The source file was a 44-page Ohio Bureau of Criminal Investigation prosecutor summary concerning an officer-involved critical incident in Findlay, Ohio. The exercise used a counterfactual prompt: assume the subject survived, was charged with felonious assault on a peace officer, and the prosecutor was preparing the State’s case-in-chief.

The offline AI system was asked to identify the strongest direct-examination witness and explain why, based only on the BCI report. Its answer selected the officer with the cleanest observation of the act: the officer who saw the subject turn and point the handgun directly at him. The typewriter's redacted output is here.

A separate citation check found no hallucinated support. The quoted lines and page references were verified against the report. Each item held up: the muzzle observation, the limitation in the first officer's account, the second officer's view of the subject's body mechanics, the direct-pointing language, the recovery of the firearm, and the timing witness.

The run took 3 minutes and 25 seconds. The model used roughly 22 percent of its available context.

This work does not require the cloud. It does not fail when the internet fails. It does not meter every question. It sits inside the office, making review faster while preserving human judgment.

The first Labs video put two offline machines against the trial clock. This one is quieter. A 162-page Ohio Bureau of Criminal Investigation prosecutor summary is dropped into VoraSync, and the machine is asked to do something less theatrical and more useful: write the first case note.

The prompt was deliberately constrained. Use only the attached document. Do not guess. Do not decide fault, guilt, justification, or policy compliance. If the report does not clearly support an answer, say so.

The model answer returned exactly that kind of note: case numbers, agencies involved, subject background, event sequence, involved persons, evidence categories, and two limits on interpretation. It identified the federal warrant, the motel-room standoff, the drone being shot down, the injured officer, the nine officers who fired, and the later recovery of the subject inside the room with a firearm. It also surfaced a separate 1981 cold-case homicide link from deep in the background section.

A separate source check verified the model’s answer against the PDF.

VoraSync does not require the cloud. It does not fail when the internet fails. It does not commoditize every question. It sits inside the office, making review faster while preserving human judgment.

VoraSync ran onsite against a public case file and returned case-file answers from the source document. The point of the run was simple: when connection, coverage, or policy makes cloud tools unavailable, investigators and prosecutors still need fast retrieval from the record in front of them.

VoraSync is built for moments when retrieval speed matters: names, exhibits, facts, contacts, inconsistencies, and case-file orientation. This field test shows the system running onsite, disconnected from the internet, against a public-record PDF in a field setting.

• Ran onsite from a laptop in a remote field setting.
• Used a clean public-record PDF.
• Answered urgent case-file questions without internet access.

This is not a model comparison and not a benchmark. It is a field test: no-internet intelligence for case-file orientation when the work has to happen onsite.

In this run, VoraSync processed three public-record PDFs one at a time. Each file produced its own first-pass orientation brief with case identifiers, agencies, subjects, summary, prior-record notes, and open uncertainty — as prompted by humans in control.

For source visibility, the run included an Akron / BCI scene-processing report, a second public-record PDF that is not linked here, and an Avon PD investigative report.

Intake is the first read of a case. The point is to get oriented on what is in the file faster than a manual review can do it, at any point when a new file lands on a desk. The briefs as the model returned them are linked here.

The run is onsite. No internet, no external API. Ongoing cost is electricity. The workflow does not require a frontier-scale cloud model or a long-term contract with a single large provider.

Public records are used here as a demonstration set. The workflow is built to show what intake could look like inside an office handling sensitive records: many files selected at once, one intake note at a time, with human review still required.

The model answers linked below are raw output from the run. They were not tailored, tweaked, or tuned for a particular agency purpose. The PDFs are real public documents, but they are public snapshots rather than comprehensive case files. Some details may therefore look incomplete or unresolved because the public record does not contain the full office file.

In real office use, this workflow would be customized for the agency, the record type, and the review purpose. VoraSync does not automate legal judgment or replace human review. This demo shows the shape of a no-internet intake workflow that could be applied to private, sensitive, agency-controlled records.