Add local Letta server setup documentation and configuration

[cpfiffer]

Summary

Adds documentation and configuration files for running a local Letta server on boot, enabling the Subconscious agent to operate against self-hosted infrastructure instead of the hosted api.letta.com.

Changes

• docs/REMOTE_SERVER.md: Comprehensive guide for setting up a local Letta server

  • macOS Launch Agent setup instructions
  • Linux systemd service configuration
  • Troubleshooting section
  • Multi-environment usage patterns (hosted vs local)

• scripts/com.letta.server.plist: macOS Launch Agent configuration

  • Starts letta server on boot
  • Automatic restart on failure
  • Log output to /tmp/letta-server.log and /tmp/letta-server.error.log

• README.md: Link to new documentation from LETTA_BASE_URL configuration

Use Cases

• Data privacy: Keep all agent data on local machine
• Offline capability: Work without internet (with local models)
• Custom models: Use any models configured on local server (Ollama, etc.)
• No API costs: Avoid API costs when using local models

Testing Needed

• Verify plist path needs updating for different install locations (/opt/homebrew/bin/letta on Apple Silicon)
• Test Launch Agent actually starts server on boot
• Confirm health check endpoint (http://localhost:8283/v1/health)
• Verify hooks connect to local server when LETTA_BASE_URL is set
• Test Linux systemd service configuration

Questions

• Should the plist path be templated or dynamically detected?
• Is port 8283 the correct default for letta server?
• Any additional troubleshooting scenarios to document?

Related to #27 (bidirectional conversation issue mentions remote server needs)

[cpfiffer]

Nonsense PR from an agent, closing.