Self-hosted · Local inference by default
From scanner evidence to reviewed reports today, then validation, collaboration, and human-supervised assessment on the same foundation you run yourself.
Reads evidence from
Workflow
Nullsid keeps mechanical execution, model interpretation, and human judgment as separate, visible stages instead of blurring them into “autonomy”.
Automation
Normalize scanner exports and plain notes into one findings workspace, with each claim tied back to its source file.
Raw evidence structured state
Local model
Correlate duplicates, suggest CVSS metrics, and draft descriptions, remediation, and summaries, with provenance attached to every field.
Structured state reviewable proposal
Operator
Trace any sentence to the evidence behind it, edit every field, then export a client-ready DOCX or PDF on your own template.
Reviewable proposal signed deliverable
Product
Reporting is the first operator loop, not the ceiling. Evidence, assets, severity, provenance, remediation, status, and audit history stay connected as new workflows land.
Every normalized finding carries the parser, host, raw record, and the operator decisions applied to it. Nothing becomes an unsourced assertion in a client deliverable.
AI-generated text is marked, editable, and traceable to the evidence that produced it.
Docker Compose, your host, your storage. No vendor tenancy holding client engagement data.
Export to the formats and templates you already use, without a manual reformat pass.
Model routing
Nullsid ships pointed at Ollama. Cloud providers stay off until an engagement permits them and you supply your own key, so the default path is the private one.
Self-hosted workspace Ollama endpoint
The workspace and the reporting pipeline live in infrastructure you control.
Model requests go to your configured Ollama endpoint, on the box or on your network.
Generated claims stay editable until a human signs them into the deliverable.
Workspace Ollama Reviewed output
Roadmap
Each stage reuses the same evidence, findings, provenance, and inference layer, so the privacy model does not change as the surface grows.
Import, normalize, draft, review, and export client-ready deliverables.
Scanner output signed-off report
Ground claims in raw evidence, flag likely false positives, surface review priorities.
Evidence verified finding
Ownership, comments, approval states, and reusable delivery standards.
Finding team decision
Human-supervised workflows that operate only inside an explicitly authorized scope.
Approved action auditable result
FAQ
Anything else: [email protected].
Not on the default configuration. Nullsid runs on your infrastructure and sends model requests to the Ollama endpoint you configure. A cloud provider is only contacted if you explicitly enable one and supply a key.
Anything your Ollama instance serves. Larger models produce better narrative sections; the structured work (parsing, dedup, scoring, export) does not depend on the model.
Generated fields are marked with their provenance and remain editable. You can trace a sentence back to the finding and the raw scanner record behind it before signing off.
A Docker host and an Ollama endpoint. Storage stays local, and the workspace is API-first so it can sit behind your existing authentication.
No. It removes the mechanical hours around the engagement. Severity, scope, and every claim in the deliverable stay a human decision.
The report workspace is in active development. Join the list and you get one message when operator access opens. No drip campaign.
Early access
Join the list for the self-hosted reporting workspace: Ollama as the default provider, BYO-key cloud when you decide otherwise.
