HivemindOS manual
SRE Incident Investigations
When something in your fleet breaks and can’t be fixed automatically, HivemindOS saves a redacted record of the incident and can ask an optional local tool to work out the root cause. Fleet Watchdog uses this only after its normal health checks and restart attempts have failed.
Before you start
Incidents are always recorded, with or without the root-cause tool. To get root-cause reports, install the optional OpenSRE sidecar during interactive setup. OpenSRE is a third-party Apache-2.0 project. By default it uses a local Ollama model.
What gets recorded
Each incident includes a short summary and severity, the type of service that failed, symptoms and checks, any fixes already tried, and a timeline.
Before anything is saved or sent to OpenSRE, HivemindOS removes credentials and secrets, masks IP addresses, replaces home-folder paths, hashes machine names, and limits the size of the data.
Investigate an incident
Ask an agent to investigate an incident or check investigation status. Agents can list and read past reports, retry a failed analysis, or submit a new incident.
When a diagnosis finishes, you get an alert with the root cause, confidence, and suggested fixes, marked for review. Nothing is acted on or saved to Shared Brain Memory automatically.
If no usable model is available, HivemindOS reports the tool as degraded and keeps the incident for a later retry instead of guessing.
What stays in your control
OpenSRE only investigates; it never acts.
- It runs only on your machine and accepts only its reviewed, pinned version.
- Telemetry, prompt logging, and history are turned off.
- It doesn’t get your Shared Env credentials.
- Its suggested fixes are just suggestions. It can’t restart services, change infrastructure, or spend money. Any follow-up goes through HivemindOS’s normal permissions and approvals.
Storage and removal
Incident records stay on your machine under ~/.hivemindos/ops/incidents. Removing OpenSRE doesn’t delete them. The uninstaller offers to remove the service and its runtime separately, and keeps your incident history unless you delete it yourself.