Local LLM

serviceupdated 2026-10-09edit the YAML

The self-hosted model server on the home server, metered by a pass-through proxy (the ollama-exporter) for time to first token, decode speed, token counts and resident models. The only service with defined SLOs and burn-rate alerts.

Owner: Stefan Coetzee
Tier: Tier 3 Tooling a reader does not meet directly. Waits for the next working session.
Lifecycle: Production
System: Research tooling
Repository: uncovertechtalent/agent-observability
Dashboard: Local LLM (Grafana)
Current state: on the status page

Docs

Runbooks

No runbook yet.

SRE principles applied

Each principle, how this platform applies it and where it falls short: Principles in practice.

Service level objectives

SLO	Target	Good event
Availability	99%	A request without an upstream error
Latency	95%	A streamed chat request with a first token within 4 s

Defined and measured in Prometheus; see Alerts and SLOs.

Dependencies

Depends on

Observability stack

Outside the catalog: Ollama.

Used by

Agent sessions

Service catalog · built from services/local-llm.yml by build_inside.py · services.json