AI and Agents•v2.0.0•1 min read
Valstorm Agent (vsagent) Overview
Universal overview of the vsagent autonomous execution engine: REPL environment, multi-model tier routing, and tool capabilities.
The Valstorm Agent (vsagent) is an autonomous terminal agent and execution kernel designed for software engineers, operations teams, and executives. It bridges natural language instructions with low-level operating system tools, database queries, knowledge vaults, and multi-model AI routing.
1. Core Runtime Highlights
- Cognitive Tier Routing: Seamlessly dispatches tasks across Gemini 3.7 Flash, Claude 3.7 Sonnet, OpenAI o3-mini, and local Ollama instances.
- In-Memory LSP Interceptor: Catches Python syntax errors, TypeScript compile failures, and type mismatches instantly in memory before committing to disk.
- Transactional State Engine: Backed by SQLite FTS5 conversation persistence and Git working-tree snapshot rollbacks (
/undo). - Multi-Tenant Platform Kernel: Direct execution tools for database mutations (
valstorm_record_cud), SQL queries (valstorm_sql_query), and dense vector VFS document search (valstorm_vfs_search). - 170+ Indexed Procedural Skills: Structured SOP markdown guides injected dynamically to standardize multi-step operations.
2. CLI Quickstart
Install and run vsagent directly in your terminal:
BASH
# Launch interactive REPL session
vsagent chat
# Run single-prompt execution mode
vsagent chat -p "Run the test suite and fix any failing auth tests"
# Select a specialized agent profile
vsagent chat --profile software-developer