gpt/TOML_MIGRATION.md

TOML Configuration Migration Guide

Overview

The ai.gpt Rust implementation (aigpt-rs) now uses TOML format for configuration instead of JSON. This provides better readability and is more idiomatic for Rust applications.

Configuration Location

The configuration file is stored at:

• macOS: ~/Library/Application Support/syui/ai/gpt/config.toml
• Linux: ~/.config/syui/ai/gpt/config.toml
• Windows: %APPDATA%\syui\ai\gpt\config.toml

Automatic Migration

When you run the Rust implementation for the first time, it will automatically:

1. Check if config.toml exists
2. If not, look for config.json in various locations:
   • ../config.json (relative to aigpt-rs directory)
   • config.json (current directory)
   • gpt/config.json (from project root)
   • /Users/syui/ai/ai/gpt/config.json (absolute path)
3. If found, automatically convert the JSON to TOML format
4. Save the converted configuration to the appropriate location

TOML Configuration Structure

# Default AI provider
default_provider = "openai"

# Provider configurations
[providers.openai]
default_model = "gpt-4o-mini"
api_key = "your-api-key-here"  # Optional, can use OPENAI_API_KEY env var
system_prompt = """
Multi-line system prompt
goes here
"""

[providers.ollama]
default_model = "qwen3"
host = "http://127.0.0.1:11434"

# AT Protocol configuration (optional)
[atproto]
host = "https://bsky.social"
handle = "your-handle.bsky.social"  # Optional
password = "your-app-password"      # Optional

# MCP (Model Context Protocol) configuration
[mcp]
enabled = true
auto_detect = true

# MCP Server definitions
[mcp.servers.ai_gpt]
base_url = "http://localhost:8001"
name = "ai.gpt MCP Server"
timeout = 10.0

# MCP endpoints
[mcp.servers.ai_gpt.endpoints]
get_memories = "/get_memories"
search_memories = "/search_memories"
# ... other endpoints ...

Manual Migration

If automatic migration doesn't work, you can manually convert your config.json:

1. Copy the example configuration from gpt/config.toml.example
2. Fill in your specific values from config.json
3. Save it to the configuration location mentioned above

Testing Configuration

To test if your configuration is working:

cd gpt/aigpt-rs
cargo run --bin test-config

This will show:

• Loaded configuration values
• Available providers
• MCP and ATProto settings
• Configuration file path

Differences from JSON

Key differences in TOML format:

• Multi-line strings use triple quotes (""")
• Comments start with #
• Tables (objects) use [table.name] syntax
• Arrays of tables use [[array.name]] syntax
• More readable for configuration files

Backward Compatibility

The Python implementation still uses JSON format. Both implementations can coexist:

• Python: Uses config.json
• Rust: Uses config.toml (with automatic migration from JSON)

The Rust implementation will only perform the migration once. After config.toml is created, it will use that file exclusively.