Soma
AI orchestration layer that routes development work across multiple LLM providers.
Si · Silicon
Soma is the intelligence layer. It orchestrates multiple AI providers (Claude, GPT, DeepSeek, local models) into workflows that automate development tasks while respecting cost and rate-limit budgets.
Coming soon
Documentation for Soma is in active migration. The first sections to land will cover the workflow engine, provider selection, and the budget system.
Roadmap for this section
- Workflow definitions and the YAML schema
- Providers and agents (Claude, DeepSeek, GPT, Claude Code, Qwen Code)
- Budgets and rate-limit awareness
- Git-aware context injection
- Workspace integration