# ElizaOS Developer Update - 2025-10-16

## 1. Core Framework

The ElizaOS framework has undergone significant architectural changes this week, with a complete migration of the CLI deployment system from traditional Docker image builds to a modern bootstrapper architecture. PR [#6058](https://github.com/elizaOS/eliza/pull/6058) introduces a lightweight artifact-based deployment approach that significantly improves deployment speed and reduces resource consumption:

- **New Bootstrapper Architecture**: Creates lightweight tar.gz artifacts (typically <50MB vs 500MB+ Docker images)
- **Streamlined Deployment**: Reduced deployment time from 5-10 minutes to 30-60 seconds
- **Eigen TEE Integration**: Added a wrapper for Trusted Execution Environment deployments in PR [#6065](https://github.com/elizaos/eliza/pull/6065)

Core functionality was extended with a new digital twin implementation as proposed by Odilitime, enabling more efficient context management by tracking relationship-relevant information between agents:

```typescript
// Example implementation for digital twin tracking
const twinManager = new DigitalTwinManager();
twinManager.onTwinChange((twinId, changes) => {
  // Background process updates relationship data
  const relationshipData = precompileRelationshipData(agent, twin);
  stateCache.set(`relationship:${agentId}:${twinId}`, relationshipData);
});
```

## 2. New Features

### GenerateText() API

A new Promise-based `generateText()` API has been implemented in PR [#6062](https://github.com/elizaos/eliza/pull/6062) to provide a more direct method for simple text generation:

```typescript
// Example usage of new generateText() API
import { ElizaOS } from '@elizaos/core';

const eliza = new ElizaOS();
const agent = await eliza.loadAgent('./my-agent');

// Simple Promise-based text generation
const result = await agent.generateText("Explain quantum computing");
console.log(result);
```

### Digital Twin System

A new digital twin implementation has been proposed for more efficient context management:

```typescript
// Creating a digital twin from RAG data
const twinEvaluator = new CharacterEvaluator({
  ragSource: conversationHistory,
  outputFormat: 'character-file'
});

// Using relationship data in conversations
const relationshipContext = await getRelationshipContext(agent.id, twin.id);
const response = await agent.chat(message, {
  contextReduction: true,
  relationshipContext
});
```

## 3. Bug Fixes

Several critical issues were resolved this week:

- **Plugin-EVM Transaction Execution**: User "ole" reported issues with approval transactions failing silently when using `@elizaos/plugin-evm`. The team identified that while token sending worked, more complex operations like checking balances or executing approval transactions failed. A fix is being implemented to support multi-step execution for complex transactions.

- **Chain of Thought Disabled in Anthropic Plugin**: Odilitime discovered that Chain of Thought (CoT) is disabled in `plugin-anthropic`, potentially making the model "smarter and more expensive." This unexpected configuration affects how the Anthropic model processes requests.

- **Docker Image Building**: User TH3H4RM1N4T0R reported issues with environment variables and volume mounting when building Eliza Docker images. Documentation is being created to better explain the proper approach.

## 4. API Changes

The core API has been extended with several important additions:

- **MessageService Interface**: A new interface has been introduced for more consistent message handling across plugins.

- **Database Pagination**: Added `limit` and `offset` parameters to the `getMemories` function, introducing database-level pagination for more efficient memory retrieval with large memory stores.

- **State Cache Exposure**: The runtime's state cache is now exposed and accessible to plugins, enabling features like the digital twin relationship caching.

## 5. Social Media Integrations

### Instagram Plugin

The Instagram plugin is now officially available and has been published to GitHub:
```bash
# Install the Instagram plugin
npm install @elizaos/plugin-instagram

# Add to your agent configuration
plugins:
  - name: '@elizaos/plugin-instagram'
    config:
      username: 'your_username'
      password: 'your_password'
```
Repository: https://github.com/elizaos-plugins/plugin-instagram

### Farcaster Integration

As part of the upcoming Jeju L3 blockchain launch (announced on October 14th), a Farcaster client is being developed to drive speculation and volume. This integration will allow agents to interact with the Farcaster social protocol.

## 6. Model Provider Updates

### Anthropic

Odilitime discovered that Chain of Thought (CoT) is disabled in the Anthropic plugin, which affects both cost and reasoning capabilities. This means the model may provide more direct, potentially higher-quality responses but at a higher cost. The team is evaluating whether to make this configurable.

### Eigencloud Integration

The proof-of-concept for Eigencloud, a deterministic OpenAI-compliant LLM inference endpoint, was formally concluded and closed (Issue [#5768](https://github.com/elizaos/eliza/issues/5768)). This service offers deterministic responses, ensuring the same output for identical inputs with controlled seeds.

## 7. Breaking Changes

Several important V1 to V2 migration issues have been identified:

### CLI Deployment System

The migration to the new bootstrapper-based deployment system has removed several CLI options:

```bash
# Old (no longer works)
elizaos deploy --use-docker --tag my-image:v1

# New (default behavior)
elizaos deploy

# With existing artifact
elizaos deploy --skip-artifact --artifact-path ./dist/artifact.tar.gz
```

### Token Migration

A token migration from AI16z to ElizaOS is scheduled for October 21st. Important notes:

- ElizaOS tokens will not be available for purchase until the migration date
- Current AI16z token holders can still purchase tokens before migration
- The migration is separate from options trading, which requires traders to manage their positions independently
- 15% SAFT in new tokenomics will not be available to the general community ("it's closed")

Developers using tokens for API access or building token-integrated applications should prepare for the switchover to avoid service disruptions.