Offline Runbook Automation With TinyLM and smoltext

The Worst Time to Lose Your Tools

Outages are exactly when teams most need their tooling and most often lose it. If your runbook assistant depends on a cloud API, a network incident can take it down with everything else. The SPRAPP suite has an answer: TinyLM runs entirely on-device.

On-Device Runbooks With TinyLM

TinyLM's eeny and meeny models run locally, so an on-call engineer can query and update runbooks with no connectivity. During a network partition, the assistant that helps you fix the network is not itself offline. Drafts are a starting point the engineer verifies before acting.

Keeping Runbooks Compact With smoltext

Runbooks and their associated metadata are full of short strings — step IDs, status labels, command snippets, tags. smoltext compresses these short strings efficiently, where gzip wastes bytes on small-payload overhead. A compact runbook store syncs faster when connectivity returns and costs less to keep on every node.

Screening Anything That Comes From Outside

When connectivity is restored and external alert content flows back in, Sprappy Filter scores it across its 25 categories before any automated responder acts on it. An automated remediation should never be steered by an untrusted alert body.

Postmortem Reasoning With Panel

After the incident, SPRAPP Panel helps the team reason about cause and prevention. Asking a panel once and reading where models converge gives a stronger root-cause hypothesis than any single model, and divergence flags ambiguous evidence.

A Resilient Loop

TinyLM keeps the team working offline, smoltext keeps the runbook lean, Filter guards inbound alerts, and Panel powers the postmortem. The pieces are independent, so each can be adopted alone.

Start Here

Put TinyLM on-device runbook access on your on-call laptops first — it is the piece that pays off precisely when everything else is down.