AgentOS productization blueprint¶
Scope¶
This blueprint defines how to package the AgentOS + template engine + omnichannel server into an enterprise-ready offering with deterministic artifacts, repeatable governance controls, and operational handoff guidance.
Deployment modes¶
1) Local-only mode¶
- Execution model: developers and QA teams run
sdetkitinside their repositories. - Control plane: none; no central API required.
- Data path:
.sdetkit/stays in-repo or on local workstation storage. - Best fit: regulated teams with strict source-code locality requirements.
- Operational deliverables:
- standardized
sdetkit agent demo --scenario repo-enterprise-auditsmoke check - local runbooks for template and dashboard generation
- workstation onboarding and deterministic cache policies
2) Server mode¶
- Execution model:
sdetkit agent servehosted in a private network segment. - Control plane: REST channel adapters (Slack/Telegram simulation + webhook ingress).
- Data path: artifact roots mapped to dedicated storage classes (workspace, history, dashboards).
- Best fit: shared platform teams, internal enablement groups, and multi-team rollouts.
- Operational deliverables:
- network boundary diagram and ingress policy
- service account and token management matrix
- retention policy for conversation logs, run history, and audit artifacts
Data ownership and controls¶
- Ownership default: customer owns all generated files, logs, and dashboards.
- Storage boundaries:
- run history:
.sdetkit/agent/history/ - dashboard artifacts:
.sdetkit/agent/workdir/or scenario-specific outputs - templates and packs: repository
templates/automations/ - Controls:
- write allowlists for filesystem actions
- shell allowlists for command execution
- provider selection (
nonefor deterministic offline workflows) - deterministic, atomic artifact writes for dashboard/export outputs
Governance and auditability¶
- Deterministic run records with stable hashes enable replay validation.
- Agent dashboard summarizes:
- run count and success rate
- top templates and top actions
- failed runs with timestamps and hashes
- Export channels:
- dashboard output formats:
json,md,html,csv - history summary export:
sdetkit agent history export --format csv - Recommended governance bundle:
- weekly dashboard snapshot (HTML + Markdown)
- CSV exports attached to audit evidence tickets
- template pack checksum archive per release
Enterprise adoption path¶
Phase 1: Pilot (1-2 squads)¶
- Baseline templates activated.
- Provider pinned to
none. - Dashboard cadence: per pull request and nightly.
Phase 2: Platform alignment¶
- Shared template catalog governance owners assigned.
- Omnichannel ingress and rate-limits enabled in controlled channels.
- Role-based operational runbooks published.
Phase 3: Program rollout¶
- Workspace bootstrap scripted with demo scenario for acceptance testing.
- Dashboard and export artifacts integrated into quality gates.
- Standardized policy exceptions and waiver process formalized.
Packaging examples (agency / micro-SaaS)¶
Agency package: Managed enablement¶
- Deliverables:
- repository bootstrap and policy baseline
- 8 curated templates aligned to client SDLC
- omnichannel server deployment and hardening playbook
- dashboard/report governance SOP and handoff session
- Commercial model: fixed setup fee + monthly operations retainer.
Micro-SaaS package: Managed platform¶
- Deliverables:
- hosted server mode with tenant-isolated workspaces
- template catalog updates and semantic version channel
- scheduled dashboard generation + CSV exports to customer bucket
- quarterly governance evidence pack (runs, failures, controls)
- Commercial model: per-tenant platform fee + usage tier based on run volume.