Oncall

Hey there, ready to ship calm through the chaos today?

Total
0
Active
0
Solved
0
Failed
0

All Tickets

Recently solved

Resolved tickets with the investigation chat stored on each ticket. Select a row for full analysis and Slack reply.

— Oncall Ops —
Helps you respond and ship fixes through
a trusted Oncall Bot

Step in when something breaks. Hand off cleanly when your shift ends. Paging, runbooks, and postmortems — in one calm place.

Allo Health Oncall Oncall Bot

What do you want to know?

Oncall skill

Three-layer memory

Structured context the agent uses before investigations. Maintained outside this UI (pipelines, incidents, merges). This tab is informational only—it does not run learning or capture.

Layer 1 — Digest & retrieval

Compressed incident patterns: symptoms, how they were recognized, and proven resolutions. Seeded from historical incidents so new issues can be matched quickly.

  • layer1-digest.json — digest records (pattern → what to check → resolution).
  • tag-index.json — tag and keyword routing into those records (hundreds of index entries).

In Claude Code, check Layer 1 before improvising: if the issue matches a digest entry, follow that path.

Layer 2 — Working memory

Short-lived investigation state: current hypothesis, queries executed, entity IDs in play. Lives in-session (scratch files, message context) rather than as one large static artifact.

Layer 3 — Deep references

Authoritative material loaded when Layer 1 is not enough: decision trees, SOPs, service maps, SQL templates, Slack style guides—typically under references/, config/, and nested skill folders.

Paths in the repo

Relative to the oncall skill root:

.claude/memory/layer1-digest.json .claude/memory/tag-index.json

Bootstrap stats (for example incidents merged into the digest and tag-index size) change on the branch that owns memory; use git history for the exact snapshot.