diff --git a/helpers/env-variables.ts b/helpers/env-variables.ts
new file mode 100644
index 0000000000000000000000000000000000000000..327fb27ad3913080a1e94ac122749d254706bd9d
--- /dev/null
+++ b/helpers/env-variables.ts
@@ -0,0 +1,239 @@
+import fs from "fs/promises";
+import path from "path";
+import {
+ FileSourceConfig,
+ TemplateDataSource,
+ TemplateFramework,
+ TemplateVectorDB,
+} from "./types";
+
+type EnvVar = {
+ name?: string;
+ description?: string;
+ value?: string;
+};
+
+const renderEnvVar = (envVars: EnvVar[]): string => {
+ return envVars.reduce(
+ (prev, env) =>
+ prev +
+ (env.description
+ ? `# ${env.description.replaceAll("\n", "\n# ")}\n`
+ : "") +
+ (env.name && env.value
+ ? `${env.name}=${env.value}\n\n`
+ : `# ${env.name}=\n\n`),
+ "",
+ );
+};
+
+const getVectorDBEnvs = (vectorDb: TemplateVectorDB) => {
+ switch (vectorDb) {
+ case "mongo":
+ return [
+ {
+ name: "MONGO_URI",
+ description:
+ "For generating a connection URI, see https://docs.timescale.com/use-timescale/latest/services/create-a-service\nThe MongoDB connection URI.",
+ },
+ {
+ name: "MONGODB_DATABASE",
+ },
+ {
+ name: "MONGODB_VECTORS",
+ },
+ {
+ name: "MONGODB_VECTOR_INDEX",
+ },
+ ];
+ case "pg":
+ return [
+ {
+ name: "PG_CONNECTION_STRING",
+ description:
+ "For generating a connection URI, see https://docs.timescale.com/use-timescale/latest/services/create-a-service\nThe PostgreSQL connection string.",
+ },
+ ];
+
+ case "pinecone":
+ return [
+ {
+ name: "PINECONE_API_KEY",
+ description:
+ "Configuration for Pinecone vector store\nThe Pinecone API key.",
+ },
+ {
+ name: "PINECONE_ENVIRONMENT",
+ },
+ {
+ name: "PINECONE_INDEX_NAME",
+ },
+ ];
+ default:
+ return [];
+ }
+};
+
+const getDataSourceEnvs = (dataSource: TemplateDataSource) => {
+ switch (dataSource.type) {
+ case "web":
+ return [
+ {
+ name: "BASE_URL",
+ description: "The base URL to start web scraping.",
+ },
+ {
+ name: "URL_PREFIX",
+ description: "The prefix of the URL to start web scraping.",
+ },
+ {
+ name: "MAX_DEPTH",
+ description: "The maximum depth to scrape.",
+ },
+ ];
+ default:
+ return [];
+ }
+};
+
+export const createBackendEnvFile = async (
+ root: string,
+ opts: {
+ openAiKey?: string;
+ llamaCloudKey?: string;
+ vectorDb?: TemplateVectorDB;
+ model?: string;
+ embeddingModel?: string;
+ framework?: TemplateFramework;
+ dataSource?: TemplateDataSource;
+ port?: number;
+ },
+) => {
+ // 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,
+ },
+ // Add vector database environment variables
+ ...(opts.vectorDb ? getVectorDBEnvs(opts.vectorDb) : []),
+ // Add data source environment variables
+ ...(opts.dataSource ? getDataSourceEnvs(opts.dataSource) : []),
+ ];
+ 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}
+"`,
+ },
+ (opts?.dataSource?.config as FileSourceConfig).useLlamaParse
+ ? {
+ name: "LLAMA_CLOUD_API_KEY",
+ description: `The Llama Cloud API key.`,
+ value: opts.llamaCloudKey,
+ }
+ : {},
+ ],
+ ];
+ } else {
+ envVars = [
+ ...defaultEnvs,
+ ...[
+ opts.framework === "nextjs"
+ ? {
+ name: "NEXT_PUBLIC_MODEL",
+ description:
+ "The LLM model to use (hardcode to front-end artifact).",
+ }
+ : {},
+ ],
+ ];
+ }
+ // Render and write env file
+ const content = renderEnvVar(envVars);
+ await fs.writeFile(path.join(root, envFileName), content);
+ console.log(`Created '${envFileName}' file. Please check the settings.`);
+};
+
+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.",
+ value: opts.customApiPath
+ ? opts.customApiPath
+ : "http://localhost:8000/api/chat",
+ },
+ ];
+ const content = renderEnvVar(defaultFrontendEnvs);
+ await fs.writeFile(path.join(root, ".env"), content);
+};
diff --git a/helpers/index.ts b/helpers/index.ts
index b81bb59d222e1acaa22ef91b2a1bf59fa2a1b0c8..91280af35319e501b99325c9152a849aa2e536af 100644
--- a/helpers/index.ts
+++ b/helpers/index.ts
@@ -7,6 +7,7 @@ import { cyan } from "picocolors";
import { COMMUNITY_OWNER, COMMUNITY_REPO } from "./constant";
import { templatesDir } from "./dir";
+import { createBackendEnvFile, createFrontendEnvFile } from "./env-variables";
import { PackageManager } from "./get-pkg-manager";
import { installLlamapackProject } from "./llama-pack";
import { isHavingPoetryLockFile, tryPoetryRun } from "./poetry";
@@ -18,89 +19,9 @@ import {
TemplateDataSource,
TemplateFramework,
TemplateVectorDB,
- WebSourceConfig,
} from "./types";
import { installTSTemplate } from "./typescript";
-const createEnvLocalFile = async (
- root: string,
- opts?: {
- openAiKey?: string;
- llamaCloudKey?: string;
- vectorDb?: TemplateVectorDB;
- model?: string;
- embeddingModel?: string;
- framework?: TemplateFramework;
- dataSource?: TemplateDataSource;
- },
-) => {
- const envFileName = ".env";
- let content = "";
-
- const model = opts?.model || "gpt-3.5-turbo";
- content += `MODEL=${model}\n`;
- if (opts?.framework === "nextjs") {
- content += `NEXT_PUBLIC_MODEL=${model}\n`;
- }
- console.log("\nUsing OpenAI model: ", model, "\n");
-
- if (opts?.openAiKey) {
- content += `OPENAI_API_KEY=${opts?.openAiKey}\n`;
- }
-
- if (opts?.embeddingModel) {
- content += `EMBEDDING_MODEL=${opts?.embeddingModel}\n`;
- }
-
- if ((opts?.dataSource?.config as FileSourceConfig).useLlamaParse) {
- if (opts?.llamaCloudKey) {
- content += `LLAMA_CLOUD_API_KEY=${opts?.llamaCloudKey}\n`;
- } else {
- content += `# Please obtain the Llama Cloud API key from https://cloud.llamaindex.ai/api-key
-# and set it to the LLAMA_CLOUD_API_KEY variable below.
-# LLAMA_CLOUD_API_KEY=`;
- }
- }
-
- switch (opts?.vectorDb) {
- case "mongo": {
- content += `# For generating a connection URI, see https://www.mongodb.com/docs/guides/atlas/connection-string\n`;
- content += `MONGO_URI=\n`;
- content += `MONGODB_DATABASE=\n`;
- content += `MONGODB_VECTORS=\n`;
- content += `MONGODB_VECTOR_INDEX=\n`;
- break;
- }
- case "pg": {
- content += `# For generating a connection URI, see https://docs.timescale.com/use-timescale/latest/services/create-a-service\n`;
- content += `PG_CONNECTION_STRING=\n`;
- break;
- }
- case "pinecone": {
- content += `PINECONE_API_KEY=\n`;
- content += `PINECONE_ENVIRONMENT=\n`;
- content += `PINECONE_INDEX_NAME=\n`;
- break;
- }
- }
-
- switch (opts?.dataSource?.type) {
- case "web": {
- const webConfig = opts?.dataSource.config as WebSourceConfig;
- content += `# web loader config\n`;
- content += `BASE_URL=${webConfig.baseUrl}\n`;
- content += `URL_PREFIX=${webConfig.baseUrl}\n`;
- content += `MAX_DEPTH=${webConfig.depth}\n`;
- break;
- }
- }
-
- if (content) {
- await fs.writeFile(path.join(root, envFileName), content);
- console.log(`Created '${envFileName}' file. Please check the settings.`);
- }
-};
-
// eslint-disable-next-line max-params
async function generateContextData(
framework: TemplateFramework,
@@ -232,7 +153,7 @@ export const installTemplate = async (
// This is a backend, so we need to copy the test data and create the env file.
// Copy the environment file to the target directory.
- await createEnvLocalFile(props.root, {
+ await createBackendEnvFile(props.root, {
openAiKey: props.openAiKey,
llamaCloudKey: props.llamaCloudKey,
vectorDb: props.vectorDb,
@@ -240,6 +161,7 @@ export const installTemplate = async (
embeddingModel: props.embeddingModel,
framework: props.framework,
dataSource: props.dataSource,
+ port: props.externalPort,
});
if (props.engine === "context") {
@@ -260,8 +182,10 @@ export const installTemplate = async (
}
} else {
// this is a frontend for a full-stack app, create .env file with model information
- const content = `MODEL=${props.model}\nNEXT_PUBLIC_MODEL=${props.model}\n`;
- await fs.writeFile(path.join(props.root, ".env"), content);
+ createFrontendEnvFile(props.root, {
+ model: props.model,
+ customApiPath: props.customApiPath,
+ });
}
};
diff --git a/helpers/typescript.ts b/helpers/typescript.ts
index ae97c2e23f7a5dd7a8cf50ebd89448d6ed5cbb8e..e607f070bd3cde4ef9f787f990d9974e3c6c163f 100644
--- a/helpers/typescript.ts
+++ b/helpers/typescript.ts
@@ -171,10 +171,6 @@ export const installTSTemplate = async ({
const apiPath = path.join(root, "app", "api");
await fs.rm(apiPath, { recursive: true });
// modify the dev script to use the custom api path
- packageJson.scripts = {
- ...packageJson.scripts,
- dev: `cross-env NEXT_PUBLIC_CHAT_API=${customApiPath} next dev`,
- };
}
if (engine === "context" && relativeEngineDestPath) {
diff --git a/templates/components/engines/python/agent/__init__.py b/templates/components/engines/python/agent/__init__.py
index 1826d0e0a9f0fccb38b3a27a5098870286988ada..52e363941fcd64238f06fba4d8feaf386dce70fd 100644
--- a/templates/components/engines/python/agent/__init__.py
+++ b/templates/components/engines/python/agent/__init__.py
@@ -1,3 +1,4 @@
+import os
from llama_index.core.settings import Settings
from llama_index.core.agent import AgentRunner
from llama_index.core.tools.query_engine import QueryEngineTool
@@ -6,11 +7,13 @@ from app.engine.index import get_index
def get_chat_engine():
+ system_prompt = os.getenv("SYSTEM_PROMPT")
+ top_k = os.getenv("TOP_K", "3")
tools = []
# Add query tool
index = get_index()
- query_engine = index.as_query_engine(similarity_top_k=3)
+ 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)
@@ -20,5 +23,6 @@ def get_chat_engine():
return AgentRunner.from_llm(
llm=Settings.llm,
tools=tools,
+ system_prompt=system_prompt,
verbose=True,
)
diff --git a/templates/components/engines/python/chat/__init__.py b/templates/components/engines/python/chat/__init__.py
index 95de61cd825b2ac299caf9faafe98c529c4fe5a7..65624c32c63c484958f2c566574ea029f731af39 100644
--- a/templates/components/engines/python/chat/__init__.py
+++ b/templates/components/engines/python/chat/__init__.py
@@ -1,7 +1,13 @@
+import os
from app.engine.index import get_index
def get_chat_engine():
+ system_prompt = os.getenv("SYSTEM_PROMPT")
+ top_k = os.getenv("TOP_K", 3)
+
return get_index().as_chat_engine(
- similarity_top_k=3, chat_mode="condense_plus_context"
+ similarity_top_k=int(top_k),
+ system_prompt=system_prompt,
+ chat_mode="condense_plus_context",
)
diff --git a/templates/components/loaders/python/llama_parse/loader.py b/templates/components/loaders/python/llama_parse/loader.py
index d74d7c03be73c3f2116e7d4ed648d401499e2b15..4c9fbf109f32918d70c06d08e0a671c5377b4f0b 100644
--- a/templates/components/loaders/python/llama_parse/loader.py
+++ b/templates/components/loaders/python/llama_parse/loader.py
@@ -1,3 +1,4 @@
+import os
from llama_parse import LlamaParse
from llama_index.core import SimpleDirectoryReader
@@ -5,6 +6,11 @@ DATA_DIR = "data" # directory containing the documents
def get_documents():
+ if os.getenv("LLAMA_CLOUD_API_KEY") is None:
+ raise ValueError(
+ "LLAMA_CLOUD_API_KEY environment variable is not set. "
+ "Please set it in .env file or in your shell environment then run again!"
+ )
parser = LlamaParse(result_type="markdown", verbose=True, language="en")
reader = SimpleDirectoryReader(DATA_DIR, file_extractor={".pdf": parser})
diff --git a/templates/types/simple/fastapi/app/settings.py b/templates/types/simple/fastapi/app/settings.py
index bd49f9452d299565e8688d5d8881d9ff65a76485..3f2c5e078e0fddb01d433ee0147de5a207f23e36 100644
--- a/templates/types/simple/fastapi/app/settings.py
+++ b/templates/types/simple/fastapi/app/settings.py
@@ -1,14 +1,41 @@
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
-from llama_index.core.settings import Settings
+
+
+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_model = os.getenv("MODEL", "gpt-3.5-turbo")
- embedding_model = os.getenv("EMBEDDING_MODEL", "text-embedding-ada-002")
+ llm_configs = llm_config_from_env()
+ embedding_configs = embedding_config_from_env()
- Settings.llm = OpenAI(model=llm_model)
- Settings.embed_model = OpenAIEmbedding(model=embedding_model)
- Settings.chunk_size = 1024
- Settings.chunk_overlap = 20
+ Settings.llm = OpenAI(**llm_configs)
+ Settings.embed_model = OpenAIEmbedding(**embedding_configs)
+ Settings.chunk_size = int(os.getenv("CHUNK_SIZE", "1024"))
+ Settings.chunk_overlap = int(os.getenv("CHUNK_OVERLAP", "20"))
diff --git a/templates/types/simple/fastapi/main.py b/templates/types/simple/fastapi/main.py
index d9291060d141f06f7de78b1e91cea2330e94ee48..691f543d6ffddf74cca975b1fce800e4b887431a 100644
--- a/templates/types/simple/fastapi/main.py
+++ b/templates/types/simple/fastapi/main.py
@@ -32,4 +32,7 @@ app.include_router(chat_router, prefix="/api/chat")
if __name__ == "__main__":
- uvicorn.run(app="main:app", host="0.0.0.0", reload=True)
+ app_host = os.getenv("APP_HOST", "0.0.0.0")
+ app_port = int(os.getenv("APP_PORT", "8000"))
+
+ uvicorn.run(app="main:app", host=app_host, port=app_port, reload=True)
diff --git a/templates/types/streaming/fastapi/app/settings.py b/templates/types/streaming/fastapi/app/settings.py
index bd49f9452d299565e8688d5d8881d9ff65a76485..3f2c5e078e0fddb01d433ee0147de5a207f23e36 100644
--- a/templates/types/streaming/fastapi/app/settings.py
+++ b/templates/types/streaming/fastapi/app/settings.py
@@ -1,14 +1,41 @@
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
-from llama_index.core.settings import Settings
+
+
+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_model = os.getenv("MODEL", "gpt-3.5-turbo")
- embedding_model = os.getenv("EMBEDDING_MODEL", "text-embedding-ada-002")
+ llm_configs = llm_config_from_env()
+ embedding_configs = embedding_config_from_env()
- Settings.llm = OpenAI(model=llm_model)
- Settings.embed_model = OpenAIEmbedding(model=embedding_model)
- Settings.chunk_size = 1024
- Settings.chunk_overlap = 20
+ Settings.llm = OpenAI(**llm_configs)
+ Settings.embed_model = OpenAIEmbedding(**embedding_configs)
+ Settings.chunk_size = int(os.getenv("CHUNK_SIZE", "1024"))
+ Settings.chunk_overlap = int(os.getenv("CHUNK_OVERLAP", "20"))
diff --git a/templates/types/streaming/fastapi/main.py b/templates/types/streaming/fastapi/main.py
index ed73c268cf119cd453144f198653224b2c57b3aa..7e7d42496b4a6baa485051ab926d7bb254e7e7d7 100644
--- a/templates/types/streaming/fastapi/main.py
+++ b/templates/types/streaming/fastapi/main.py
@@ -33,4 +33,7 @@ app.include_router(chat_router, prefix="/api/chat")
if __name__ == "__main__":
- uvicorn.run(app="main:app", host="0.0.0.0", reload=True)
+ app_host = os.getenv("APP_HOST", "0.0.0.0")
+ app_port = int(os.getenv("APP_PORT", "8000"))
+
+ uvicorn.run(app="main:app", host=app_host, port=app_port, reload=True)