# ElizaOS Developer Update - October 3, 2025

## 1. Core Framework

The ElizaOS framework has undergone significant architectural improvements this week focused on message handling and core stability:

- **Agent Runtime Improvements**: PR #6030 introduces a platform-agnostic mention detection system through the new `MentionContext` interface in the `Content` type. This enables universal structure for platform mention metadata (isMention, isReply, isThread) across all communication channels. ([PR #6030](https://github.com/elizaOS/eliza/pull/6030))

- **Message Bus Enhancements**: The core message bus implementation is undergoing active development with 0xbbjoker leading efforts to complete core development and testing in browser environments. PRs are being structured to avoid breaking changes, with redundant tables being removed to streamline the architecture.

- **Bootstrap Plugin Enhancement**: Fixed a critical issue where `shouldRespondProvider` was not properly registered or exported, restoring crucial message processing functionality that affects all agent response behaviors. ([PR #6024](https://github.com/elizaOS/eliza/pull/6024))

- **Dependency Updates**: Major dependency versions have been updated across the monorepo, including TypeScript, ESLint, Vite, Puppeteer, dotenv, langchain, uuid, and Sentry libraries to ensure framework stability. ([PR #6025](https://github.com/elizaOS/eliza/pull/6025))

## 2. New Features

### Eliza Cloud Platform

```typescript
// Upcoming Eliza Cloud system will provide infrastructure 
// for running AI agents in a managed environment
import { CloudAgent } from '@elizaos/cloud';

// Create a cloud-hosted agent with automatic scaling
const agent = new CloudAgent({
  name: 'CustomerSupport',
  model: 'anthropic/claude-3-haiku',
  memory: {
    persistent: true,
    encryption: 'tee'  // Uses Trusted Execution Environment
  }
});

// Register with the cloud platform for monitoring and billing
await agent.register();
```

ElizaOS is developing Eliza Cloud, a comprehensive system for running AI agents targeted for release by end of year. The system will feature:

- **Auto-scaling infrastructure** for agent hosting
- **TEE-based security** for sensitive operations, with the team choosing TDX over SGX implementations for better security and developer experience
- **Integrated monitoring and billing** for enterprise deployments

### Message Context Handling

The Discord plugin is being enhanced with improved context handling through new boolean flags:

```typescript
// New MentionContext interface in core/src/types/primitives.ts
interface MentionContext {
  isMention: boolean;  // Direct @mention of the agent
  isReply: boolean;    // Reply to agent's message
  isThread: boolean;   // Message is in a thread the agent is participating in
}

// Enhanced shouldRespondTemplate in core/src/prompts.ts
// Now has explicit rules for determining response context
const shouldRespondTemplate = `
IMPORTANT RULES FOR RESPONDING:
1. Only respond when someone is talking TO you (not ABOUT you)
2. Respond if someone directly asks you a question
3. Respond if your name appears at the start of a message
...
`;
```

## 3. Bug Fixes

Several critical bugs have been addressed this week:

- **Core Module Exports**: An issue has been identified where the `@os/core` module is missing critical exports including 'logger', 'IAgentRuntime', and 'ProjectAgent', preventing proper TypeScript imports in new projects. This has been logged as Issue #6031 and is currently being investigated. ([Issue #6031](https://github.com/elizaOS/eliza/issues/6031))

- **Security Vulnerability**: A security issue was discovered where unencrypted file contents in `.claude` and `.codex` folders could expose secrets from environment files, making them vulnerable to malicious npm packages. The team is recommending users check these directories for exposed secrets that may need rotation.

- **Venice Plugin**: A fix has been applied to the Venice plugin to comply with recent API changes, resolving compatibility issues with the latest core framework.

- **Schema Transformer**: Fixed a variable declaration in the SQL plugin's schema transformer, changing `let result` to `const result` as the variable is never reassigned, improving code safety. ([PR #6027](https://github.com/elizaOS/eliza/pull/6027))

## 4. API Changes

### MentionContext Interface

A new `MentionContext` interface has been added to the `Content` type in the core framework:

```typescript
// Before - Content only contained text and optional metadata
interface Content {
  text: string;
  metadata?: Record<string, any>;
}

// After - Content now includes structured mention context
interface Content {
  text: string;
  metadata?: Record<string, any>;
  mentionContext?: MentionContext;
}

interface MentionContext {
  isMention: boolean;
  isReply: boolean;
  isThread: boolean;
}
```

This allows plugins to provide platform-specific mention information in a standardized way, enabling the core framework to make more intelligent decisions about when agents should respond.

### shouldBypassShouldRespond Changes

The `shouldBypassShouldRespond` function in the bootstrap plugin has been enhanced to accept an optional `mentionContext` parameter:

```typescript
// Before
function shouldBypassShouldRespond(content: Content, source: string): boolean {
  // Simple checks based on content text and source
}

// After
function shouldBypassShouldRespond(
  content: Content, 
  source: string
): boolean {
  // Fast-path optimization using mentionContext when available
  if (content.mentionContext) {
    if (content.mentionContext.isMention || content.mentionContext.isReply) {
      return true; // Skip LLM call for obvious mentions
    }
  }
  
  // Fall back to existing logic for other cases
}
```

This implementation creates a fast-path optimization that skips expensive LLM calls for platform-native mentions, saving tokens and improving response time.

## 5. Social Media Integrations

The Discord plugin is receiving significant updates to improve message handling and responsiveness:

- PR #6030 introduces boolean flags for better message context handling (isMention, isReply, isThread)
- Enhanced `shouldRespond` logic now differentiates between someone talking TO the agent versus ABOUT the agent
- The plugin will be configurable via environment variables including:
  - `allowedChannelIds`
  - `shouldIgnoreDirectMessages`
  - `shouldRespondOnlyToMentions`

Additionally, DegenAI streaming capabilities on platforms like pumpfun are being developed, with future plans for expanded livestreaming features.

## 6. Model Provider Updates

OpenRouter has announced two important updates:

1. A new Performance tab for visualizing provider performance across models
2. The end of the free period for the Grok-4-Fast model

GLM 4.6 is now available on OpenRouter with improved capabilities, offering a 200k context length and 128k max tokens, making it suitable for complex, multi-turn conversations.

## 7. Breaking Changes

### Token Migration (AI16Z to ElizaOS)

```
# Migration Process (Upcoming)
1. Transfer AI16Z tokens from exchanges to personal wallet
   (Phantom, Solflare, or Metamask with Solana support)
2. Connect wallet to migration portal (to be announced)
3. Execute migration at 1:10 ratio (1 AI16Z → 10 ElizaOS)
4. Note: 20% dilution will be applied for development funding
```

The ElizaOS ecosystem is undergoing a significant token migration from AI16Z to ElizaOS tokens:

- **Migration Ratio**: 1:10 (each AI16Z token becomes 10 ElizaOS tokens)
- **Dilution**: 20% increase in supply (10% for team with vesting, 10% for liquidity)
- **Total Supply**: Approximately 12-13.2 billion tokens after migration
- **Process**: Users must manually transfer tokens from exchanges to personal wallets
- **Timeline**: No firm date announced yet, but appears imminent

The team is developing a migration portal with clear documentation to guide users through the process. Token buybacks are planned once Eliza Cloud generates revenue to mitigate dilution impacts.

Once migration is complete, the project will focus on developing L2/L3 chain with productive asset tokenomics featuring issuance/burn mechanics for agent building and usage.