# elizaOS Development Discord - 2025-05-16

## Overall Discussion Highlights

### LinkedIn Automation
- A user inquired about LinkedIn automation tools similar to ElizaOS, specifically for commenting on others' posts with personality features
- There may have been a LinkedIn client in version 0.x, though this wasn't confirmed
- Browser-based automation tools were suggested as an alternative, though the user found them less reliable than desired

### Token Limitation & API Configuration
- Extensive troubleshooting session regarding token limitations when using OpenRouter's free tier
- User encountered persistent issues with the system defaulting to 8192 tokens despite configuration attempts to limit to 512 tokens
- Multiple model alternatives were suggested when original choices returned errors
- The discussion highlighted challenges with free tier API usage and proper configuration of character JSON files

### Local Knowledge Processing
- Clarification that the folder2knowledge command works with the default mxbai-embed-large model through Ollama locally
- No need for OpenAI or Anthropic API keys for this functionality

## Key Questions & Answers

**Q: Does anyone know of something like ElizaOS but for LinkedIn?**  
A: There might have been a LinkedIn client in version 0.x. Browser-use tools could accomplish this, though they may lack personality features.

**Q: If I want to use the folder2knowledge command to create a character, do I need an OpenAI or Anthropic API?**  
A: No, the folder2knowledge command works with the default mxbai-embed-large model through Ollama locally.

**Q: Can I limit max_tokens so my AI doesn't fail?**  
A: Yes, you can set max_tokens in your character config or use the RESPONSE_LENGTH env var as a global limit.

**Q: Is there any free AI API keys that you recommend using for my agent?**  
A: Consider Mistral's free tier, Ollama's local models, or OpenRouter's free quotas, but be cautious of reliability and data privacy with free options.

## Community Help & Collaboration

### OpenRouter Configuration Troubleshooting
- **Helper:** Ruby
- **Helpee:** Ramiroo
- **Context:** User struggling with token limitations and model selection on OpenRouter's free tier
- **Resolution:** Ruby provided multiple configuration approaches including:
  - Adding max_tokens settings in both character JSON and .env file
  - Suggesting alternative models when original choices returned errors
  - Explaining proper JSON structure for Twitter API configuration
  - Confirming correct syntax for character configuration

### LinkedIn Automation Inquiry
- **Helper:** sayonara
- **Helpee:** aith
- **Context:** Looking for LinkedIn automation tool for commenting on posts with personality features
- **Resolution:** Partial - suggested browser-use tools and possible existence of a LinkedIn client in version 0.x

## Action Items

### Technical
- Verify if useEnvParameters is set to true in character config to ensure environment variables are loaded (Mentioned by Ruby)
- Try adding max_tokens in both modelSettings and parameters fields to override default token limits (Mentioned by Ruby)
- Consider implementing rate limiting for free API usage (Mentioned by Ruby)
- Reconsider crypto evangelist prompt that might trigger content filters (Mentioned by Ruby)

### Documentation
- Document the correct model names for OpenRouter's free tier as they change over time (Mentioned by Ruby)

### Feature Requests
- LinkedIn automation tool with personality features similar to Eliza characters for commenting on LinkedIn posts (Mentioned by aith)