Deprecated Features

List of deprecated APIs and their recommended replacements

This page tracks deprecated features in LlamaIndex.TS and provides migration guidance.

Currently Deprecated

Agents (ReAct & Function Calling)

Deprecated in: v0.11.0
Status: Showing deprecation warnings
Removal Timeline: To be determined

Deprecated APIs:

• llamaindex/agent
• ReActAgent
• OpenAIAgent
• Legacy agent implementations in @llamaindex/core/agent

Replacement:

Use the new @llamaindex/workflow package for building agentic workflows.

Before:

import { ReActAgent } from "llamaindex/agent";
import { OpenAI } from "@llamaindex/openai";

const agent = new ReActAgent({
  llm: new OpenAI({ model: "gpt-4" }),
  tools: [myTool],
});

const response = await agent.chat({ message: "Hello" });

After:

import { AgentWorkflow } from "@llamaindex/workflow";
import { OpenAI } from "@llamaindex/openai";

const workflow = new AgentWorkflow({
  llm: new OpenAI({ model: "gpt-4" }),
  tools: [myTool],
});

const response = await workflow.run({ message: "Hello" });

Why the change:

• More flexible workflow orchestration
• Better state management
• Improved error handling
• Support for complex multi-agent systems

ServiceContext

Deprecated in: v0.8.0
Removed in: v0.9.0
Status: Completely removed

Deprecated APIs:

• ServiceContext
• serviceContextFromDefaults()
• All ServiceContext-related methods

Replacement:

Use the Settings global configuration object.

Before:

import { serviceContextFromDefaults } from "llamaindex";
import { OpenAI, OpenAIEmbedding } from "@llamaindex/openai";

const serviceContext = serviceContextFromDefaults({
  llm: new OpenAI({ model: "gpt-4" }),
  embedModel: new OpenAIEmbedding(),
  chunkSize: 512,
});

const index = await VectorStoreIndex.fromDocuments(documents, {
  serviceContext,
});

After:

import { Settings, VectorStoreIndex } from "llamaindex";
import { OpenAI, OpenAIEmbedding } from "@llamaindex/openai";

Settings.llm = new OpenAI({ model: "gpt-4" });
Settings.embedModel = new OpenAIEmbedding();
Settings.chunkSize = 512;

const index = await VectorStoreIndex.fromDocuments(documents);

Why the change:

• Simpler, more intuitive API
• Reduced boilerplate code
• Better alignment with modern patterns

Planned Deprecations

Legacy Workflows

Deprecation Notice: The old workflow system in the main package will be removed in a future version.

Migration Path:

Migrate to @llamaindex/workflow package:

npm install @llamaindex/workflow

import { Workflow, StartEvent, StopEvent } from "@llamaindex/workflow";

class MyWorkflow extends Workflow {
  async run(ctx: Context, ev: StartEvent) {
    // Your workflow logic
    return new StopEvent({ result: "done" });
  }
}

Old Readers in Main Package

Status: Readers are being migrated to separate packages for better modularity.

Current Approach:

import { PDFReader } from "llamaindex/readers/PDFReader";

Recommended Approach:

npm install @llamaindex/readers

import { PDFReader } from "@llamaindex/readers/pdf";

Removed Features

Provider Re-exports (Removed in 0.9.0)

What was removed:

// These no longer work
import { OpenAI, Anthropic, Pinecone } from "llamaindex";

How to update:

import { OpenAI } from "@llamaindex/openai";
import { Anthropic } from "@llamaindex/anthropic";
import { PineconeVectorStore } from "@llamaindex/pinecone";

gpt-tokenizer in Main Package (Removed in 0.9.16)

What changed:

The gpt-tokenizer dependency is now optional to reduce package size.

Impact:

If you need tokenization utilities, install separately:

npm install gpt-tokenizer

Cloud Package Export (Removed in 0.12.0)

What was removed:

import { LlamaCloudIndex } from "llamaindex";

How to update:

npm install @llamaindex/cloud

import { LlamaCloudIndex } from "@llamaindex/cloud";

API Naming Changes

These aren’t deprecations but important naming changes to be aware of:

Embed Model

Old: embedModel (still works)
Preferred: embedModel (no change needed)

Chat vs Query

Before (some older examples):

engine.query(message); // Inconsistent usage

Current (standardized):

queryEngine.query({ query: "..." });
chatEngine.chat({ message: "..." });

Best Practices for Handling Deprecations

1. Monitor Console Warnings

Deprecated features will show warnings in development:

// Enable verbose logging to see deprecation warnings
Settings.debug = true;

2. Check Your Dependencies

Regularly audit your imports:

# Check for deprecated imports
grep -r "from 'llamaindex/agent'" .
grep -r "ServiceContext" .

3. Update Incrementally

Don’t wait until removal:

// ✗ Bad: Ignoring deprecation warnings
import { ReActAgent } from "llamaindex/agent"; // Shows warning

// ✓ Good: Update when you see the warning
import { AgentWorkflow } from "@llamaindex/workflow";

4. Read the Changelog

Before upgrading, review the CHANGELOG.md.

5. Test After Migration

Always test your application after updating deprecated APIs:

npm test
npm run build

Getting Help with Deprecated Features

If you need help migrating from deprecated features:

Discord Community

Ask questions in our Discord server

GitHub Discussions

Search for or start a discussion

Version Upgrades

See version-specific migration guides

Troubleshooting

Common migration issues and solutions

Deprecation Policy

LlamaIndex.TS follows semantic versioning:

• Deprecation warnings: Added in minor versions (e.g., 0.11.0)
• Removal: Occurs in major versions (e.g., 1.0.0) or specified minor versions
• Notice period: At least one minor version before removal

We aim to provide clear migration paths and adequate notice for all deprecations.