          });
+          break;
+        case "db":
+          dependencies.push({
+            name: "llama-index-readers-database",
+            version: "^0.1.3",
+          });
+          dependencies.push({
+            name: "pymysql",
+            version: "^1.1.0",
+            extras: ["rsa"],
+          });
+          dependencies.push({
+            name: "psycopg2",
+            version: "^2.9.9",
+          });
+          break;
+      }
+    }
   }
 
   // Add tools dependencies
+  console.log("Adding tools dependencies");
   tools?.forEach((tool) => {
     tool.dependencies?.forEach((dep) => {
       dependencies.push(dep);
     });
   });
 
+  switch (modelConfig.provider) {
+    case "ollama":
+      dependencies.push({
+        name: "llama-index-llms-ollama",
+        version: "0.1.2",
+      });
+      dependencies.push({
+        name: "llama-index-embeddings-ollama",
+        version: "0.1.2",
+      });
+      break;
+    case "openai":
+      dependencies.push({
+        name: "llama-index-agent-openai",
+        version: "0.2.2",
+      });
+      break;
+    case "anthropic":
+      dependencies.push({
+        name: "llama-index-llms-anthropic",
+        version: "0.1.10",
+      });
+      dependencies.push({
+        name: "llama-index-embeddings-huggingface",
+        version: "0.2.0",
+      });
+      break;
+    case "gemini":
+      dependencies.push({
+        name: "llama-index-llms-gemini",
+        version: "0.1.7",
+      });
+      dependencies.push({
+        name: "llama-index-embeddings-gemini",
+        version: "0.1.6",
+      });
+      break;
+  }
+
   return dependencies;
 };
 
@@ -182,7 +252,8 @@ export const installPythonTemplate = async ({
   dataSources,
   tools,
   postInstallAction,
-  useLlamaParse,
+  observability,
+  modelConfig,
 }: Pick<
   InstallTemplateArgs,
   | "root"
@@ -191,133 +262,75 @@ export const installPythonTemplate = async ({
   | "vectorDb"
   | "dataSources"
   | "tools"
-  | "useLlamaParse"
   | "postInstallAction"
+  | "observability"
+  | "modelConfig"
 >) => {
   console.log("\nInitializing Python project with template:", template, "\n");
   const templatePath = path.join(templatesDir, "types", template, framework);
   await copy("**", root, {
     parents: true,
     cwd: templatePath,
-    rename(name) {
-      switch (name) {
-        case "gitignore": {
-          return `.${name}`;
-        }
-        // README.md is ignored by webpack-asset-relocator-loader used by ncc:
-        // https://github.com/vercel/webpack-asset-relocator-loader/blob/e9308683d47ff507253e37c9bcbb99474603192b/src/asset-relocator.js#L227
-        case "README-template.md": {
-          return "README.md";
-        }
-        default: {
-          return name;
-        }
-      }
-    },
+    rename: assetRelocator,
   });
 
   const compPath = path.join(templatesDir, "components");
+  const enginePath = path.join(root, "app", "engine");
 
-  if (dataSources.length > 0) {
-    const enginePath = path.join(root, "app", "engine");
+  // Copy selected vector DB
+  await copy("**", enginePath, {
+    parents: true,
+    cwd: path.join(compPath, "vectordbs", "python", vectorDb ?? "none"),
+  });
 
-    const vectorDbDirName = vectorDb ?? "none";
-    const VectorDBPath = path.join(
-      compPath,
-      "vectordbs",
+  // Copy all loaders to enginePath
+  const loaderPath = path.join(enginePath, "loaders");
+  await copy("**", loaderPath, {
+    parents: true,
+    cwd: path.join(compPath, "loaders", "python"),
+  });
+
+  // Select and copy engine code based on data sources and tools
+  let engine;
+  tools = tools ?? [];
+  if (dataSources.length > 0 && tools.length === 0) {
+    console.log("\nNo tools selected - use optimized context chat engine\n");
+    engine = "chat";
+  } else {
+    engine = "agent";
+  }
+  await copy("**", enginePath, {
+    parents: true,
+    cwd: path.join(compPath, "engines", "python", engine),
+  });
+
+  console.log("Adding additional dependencies");
+
+  const addOnDependencies = getAdditionalDependencies(
+    modelConfig,
+    vectorDb,
+    dataSources,
+    tools,
+  );
+
+  if (observability === "opentelemetry") {
+    addOnDependencies.push({
+      name: "traceloop-sdk",
+      version: "^0.15.11",
+    });
+
+    const templateObservabilityPath = path.join(
+      templatesDir,
+      "components",
+      "observability",
       "python",
-      vectorDbDirName,
+      "opentelemetry",
     );
-    await copy("**", enginePath, {
-      parents: true,
-      cwd: VectorDBPath,
+    await copy("**", path.join(root, "app"), {
+      cwd: templateObservabilityPath,
     });
-
-    // Copy engine code
-    if (tools !== undefined && tools.length > 0) {
-      await copy("**", enginePath, {
-        parents: true,
-        cwd: path.join(compPath, "engines", "python", "agent"),
-      });
-      // Write tool configs
-      const configContent: Record<string, any> = {};
-      tools.forEach((tool) => {
-        configContent[tool.name] = tool.config ?? {};
-      });
-      const configFilePath = path.join(root, "config/tools.yaml");
-      await fs.mkdir(path.join(root, "config"), { recursive: true });
-      await fs.writeFile(configFilePath, yaml.stringify(configContent));
-    } else {
-      await copy("**", enginePath, {
-        parents: true,
-        cwd: path.join(compPath, "engines", "python", "chat"),
-      });
-    }
-
-    const loaderConfig = new Document({});
-    const loaderPath = path.join(enginePath, "loaders");
-
-    // Copy loaders to enginePath
-    await copy("**", loaderPath, {
-      parents: true,
-      cwd: path.join(compPath, "loaders", "python"),
-    });
-
-    // Generate loaders config
-    // Web loader config
-    if (dataSources.some((ds) => ds.type === "web")) {
-      const webLoaderConfig = new Document({});
-
-      // Create config for browser driver arguments
-      const driverArgNodeValue = webLoaderConfig.createNode([
-        "--no-sandbox",
-        "--disable-dev-shm-usage",
-      ]);
-      driverArgNodeValue.commentBefore =
-        " The arguments to pass to the webdriver. E.g.: add --headless to run in headless mode";
-      webLoaderConfig.set("driver_arguments", driverArgNodeValue);
-
-      // Create config for urls
-      const urlConfigs = dataSources
-        .filter((ds) => ds.type === "web")
-        .map((ds) => {
-          const dsConfig = ds.config as WebSourceConfig;
-          return {
-            base_url: dsConfig.baseUrl,
-            prefix: dsConfig.prefix,
-            depth: dsConfig.depth,
-          };
-        });
-      const urlConfigNode = webLoaderConfig.createNode(urlConfigs);
-      urlConfigNode.commentBefore = ` base_url: The URL to start crawling with
- prefix: Only crawl URLs matching the specified prefix
- depth: The maximum depth for BFS traversal
- You can add more websites by adding more entries (don't forget the - prefix from YAML)`;
-      webLoaderConfig.set("urls", urlConfigNode);
-
-      // Add web config to the loaders config
-      loaderConfig.set("web", webLoaderConfig);
-    }
-    // File loader config
-    if (dataSources.some((ds) => ds.type === "file")) {
-      // Add documentation to web loader config
-      const node = loaderConfig.createNode({
-        use_llama_parse: useLlamaParse,
-      });
-      node.commentBefore = ` use_llama_parse: Use LlamaParse if \`true\`. Needs a \`LLAMA_CLOUD_API_KEY\` from https://cloud.llamaindex.ai set as environment variable`;
-      loaderConfig.set("file", node);
-    }
-    // Write loaders config
-    if (Object.keys(loaderConfig).length > 0) {
-      const loaderConfigPath = path.join(root, "config/loaders.yaml");
-      await fs.mkdir(path.join(root, "config"), { recursive: true });
-      await fs.writeFile(loaderConfigPath, yaml.stringify(loaderConfig));
-    }
   }
 
-  const addOnDependencies = dataSources
-    .map((ds) => getAdditionalDependencies(vectorDb, ds, tools))
-    .flat();
   await addDependencies(root, addOnDependencies);
 
   if (postInstallAction === "runApp" || postInstallAction === "dependencies") {

helpers/tools.ts

@@ -1,13 +1,24 @@
+import fs from "fs/promises";
+import path from "path";
 import { red } from "picocolors";
+import yaml from "yaml";
+import { makeDir } from "./make-dir";
 import { TemplateFramework } from "./types";
 
+export enum ToolType {
+  LLAMAHUB = "llamahub",
+  LOCAL = "local",
+}
+
 export type Tool = {
   display: string;
   name: string;
   config?: Record<string, any>;
   dependencies?: ToolDependencies[];
   supportedFrameworks?: Array<TemplateFramework>;
+  type: ToolType;
 };
+
 export type ToolDependencies = {
   name: string;
   version?: string;
@@ -30,6 +41,7 @@ export const supportedTools: Tool[] = [
       },
     ],
     supportedFrameworks: ["fastapi"],
+    type: ToolType.LLAMAHUB,
   },
   {
     display: "Wikipedia",
@@ -41,6 +53,14 @@ export const supportedTools: Tool[] = [
       },
     ],
     supportedFrameworks: ["fastapi", "express", "nextjs"],
+    type: ToolType.LLAMAHUB,
+  },
+  {
+    display: "Weather",
+    name: "weather",
+    dependencies: [],
+    supportedFrameworks: ["fastapi", "express", "nextjs"],
+    type: ToolType.LOCAL,
   },
 ];
 
@@ -73,3 +93,43 @@ export const toolsRequireConfig = (tools?: Tool[]): boolean => {
   }
   return false;
 };
+
+export enum ConfigFileType {
+  YAML = "yaml",
+  JSON = "json",
+}
+
+export const writeToolsConfig = async (
+  root: string,
+  tools: Tool[] = [],
+  type: ConfigFileType = ConfigFileType.YAML,
+) => {
+  if (tools.length === 0) return; // no tools selected, no config need
+  const configContent: {
+    [key in ToolType]: Record<string, any>;
+  } = {
+    local: {},
+    llamahub: {},
+  };
+  tools.forEach((tool) => {
+    if (tool.type === ToolType.LLAMAHUB) {
+      configContent.llamahub[tool.name] = tool.config ?? {};
+    }
+    if (tool.type === ToolType.LOCAL) {
+      configContent.local[tool.name] = tool.config ?? {};
+    }
+  });
+  const configPath = path.join(root, "config");
+  await makeDir(configPath);
+  if (type === ConfigFileType.YAML) {
+    await fs.writeFile(
+      path.join(configPath, "tools.yaml"),
+      yaml.stringify(configContent),
+    );
+  } else {
+    await fs.writeFile(
+      path.join(configPath, "tools.json"),
+      JSON.stringify(configContent, null, 2),
+    );
+  }
+};

helpers/types.ts

@@ -1,10 +1,26 @@
 import { PackageManager } from "../helpers/get-pkg-manager";
 import { Tool } from "./tools";
 
+export type ModelProvider = "openai" | "ollama" | "anthropic" | "gemini";
+export type ModelConfig = {
+  provider: ModelProvider;
+  apiKey?: string;
+  model: string;
+  embeddingModel: string;
+  dimensions: number;
+  isConfigured(): boolean;
+};
 export type TemplateType = "streaming" | "community" | "llamapack";
 export type TemplateFramework = "nextjs" | "express" | "fastapi";
 export type TemplateUI = "html" | "shadcn";
-export type TemplateVectorDB = "none" | "mongo" | "pg" | "pinecone" | "milvus";
+export type TemplateVectorDB =
+  | "none"
+  | "mongo"
+  | "pg"
+  | "pinecone"
+  | "milvus"
+  | "astra"
+  | "qdrant";
 export type TemplatePostInstallAction =
   | "none"
   | "VSCode"
@@ -14,7 +30,7 @@ export type TemplateDataSource = {
   type: TemplateDataSourceType;
   config: TemplateDataSourceConfig;
 };
-export type TemplateDataSourceType = "file" | "web";
+export type TemplateDataSourceType = "file" | "web" | "db";
 export type TemplateObservability = "none" | "opentelemetry";
 // Config for both file and folder
 export type FileSourceConfig = {
@@ -25,8 +41,15 @@ export type WebSourceConfig = {
   prefix?: string;
   depth?: number;
 };
+export type DbSourceConfig = {
+  uri?: string;
+  queries?: string;
+};
 
-export type TemplateDataSourceConfig = FileSourceConfig | WebSourceConfig;
+export type TemplateDataSourceConfig =
+  | FileSourceConfig
+  | WebSourceConfig
+  | DbSourceConfig;
 
 export type CommunityProjectConfig = {
   owner: string;
@@ -45,11 +68,9 @@ export interface InstallTemplateArgs {
   ui: TemplateUI;
   dataSources: TemplateDataSource[];
   customApiPath?: string;
-  openAiKey?: string;
+  modelConfig: ModelConfig;
   llamaCloudKey?: string;
   useLlamaParse?: boolean;
-  model: string;
-  embeddingModel: string;
   communityProjectConfig?: CommunityProjectConfig;
   llamapack?: string;
   vectorDb?: TemplateVectorDB;

helpers/typescript.ts

@@ -2,51 +2,12 @@ import fs from "fs/promises";
 import os from "os";
 import path from "path";
 import { bold, cyan } from "picocolors";
-import { copy } from "../helpers/copy";
+import { assetRelocator, copy } from "../helpers/copy";
 import { callPackageManager } from "../helpers/install";
 import { templatesDir } from "./dir";
 import { PackageManager } from "./get-pkg-manager";
-import { makeDir } from "./make-dir";
 import { InstallTemplateArgs } from "./types";
 
-const rename = (name: string) => {
-  switch (name) {
-    case "gitignore":
-    case "eslintrc.json": {
-      return `.${name}`;
-    }
-    // README.md is ignored by webpack-asset-relocator-loader used by ncc:
-    // https://github.com/vercel/webpack-asset-relocator-loader/blob/e9308683d47ff507253e37c9bcbb99474603192b/src/asset-relocator.js#L227
-    case "README-template.md": {
-      return "README.md";
-    }
-    default: {
-      return name;
-    }
-  }
-};
-
-export const installTSDependencies = async (
-  packageJson: any,
-  packageManager: PackageManager,
-  isOnline: boolean,
-): Promise<void> => {
-  console.log("\nInstalling dependencies:");
-  for (const dependency in packageJson.dependencies)
-    console.log(`- ${cyan(dependency)}`);
-
-  console.log("\nInstalling devDependencies:");
-  for (const dependency in packageJson.devDependencies)
-    console.log(`- ${cyan(dependency)}`);
-
-  console.log();
-
-  await callPackageManager(packageManager, isOnline).catch((error) => {
-    console.error("Failed to install TS dependencies. Exiting...");
-    process.exit(1);
-  });
-};
-
 /**
  * Install a LlamaIndex internal template to a given `root` directory.
  */
@@ -58,7 +19,6 @@ export const installTSTemplate = async ({
   template,
   framework,
   ui,
-  customApiPath,
   vectorDb,
   postInstallAction,
   backend,
@@ -79,7 +39,7 @@ export const installTSTemplate = async ({
   await copy(copySource, root, {
     parents: true,
     cwd: templatePath,
-    rename,
+    rename: assetRelocator,
   });
 
   /**
@@ -137,9 +97,6 @@ export const installTSTemplate = async ({
     );
   }
 
-  /**
-   * Copy the selected chat engine files to the target directory and reference it.
-   */
   const compPath = path.join(templatesDir, "components");
   const relativeEngineDestPath =
     framework === "nextjs"
@@ -147,59 +104,33 @@ export const installTSTemplate = async ({
       : path.join("src", "controllers");
   const enginePath = path.join(root, relativeEngineDestPath, "engine");
 
-  if (dataSources.length === 0) {
-    // use simple hat engine if user neither select tools nor a data source
-    console.log("\nUsing simple chat engine\n");
-  } else {
-    if (vectorDb) {
-      // copy vector db component
-      console.log("\nUsing vector DB:", vectorDb, "\n");
-      const vectorDBPath = path.join(
-        compPath,
-        "vectordbs",
-        "typescript",
-        vectorDb,
-      );
-      await copy("**", enginePath, {
-        parents: true,
-        cwd: vectorDBPath,
-      });
-    }
-    // copy loader component (TS only supports llama_parse and file for now)
-    let loaderFolder: string;
-    loaderFolder = useLlamaParse ? "llama_parse" : "file";
-    await copy("**", enginePath, {
-      parents: true,
-      cwd: path.join(compPath, "loaders", "typescript", loaderFolder),
-    });
-    if (tools?.length) {
-      // use agent chat engine if user selects tools
-      console.log("\nUsing agent chat engine\n");
-      await copy("**", enginePath, {
-        parents: true,
-        cwd: path.join(compPath, "engines", "typescript", "agent"),
-      });
+  // copy vector db component
+  console.log("\nUsing vector DB:", vectorDb ?? "none", "\n");
+  await copy("**", enginePath, {
+    parents: true,
+    cwd: path.join(compPath, "vectordbs", "typescript", vectorDb ?? "none"),
+  });
 
-      // Write config/tools.json
-      const configContent: Record<string, any> = {};
-      tools.forEach((tool) => {
-        configContent[tool.name] = tool.config ?? {};
-      });
-      const configPath = path.join(root, "config");
-      await makeDir(configPath);
-      await fs.writeFile(
-        path.join(configPath, "tools.json"),
-        JSON.stringify(configContent, null, 2),
-      );
-    } else {
-      // use context chat engine if user does not select tools
-      console.log("\nUsing context chat engine\n");
-      await copy("**", enginePath, {
-        parents: true,
-        cwd: path.join(compPath, "engines", "typescript", "chat"),
-      });
-    }
+  // copy loader component (TS only supports llama_parse and file for now)
+  const loaderFolder = useLlamaParse ? "llama_parse" : "file";
+  await copy("**", enginePath, {
+    parents: true,
+    cwd: path.join(compPath, "loaders", "typescript", loaderFolder),
+  });
+
+  // Select and copy engine code based on data sources and tools
+  let engine;
+  tools = tools ?? [];
+  if (dataSources.length > 0 && tools.length === 0) {
+    console.log("\nNo tools selected - use optimized context chat engine\n");
+    engine = "chat";
+  } else {
+    engine = "agent";
   }
+  await copy("**", enginePath, {
+    parents: true,
+    cwd: path.join(compPath, "engines", "typescript", engine),
+  });
 
   /**
    * Copy the selected UI files to the target directory and reference it.
@@ -214,13 +145,54 @@ export const installTSTemplate = async ({
     await copy("**", destUiPath, {
       parents: true,
       cwd: uiPath,
-      rename,
+      rename: assetRelocator,
     });
   }
 
-  /**
-   * Update the package.json scripts.
-   */
+  /** Modify frontend code to use custom API path */
+  if (framework === "nextjs" && !backend) {
+    console.log(
+      "\nUsing external API for frontend, removing API code and configuration\n",
+    );
+    // remove the default api folder and config folder
+    await fs.rm(path.join(root, "app", "api"), { recursive: true });
+    await fs.rm(path.join(root, "config"), { recursive: true, force: true });
+  }
+
+  const packageJson = await updatePackageJson({
+    root,
+    appName,
+    dataSources,
+    relativeEngineDestPath,
+    framework,
+    ui,
+    observability,
+  });
+
+  if (postInstallAction === "runApp" || postInstallAction === "dependencies") {
+    await installTSDependencies(packageJson, packageManager, isOnline);
+  }
+
+  // Copy deployment files for typescript
+  await copy("**", root, {
+    cwd: path.join(compPath, "deployments", "typescript"),
+  });
+};
+
+async function updatePackageJson({
+  root,
+  appName,
+  dataSources,
+  relativeEngineDestPath,
+  framework,
+  ui,
+  observability,
+}: Pick<
+  InstallTemplateArgs,
+  "root" | "appName" | "dataSources" | "framework" | "ui" | "observability"
+> & {
+  relativeEngineDestPath: string;
+}): Promise<any> {
   const packageJsonFile = path.join(root, "package.json");
   const packageJson: any = JSON.parse(
     await fs.readFile(packageJsonFile, "utf8"),
@@ -228,26 +200,15 @@ export const installTSTemplate = async ({
   packageJson.name = appName;
   packageJson.version = "0.1.0";
 
-  if (framework === "nextjs" && customApiPath) {
-    console.log(
-      "\nUsing external API with custom API path:",
-      customApiPath,
-      "\n",
-    );
-    // remove the default api folder
-    const apiPath = path.join(root, "app", "api");
-    await fs.rm(apiPath, { recursive: true });
-    // modify the dev script to use the custom api path
-  }
-
-  if (dataSources.length > 0 && relativeEngineDestPath) {
+  if (relativeEngineDestPath) {
+    // TODO: move script to {root}/scripts for all frameworks
     // add generate script if using context engine
     packageJson.scripts = {
       ...packageJson.scripts,
-      generate: `node ${path.join(
+      generate: `tsx ${path.join(
         relativeEngineDestPath,
         "engine",
-        "generate.mjs",
+        "generate.ts",
       )}`,
     };
   }
@@ -292,12 +253,26 @@ export const installTSTemplate = async ({
     JSON.stringify(packageJson, null, 2) + os.EOL,
   );
 
-  if (postInstallAction === "runApp" || postInstallAction === "dependencies") {
-    await installTSDependencies(packageJson, packageManager, isOnline);
-  }
+  return packageJson;
+}
 
-  // Copy deployment files for typescript
-  await copy("**", root, {
-    cwd: path.join(compPath, "deployments", "typescript"),
+async function installTSDependencies(
+  packageJson: any,
+  packageManager: PackageManager,
+  isOnline: boolean,
+): Promise<void> {
+  console.log("\nInstalling dependencies:");
+  for (const dependency in packageJson.dependencies)
+    console.log(`- ${cyan(dependency)}`);
+
+  console.log("\nInstalling devDependencies:");
+  for (const dependency in packageJson.devDependencies)
+    console.log(`- ${cyan(dependency)}`);
+
+  console.log();
+
+  await callPackageManager(packageManager, isOnline).catch((error) => {
+    console.error("Failed to install TS dependencies. Exiting...");
+    process.exit(1);
   });
-};
+}

index.ts

@@ -78,7 +78,7 @@ const program = new Commander.Command(packageJson.name)
     "--files <path>",
     `
   
-    Specify the path to a local file or folder for chatting.
+  Specify the path to a local file or folder for chatting.
 `,
   )
   .option(
@@ -107,19 +107,6 @@ const program = new Commander.Command(packageJson.name)
     `
 
   Whether to generate a frontend for your backend.
-`,
-  )
-  .option(
-    "--model <model>",
-    `
-
-  Select OpenAI model to use. E.g. gpt-3.5-turbo.
-`,
-  )
-  .option(
-    "--embedding-model <embeddingModel>",
-    `
-  Select OpenAI embedding model to use. E.g. text-embedding-ada-002.
 `,
   )
   .option(
@@ -160,22 +147,30 @@ const program = new Commander.Command(packageJson.name)
   .option(
     "--use-llama-parse",
     `
-    Enable LlamaParse.
+
+  Enable LlamaParse.
 `,
   )
   .option(
     "--llama-cloud-key <key>",
     `
+  
   Provide a LlamaCloud API key.
 `,
   )
   .option(
-    "--list-server-models",
-    "Fetch available LLM and embedding models from OpenAI API.",
+    "--observability <observability>",
+    `
+    
+  Specify observability tools to use. Eg: none, opentelemetry
+`,
   )
   .option(
-    "--observability <observability>",
-    "Specify observability tools to use. Eg: none, opentelemetry",
+    "--ask-models",
+    `
+
+  Select LLM and embedding models.
+`,
   )
   .allowUnknownOption()
   .parse(process.argv);
@@ -192,6 +187,7 @@ if (process.argv.includes("--tools")) {
 if (process.argv.includes("--no-llama-parse")) {
   program.useLlamaParse = false;
 }
+program.askModels = process.argv.includes("--ask-models");
 if (process.argv.includes("--no-files")) {
   program.dataSources = [];
 } else {
@@ -278,7 +274,11 @@ async function run(): Promise<void> {
   }
 
   const preferences = (conf.get("preferences") || {}) as QuestionArgs;
-  await askQuestions(program as unknown as QuestionArgs, preferences);
+  await askQuestions(
+    program as unknown as QuestionArgs,
+    preferences,
+    program.openAiKey,
+  );
 
   await createApp({
     template: program.template,
@@ -287,10 +287,8 @@ async function run(): Promise<void> {
     appPath: resolvedProjectPath,
     packageManager,
     frontend: program.frontend,
-    openAiKey: program.openAiKey,
+    modelConfig: program.modelConfig,
     llamaCloudKey: program.llamaCloudKey,
-    model: program.model,
-    embeddingModel: program.embeddingModel,
     communityProjectConfig: program.communityProjectConfig,
     llamapack: program.llamapack,
     vectorDb: program.vectorDb,

package.json

@@ -1,12 +1,12 @@
 {
   "name": "create-llama",
-  "version": "0.0.31",
+  "version": "0.1.3",
+  "description": "Create LlamaIndex-powered apps with one command",
   "keywords": [
     "rag",
     "llamaindex",
     "next.js"
   ],
-  "description": "Create LlamaIndex-powered apps with one command",
   "repository": {
     "type": "git",
     "url": "https://github.com/run-llama/LlamaIndexTS",
@@ -20,32 +20,30 @@
     "dist"
   ],
   "scripts": {
-    "clean": "rimraf --glob ./dist ./templates/**/__pycache__ ./templates/**/node_modules ./templates/**/poetry.lock",
-    "format": "prettier --ignore-unknown --cache --check .",
-    "format:write": "prettier --ignore-unknown --write .",
-    "dev": "ncc build ./index.ts -w -o dist/",
     "build": "bash ./scripts/build.sh",
     "build:ncc": "pnpm run clean && ncc build ./index.ts -o ./dist/ --minify --no-cache --no-source-map-register",
-    "lint": "eslint . --ignore-pattern dist --ignore-pattern e2e/cache",
+    "clean": "rimraf --glob ./dist ./templates/**/__pycache__ ./templates/**/node_modules ./templates/**/poetry.lock",
+    "dev": "ncc build ./index.ts -w -o dist/",
     "e2e": "playwright test",
+    "format": "prettier --ignore-unknown --cache --check .",
+    "format:write": "prettier --ignore-unknown --write .",
+    "lint": "eslint . --ignore-pattern dist --ignore-pattern e2e/cache",
+    "new-snapshot": "pnpm run build && changeset version --snapshot",
+    "new-version": "pnpm run build && changeset version",
+    "pack-install": "bash ./scripts/pack.sh",
     "prepare": "husky",
     "release": "pnpm run build && changeset publish",
-    "new-version": "pnpm run build && changeset version",
-    "release-snapshot": "pnpm run build && changeset publish --tag snapshot",
-    "new-snapshot": "pnpm run build && changeset version --snapshot",
-    "pack-install": "bash ./scripts/pack.sh"
+    "release-snapshot": "pnpm run build && changeset publish --tag snapshot"
   },
-  "devDependencies": {
-    "@playwright/test": "^1.41.1",
+  "dependencies": {
     "@types/async-retry": "1.4.2",
     "@types/ci-info": "2.0.0",
     "@types/cross-spawn": "6.0.0",
+    "@types/fs-extra": "11.0.4",
     "@types/node": "^20.11.7",
     "@types/prompts": "2.0.1",
     "@types/tar": "6.1.5",
     "@types/validate-npm-package-name": "3.0.0",
-    "@types/fs-extra": "11.0.4",
-    "@vercel/ncc": "0.38.1",
     "async-retry": "1.3.1",
     "async-sema": "3.0.1",
     "ci-info": "github:watson/ci-info#f43f6a1cefff47fb361c88cf4b943fdbcaafe540",
@@ -53,29 +51,34 @@
     "conf": "10.2.0",
     "cross-spawn": "7.0.3",
     "fast-glob": "3.3.1",
+    "fs-extra": "11.2.0",
     "got": "10.7.0",
+    "ollama": "^0.5.0",
+    "ora": "^8.0.1",
     "picocolors": "1.0.0",
     "prompts": "2.1.0",
-    "rimraf": "^5.0.5",
     "smol-toml": "^1.1.4",
     "tar": "6.1.15",
     "terminal-link": "^3.0.0",
     "update-check": "1.5.4",
     "validate-npm-package-name": "3.0.0",
-    "wait-port": "^1.1.0",
+    "yaml": "2.4.1"
+  },
+  "devDependencies": {
     "@changesets/cli": "^2.27.1",
+    "@playwright/test": "^1.41.1",
+    "@vercel/ncc": "0.38.1",
     "eslint": "^8.56.0",
+    "eslint-config-prettier": "^8.10.0",
     "husky": "^9.0.10",
     "prettier": "^3.2.5",
     "prettier-plugin-organize-imports": "^3.2.4",
+    "rimraf": "^5.0.5",
     "typescript": "^5.3.3",
-    "eslint-config-prettier": "^8.10.0",
-    "ora": "^8.0.1",
-    "fs-extra": "11.2.0",
-    "yaml": "2.4.1"
+    "wait-port": "^1.1.0"
   },
+  "packageManager": "pnpm@9.0.5",
   "engines": {
     "node": ">=16.14.0"
-  },
-  "packageManager": "pnpm@8.15.1"
+  }
 }

questions.ts

@@ -1,8 +1,6 @@
 import { execSync } from "child_process";
 import ciInfo from "ci-info";
 import fs from "fs";
-import got from "got";
-import ora from "ora";
 import path from "path";
 import { blue, green, red } from "picocolors";
 import prompts from "prompts";
@@ -16,16 +14,15 @@ import { COMMUNITY_OWNER, COMMUNITY_REPO } from "./helpers/constant";
 import { EXAMPLE_FILE } from "./helpers/datasources";
 import { templatesDir } from "./helpers/dir";
 import { getAvailableLlamapackOptions } from "./helpers/llama-pack";
+import { askModelConfig } from "./helpers/providers";
 import { getProjectOptions } from "./helpers/repo";
 import { supportedTools, toolsRequireConfig } from "./helpers/tools";
 
-const OPENAI_API_URL = "https://api.openai.com/v1";
-
 export type QuestionArgs = Omit<
   InstallAppArgs,
   "appPath" | "packageManager"
 > & {
-  listServerModels?: boolean;
+  askModels?: boolean;
 };
 const supportedContextFileTypes = [
   ".pdf",
@@ -67,16 +64,13 @@ if ($dialogResult -eq [System.Windows.Forms.DialogResult]::OK)
 }
 `;
 
-const defaults: QuestionArgs = {
+const defaults: Omit<QuestionArgs, "modelConfig"> = {
   template: "streaming",
   framework: "nextjs",
   ui: "shadcn",
   frontend: false,
-  openAiKey: "",
   llamaCloudKey: "",
   useLlamaParse: false,
-  model: "gpt-3.5-turbo",
-  embeddingModel: "text-embedding-ada-002",
   communityProjectConfig: undefined,
   llamapack: "",
   postInstallAction: "dependencies",
@@ -84,7 +78,7 @@ const defaults: QuestionArgs = {
   tools: [],
 };
 
-const handlers = {
+export const questionHandlers = {
   onCancel: () => {
     console.error("Exiting.");
     process.exit(1);
@@ -101,6 +95,8 @@ const getVectorDbChoices = (framework: TemplateFramework) => {
     { title: "PostgreSQL", value: "pg" },
     { title: "Pinecone", value: "pinecone" },
     { title: "Milvus", value: "milvus" },
+    { title: "Astra", value: "astra" },
+    { title: "Qdrant", value: "qdrant" },
   ];
 
   const vectordbLang = framework === "fastapi" ? "python" : "typescript";
@@ -131,7 +127,7 @@ export const getDataSourceChoices = (
   }
   if (selectedDataSource === undefined || selectedDataSource.length === 0) {
     choices.push({
-      title: "No data, just a simple chat",
+      title: "No data, just a simple chat or agent",
       value: "none",
     });
     choices.push({
@@ -159,6 +155,10 @@ export const getDataSourceChoices = (
       title: "Use website content (requires Chrome)",
       value: "web",
     });
+    choices.push({
+      title: "Use data from a database (Mysql, PostgreSQL)",
+      value: "db",
+    });
   }
   return choices;
 };
@@ -226,85 +226,15 @@ export const onPromptState = (state: any) => {
   }
 };
 
-const getAvailableModelChoices = async (
-  selectEmbedding: boolean,
-  apiKey?: string,
-  listServerModels?: boolean,
-) => {
-  const defaultLLMModels = [
-    "gpt-3.5-turbo-0125",
-    "gpt-4-turbo-preview",
-    "gpt-4",
-    "gpt-4-vision-preview",
-  ];
-  const defaultEmbeddingModels = [
-    "text-embedding-ada-002",
-    "text-embedding-3-small",
-    "text-embedding-3-large",
-  ];
-
-  const isLLMModels = (model_id: string) => {
-    return model_id.startsWith("gpt");
-  };
-
-  const isEmbeddingModel = (model_id: string) => {
-    return (
-      model_id.includes("embedding") ||
-      defaultEmbeddingModels.includes(model_id)
-    );
-  };
-
-  if (apiKey && listServerModels) {
-    const spinner = ora("Fetching available models").start();
-    try {
-      const response = await got(`${OPENAI_API_URL}/models`, {
-        headers: {
-          Authorization: "Bearer " + apiKey,
-        },
-        timeout: 5000,
-        responseType: "json",
-      });
-      const data: any = await response.body;
-      spinner.stop();
-      return data.data
-        .filter((model: any) =>
-          selectEmbedding ? isEmbeddingModel(model.id) : isLLMModels(model.id),
-        )
-        .map((el: any) => {
-          return {
-            title: el.id,
-            value: el.id,
-          };
-        });
-    } catch (error) {
-      spinner.stop();
-      if ((error as any).response?.statusCode === 401) {
-        console.log(
-          red(
-            "Invalid OpenAI API key provided! Please provide a valid key and try again!",
-          ),
-        );
-      } else {
-        console.log(red("Request failed: " + error));
-      }
-      process.exit(1);
-    }
-  } else {
-    const data = selectEmbedding ? defaultEmbeddingModels : defaultLLMModels;
-    return data.map((model) => ({
-      title: model,
-      value: model,
-    }));
-  }
-};
-
 export const askQuestions = async (
   program: QuestionArgs,
   preferences: QuestionArgs,
+  openAiKey?: string,
 ) => {
-  const getPrefOrDefault = <K extends keyof QuestionArgs>(
+  const getPrefOrDefault = <K extends keyof Omit<QuestionArgs, "modelConfig">>(
     field: K,
-  ): QuestionArgs[K] => preferences[field] ?? defaults[field];
+  ): Omit<QuestionArgs, "modelConfig">[K] =>
+    preferences[field] ?? defaults[field];
 
   // Ask for next action after installation
   async function askPostInstallAction() {
@@ -327,8 +257,8 @@ export const askQuestions = async (
           },
         ];
 
-        const openAiKeyConfigured =
-          program.openAiKey || process.env["OPENAI_API_KEY"];
+        const modelConfigured =
+          !program.llamapack && program.modelConfig.isConfigured();
         // If using LlamaParse, require LlamaCloud API key
         const llamaCloudKeyConfigured = program.useLlamaParse
           ? program.llamaCloudKey || process.env["LLAMA_CLOUD_API_KEY"]
@@ -337,10 +267,9 @@ export const askQuestions = async (
         // Can run the app if all tools do not require configuration
         if (
           !hasVectorDb &&
-          openAiKeyConfigured &&
+          modelConfigured &&
           llamaCloudKeyConfigured &&
-          !toolsRequireConfig(program.tools) &&
-          !program.llamapack
+          !toolsRequireConfig(program.tools)
         ) {
           actionChoices.push({
             title:
@@ -357,7 +286,7 @@ export const askQuestions = async (
             choices: actionChoices,
             initial: 1,
           },
-          handlers,
+          questionHandlers,
         );
 
         program.postInstallAction = action;
@@ -390,7 +319,7 @@ export const askQuestions = async (
           ],
           initial: 0,
         },
-        handlers,
+        questionHandlers,
       );
       program.template = template;
       preferences.template = template;
@@ -413,7 +342,7 @@ export const askQuestions = async (
         })),
         initial: 0,
       },
-      handlers,
+      questionHandlers,
     );
     const projectConfig = JSON.parse(communityProjectConfig);
     program.communityProjectConfig = projectConfig;
@@ -434,7 +363,7 @@ export const askQuestions = async (
         })),
         initial: 0,
       },
-      handlers,
+      questionHandlers,
     );
     program.llamapack = llamapack;
     preferences.llamapack = llamapack;
@@ -460,7 +389,7 @@ export const askQuestions = async (
           choices,
           initial: 0,
         },
-        handlers,
+        questionHandlers,
       );
       program.framework = framework;
       preferences.framework = framework;
@@ -469,7 +398,6 @@ export const askQuestions = async (
 
   if (program.framework === "express" || program.framework === "fastapi") {
     // if a backend-only framework is selected, ask whether we should create a frontend
-    // (only for streaming backends)
     if (program.frontend === undefined) {
       if (ciInfo.isCI) {
         program.frontend = getPrefOrDefault("frontend");
@@ -501,104 +429,40 @@ export const askQuestions = async (
 
   if (program.framework === "nextjs" || program.frontend) {
     if (!program.ui) {
-      program.ui = getPrefOrDefault("ui");
+      program.ui = defaults.ui;
     }
   }
 
-  if (program.framework === "express" || program.framework === "nextjs") {
-    if (!program.observability) {
-      if (ciInfo.isCI) {
-        program.observability = getPrefOrDefault("observability");
-      } else {
-        const { observability } = await prompts(
-          {
-            type: "select",
-            name: "observability",
-            message: "Would you like to set up observability?",
-            choices: [
-              { title: "No", value: "none" },
-              { title: "OpenTelemetry", value: "opentelemetry" },
-            ],
-            initial: 0,
-          },
-          handlers,
-        );
-
-        program.observability = observability;
-        preferences.observability = observability;
-      }
-    }
-  }
-
-  if (!program.openAiKey) {
-    const { key } = await prompts(
-      {
-        type: "text",
-        name: "key",
-        message: program.listServerModels
-          ? "Please provide your OpenAI API key (or reuse OPENAI_API_KEY env variable):"
-          : "Please provide your OpenAI API key (leave blank to skip):",
-        validate: (value: string) => {
-          if (program.listServerModels && !value) {
-            if (process.env.OPENAI_API_KEY) {
-              return true;
-            }
-            return "OpenAI API key is required";
-          }
-          return true;
-        },
-      },
-      handlers,
-    );
-
-    program.openAiKey = key || process.env.OPENAI_API_KEY;
-    preferences.openAiKey = key || process.env.OPENAI_API_KEY;
-  }
-
-  if (!program.model) {
+  if (!program.observability) {
     if (ciInfo.isCI) {
-      program.model = getPrefOrDefault("model");
+      program.observability = getPrefOrDefault("observability");
     } else {
-      const { model } = await prompts(
+      const { observability } = await prompts(
         {
           type: "select",
-          name: "model",
-          message: "Which model would you like to use?",
-          choices: await getAvailableModelChoices(
-            false,
-            program.openAiKey,
-            program.listServerModels,
-          ),
+          name: "observability",
+          message: "Would you like to set up observability?",
+          choices: [
+            { title: "No", value: "none" },
+            { title: "OpenTelemetry", value: "opentelemetry" },
+          ],
           initial: 0,
         },
-        handlers,
+        questionHandlers,
       );
-      program.model = model;
-      preferences.model = model;
+
+      program.observability = observability;
+      preferences.observability = observability;
     }
   }
 
-  if (!program.embeddingModel && program.framework === "fastapi") {
-    if (ciInfo.isCI) {
-      program.embeddingModel = getPrefOrDefault("embeddingModel");
-    } else {
-      const { embeddingModel } = await prompts(
-        {
-          type: "select",
-          name: "embeddingModel",
-          message: "Which embedding model would you like to use?",
-          choices: await getAvailableModelChoices(
-            true,
-            program.openAiKey,
-            program.listServerModels,
-          ),
-          initial: 0,
-        },
-        handlers,
-      );
-      program.embeddingModel = embeddingModel;
-      preferences.embeddingModel = embeddingModel;
-    }
+  if (!program.modelConfig) {
+    const modelConfig = await askModelConfig({
+      openAiKey,
+      askModels: program.askModels ?? false,
+    });
+    program.modelConfig = modelConfig;
+    preferences.modelConfig = modelConfig;
   }
 
   if (!program.dataSources) {
@@ -622,59 +486,100 @@ export const askQuestions = async (
             ),
             initial: firstQuestion ? 1 : 0,
           },
-          handlers,
+          questionHandlers,
         );
 
         if (selectedSource === "no" || selectedSource === "none") {
           // user doesn't want another data source or any data source
           break;
         }
-        if (selectedSource === "exampleFile") {
-          program.dataSources.push(EXAMPLE_FILE);
-        } else if (selectedSource === "file" || selectedSource === "folder") {
-          // Select local data source
-          const selectedPaths = await selectLocalContextData(selectedSource);
-          for (const p of selectedPaths) {
+        switch (selectedSource) {
+          case "exampleFile": {
+            program.dataSources.push(EXAMPLE_FILE);
+            break;
+          }
+          case "file":
+          case "folder": {
+            const selectedPaths = await selectLocalContextData(selectedSource);
+            for (const p of selectedPaths) {
+              program.dataSources.push({
+                type: "file",
+                config: {
+                  path: p,
+                },
+              });
+            }
+            break;
+          }
+          case "web": {
+            const { baseUrl } = await prompts(
+              {
+                type: "text",
+                name: "baseUrl",
+                message: "Please provide base URL of the website: ",
+                initial: "https://www.llamaindex.ai",
+                validate: (value: string) => {
+                  if (!value.includes("://")) {
+                    value = `https://${value}`;
+                  }
+                  const urlObj = new URL(value);
+                  if (
+                    urlObj.protocol !== "https:" &&
+                    urlObj.protocol !== "http:"
+                  ) {
+                    return `URL=${value} has invalid protocol, only allow http or https`;
+                  }
+                  return true;
+                },
+              },
+              questionHandlers,
+            );
+
             program.dataSources.push({
-              type: "file",
+              type: "web",
               config: {
-                path: p,
+                baseUrl,
+                prefix: baseUrl,
+                depth: 1,
               },
             });
+            break;
           }
-        } else if (selectedSource === "web") {
-          // Selected web data source
-          const { baseUrl } = await prompts(
-            {
-              type: "text",
-              name: "baseUrl",
-              message: "Please provide base URL of the website: ",
-              initial: "https://www.llamaindex.ai",
-              validate: (value: string) => {
-                if (!value.includes("://")) {
-                  value = `https://${value}`;
-                }
-                const urlObj = new URL(value);
-                if (
-                  urlObj.protocol !== "https:" &&
-                  urlObj.protocol !== "http:"
-                ) {
-                  return `URL=${value} has invalid protocol, only allow http or https`;
-                }
-                return true;
+          case "db": {
+            const dbPrompts: prompts.PromptObject<string>[] = [
+              {
+                type: "text",
+                name: "uri",
+                message:
+                  "Please enter the connection string (URI) for the database.",
+                initial: "mysql+pymysql://user:pass@localhost:3306/mydb",
+                validate: (value: string) => {
+                  if (!value) {
+                    return "Please provide a valid connection string";
+                  } else if (
+                    !(
+                      value.startsWith("mysql+pymysql://") ||
+                      value.startsWith("postgresql+psycopg://")
+                    )
+                  ) {
+                    return "The connection string must start with 'mysql+pymysql://' for MySQL or 'postgresql+psycopg://' for PostgreSQL";
+                  }
+                  return true;
+                },
               },
-            },
-            handlers,
-          );
-
-          program.dataSources.push({
-            type: "web",
-            config: {
-              baseUrl,
-              prefix: baseUrl,
-              depth: 1,
-            },
-          });
+              // Only ask for a query, user can provide more complex queries in the config file later
+              {
+                type: (prev) => (prev ? "text" : null),
+                name: "queries",
+                message: "Please enter the SQL query to fetch data:",
+                initial: "SELECT * FROM mytable",
+              },
+            ];
+            program.dataSources.push({
+              type: "db",
+              config: await prompts(dbPrompts, questionHandlers),
+            });
+          }
         }
       }
     }
@@ -699,7 +604,7 @@ export const askQuestions = async (
           active: "yes",
           inactive: "no",
         },
-        handlers,
+        questionHandlers,
       );
       program.useLlamaParse = useLlamaParse;
 
@@ -712,7 +617,7 @@ export const askQuestions = async (
             message:
               "Please provide your LlamaIndex Cloud API key (leave blank to skip):",
           },
-          handlers,
+          questionHandlers,
         );
         program.llamaCloudKey = llamaCloudKey;
       }
@@ -731,15 +636,14 @@ export const askQuestions = async (
           choices: getVectorDbChoices(program.framework),
           initial: 0,
         },
-        handlers,
+        questionHandlers,
       );
       program.vectorDb = vectorDb;
       preferences.vectorDb = vectorDb;
     }
   }
 
-  // TODO: allow tools also without datasources
-  if (!program.tools && program.dataSources.length > 0) {
+  if (!program.tools) {
     if (ciInfo.isCI) {
       program.tools = getPrefOrDefault("tools");
     } else {
@@ -767,3 +671,7 @@ export const askQuestions = async (
 
   await askPostInstallAction();
 };
+
+export const toChoice = (value: string) => {
+  return { title: value, value };
+};

templates/components/engines/python/agent/__init__.py

@@ -11,11 +11,12 @@ def get_chat_engine():
     top_k = os.getenv("TOP_K", "3")
     tools = []
 
-    # Add query tool
+    # Add query tool if index exists
     index = get_index()
-    query_engine = index.as_query_engine(similarity_top_k=int(top_k))
-    query_engine_tool = QueryEngineTool.from_defaults(query_engine=query_engine)
-    tools.append(query_engine_tool)
+    if index is not None:
+        query_engine = index.as_query_engine(similarity_top_k=int(top_k))
+        query_engine_tool = QueryEngineTool.from_defaults(query_engine=query_engine)
+        tools.append(query_engine_tool)
 
     # Add additional tools
     tools += ToolFactory.from_env()

templates/components/engines/python/agent/tools.py

@@ -1,33 +0,0 @@
-import yaml
-import importlib
-
-from llama_index.core.tools.tool_spec.base import BaseToolSpec
-from llama_index.core.tools.function_tool import FunctionTool
-
-
-class ToolFactory:
-
-    @staticmethod
-    def create_tool(tool_name: str, **kwargs) -> list[FunctionTool]:
-        try:
-            tool_package, tool_cls_name = tool_name.split(".")
-            module_name = f"llama_index.tools.{tool_package}"
-            module = importlib.import_module(module_name)
-            tool_class = getattr(module, tool_cls_name)
-            tool_spec: BaseToolSpec = tool_class(**kwargs)
-            return tool_spec.to_tool_list()
-        except (ImportError, AttributeError) as e:
-            raise ValueError(f"Unsupported tool: {tool_name}") from e
-        except TypeError as e:
-            raise ValueError(
-                f"Could not create tool: {tool_name}. With config: {kwargs}"
-            ) from e
-
-    @staticmethod
-    def from_env() -> list[FunctionTool]:
-        tools = []
-        with open("config/tools.yaml", "r") as f:
-            tool_configs = yaml.safe_load(f)
-            for name, config in tool_configs.items():
-                tools += ToolFactory.create_tool(name, **config)
-        return tools

templates/components/engines/python/agent/tools/__init__.py

@@ -0,0 +1,56 @@
+import os
+import yaml
+import importlib
+
+from llama_index.core.tools.tool_spec.base import BaseToolSpec
+from llama_index.core.tools.function_tool import FunctionTool
+
+
+class ToolType:
+    LLAMAHUB = "llamahub"
+    LOCAL = "local"
+
+
+class ToolFactory:
+
+    TOOL_SOURCE_PACKAGE_MAP = {
+        ToolType.LLAMAHUB: "llama_index.tools",
+        ToolType.LOCAL: "app.engine.tools",
+    }
+
+    @staticmethod
+    def load_tools(tool_type: str, tool_name: str, config: dict) -> list[FunctionTool]:
+        source_package = ToolFactory.TOOL_SOURCE_PACKAGE_MAP[tool_type]
+        try:
+            if "ToolSpec" in tool_name:
+                tool_package, tool_cls_name = tool_name.split(".")
+                module_name = f"{source_package}.{tool_package}"
+                module = importlib.import_module(module_name)
+                tool_class = getattr(module, tool_cls_name)
+                tool_spec: BaseToolSpec = tool_class(**config)
+                return tool_spec.to_tool_list()
+            else:
+                module = importlib.import_module(f"{source_package}.{tool_name}")
+                tools = getattr(module, "tools")
+                if not all(isinstance(tool, FunctionTool) for tool in tools):
+                    raise ValueError(
+                        f"The module {module} does not contain valid tools"
+                    )
+                return tools
+        except ImportError as e:
+            raise ValueError(f"Failed to import tool {tool_name}: {e}")
+        except AttributeError as e:
+            raise ValueError(f"Failed to load tool {tool_name}: {e}")
+
+    @staticmethod
+    def from_env() -> list[FunctionTool]:
+        tools = []
+        if os.path.exists("config/tools.yaml"):
+            with open("config/tools.yaml", "r") as f:
+                tool_configs = yaml.safe_load(f)
+                for tool_type, config_entries in tool_configs.items():
+                    for tool_name, config in config_entries.items():
+                        tools.extend(
+                            ToolFactory.load_tools(tool_type, tool_name, config)
+                        )
+        return tools

templates/components/engines/python/agent/tools/weather.py

@@ -0,0 +1,72 @@
+"""Open Meteo weather map tool spec."""
+
+import logging
+import requests
+import pytz
+from llama_index.core.tools import FunctionTool
+
+logger = logging.getLogger(__name__)
+
+
+class OpenMeteoWeather:
+    geo_api = "https://geocoding-api.open-meteo.com/v1"
+    weather_api = "https://api.open-meteo.com/v1"
+
+    @classmethod
+    def _get_geo_location(cls, location: str) -> dict:
+        """Get geo location from location name."""
+        params = {"name": location, "count": 10, "language": "en", "format": "json"}
+        response = requests.get(f"{cls.geo_api}/search", params=params)
+        if response.status_code != 200:
+            raise Exception(f"Failed to fetch geo location: {response.status_code}")
+        else:
+            data = response.json()
+            result = data["results"][0]
+            geo_location = {
+                "id": result["id"],
+                "name": result["name"],
+                "latitude": result["latitude"],
+                "longitude": result["longitude"],
+            }
+            return geo_location
+
+    @classmethod
+    def get_weather_information(cls, location: str) -> dict:
+        """Use this function to get the weather of any given location.
+        Note that the weather code should follow WMO Weather interpretation codes (WW):
+        0: Clear sky
+        1, 2, 3: Mainly clear, partly cloudy, and overcast
+        45, 48: Fog and depositing rime fog
+        51, 53, 55: Drizzle: Light, moderate, and dense intensity
+        56, 57: Freezing Drizzle: Light and dense intensity
+        61, 63, 65: Rain: Slight, moderate and heavy intensity
+        66, 67: Freezing Rain: Light and heavy intensity
+        71, 73, 75: Snow fall: Slight, moderate, and heavy intensity
+        77: Snow grains
+        80, 81, 82: Rain showers: Slight, moderate, and violent
+        85, 86: Snow showers slight and heavy
+        95: Thunderstorm: Slight or moderate
+        96, 99: Thunderstorm with slight and heavy hail
+        """
+        logger.info(
+            f"Calling open-meteo api to get weather information of location: {location}"
+        )
+        geo_location = cls._get_geo_location(location)
+        timezone = pytz.timezone("UTC").zone
+        params = {
+            "latitude": geo_location["latitude"],
+            "longitude": geo_location["longitude"],
+            "current": "temperature_2m,weather_code",
+            "hourly": "temperature_2m,weather_code",
+            "daily": "weather_code",
+            "timezone": timezone,
+        }
+        response = requests.get(f"{cls.weather_api}/forecast", params=params)
+        if response.status_code != 200:
+            raise Exception(
+                f"Failed to fetch weather information: {response.status_code}"
+            )
+        return response.json()
+
+
+tools = [FunctionTool.from_defaults(OpenMeteoWeather.get_weather_information)]

templates/components/engines/python/chat/__init__.py

@@ -1,12 +1,22 @@
 import os
 from app.engine.index import get_index
+from fastapi import HTTPException
 
 
 def get_chat_engine():
     system_prompt = os.getenv("SYSTEM_PROMPT")
     top_k = os.getenv("TOP_K", 3)
 
-    return get_index().as_chat_engine(
+    index = get_index()
+    if index is None:
+        raise HTTPException(
+            status_code=500,
+            detail=str(
+                "StorageContext is empty - call 'poetry run generate' to generate the storage first"
+            ),
+        )
+
+    return index.as_chat_engine(
         similarity_top_k=int(top_k),
         system_prompt=system_prompt,
         chat_mode="condense_plus_context",

templates/components/engines/typescript/agent/chat.ts

@@ -1,26 +1,45 @@
-import config from "@/config/tools.json";
-import { OpenAI, OpenAIAgent, QueryEngineTool, ToolFactory } from "llamaindex";
-import { STORAGE_CACHE_DIR } from "./constants.mjs";
+import { BaseToolWithCall, OpenAIAgent, QueryEngineTool } from "llamaindex";
+import { ToolsFactory } from "llamaindex/tools/ToolsFactory";
+import fs from "node:fs/promises";
+import path from "node:path";
 import { getDataSource } from "./index";
+import { STORAGE_CACHE_DIR } from "./shared";
+import { createLocalTools } from "./tools";
 
-export async function createChatEngine(llm: OpenAI) {
-  const index = await getDataSource(llm);
-  const queryEngine = index.asQueryEngine();
-  const queryEngineTool = new QueryEngineTool({
-    queryEngine: queryEngine,
-    metadata: {
-      name: "data_query_engine",
-      description: `A query engine for documents in storage folder: ${STORAGE_CACHE_DIR}`,
-    },
+export async function createChatEngine() {
+  const tools: BaseToolWithCall[] = [];
+
+  // Add a query engine tool if we have a data source
+  // Delete this code if you don't have a data source
+  const index = await getDataSource();
+  if (index) {
+    tools.push(
+      new QueryEngineTool({
+        queryEngine: index.asQueryEngine(),
+        metadata: {
+          name: "data_query_engine",
+          description: `A query engine for documents in storage folder: ${STORAGE_CACHE_DIR}`,
+        },
+      }),
+    );
+  }
+
+  try {
+    // add tools from config file if it exists
+    const config = JSON.parse(
+      await fs.readFile(path.join("config", "tools.json"), "utf8"),
+    );
+
+    // add local tools from the 'tools' folder (if configured)
+    const localTools = createLocalTools(config.local);
+    tools.push(...localTools);
+
+    // add tools from LlamaIndexTS (if configured)
+    const llamaTools = await ToolsFactory.createTools(config.llamahub);
+    tools.push(...llamaTools);
+  } catch {}
+
+  return new OpenAIAgent({
+    tools,
   });
-
-  const externalTools = await ToolFactory.createTools(config);
-
-  const agent = new OpenAIAgent({
-    tools: [queryEngineTool, ...externalTools],
-    verbose: true,
-    llm,
-  });
-
-  return agent;
 }

templates/components/engines/typescript/agent/tools/index.ts

@@ -0,0 +1,26 @@
+import { BaseToolWithCall } from "llamaindex";
+import { WeatherTool, WeatherToolParams } from "./weather";
+
+type ToolCreator = (config: unknown) => BaseToolWithCall;
+
+const toolFactory: Record<string, ToolCreator> = {
+  weather: (config: unknown) => {
+    return new WeatherTool(config as WeatherToolParams);
+  },
+};
+
+export function createLocalTools(
+  localConfig: Record<string, unknown>,
+): BaseToolWithCall[] {
+  const tools: BaseToolWithCall[] = [];
+
+  Object.keys(localConfig).forEach((key) => {
+    if (key in toolFactory) {
+      const toolConfig = localConfig[key];
+      const tool = toolFactory[key](toolConfig);
+      tools.push(tool);
+    }
+  });
+
+  return tools;
+}

templates/components/engines/typescript/agent/tools/weather.ts

@@ -0,0 +1,81 @@
+import type { JSONSchemaType } from "ajv";
+import { BaseTool, ToolMetadata } from "llamaindex";
+
+interface GeoLocation {
+  id: string;
+  name: string;
+  latitude: number;
+  longitude: number;
+}
+
+export type WeatherParameter = {
+  location: string;
+};
+
+export type WeatherToolParams = {
+  metadata?: ToolMetadata<JSONSchemaType<WeatherParameter>>;
+};
+
+const DEFAULT_META_DATA: ToolMetadata<JSONSchemaType<WeatherParameter>> = {
+  name: "get_weather_information",
+  description: `
+    Use this function to get the weather of any given location.
+    Note that the weather code should follow WMO Weather interpretation codes (WW):
+    0: Clear sky
+    1, 2, 3: Mainly clear, partly cloudy, and overcast
+    45, 48: Fog and depositing rime fog
+    51, 53, 55: Drizzle: Light, moderate, and dense intensity
+    56, 57: Freezing Drizzle: Light and dense intensity
+    61, 63, 65: Rain: Slight, moderate and heavy intensity
+    66, 67: Freezing Rain: Light and heavy intensity
+    71, 73, 75: Snow fall: Slight, moderate, and heavy intensity
+    77: Snow grains
+    80, 81, 82: Rain showers: Slight, moderate, and violent
+    85, 86: Snow showers slight and heavy
+    95: Thunderstorm: Slight or moderate
+    96, 99: Thunderstorm with slight and heavy hail
+  `,
+  parameters: {
+    type: "object",
+    properties: {
+      location: {
+        type: "string",
+        description: "The location to get the weather information",
+      },
+    },
+    required: ["location"],
+  },
+};
+
+export class WeatherTool implements BaseTool<WeatherParameter> {
+  metadata: ToolMetadata<JSONSchemaType<WeatherParameter>>;
+
+  private getGeoLocation = async (location: string): Promise<GeoLocation> => {
+    const apiUrl = `https://geocoding-api.open-meteo.com/v1/search?name=${location}&count=10&language=en&format=json`;
+    const response = await fetch(apiUrl);
+    const data = await response.json();
+    const { id, name, latitude, longitude } = data.results[0];
+    return { id, name, latitude, longitude };
+  };
+
+  private getWeatherByLocation = async (location: string) => {
+    console.log(
+      "Calling open-meteo api to get weather information of location:",
+      location,
+    );
+    const { latitude, longitude } = await this.getGeoLocation(location);
+    const timezone = Intl.DateTimeFormat().resolvedOptions().timeZone;
+    const apiUrl = `https://api.open-meteo.com/v1/forecast?latitude=${latitude}&longitude=${longitude}&current=temperature_2m,weather_code&hourly=temperature_2m,weather_code&daily=weather_code&timezone=${timezone}`;
+    const response = await fetch(apiUrl);
+    const data = await response.json();
+    return data;
+  };
+
+  constructor(params?: WeatherToolParams) {
+    this.metadata = params?.metadata || DEFAULT_META_DATA;
+  }
+
+  async call(input: WeatherParameter) {
+    return await this.getWeatherByLocation(input.location);
+  }
+}

templates/components/engines/typescript/chat/chat.ts

@@ -1,13 +1,20 @@
-import { ContextChatEngine, LLM } from "llamaindex";
+import { ContextChatEngine, Settings } from "llamaindex";
 import { getDataSource } from "./index";
 
-export async function createChatEngine(llm: LLM) {
-  const index = await getDataSource(llm);
+export async function createChatEngine() {
+  const index = await getDataSource();
+  if (!index) {
+    throw new Error(
+      `StorageContext is empty - call 'npm run generate' to generate the storage first`,
+    );
+  }
   const retriever = index.asRetriever();
-  retriever.similarityTopK = 3;
+  retriever.similarityTopK = process.env.TOP_K
+    ? parseInt(process.env.TOP_K)
+    : 3;
 
   return new ContextChatEngine({
-    chatModel: llm,
+    chatModel: Settings.llm,
     retriever,
   });
 }

templates/components/loaders/python/__init__.py

@@ -5,6 +5,7 @@ import logging
 from typing import Dict
 from app.engine.loaders.file import FileLoaderConfig, get_file_documents
 from app.engine.loaders.web import WebLoaderConfig, get_web_documents
+from app.engine.loaders.db import DBLoaderConfig, get_db_documents
 
 logger = logging.getLogger(__name__)
 
@@ -22,11 +23,17 @@ def get_documents():
         logger.info(
             f"Loading documents from loader: {loader_type}, config: {loader_config}"
         )
-        if loader_type == "file":
-            document = get_file_documents(FileLoaderConfig(**loader_config))
-            documents.extend(document)
-        elif loader_type == "web":
-            document = get_web_documents(WebLoaderConfig(**loader_config))
-            documents.extend(document)
+        match loader_type:
+            case "file":
+                document = get_file_documents(FileLoaderConfig(**loader_config))
+            case "web":
+                document = get_web_documents(WebLoaderConfig(**loader_config))
+            case "db":
+                document = get_db_documents(
+                    configs=[DBLoaderConfig(**cfg) for cfg in loader_config]
+                )
+            case _:
+                raise ValueError(f"Invalid loader type: {loader_type}")
+        documents.extend(document)
 
     return documents

templates/components/loaders/python/db.py

@@ -0,0 +1,26 @@
+import os
+import logging
+from typing import List
+from pydantic import BaseModel, validator
+from llama_index.core.indices.vector_store import VectorStoreIndex
+
+logger = logging.getLogger(__name__)
+
+
+class DBLoaderConfig(BaseModel):
+    uri: str
+    queries: List[str]
+
+
+def get_db_documents(configs: list[DBLoaderConfig]):
+    from llama_index.readers.database import DatabaseReader
+
+    docs = []
+    for entry in configs:
+        loader = DatabaseReader(uri=entry.uri)
+        for query in entry.queries:
+            logger.info(f"Loading data from database with query: {query}")
+            documents = loader.load_data(query=query)
+            docs.extend(documents)
+
+    return documents

templates/components/observability/python/opentelemetry/observability.py

@@ -0,0 +1,5 @@
+from traceloop.sdk import Traceloop
+
+
+def init_observability():
+    Traceloop.init()

templates/components/vectordbs/python/astra/generate.py

@@ -0,0 +1,37 @@
+from dotenv import load_dotenv
+
+load_dotenv()
+
+import os
+import logging
+from llama_index.core.storage import StorageContext
+from llama_index.core.indices import VectorStoreIndex
+from llama_index.vector_stores.astra_db import AstraDBVectorStore
+from app.settings import init_settings
+from app.engine.loaders import get_documents
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger()
+
+
+def generate_datasource():
+    init_settings()
+    logger.info("Creating new index")
+    documents = get_documents()
+    store = AstraDBVectorStore(
+        token=os.environ["ASTRA_DB_APPLICATION_TOKEN"],
+        api_endpoint=os.environ["ASTRA_DB_ENDPOINT"],
+        collection_name=os.environ["ASTRA_DB_COLLECTION"],
+        embedding_dimension=int(os.environ["EMBEDDING_DIM"]),
+    )
+    storage_context = StorageContext.from_defaults(vector_store=store)
+    VectorStoreIndex.from_documents(
+        documents,
+        storage_context=storage_context,
+        show_progress=True,  # this will show you a progress bar as the embeddings are created
+    )
+    logger.info(f"Successfully created embeddings in the AstraDB")
+
+
+if __name__ == "__main__":
+    generate_datasource()

templates/components/vectordbs/python/astra/index.py

@@ -0,0 +1,21 @@
+import logging
+import os
+
+from llama_index.core.indices import VectorStoreIndex
+from llama_index.vector_stores.astra_db import AstraDBVectorStore
+
+
+logger = logging.getLogger("uvicorn")
+
+
+def get_index():
+    logger.info("Connecting to index from AstraDB...")
+    store = AstraDBVectorStore(
+        token=os.environ["ASTRA_DB_APPLICATION_TOKEN"],
+        api_endpoint=os.environ["ASTRA_DB_ENDPOINT"],
+        collection_name=os.environ["ASTRA_DB_COLLECTION"],
+        embedding_dimension=int(os.environ["EMBEDDING_DIM"]),
+    )
+    index = VectorStoreIndex.from_vector_store(store)
+    logger.info("Finished connecting to index from AstraDB.")
+    return index

templates/components/vectordbs/python/milvus/generate.py

@@ -15,6 +15,7 @@ logger = logging.getLogger()
 
 
 def generate_datasource():
+    init_settings()
     logger.info("Creating new index")
     # load the documents and create the index
     documents = get_documents()
@@ -23,7 +24,7 @@ def generate_datasource():
         user=os.getenv("MILVUS_USERNAME"),
         password=os.getenv("MILVUS_PASSWORD"),
         collection_name=os.getenv("MILVUS_COLLECTION"),
-        dim=int(os.getenv("MILVUS_DIMENSION", "1536")),
+        dim=int(os.getenv("EMBEDDING_DIM")),
     )
     storage_context = StorageContext.from_defaults(vector_store=store)
     VectorStoreIndex.from_documents(
@@ -35,5 +36,4 @@ def generate_datasource():
 
 
 if __name__ == "__main__":
-    init_settings()
     generate_datasource()

templates/components/vectordbs/python/milvus/index.py

@@ -15,7 +15,7 @@ def get_index():
         user=os.getenv("MILVUS_USERNAME"),
         password=os.getenv("MILVUS_PASSWORD"),
         collection_name=os.getenv("MILVUS_COLLECTION"),
-        dim=int(os.getenv("EMBEDDING_DIM", "1536")),
+        dim=int(os.getenv("EMBEDDING_DIM")),
     )
     index = VectorStoreIndex.from_vector_store(store)
     logger.info("Finished connecting to index from Milvus.")

templates/components/vectordbs/python/mongo/generate.py

@@ -15,6 +15,7 @@ logger = logging.getLogger()
 
 
 def generate_datasource():
+    init_settings()
     logger.info("Creating new index")
     # load the documents and create the index
     documents = get_documents()
@@ -39,5 +40,4 @@ See https://github.com/run-llama/mongodb-demo/tree/main?tab=readme-ov-file#creat
 
 
 if __name__ == "__main__":
-    init_settings()
     generate_datasource()

templates/components/vectordbs/python/none/generate.py

@@ -16,6 +16,7 @@ logger = logging.getLogger()
 
 
 def generate_datasource():
+    init_settings()
     logger.info("Creating new index")
     # load the documents and create the index
     documents = get_documents()
@@ -28,5 +29,4 @@ def generate_datasource():
 
 
 if __name__ == "__main__":
-    init_settings()
     generate_datasource()

templates/components/vectordbs/python/none/index.py

@@ -11,10 +11,7 @@ logger = logging.getLogger("uvicorn")
 def get_index():
     # check if storage already exists
     if not os.path.exists(STORAGE_DIR):
-        raise Exception(
-            "StorageContext is empty - call 'python app/engine/generate.py' to generate the storage first"
-        )
-
+        return None
     # load the existing index
     logger.info(f"Loading index from {STORAGE_DIR}...")
     storage_context = StorageContext.from_defaults(persist_dir=STORAGE_DIR)

templates/components/vectordbs/python/pg/generate.py

@@ -15,6 +15,7 @@ logger = logging.getLogger()
 
 
 def generate_datasource():
+    init_settings()
     logger.info("Creating new index")
     # load the documents and create the index
     documents = get_documents()
@@ -31,5 +32,4 @@ def generate_datasource():
 
 
 if __name__ == "__main__":
-    init_settings()
     generate_datasource()

templates/components/vectordbs/python/pinecone/generate.py

@@ -15,6 +15,7 @@ logger = logging.getLogger()
 
 
 def generate_datasource():
+    init_settings()
     logger.info("Creating new index")
     # load the documents and create the index
     documents = get_documents()
@@ -35,5 +36,4 @@ def generate_datasource():
 
 
 if __name__ == "__main__":
-    init_settings()
     generate_datasource()

templates/components/vectordbs/python/qdrant/generate.py

@@ -0,0 +1,37 @@
+import logging
+import os
+from app.engine.loaders import get_documents
+from app.settings import init_settings
+from dotenv import load_dotenv
+from llama_index.core.indices import VectorStoreIndex
+from llama_index.core.storage import StorageContext
+from llama_index.vector_stores.qdrant import QdrantVectorStore
+load_dotenv()
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger()
+
+
+def generate_datasource():
+    init_settings()
+    logger.info("Creating new index with Qdrant")
+    # load the documents and create the index
+    documents = get_documents()
+    store = QdrantVectorStore(
+        collection_name=os.getenv("QDRANT_COLLECTION"),
+        url=os.getenv("QDRANT_URL"),
+        api_key=os.getenv("QDRANT_API_KEY"),
+    )
+    storage_context = StorageContext.from_defaults(vector_store=store)
+    VectorStoreIndex.from_documents(
+        documents,
+        storage_context=storage_context,
+        show_progress=True,  # this will show you a progress bar as the embeddings are created
+    )
+    logger.info(
+        f"Successfully uploaded documents to the {os.getenv('QDRANT_COLLECTION')} collection."
+    )
+
+
+if __name__ == "__main__":
+    generate_datasource()

templates/components/vectordbs/python/qdrant/index.py

@@ -0,0 +1,20 @@
+import logging
+import os
+
+from llama_index.core.indices import VectorStoreIndex
+from llama_index.vector_stores.qdrant import QdrantVectorStore
+
+
+logger = logging.getLogger("uvicorn")
+
+
+def get_index():
+    logger.info("Connecting to Qdrant collection..")
+    store = QdrantVectorStore(
+        collection_name=os.getenv("QDRANT_COLLECTION"),
+        url=os.getenv("QDRANT_URL"),
+        api_key=os.getenv("QDRANT_API_KEY"),
+    )
+    index = VectorStoreIndex.from_vector_store(store)
+    logger.info("Finished connecting to Qdrant collection.")
+    return index

templates/components/vectordbs/typescript/astra/generate.ts

@@ -0,0 +1,40 @@
+/* eslint-disable turbo/no-undeclared-env-vars */
+import * as dotenv from "dotenv";
+import { VectorStoreIndex, storageContextFromDefaults } from "llamaindex";
+import { AstraDBVectorStore } from "llamaindex/storage/vectorStore/AstraDBVectorStore";
+import { getDocuments } from "./loader";
+import { initSettings } from "./settings";
+import { checkRequiredEnvVars } from "./shared";
+
+dotenv.config();
+
+async function loadAndIndex() {
+  // load objects from storage and convert them into LlamaIndex Document objects
+  const documents = await getDocuments();
+
+  // create vector store and a collection
+  const collectionName = process.env.ASTRA_DB_COLLECTION!;
+  const vectorStore = new AstraDBVectorStore();
+  await vectorStore.create(collectionName, {
+    vector: {
+      dimension: parseInt(process.env.EMBEDDING_DIM!),
+      metric: "cosine",
+    },
+  });
+  await vectorStore.connect(collectionName);
+
+  // create index from documents and store them in Astra
+  console.log("Start creating embeddings...");
+  const storageContext = await storageContextFromDefaults({ vectorStore });
+  await VectorStoreIndex.fromDocuments(documents, { storageContext });
+  console.log(
+    "Successfully created embeddings and save to your Astra database.",
+  );
+}
+
+(async () => {
+  checkRequiredEnvVars();
+  initSettings();
+  await loadAndIndex();
+  console.log("Finished generating storage.");
+})();

templates/components/vectordbs/typescript/astra/index.ts

@@ -0,0 +1,11 @@
+/* eslint-disable turbo/no-undeclared-env-vars */
+import { VectorStoreIndex } from "llamaindex";
+import { AstraDBVectorStore } from "llamaindex/storage/vectorStore/AstraDBVectorStore";
+import { checkRequiredEnvVars } from "./shared";
+
+export async function getDataSource() {
+  checkRequiredEnvVars();
+  const store = new AstraDBVectorStore();
+  await store.connect(process.env.ASTRA_DB_COLLECTION!);
+  return await VectorStoreIndex.fromVectorStore(store);
+}

templates/components/vectordbs/typescript/astra/shared.ts

@@ -0,0 +1,23 @@
+const REQUIRED_ENV_VARS = [
+  "ASTRA_DB_APPLICATION_TOKEN",
+  "ASTRA_DB_ENDPOINT",
+  "ASTRA_DB_COLLECTION",
+  "EMBEDDING_DIM",
+];
+
+export function checkRequiredEnvVars() {
+  const missingEnvVars = REQUIRED_ENV_VARS.filter((envVar) => {
+    return !process.env[envVar];
+  });
+
+  if (missingEnvVars.length > 0) {
+    console.log(
+      `The following environment variables are required but missing: ${missingEnvVars.join(
+        ", ",
+      )}`,
+    );
+    throw new Error(
+      `Missing environment variables: ${missingEnvVars.join(", ")}`,
+    );
+  }
+}

templates/components/vectordbs/typescript/milvus/generate.ts

@@ -1,12 +1,10 @@
 /* eslint-disable turbo/no-undeclared-env-vars */
 import * as dotenv from "dotenv";
-import {
-  MilvusVectorStore,
-  VectorStoreIndex,
-  storageContextFromDefaults,
-} from "llamaindex";
-import { getDocuments } from "./loader.mjs";
-import { checkRequiredEnvVars, getMilvusClient } from "./shared.mjs";
+import { VectorStoreIndex, storageContextFromDefaults } from "llamaindex";
+import { MilvusVectorStore } from "llamaindex/storage/vectorStore/MilvusVectorStore";
+import { getDocuments } from "./loader";
+import { initSettings } from "./settings";
+import { checkRequiredEnvVars, getMilvusClient } from "./shared";
 
 dotenv.config();
 
@@ -32,6 +30,7 @@ async function loadAndIndex() {
 
 (async () => {
   checkRequiredEnvVars();
+  initSettings();
   await loadAndIndex();
   console.log("Finished generating storage.");
 })();

templates/components/vectordbs/typescript/milvus/index.ts

@@ -1,35 +1,11 @@
-import {
-  ContextChatEngine,
-  LLM,
-  MilvusVectorStore,
-  serviceContextFromDefaults,
-  VectorStoreIndex,
-} from "llamaindex";
-import {
-  checkRequiredEnvVars,
-  CHUNK_OVERLAP,
-  CHUNK_SIZE,
-  getMilvusClient,
-} from "./shared.mjs";
+import { VectorStoreIndex } from "llamaindex";
+import { MilvusVectorStore } from "llamaindex/storage/vectorStore/MilvusVectorStore";
+import { checkRequiredEnvVars, getMilvusClient } from "./shared";
 
-async function getDataSource(llm: LLM) {
+export async function getDataSource() {
   checkRequiredEnvVars();
-  const serviceContext = serviceContextFromDefaults({
-    llm,
-    chunkSize: CHUNK_SIZE,
-    chunkOverlap: CHUNK_OVERLAP,
-  });
   const milvusClient = getMilvusClient();
   const store = new MilvusVectorStore({ milvusClient });
 
-  return await VectorStoreIndex.fromVectorStore(store, serviceContext);
-}
-
-export async function createChatEngine(llm: LLM) {
-  const index = await getDataSource(llm);
-  const retriever = index.asRetriever({ similarityTopK: 3 });
-  return new ContextChatEngine({
-    chatModel: llm,
-    retriever,
-  });
+  return await VectorStoreIndex.fromVectorStore(store);
 }

templates/components/vectordbs/typescript/milvus/shared.ts

@@ -1,8 +1,5 @@
 import { MilvusClient } from "@zilliz/milvus2-sdk-node";
 
-export const CHUNK_SIZE = 512;
-export const CHUNK_OVERLAP = 20;
-
 const REQUIRED_ENV_VARS = [
   "MILVUS_ADDRESS",
   "MILVUS_USERNAME",
@@ -16,7 +13,7 @@ export function getMilvusClient() {
     throw new Error("MILVUS_ADDRESS environment variable is required");
   }
   return new MilvusClient({
-    address: process.env.MILVUS_ADDRESS,
+    address: process.env.MILVUS_ADDRESS!,
     username: process.env.MILVUS_USERNAME,
     password: process.env.MILVUS_PASSWORD,
   });

templates/components/vectordbs/typescript/mongo/generate.ts

@@ -1,19 +1,17 @@
 /* eslint-disable turbo/no-undeclared-env-vars */
 import * as dotenv from "dotenv";
-import {
-  MongoDBAtlasVectorSearch,
-  VectorStoreIndex,
-  storageContextFromDefaults,
-} from "llamaindex";
+import { VectorStoreIndex, storageContextFromDefaults } from "llamaindex";
+import { MongoDBAtlasVectorSearch } from "llamaindex/storage/vectorStore/MongoDBAtlasVectorSearch";
 import { MongoClient } from "mongodb";
-import { getDocuments } from "./loader.mjs";
-import { checkRequiredEnvVars } from "./shared.mjs";
+import { getDocuments } from "./loader";
+import { initSettings } from "./settings";
+import { checkRequiredEnvVars } from "./shared";
 
 dotenv.config();
 
-const mongoUri = process.env.MONGO_URI;
-const databaseName = process.env.MONGODB_DATABASE;
-const vectorCollectionName = process.env.MONGODB_VECTORS;
+const mongoUri = process.env.MONGO_URI!;
+const databaseName = process.env.MONGODB_DATABASE!;
+const vectorCollectionName = process.env.MONGODB_VECTORS!;
 const indexName = process.env.MONGODB_VECTOR_INDEX;
 
 async function loadAndIndex() {
@@ -42,6 +40,7 @@ async function loadAndIndex() {
 
 (async () => {
   checkRequiredEnvVars();
+  initSettings();
   await loadAndIndex();
   console.log("Finished generating storage.");
 })();

templates/components/vectordbs/typescript/mongo/index.ts

@@ -1,37 +1,18 @@
 /* eslint-disable turbo/no-undeclared-env-vars */
-import {
-  ContextChatEngine,
-  LLM,
-  MongoDBAtlasVectorSearch,
-  serviceContextFromDefaults,
-  VectorStoreIndex,
-} from "llamaindex";
+import { VectorStoreIndex } from "llamaindex";
+import { MongoDBAtlasVectorSearch } from "llamaindex/storage/vectorStore/MongoDBAtlasVectorSearch";
 import { MongoClient } from "mongodb";
-import { checkRequiredEnvVars, CHUNK_OVERLAP, CHUNK_SIZE } from "./shared.mjs";
+import { checkRequiredEnvVars } from "./shared";
 
-async function getDataSource(llm: LLM) {
+export async function getDataSource() {
   checkRequiredEnvVars();
   const client = new MongoClient(process.env.MONGO_URI!);
-  const serviceContext = serviceContextFromDefaults({
-    llm,
-    chunkSize: CHUNK_SIZE,
-    chunkOverlap: CHUNK_OVERLAP,
-  });
   const store = new MongoDBAtlasVectorSearch({
     mongodbClient: client,
-    dbName: process.env.MONGODB_DATABASE,
-    collectionName: process.env.MONGODB_VECTORS,
+    dbName: process.env.MONGODB_DATABASE!,
+    collectionName: process.env.MONGODB_VECTORS!,
     indexName: process.env.MONGODB_VECTOR_INDEX,
   });
 
-  return await VectorStoreIndex.fromVectorStore(store, serviceContext);
-}
-
-export async function createChatEngine(llm: LLM) {
-  const index = await getDataSource(llm);
-  const retriever = index.asRetriever({ similarityTopK: 3 });
-  return new ContextChatEngine({
-    chatModel: llm,
-    retriever,
-  });
+  return await VectorStoreIndex.fromVectorStore(store);
 }

templates/components/vectordbs/typescript/mongo/shared.ts

@@ -1,6 +1,3 @@
-export const CHUNK_SIZE = 512;
-export const CHUNK_OVERLAP = 20;
-
 const REQUIRED_ENV_VARS = [
   "MONGO_URI",
   "MONGODB_DATABASE",

templates/components/vectordbs/typescript/none/constants.mjs

@@ -1,3 +0,0 @@
-export const STORAGE_CACHE_DIR = "./cache";
-export const CHUNK_SIZE = 512;
-export const CHUNK_OVERLAP = 20;

templates/components/vectordbs/typescript/none/generate.ts

@@ -1,25 +1,23 @@
-import {
-  serviceContextFromDefaults,
-  storageContextFromDefaults,
-  VectorStoreIndex,
-} from "llamaindex";
+import { VectorStoreIndex } from "llamaindex";
+import { storageContextFromDefaults } from "llamaindex/storage/StorageContext";
 
 import * as dotenv from "dotenv";
 
-import { CHUNK_OVERLAP, CHUNK_SIZE, STORAGE_CACHE_DIR } from "./constants.mjs";
-import { getDocuments } from "./loader.mjs";
+import { getDocuments } from "./loader";
+import { initSettings } from "./settings";
+import { STORAGE_CACHE_DIR } from "./shared";
 
 // Load environment variables from local .env file
 dotenv.config();
 
-async function getRuntime(func) {
+async function getRuntime(func: any) {
   const start = Date.now();
   await func();
   const end = Date.now();
   return end - start;
 }
 
-async function generateDatasource(serviceContext) {
+async function generateDatasource() {
   console.log(`Generating storage context...`);
   // Split documents, create embeddings and store them in the storage context
   const ms = await getRuntime(async () => {
@@ -29,18 +27,13 @@ async function generateDatasource(serviceContext) {
     const documents = await getDocuments();
     await VectorStoreIndex.fromDocuments(documents, {
       storageContext,
-      serviceContext,
     });
   });
   console.log(`Storage context successfully generated in ${ms / 1000}s.`);
 }
 
 (async () => {
-  const serviceContext = serviceContextFromDefaults({
-    chunkSize: CHUNK_SIZE,
-    chunkOverlap: CHUNK_OVERLAP,
-  });
-
-  await generateDatasource(serviceContext);
+  initSettings();
+  await generateDatasource();
   console.log("Finished generating storage.");
 })();

templates/components/vectordbs/typescript/none/index.ts

@@ -1,18 +1,8 @@
-import {
-  LLM,
-  serviceContextFromDefaults,
-  SimpleDocumentStore,
-  storageContextFromDefaults,
-  VectorStoreIndex,
-} from "llamaindex";
-import { CHUNK_OVERLAP, CHUNK_SIZE, STORAGE_CACHE_DIR } from "./constants.mjs";
+import { SimpleDocumentStore, VectorStoreIndex } from "llamaindex";
+import { storageContextFromDefaults } from "llamaindex/storage/StorageContext";
+import { STORAGE_CACHE_DIR } from "./shared";
 
-export async function getDataSource(llm: LLM) {
-  const serviceContext = serviceContextFromDefaults({
-    llm,
-    chunkSize: CHUNK_SIZE,
-    chunkOverlap: CHUNK_OVERLAP,
-  });
+export async function getDataSource() {
   const storageContext = await storageContextFromDefaults({
     persistDir: `${STORAGE_CACHE_DIR}`,
   });
@@ -21,12 +11,9 @@ export async function getDataSource(llm: LLM) {
     (storageContext.docStore as SimpleDocumentStore).toDict(),
   ).length;
   if (numberOfDocs === 0) {
-    throw new Error(
-      `StorageContext is empty - call 'npm run generate' to generate the storage first`,
-    );
+    return null;
   }
   return await VectorStoreIndex.init({
     storageContext,
-    serviceContext,
   });
 }

templates/components/vectordbs/typescript/none/shared.ts

@@ -0,0 +1 @@
+export const STORAGE_CACHE_DIR = "./cache";

templates/components/vectordbs/typescript/pg/generate.ts

@@ -1,17 +1,15 @@
 /* eslint-disable turbo/no-undeclared-env-vars */
 import * as dotenv from "dotenv";
-import {
-  PGVectorStore,
-  VectorStoreIndex,
-  storageContextFromDefaults,
-} from "llamaindex";
-import { getDocuments } from "./loader.mjs";
+import { VectorStoreIndex, storageContextFromDefaults } from "llamaindex";
+import { PGVectorStore } from "llamaindex/storage/vectorStore/PGVectorStore";
+import { getDocuments } from "./loader";
+import { initSettings } from "./settings";
 import {
   PGVECTOR_COLLECTION,
   PGVECTOR_SCHEMA,
   PGVECTOR_TABLE,
   checkRequiredEnvVars,
-} from "./shared.mjs";
+} from "./shared";
 
 dotenv.config();
 
@@ -37,6 +35,7 @@ async function loadAndIndex() {
 
 (async () => {
   checkRequiredEnvVars();
+  initSettings();
   await loadAndIndex();
   console.log("Finished generating storage.");
   process.exit(0);

templates/components/vectordbs/typescript/pg/index.ts

@@ -1,39 +1,18 @@
 /* eslint-disable turbo/no-undeclared-env-vars */
+import { VectorStoreIndex } from "llamaindex";
+import { PGVectorStore } from "llamaindex/storage/vectorStore/PGVectorStore";
 import {
-  ContextChatEngine,
-  LLM,
-  PGVectorStore,
-  VectorStoreIndex,
-  serviceContextFromDefaults,
-} from "llamaindex";
-import {
-  CHUNK_OVERLAP,
-  CHUNK_SIZE,
   PGVECTOR_SCHEMA,
   PGVECTOR_TABLE,
   checkRequiredEnvVars,
-} from "./shared.mjs";
+} from "./shared";
 
-async function getDataSource(llm: LLM) {
+export async function getDataSource() {
   checkRequiredEnvVars();
   const pgvs = new PGVectorStore({
     connectionString: process.env.PG_CONNECTION_STRING,
     schemaName: PGVECTOR_SCHEMA,
     tableName: PGVECTOR_TABLE,
   });
-  const serviceContext = serviceContextFromDefaults({
-    llm,
-    chunkSize: CHUNK_SIZE,
-    chunkOverlap: CHUNK_OVERLAP,
-  });
-  return await VectorStoreIndex.fromVectorStore(pgvs, serviceContext);
-}
-
-export async function createChatEngine(llm: LLM) {
-  const index = await getDataSource(llm);
-  const retriever = index.asRetriever({ similarityTopK: 3 });
-  return new ContextChatEngine({
-    chatModel: llm,
-    retriever,
-  });
+  return await VectorStoreIndex.fromVectorStore(pgvs);
 }

templates/components/vectordbs/typescript/pg/shared.ts

@@ -1,10 +1,8 @@
 export const PGVECTOR_COLLECTION = "data";
-export const CHUNK_SIZE = 512;
-export const CHUNK_OVERLAP = 20;
 export const PGVECTOR_SCHEMA = "public";
 export const PGVECTOR_TABLE = "llamaindex_embedding";
 
-const REQUIRED_ENV_VARS = ["PG_CONNECTION_STRING", "OPENAI_API_KEY"];
+const REQUIRED_ENV_VARS = ["PG_CONNECTION_STRING"];
 
 export function checkRequiredEnvVars() {
   const missingEnvVars = REQUIRED_ENV_VARS.filter((envVar) => {

templates/components/vectordbs/typescript/pinecone/generate.ts

@@ -1,12 +1,10 @@
 /* eslint-disable turbo/no-undeclared-env-vars */
 import * as dotenv from "dotenv";
-import {
-  PineconeVectorStore,
-  VectorStoreIndex,
-  storageContextFromDefaults,
-} from "llamaindex";
-import { getDocuments } from "./loader.mjs";
-import { checkRequiredEnvVars } from "./shared.mjs";
+import { VectorStoreIndex, storageContextFromDefaults } from "llamaindex";
+import { PineconeVectorStore } from "llamaindex/storage/vectorStore/PineconeVectorStore";
+import { getDocuments } from "./loader";
+import { initSettings } from "./settings";
+import { checkRequiredEnvVars } from "./shared";
 
 dotenv.config();
 
@@ -28,6 +26,7 @@ async function loadAndIndex() {
 
 (async () => {
   checkRequiredEnvVars();
+  initSettings();
   await loadAndIndex();
   console.log("Finished generating storage.");
 })();

templates/components/vectordbs/typescript/pinecone/index.ts

@@ -1,29 +1,10 @@
 /* eslint-disable turbo/no-undeclared-env-vars */
-import {
-  ContextChatEngine,
-  LLM,
-  PineconeVectorStore,
-  VectorStoreIndex,
-  serviceContextFromDefaults,
-} from "llamaindex";
-import { CHUNK_OVERLAP, CHUNK_SIZE, checkRequiredEnvVars } from "./shared.mjs";
+import { VectorStoreIndex } from "llamaindex";
+import { PineconeVectorStore } from "llamaindex/storage/vectorStore/PineconeVectorStore";
+import { checkRequiredEnvVars } from "./shared";
 
-async function getDataSource(llm: LLM) {
+export async function getDataSource() {
   checkRequiredEnvVars();
-  const serviceContext = serviceContextFromDefaults({
-    llm,
-    chunkSize: CHUNK_SIZE,
-    chunkOverlap: CHUNK_OVERLAP,
-  });
   const store = new PineconeVectorStore();
-  return await VectorStoreIndex.fromVectorStore(store, serviceContext);
-}
-
-export async function createChatEngine(llm: LLM) {
-  const index = await getDataSource(llm);
-  const retriever = index.asRetriever({ similarityTopK: 5 });
-  return new ContextChatEngine({
-    chatModel: llm,
-    retriever,
-  });
+  return await VectorStoreIndex.fromVectorStore(store);
 }

templates/components/vectordbs/typescript/pinecone/shared.ts

@@ -1,6 +1,3 @@
-export const CHUNK_SIZE = 512;
-export const CHUNK_OVERLAP = 20;
-
 const REQUIRED_ENV_VARS = ["PINECONE_ENVIRONMENT", "PINECONE_API_KEY"];
 
 export function checkRequiredEnvVars() {

templates/components/vectordbs/typescript/qdrant/generate.ts

@@ -0,0 +1,37 @@
+/* eslint-disable turbo/no-undeclared-env-vars */
+import * as dotenv from "dotenv";
+import { VectorStoreIndex, storageContextFromDefaults } from "llamaindex";
+import { QdrantVectorStore } from "llamaindex/storage/vectorStore/QdrantVectorStore";
+import { getDocuments } from "./loader";
+import { initSettings } from "./settings";
+import { checkRequiredEnvVars, getQdrantClient } from "./shared";
+
+dotenv.config();
+
+const collectionName = process.env.QDRANT_COLLECTION;
+
+async function loadAndIndex() {
+  // load objects from storage and convert them into LlamaIndex Document objects
+  const documents = await getDocuments();
+
+  // Connect to Qdrant
+  const vectorStore = new QdrantVectorStore({
+    collectionName,
+    client: getQdrantClient(),
+  });
+
+  const storageContext = await storageContextFromDefaults({ vectorStore });
+  await VectorStoreIndex.fromDocuments(documents, {
+    storageContext: storageContext,
+  });
+  console.log(
+    `Successfully upload embeddings to Qdrant collection ${collectionName}.`,
+  );
+}
+
+(async () => {
+  checkRequiredEnvVars();
+  initSettings();
+  await loadAndIndex();
+  console.log("Finished generating storage.");
+})();

templates/components/vectordbs/typescript/qdrant/index.ts

@@ -0,0 +1,17 @@
+import * as dotenv from "dotenv";
+import { VectorStoreIndex } from "llamaindex";
+import { QdrantVectorStore } from "llamaindex/storage/vectorStore/QdrantVectorStore";
+import { checkRequiredEnvVars, getQdrantClient } from "./shared";
+
+dotenv.config();
+
+export async function getDataSource() {
+  checkRequiredEnvVars();
+  const collectionName = process.env.QDRANT_COLLECTION;
+  const store = new QdrantVectorStore({
+    collectionName,
+    client: getQdrantClient(),
+  });
+
+  return await VectorStoreIndex.fromVectorStore(store);
+}

templates/components/vectordbs/typescript/qdrant/shared.ts

@@ -0,0 +1,32 @@
+import { QdrantClient } from "@qdrant/js-client-rest";
+
+const REQUIRED_ENV_VARS = ["QDRANT_URL", "QDRANT_COLLECTION"]; // QDRANT_API_KEY is optional
+
+export function getQdrantClient() {
+  const url = process.env.QDRANT_URL;
+  if (!url) {
+    throw new Error("QDRANT_URL environment variable is required");
+  }
+  const apiKey = process.env?.QDRANT_API_KEY;
+  return new QdrantClient({
+    url,
+    apiKey,
+  });
+}
+
+export function checkRequiredEnvVars() {
+  const missingEnvVars = REQUIRED_ENV_VARS.filter((envVar) => {
+    return !process.env[envVar];
+  });
+
+  if (missingEnvVars.length > 0) {
+    console.log(
+      `The following environment variables are required but missing: ${missingEnvVars.join(
+        ", ",
+      )}`,
+    );
+    throw new Error(
+      `Missing environment variables: ${missingEnvVars.join(", ")}`,
+    );
+  }
+}

templates/types/streaming/express/npmrc

@@ -0,0 +1 @@
+node-linker=hoisted

templates/types/streaming/express/package.json

@@ -5,16 +5,18 @@
   "scripts": {
     "format": "prettier --ignore-unknown --cache --check .",
     "format:write": "prettier --ignore-unknown --write .",
-    "build": "tsup index.ts --format esm --dts",
-    "start": "node dist/index.mjs",
-    "dev": "concurrently \"tsup index.ts --format esm --dts --watch\" \"nodemon -q dist/index.mjs\""
+    "build": "tsup index.ts --format cjs --dts",
+    "start": "node dist/index.js",
+    "dev": "concurrently \"tsup index.ts --format cjs --dts --watch\" \"nodemon -q dist/index.mjs\""
   },
   "dependencies": {
-    "ai": "^2.2.25",
+    "ai": "^3.0.21",
     "cors": "^2.8.5",
     "dotenv": "^16.3.1",
     "express": "^4.18.2",
-    "llamaindex": "latest"
+    "llamaindex": "0.3.9",
+    "pdf2json": "3.0.5",
+    "ajv": "^8.12.0"
   },
   "devDependencies": {
     "@types/cors": "^2.8.16",
@@ -22,11 +24,12 @@
     "@types/node": "^20.9.5",
     "concurrently": "^8.2.2",
     "eslint": "^8.54.0",
+    "eslint-config-prettier": "^8.10.0",
     "nodemon": "^3.0.1",
-    "tsup": "^8.0.1",
-    "typescript": "^5.3.2",
     "prettier": "^3.2.5",
     "prettier-plugin-organize-imports": "^3.2.4",
-    "eslint-config-prettier": "^8.10.0"
+    "tsx": "^4.7.2",
+    "tsup": "^8.0.1",
+    "typescript": "^5.3.2"
   }
 }

templates/types/streaming/express/src/controllers/chat-request.controller.ts

@@ -1,5 +1,5 @@
 import { Request, Response } from "express";
-import { ChatMessage, MessageContent, OpenAI } from "llamaindex";
+import { ChatMessage, MessageContent } from "llamaindex";
 import { createChatEngine } from "./engine/chat";
 
 const convertMessageContent = (
@@ -32,10 +32,6 @@ export const chatRequest = async (req: Request, res: Response) => {
       });
     }
 
-    const llm = new OpenAI({
-      model: process.env.MODEL || "gpt-3.5-turbo",
-    });
-
     // Convert message content from Vercel/AI format to LlamaIndex/OpenAI format
     // Note: The non-streaming template does not need the Vercel/AI format, we're still using it for consistency with the streaming template
     const userMessageContent = convertMessageContent(
@@ -43,7 +39,7 @@ export const chatRequest = async (req: Request, res: Response) => {
       data?.imageUrl,
     );
 
-    const chatEngine = await createChatEngine(llm);
+    const chatEngine = await createChatEngine();
 
     // Calling LlamaIndex's ChatEngine to get a response
     const response = await chatEngine.chat({
@@ -61,7 +57,7 @@ export const chatRequest = async (req: Request, res: Response) => {
   } catch (error) {
     console.error("[LlamaIndex]", error);
     return res.status(500).json({
-      error: (error as Error).message,
+      detail: (error as Error).message,
     });
   }
 };

templates/types/streaming/express/src/controllers/chat.controller.ts

@@ -1,8 +1,9 @@
-import { streamToResponse } from "ai";
+import { Message, StreamData, streamToResponse } from "ai";
 import { Request, Response } from "express";
-import { ChatMessage, MessageContent, OpenAI } from "llamaindex";
+import { ChatMessage, MessageContent, Settings } from "llamaindex";
 import { createChatEngine } from "./engine/chat";
 import { LlamaIndexStream } from "./llamaindex-stream";
+import { createCallbackManager } from "./stream-helper";
 
 const convertMessageContent = (
   textMessage: string,
@@ -25,7 +26,7 @@ const convertMessageContent = (
 
 export const chat = async (req: Request, res: Response) => {
   try {
-    const { messages, data }: { messages: ChatMessage[]; data: any } = req.body;
+    const { messages, data }: { messages: Message[]; data: any } = req.body;
     const userMessage = messages.pop();
     if (!messages || !userMessage || userMessage.role !== "user") {
       return res.status(400).json({
@@ -34,11 +35,7 @@ export const chat = async (req: Request, res: Response) => {
       });
     }
 
-    const llm = new OpenAI({
-      model: (process.env.MODEL as any) || "gpt-3.5-turbo",
-    });
-
-    const chatEngine = await createChatEngine(llm);
+    const chatEngine = await createChatEngine();
 
     // Convert message content from Vercel/AI format to LlamaIndex/OpenAI format
     const userMessageContent = convertMessageContent(
@@ -46,36 +43,34 @@ export const chat = async (req: Request, res: Response) => {
       data?.imageUrl,
     );
 
+    // Init Vercel AI StreamData
+    const vercelStreamData = new StreamData();
+
+    // Setup callbacks
+    const callbackManager = createCallbackManager(vercelStreamData);
+
     // Calling LlamaIndex's ChatEngine to get a streamed response
-    const response = await chatEngine.chat({
-      message: userMessageContent,
-      chatHistory: messages,
-      stream: true,
+    const response = await Settings.withCallbackManager(callbackManager, () => {
+      return chatEngine.chat({
+        message: userMessageContent,
+        chatHistory: messages as ChatMessage[],
+        stream: true,
+      });
     });
 
     // Return a stream, which can be consumed by the Vercel/AI client
-    const { stream, data: streamData } = LlamaIndexStream(response, {
+    const stream = LlamaIndexStream(response, vercelStreamData, {
       parserOptions: {
         image_url: data?.imageUrl,
       },
     });
+    const processedStream = stream.pipeThrough(vercelStreamData.stream);
 
-    // Pipe LlamaIndexStream to response
-    const processedStream = stream.pipeThrough(streamData.stream);
-    return streamToResponse(processedStream, res, {
-      headers: {
-        // response MUST have the `X-Experimental-Stream-Data: 'true'` header
-        // so that the client uses the correct parsing logic, see
-        // https://sdk.vercel.ai/docs/api-reference/stream-data#on-the-server
-        "X-Experimental-Stream-Data": "true",
-        "Content-Type": "text/plain; charset=utf-8",
-        "Access-Control-Expose-Headers": "X-Experimental-Stream-Data",
-      },
-    });
+    return streamToResponse(processedStream, res);
   } catch (error) {
     console.error("[LlamaIndex]", error);
     return res.status(500).json({
-      error: (error as Error).message,
+      detail: (error as Error).message,
     });
   }
 };

templates/types/streaming/express/src/controllers/engine/chat.ts

@@ -1,7 +1,7 @@
-import { LLM, SimpleChatEngine } from "llamaindex";
+import { Settings, SimpleChatEngine } from "llamaindex";
 
-export async function createChatEngine(llm: LLM) {
+export async function createChatEngine() {
   return new SimpleChatEngine({
-    llm,
+    llm: Settings.llm,
   });
 }

templates/types/streaming/express/src/controllers/engine/settings.ts

@@ -0,0 +1,87 @@
+import {
+  Anthropic,
+  GEMINI_EMBEDDING_MODEL,
+  GEMINI_MODEL,
+  Gemini,
+  GeminiEmbedding,
+  OpenAI,
+  OpenAIEmbedding,
+  Settings,
+} from "llamaindex";
+import { HuggingFaceEmbedding } from "llamaindex/embeddings/HuggingFaceEmbedding";
+import { OllamaEmbedding } from "llamaindex/embeddings/OllamaEmbedding";
+import { ALL_AVAILABLE_ANTHROPIC_MODELS } from "llamaindex/llm/anthropic";
+import { Ollama } from "llamaindex/llm/ollama";
+
+const CHUNK_SIZE = 512;
+const CHUNK_OVERLAP = 20;
+
+export const initSettings = async () => {
+  // HINT: you can delete the initialization code for unused model providers
+  console.log(`Using '${process.env.MODEL_PROVIDER}' model provider`);
+
+  if (!process.env.MODEL || !process.env.EMBEDDING_MODEL) {
+    throw new Error("'MODEL' and 'EMBEDDING_MODEL' env variables must be set.");
+  }
+
+  switch (process.env.MODEL_PROVIDER) {
+    case "ollama":
+      initOllama();
+      break;
+    case "anthropic":
+      initAnthropic();
+      break;
+    case "gemini":
+      initGemini();
+      break;
+    default:
+      initOpenAI();
+      break;
+  }
+  Settings.chunkSize = CHUNK_SIZE;
+  Settings.chunkOverlap = CHUNK_OVERLAP;
+};
+
+function initOpenAI() {
+  Settings.llm = new OpenAI({
+    model: process.env.MODEL ?? "gpt-3.5-turbo",
+    maxTokens: 512,
+  });
+  Settings.embedModel = new OpenAIEmbedding({
+    model: process.env.EMBEDDING_MODEL,
+    dimensions: process.env.EMBEDDING_DIM
+      ? parseInt(process.env.EMBEDDING_DIM)
+      : undefined,
+  });
+}
+
+function initOllama() {
+  Settings.llm = new Ollama({
+    model: process.env.MODEL ?? "",
+  });
+  Settings.embedModel = new OllamaEmbedding({
+    model: process.env.EMBEDDING_MODEL ?? "",
+  });
+}
+
+function initAnthropic() {
+  const embedModelMap: Record<string, string> = {
+    "all-MiniLM-L6-v2": "Xenova/all-MiniLM-L6-v2",
+    "all-mpnet-base-v2": "Xenova/all-mpnet-base-v2",
+  };
+  Settings.llm = new Anthropic({
+    model: process.env.MODEL as keyof typeof ALL_AVAILABLE_ANTHROPIC_MODELS,
+  });
+  Settings.embedModel = new HuggingFaceEmbedding({
+    modelType: embedModelMap[process.env.EMBEDDING_MODEL!],
+  });
+}
+
+function initGemini() {
+  Settings.llm = new Gemini({
+    model: process.env.MODEL as GEMINI_MODEL,
+  });
+  Settings.embedModel = new GeminiEmbedding({
+    model: process.env.EMBEDDING_MODEL as GEMINI_EMBEDDING_MODEL,
+  });
+}

templates/types/streaming/express/src/controllers/llamaindex-stream.ts

@@ -1,49 +1,65 @@
 import {
-  JSONValue,
+  StreamData,
   createCallbacksTransformer,
   createStreamDataTransformer,
-  experimental_StreamData,
   trimStartOfStreamHelper,
   type AIStreamCallbacksAndOptions,
 } from "ai";
-import { Response, StreamingAgentChatResponse } from "llamaindex";
+import {
+  Metadata,
+  NodeWithScore,
+  Response,
+  ToolCallLLMMessageOptions,
+} from "llamaindex";
+
+import { AgentStreamChatResponse } from "llamaindex/agent/base";
+import { appendImageData, appendSourceData } from "./stream-helper";
+
+type LlamaIndexResponse =
+  | AgentStreamChatResponse<ToolCallLLMMessageOptions>
+  | Response;
 
 type ParserOptions = {
   image_url?: string;
 };
 
 function createParser(
-  res: AsyncIterable<Response>,
-  data: experimental_StreamData,
+  res: AsyncIterable<LlamaIndexResponse>,
+  data: StreamData,
   opts?: ParserOptions,
 ) {
   const it = res[Symbol.asyncIterator]();
   const trimStartOfStream = trimStartOfStreamHelper();
+
+  let sourceNodes: NodeWithScore<Metadata>[] | undefined;
   return new ReadableStream<string>({
     start() {
-      // if image_url is provided, send it via the data stream
-      if (opts?.image_url) {
-        const message: JSONValue = {
-          type: "image_url",
-          image_url: {
-            url: opts.image_url,
-          },
-        };
-        data.append(message);
-      } else {
-        data.append({}); // send an empty image response for the user's message
-      }
+      appendImageData(data, opts?.image_url);
     },
     async pull(controller): Promise<void> {
       const { value, done } = await it.next();
       if (done) {
+        if (sourceNodes) {
+          appendSourceData(data, sourceNodes);
+        }
         controller.close();
-        data.append({}); // send an empty image response for the assistant's message
         data.close();
         return;
       }
 
-      const text = trimStartOfStream(value.response ?? "");
+      let delta;
+      if (value instanceof Response) {
+        // handle Response type
+        if (value.sourceNodes) {
+          // get source nodes from the first response
+          sourceNodes = value.sourceNodes;
+        }
+        delta = value.response ?? "";
+      } else {
+        // handle other types
+        delta = value.response.delta;
+      }
+      const text = trimStartOfStream(delta ?? "");
       if (text) {
         controller.enqueue(text);
       }
@@ -52,21 +68,14 @@ function createParser(
 }
 
 export function LlamaIndexStream(
-  response: StreamingAgentChatResponse | AsyncIterable<Response>,
+  response: AsyncIterable<LlamaIndexResponse>,
+  data: StreamData,
   opts?: {
     callbacks?: AIStreamCallbacksAndOptions;
     parserOptions?: ParserOptions;
   },
-): { stream: ReadableStream; data: experimental_StreamData } {
-  const data = new experimental_StreamData();
-  const res =
-    response instanceof StreamingAgentChatResponse
-      ? response.response
-      : response;
-  return {
-    stream: createParser(res, data, opts?.parserOptions)
-      .pipeThrough(createCallbacksTransformer(opts?.callbacks))
-      .pipeThrough(createStreamDataTransformer(true)),
-    data,
-  };
+): ReadableStream<Uint8Array> {
+  return createParser(response, data, opts?.parserOptions)
+    .pipeThrough(createCallbacksTransformer(opts?.callbacks))
+    .pipeThrough(createStreamDataTransformer());
 }

templates/types/streaming/express/src/controllers/stream-helper.ts

@@ -0,0 +1,97 @@
+import { StreamData } from "ai";
+import {
+  CallbackManager,
+  Metadata,
+  NodeWithScore,
+  ToolCall,
+  ToolOutput,
+} from "llamaindex";
+
+export function appendImageData(data: StreamData, imageUrl?: string) {
+  if (!imageUrl) return;
+  data.appendMessageAnnotation({
+    type: "image",
+    data: {
+      url: imageUrl,
+    },
+  });
+}
+
+export function appendSourceData(
+  data: StreamData,
+  sourceNodes?: NodeWithScore<Metadata>[],
+) {
+  if (!sourceNodes?.length) return;
+  data.appendMessageAnnotation({
+    type: "sources",
+    data: {
+      nodes: sourceNodes.map((node) => ({
+        ...node.node.toMutableJSON(),
+        id: node.node.id_,
+        score: node.score ?? null,
+      })),
+    },
+  });
+}
+
+export function appendEventData(data: StreamData, title?: string) {
+  if (!title) return;
+  data.appendMessageAnnotation({
+    type: "events",
+    data: {
+      title,
+    },
+  });
+}
+
+export function appendToolData(
+  data: StreamData,
+  toolCall: ToolCall,
+  toolOutput: ToolOutput,
+) {
+  data.appendMessageAnnotation({
+    type: "tools",
+    data: {
+      toolCall: {
+        id: toolCall.id,
+        name: toolCall.name,
+        input: toolCall.input,
+      },
+      toolOutput: {
+        output: toolOutput.output,
+        isError: toolOutput.isError,
+      },
+    },
+  });
+}
+
+export function createCallbackManager(stream: StreamData) {
+  const callbackManager = new CallbackManager();
+
+  callbackManager.on("retrieve", (data) => {
+    const { nodes, query } = data.detail;
+    appendEventData(stream, `Retrieving context for query: '${query}'`);
+    appendEventData(
+      stream,
+      `Retrieved ${nodes.length} sources to use as context for the query`,
+    );
+  });
+
+  callbackManager.on("llm-tool-call", (event) => {
+    const { name, input } = event.detail.payload.toolCall;
+    const inputString = Object.entries(input)
+      .map(([key, value]) => `${key}: ${value}`)
+      .join(", ");
+    appendEventData(
+      stream,
+      `Using tool: '${name}' with inputs: '${inputString}'`,
+    );
+  });
+
+  callbackManager.on("llm-tool-result", (event) => {
+    const { toolCall, toolResult } = event.detail.payload;
+    appendToolData(stream, toolCall, toolResult);
+  });
+
+  return callbackManager;
+}

templates/types/streaming/express/src/routes/chat.route.ts

@@ -1,9 +1,11 @@
 import express, { Router } from "express";
 import { chatRequest } from "../controllers/chat-request.controller";
 import { chat } from "../controllers/chat.controller";
+import { initSettings } from "../controllers/engine/settings";
 
 const llmRouter: Router = express.Router();
 
+initSettings();
 llmRouter.route("/").post(chat);
 llmRouter.route("/request").post(chatRequest);
 

templates/types/streaming/fastapi/README-template.md

@@ -11,20 +11,14 @@ poetry install
 poetry shell
 ```
 
-By default, we use the OpenAI LLM (though you can customize, see `app/settings.py`). As a result, you need to specify an `OPENAI_API_KEY` in an .env file in this directory.
-
-Example `.env` file:
-
-```
-OPENAI_API_KEY=<openai_api_key>
-```
+Then check the parameters that have been pre-configured in the `.env` file in this directory. (E.g. you might need to configure an `OPENAI_API_KEY` if you're using OpenAI as model provider).
 
 If you are using any tools or data sources, you can update their config files in the `config` folder.
 
 Second, generate the embeddings of the documents in the `./data` directory (if this folder exists - otherwise, skip this step):
 
 ```
-python app/engine/generate.py
+poetry run generate
 ```
 
 Third, run the development server:
@@ -84,7 +78,7 @@ docker run \
   -v $(pwd)/data:/app/data \ # Use your local folder to read the data
   -v $(pwd)/storage:/app/storage \ # Use your file system to store the vector database
   <your_backend_image_name> \
-  python app/engine/generate.py
+  poetry run generate
 ```
 
 3. Start the API:

templates/types/streaming/fastapi/app/api/routers/chat.py

@@ -1,11 +1,13 @@
-from typing import List
 from pydantic import BaseModel
-from fastapi.responses import StreamingResponse
+from typing import List, Any, Optional, Dict, Tuple
 from fastapi import APIRouter, Depends, HTTPException, Request, status
 from llama_index.core.chat_engine.types import BaseChatEngine
+from llama_index.core.schema import NodeWithScore
 from llama_index.core.llms import ChatMessage, MessageRole
 from app.engine import get_chat_engine
-from typing import List, Tuple
+from app.api.routers.vercel_response import VercelStreamResponse
+from app.api.routers.messaging import EventCallbackHandler
+from aiostream import stream
 
 chat_router = r = APIRouter()
 
@@ -18,9 +20,42 @@ class _Message(BaseModel):
 class _ChatData(BaseModel):
     messages: List[_Message]
 
+    class Config:
+        json_schema_extra = {
+            "example": {
+                "messages": [
+                    {
+                        "role": "user",
+                        "content": "What standards for letters exist?",
+                    }
+                ]
+            }
+        }
+
+
+class _SourceNodes(BaseModel):
+    id: str
+    metadata: Dict[str, Any]
+    score: Optional[float]
+    text: str
+
+    @classmethod
+    def from_source_node(cls, source_node: NodeWithScore):
+        return cls(
+            id=source_node.node.node_id,
+            metadata=source_node.node.metadata,
+            score=source_node.score,
+            text=source_node.node.text,  # type: ignore
+        )
+
+    @classmethod
+    def from_source_nodes(cls, source_nodes: List[NodeWithScore]):
+        return [cls.from_source_node(node) for node in source_nodes]
+
 
 class _Result(BaseModel):
     result: _Message
+    nodes: List[_SourceNodes]
 
 
 async def parse_chat_data(data: _ChatData) -> Tuple[str, List[ChatMessage]]:
@@ -56,15 +91,46 @@ async def chat(
 ):
     last_message_content, messages = await parse_chat_data(data)
 
+    event_handler = EventCallbackHandler()
+    chat_engine.callback_manager.handlers.append(event_handler)  # type: ignore
     response = await chat_engine.astream_chat(last_message_content, messages)
 
-    async def event_generator():
-        async for token in response.async_response_gen():
-            if await request.is_disconnected():
-                break
-            yield token
+    async def content_generator():
+        # Yield the text response
+        async def _text_generator():
+            async for token in response.async_response_gen():
+                yield VercelStreamResponse.convert_text(token)
+            # the text_generator is the leading stream, once it's finished, also finish the event stream
+            event_handler.is_done = True
 
-    return StreamingResponse(event_generator(), media_type="text/plain")
+        # Yield the events from the event handler
+        async def _event_generator():
+            async for event in event_handler.async_event_gen():
+                event_response = event.to_response()
+                if event_response is not None:
+                    yield VercelStreamResponse.convert_data(event_response)
+
+        combine = stream.merge(_text_generator(), _event_generator())
+        async with combine.stream() as streamer:
+            async for item in streamer:
+                if await request.is_disconnected():
+                    break
+                yield item
+
+        # Yield the source nodes
+        yield VercelStreamResponse.convert_data(
+            {
+                "type": "sources",
+                "data": {
+                    "nodes": [
+                        _SourceNodes.from_source_node(node).dict()
+                        for node in response.source_nodes
+                    ]
+                },
+            }
+        )
+
+    return VercelStreamResponse(content=content_generator())
 
 
 # non-streaming endpoint - delete if not needed
@@ -77,5 +143,6 @@ async def chat_request(
 
     response = await chat_engine.achat(last_message_content, messages)
     return _Result(
-        result=_Message(role=MessageRole.ASSISTANT, content=response.response)
+        result=_Message(role=MessageRole.ASSISTANT, content=response.response),
+        nodes=_SourceNodes.from_source_nodes(response.source_nodes),
     )

templates/types/streaming/fastapi/app/api/routers/messaging.py

@@ -0,0 +1,141 @@
+import json
+import asyncio
+from typing import AsyncGenerator, Dict, Any, List, Optional
+from llama_index.core.callbacks.base import BaseCallbackHandler
+from llama_index.core.callbacks.schema import CBEventType
+from llama_index.core.tools.types import ToolOutput
+from pydantic import BaseModel
+
+
+class CallbackEvent(BaseModel):
+    event_type: CBEventType
+    payload: Optional[Dict[str, Any]] = None
+    event_id: str = ""
+
+    def get_retrieval_message(self) -> dict | None:
+        if self.payload:
+            nodes = self.payload.get("nodes")
+            if nodes:
+                msg = f"Retrieved {len(nodes)} sources to use as context for the query"
+            else:
+                msg = f"Retrieving context for query: '{self.payload.get('query_str')}'"
+            return {
+                "type": "events",
+                "data": {"title": msg},
+            }
+        else:
+            return None
+
+    def get_tool_message(self) -> dict | None:
+        func_call_args = self.payload.get("function_call")
+        if func_call_args is not None and "tool" in self.payload:
+            tool = self.payload.get("tool")
+            return {
+                "type": "events",
+                "data": {
+                    "title": f"Calling tool: {tool.name} with inputs: {func_call_args}",
+                },
+            }
+
+    def _is_output_serializable(self, output: Any) -> bool:
+        try:
+            json.dumps(output)
+            return True
+        except TypeError:
+            return False
+
+    def get_agent_tool_response(self) -> dict | None:
+        response = self.payload.get("response")
+        if response is not None:
+            sources = response.sources
+            for source in sources:
+                # Return the tool response here to include the toolCall information
+                if isinstance(source, ToolOutput):
+                    if self._is_output_serializable(source.raw_output):
+                        output = source.raw_output
+                    else:
+                        output = source.content
+
+                    return {
+                        "type": "tools",
+                        "data": {
+                            "toolOutput": {
+                                "output": output,
+                                "isError": source.is_error,
+                            },
+                            "toolCall": {
+                                "id": None,  # There is no tool id in the ToolOutput
+                                "name": source.tool_name,
+                                "input": source.raw_input,
+                            },
+                        },
+                    }
+
+    def to_response(self):
+        match self.event_type:
+            case "retrieve":
+                return self.get_retrieval_message()
+            case "function_call":
+                return self.get_tool_message()
+            case "agent_step":
+                return self.get_agent_tool_response()
+            case _:
+                return None
+
+
+class EventCallbackHandler(BaseCallbackHandler):
+    _aqueue: asyncio.Queue
+    is_done: bool = False
+
+    def __init__(
+        self,
+    ):
+        """Initialize the base callback handler."""
+        ignored_events = [
+            CBEventType.CHUNKING,
+            CBEventType.NODE_PARSING,
+            CBEventType.EMBEDDING,
+            CBEventType.LLM,
+            CBEventType.TEMPLATING,
+        ]
+        super().__init__(ignored_events, ignored_events)
+        self._aqueue = asyncio.Queue()
+
+    def on_event_start(
+        self,
+        event_type: CBEventType,
+        payload: Optional[Dict[str, Any]] = None,
+        event_id: str = "",
+        **kwargs: Any,
+    ) -> str:
+        event = CallbackEvent(event_id=event_id, event_type=event_type, payload=payload)
+        if event.to_response() is not None:
+            self._aqueue.put_nowait(event)
+
+    def on_event_end(
+        self,
+        event_type: CBEventType,
+        payload: Optional[Dict[str, Any]] = None,
+        event_id: str = "",
+        **kwargs: Any,
+    ) -> None:
+        event = CallbackEvent(event_id=event_id, event_type=event_type, payload=payload)
+        if event.to_response() is not None:
+            self._aqueue.put_nowait(event)
+
+    def start_trace(self, trace_id: Optional[str] = None) -> None:
+        """No-op."""
+
+    def end_trace(
+        self,
+        trace_id: Optional[str] = None,
+        trace_map: Optional[Dict[str, List[str]]] = None,
+    ) -> None:
+        """No-op."""
+
+    async def async_event_gen(self) -> AsyncGenerator[CallbackEvent, None]:
+        while not self._aqueue.empty() or not self.is_done:
+            try:
+                yield await asyncio.wait_for(self._aqueue.get(), timeout=0.1)
+            except asyncio.TimeoutError:
+                pass

templates/types/streaming/fastapi/app/api/routers/vercel_response.py

@@ -0,0 +1,29 @@
+import json
+from typing import Any
+from fastapi.responses import StreamingResponse
+
+
+class VercelStreamResponse(StreamingResponse):
+    """
+    Class to convert the response from the chat engine to the streaming format expected by Vercel
+    """
+
+    TEXT_PREFIX = "0:"
+    DATA_PREFIX = "8:"
+
+    @classmethod
+    def convert_text(cls, token: str):
+        # Escape newlines and double quotes to avoid breaking the stream
+        token = json.dumps(token)
+        return f"{cls.TEXT_PREFIX}{token}\n"
+
+    @classmethod
+    def convert_data(cls, data: dict):
+        data_str = json.dumps(data)
+        return f"{cls.DATA_PREFIX}[{data_str}]\n"
+
+    def __init__(self, content: Any, **kwargs):
+        super().__init__(
+            content=content,
+            **kwargs,
+        )

templates/types/streaming/fastapi/app/observability.py

@@ -0,0 +1,2 @@
+def init_observability():
+    pass

templates/types/streaming/fastapi/app/settings.py

@@ -1,41 +1,92 @@
 import os
 from typing import Dict
 from llama_index.core.settings import Settings
-from llama_index.llms.openai import OpenAI
-from llama_index.embeddings.openai import OpenAIEmbedding
-
-
-def llm_config_from_env() -> Dict:
-    from llama_index.core.constants import DEFAULT_TEMPERATURE
-
-    model = os.getenv("MODEL")
-    temperature = os.getenv("LLM_TEMPERATURE", DEFAULT_TEMPERATURE)
-    max_tokens = os.getenv("LLM_MAX_TOKENS")
-
-    config = {
-        "model": model,
-        "temperature": float(temperature),
-        "max_tokens": int(max_tokens) if max_tokens is not None else None,
-    }
-    return config
-
-
-def embedding_config_from_env() -> Dict:
-    model = os.getenv("EMBEDDING_MODEL")
-    dimension = os.getenv("EMBEDDING_DIM")
-
-    config = {
-        "model": model,
-        "dimension": int(dimension) if dimension is not None else None,
-    }
-    return config
 
 
 def init_settings():
-    llm_configs = llm_config_from_env()
-    embedding_configs = embedding_config_from_env()
-
-    Settings.llm = OpenAI(**llm_configs)
-    Settings.embed_model = OpenAIEmbedding(**embedding_configs)
+    model_provider = os.getenv("MODEL_PROVIDER")
+    if model_provider == "openai":
+        init_openai()
+    elif model_provider == "ollama":
+        init_ollama()
+    elif model_provider == "anthropic":
+        init_anthropic()
+    elif model_provider == "gemini":
+        init_gemini()
+    else:
+        raise ValueError(f"Invalid model provider: {model_provider}")
     Settings.chunk_size = int(os.getenv("CHUNK_SIZE", "1024"))
     Settings.chunk_overlap = int(os.getenv("CHUNK_OVERLAP", "20"))
+
+
+def init_ollama():
+    from llama_index.llms.ollama import Ollama
+    from llama_index.embeddings.ollama import OllamaEmbedding
+
+    Settings.embed_model = OllamaEmbedding(model_name=os.getenv("EMBEDDING_MODEL"))
+    Settings.llm = Ollama(model=os.getenv("MODEL"))
+
+
+def init_openai():
+    from llama_index.llms.openai import OpenAI
+    from llama_index.embeddings.openai import OpenAIEmbedding
+    from llama_index.core.constants import DEFAULT_TEMPERATURE
+
+    max_tokens = os.getenv("LLM_MAX_TOKENS")
+    config = {
+        "model": os.getenv("MODEL"),
+        "temperature": float(os.getenv("LLM_TEMPERATURE", DEFAULT_TEMPERATURE)),
+        "max_tokens": int(max_tokens) if max_tokens is not None else None,
+    }
+    Settings.llm = OpenAI(**config)
+
+    dimensions = os.getenv("EMBEDDING_DIM")
+    config = {
+        "model": os.getenv("EMBEDDING_MODEL"),
+        "dimensions": int(dimensions) if dimensions is not None else None,
+    }
+    Settings.embed_model = OpenAIEmbedding(**config)
+
+
+def init_anthropic():
+    from llama_index.llms.anthropic import Anthropic
+    from llama_index.embeddings.huggingface import HuggingFaceEmbedding
+
+    model_map: Dict[str, str] = {
+        "claude-3-opus": "claude-3-opus-20240229",
+        "claude-3-sonnet": "claude-3-sonnet-20240229",
+        "claude-3-haiku": "claude-3-haiku-20240307",
+        "claude-2.1": "claude-2.1",
+        "claude-instant-1.2": "claude-instant-1.2",
+    }
+
+    embed_model_map: Dict[str, str] = {
+        "all-MiniLM-L6-v2": "sentence-transformers/all-MiniLM-L6-v2",
+        "all-mpnet-base-v2": "sentence-transformers/all-mpnet-base-v2",
+    }
+
+    Settings.llm = Anthropic(model=model_map[os.getenv("MODEL")])
+    Settings.embed_model = HuggingFaceEmbedding(
+        model_name=embed_model_map[os.getenv("EMBEDDING_MODEL")]
+    )
+
+
+def init_gemini():
+    from llama_index.llms.gemini import Gemini
+    from llama_index.embeddings.gemini import GeminiEmbedding
+
+    model_map: Dict[str, str] = {
+        "gemini-1.5-pro-latest": "models/gemini-1.5-pro-latest",
+        "gemini-pro": "models/gemini-pro",
+        "gemini-pro-vision": "models/gemini-pro-vision",
+    }
+
+    embed_model_map: Dict[str, str] = {
+        "embedding-001": "models/embedding-001",
+        "text-embedding-004": "models/text-embedding-004",
+    }
+
+    Settings.llm = Gemini(model=model_map[os.getenv("MODEL")])
+    Settings.embed_model = GeminiEmbedding(
+        model_name=embed_model_map[os.getenv("EMBEDDING_MODEL")]
+    )

templates/types/streaming/fastapi/main.py

@@ -7,13 +7,16 @@ import os
 import uvicorn
 from fastapi import FastAPI
 from fastapi.middleware.cors import CORSMiddleware
+from fastapi.responses import RedirectResponse
 from app.api.routers.chat import chat_router
 from app.settings import init_settings
+from app.observability import init_observability
 
 
 app = FastAPI()
 
 init_settings()
+init_observability()
 
 environment = os.getenv("ENVIRONMENT", "dev")  # Default to 'development' if not set
 
@@ -29,6 +32,12 @@ if environment == "dev":
         allow_headers=["*"],
     )
 
+    # Redirect to documentation page when accessing base URL
+    @app.get("/")
+    async def redirect_to_docs():
+        return RedirectResponse(url="/docs")
+
+
 app.include_router(chat_router, prefix="/api/chat")
 
 

templates/types/streaming/fastapi/pyproject.toml

@@ -5,14 +5,17 @@ description = ""
 authors = ["Marcus Schiesser <mail@marcusschiesser.de>"]
 readme = "README.md"
 
+[tool.poetry.scripts]
+generate = "app.engine.generate:generate_datasource"
+
 [tool.poetry.dependencies]
 python = "^3.11,<3.12"
 fastapi = "^0.109.1"
 uvicorn = { extras = ["standard"], version = "^0.23.2" }
 python-dotenv = "^1.0.0"
-llama-index = "0.10.15"
-llama-index-core = "0.10.15"
-llama-index-agent-openai = "0.1.5"
+aiostream = "^0.5.2"
+llama-index = "0.10.28"
+llama-index-core = "0.10.28"
 
 [build-system]
 requires = ["poetry-core"]