Skip to content

v0.1.1 — Your Rails app is already an agent runtime

Choose a tag to compare

View all tags
@danielstpaul danielstpaul released this 16 Jul 07:42
v0.1.1
a0e43f7

Silas turns the Rails app you already run into a durable AI-agent runtime. Active Job Continuations, Solid Queue, and one Postgres ledger table make every turn survive kill -9 and resume from the last completed step — with a live operator inbox at /silas/inbox and park-at-zero human-in-the-loop approvals.

First public release.

Highlights

  • Durable loop on Active Job Continuations + Solid Queue — chaos-gated: 100/100 kill -9 cycles, zero duplicate effects, byte-identical replay, on SQLite and Postgres.
  • Exactly-once tool execution via a transactional ledger (at-most-once with in-doubt→human resolution for external effects).
  • Park-at-zero human-in-the-loop approvals — resumable from the inbox, Slack buttons, or signed email links.
  • app/agent/ directory convention: tools (signature = schema), skills, instructions, schedules, channels, subagents, connections.
  • Two engines: :ruby_llm (any provider) and :agent_sdk (a claude -p subprocess; API-key auth).
  • Mountable operator inbox with live Turbo trace + per-session cost accounting; deny-by-default.
  • Self-hosted deploy with Kamal.

Honest scope

v0.1, one maintainer, zero external users; durability proven by an in-repo chaos harness, not yet production traffic. Scoped to trusted-code agents — the Docker sandbox is an interim seam, weaker than a microVM. See Why Silas and Silas vs eve.

Assets 2
Loading