Skip to content

Add Runtime observability and lightweight PostgreSQL history - #30

Merged
SaladDay merged 67 commits into
mainfrom
codex/runtime-observability-microsandbox
Sep 23, 2026
Merged

SaladDay merged 67 commits into
mainfrom
codex/runtime-observability-microsandbox

Conversation

@sam2tom

@sam2tom sam2tom commented Sep 22, 2026 •

Copy link
Copy Markdown
Collaborator

Core Web now restores recent Runtime CPU, memory and canonical Session token history after refresh or Core restart, using the existing Core PostgreSQL database. No ClickHouse or Collector deployment is required.

  • Add provider-neutral, read-only Docker and microsandbox observations through authenticated Core APIs; retain unknown values and safe failure responses.
  • Sample through the execution lease owner, persist sanitized periodic records asynchronously, and bound retention, queries and cleanup. Optional OTLP export uses an independent queue.
  • Keep the administrator console lightweight: current observations, retained resource/token charts and basic Runtime details. Compute uptime stays current/Live; CPU deltas reset across compute restarts and counter regressions.
  • Remove obsolete ClickHouse code, dependencies, configuration and tests. Retain canonical Session Usage and Core lifecycle authority without compatibility scaffolding.

Validation:

  • Full server make check, including real dedicated PostgreSQL, generated SQL consistency, Web/browser, native helper, Claude, MiniMax and Rust checks.
  • OpenAPI regeneration and focused race tests.
  • Real Codex/Kimi execution on isolated Docker and microsandbox deployments; native CPU/memory samples, exact microsandbox timestamp/counter matching, PostgreSQL persistence, fixed-range history across Core restart, tenant isolation and sanitized backend failures.
  • Independent GPT-6 Astra high review of the complete PR. Its two reproducible chart findings were corrected and self-reviewed with focused regression/browser coverage under the user's risk-based review policy.

History is best-effort operational telemetry, not execution, billing or Usage authority. Retention is seven days; queries span at most 24 hours. This batch qualifies Docker and microsandbox telemetry, not additional providers. A failed test namespace permission setup left a never-created allocation cleanup record; no compute remains and that existing lifecycle follow-up is recorded separately.

Closes #19.
Closes #27.
Closes #38.
Supersedes #15.

sam added 4 commits September 23, 2026 17:31
…bility-microsandbox

# Conflicts:
#	apps/web/e2e/agents-lifecycle.spec.ts
#	apps/web/src/App.tsx
#	packages/agents-client/src/types.ts
#	services/agents-api/cmd/server/main.go
#	services/agents-api/internal/api/handler.go
sam and others added 21 commits September 23, 2026 19:16
…osandbox' into codex/runtime-observability-clickhouse

# Conflicts:
#	apps/web/src/features/dashboard/RuntimeTrendCharts.tsx
#	contracts/agents-api/runtime-history-api.md
#	packages/agents-client/src/runtime-history-projection.ts
#	services/agents-api/internal/api/runtime_history_test.go
#	services/agents-api/internal/runtimehistory/clickhousereader/reader_test.go
@SaladDay SaladDay changed the title Add Runtime observability and durable ClickHouse history Add Runtime observability and lightweight PostgreSQL history Sep 23, 2026
@SaladDay
SaladDay merged commit e16c421 into main Sep 23, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

2 participants