# elizaOS Development Discord - 2025-05-08

## Overall Discussion Highlights

### Multi-Agent Architecture
- Ruby confirmed that multiple agents can run on a single Postgres database
- The platform handles memory isolation automatically through namespacing in the database schema
- Each agent requires a unique name/ID in their configuration to ensure proper separation

### Model Configuration
- Detailed discussion on setting up tiered models (small, medium, large) in agent configurations
- Example provided showing how to mix Claude and GPT models for different complexity levels:
  - 'small': 'claude-2.1'
  - 'medium': 'claude-2.1'
  - 'large': 'claude-3'
- Clarification on adding Anthropic API keys to .env files for Claude model usage

### Package Manager Compatibility
- Users experienced module resolution issues when mixing npm and Bun
- Problem occurred specifically when CLI was installed with npm but plugins were being installed with Bun
- Two solutions were recommended:
  1. Set `PACKAGE_MANAGER=npm` in the .env file
  2. Install the CLI with Bun instead of npm using `bun install -g @elizaos/cli@beta`

### Token Clarification
- Odilitime clarified that "EDDY" and "ELI5" are characters in the code but not official tokens
- Only `$ai16z` and `$degenai` are official tokens in the ecosystem

### Community Contributions
- User "lost" worked on adding an Indonesian translation (README_IND.md) to the project
- Ruby provided detailed guidance on proper PR submission process
- Discussion highlighted the importance of targeting the correct branch (develop vs. main)

## Key Questions & Answers

**Q: Is it ok to run multiple agents off the same postgres db? Does the platform handle storing all of their memories separately?**  
A: Yes, you can run multiple agents on one postgres db. The platform handles memory isolation by default - each agent gets their own namespace in the db schema. Just make sure you give each agent a unique name/id in their config.

**Q: How do I setup my .env for anthropic models?**  
A: Add `ANTHROPIC_API_KEY=your_key_here` in your .env file. The platform will automatically detect and use it when you specify claude models in your agent config.

**Q: How can I define small, large, etc. models to use different ones in the agent config?**  
A: In your agent's config.json, under the 'models' section, you can specify different model tiers like: 'small': 'claude-2.1', 'medium': 'claude-2.1', 'large': 'claude-3'.

**Q: Why do plugins fail to load with global CLI install but work with npx?**  
A: It's a path resolution issue. When you install the CLI globally with npm, it expects plugins in the npm global path, but the project is using bun's module resolution.

**Q: Is EDDY the V2 agent and ELI5 going to be "a character"?**  
A: They are characters in the code, yes. Are the tokens related? No. The only official tokens are $ai16z and $degenai.

**Q: How can I put a Readme file for Indonesian translation in the repository?**  
A: Create a new file called README_IND.md in the root directory of your fork, paste in your translation, commit and push.

**Q: Should I worry about integration-tests and GitHub Actions failures for documentation changes?**  
A: No, those failures are expected for doc changes and only relevant for code changes, not documentation updates.

## Community Help & Collaboration

### Multi-Agent Setup Guidance
Ruby provided comprehensive assistance to Blaze Leclairt on setting up multiple agents with a shared database, explaining memory isolation mechanisms and offering configuration examples for different model tiers.

### Package Manager Troubleshooting
Ruby helped kandizzy resolve module resolution issues between npm and Bun, offering two alternative solutions to maintain consistency in the development environment.

### Translation Contribution Support
Ruby guided user "lost" through the entire process of contributing an Indonesian translation:
- Creating the translation file in the correct location
- Adding proper language badges
- Formatting links correctly
- Targeting the correct branch for the PR
- Understanding which build errors could be safely ignored for documentation changes

SpartanDev also offered troubleshooting steps for build errors, suggesting cache clearing and clean installation approaches.

### Token Clarification
Odilitime helped Sidtoshi Vision understand the distinction between characters in the code and official tokens in the ecosystem.

## Action Items

### Technical
- Use unique AGENT_ID in each agent's config when running multiple agents on a single database (Mentioned by Ruby)
- Set PACKAGE_MANAGER=npm in .env file when CLI is installed with npm (Mentioned by Ruby)
- Install CLI with Bun instead of npm for consistency using `bun install -g @elizaos/cli@beta` (Mentioned by Ruby)
- Fix PR to target develop branch instead of main for contributions (Mentioned by Ruby)

### Documentation
- Add Indonesian translation README (README_IND.md) to the repository (Mentioned by lost)
- Ensure all documentation links in README_IND.md point to the correct locations (Mentioned by Ruby)
- Add language badge to Indonesian README (Mentioned by Ruby)
- Update documentation regarding CLI installation with Bun as an alternative to npm (Mentioned by sayonara)