Primary fit
Cyber & Networks
For AI/ML, DevSecOps, cyber, and enterprise IT workflows where context, memory, and evidence must stay inside the controlled network.Cyber / IT
Download PDFGovernment · primes · technical partners
Capability statement.
Infrastructure for AI agents that need their context, memory, and evidence to stay inside the mission boundary.
Perseus Computing builds local-first infrastructure for controlled, disconnected, private-VPC, and regulated environments. Start with the document; choose a tailored version only when the conversation calls for it.
Three send-ready versions
Choose the version for the conversation.
Each version keeps the same evidence and procurement facts while changing the front door and fit language.
Enabling fit
C3BM
For software factories and distributed decision workflows where Perseus can sit beside—not replace—the mission platform.Software factory / decisions
Download PDFPartner-led
Electronic Systems
For a prime, integrator, or technical partner exploring local, reviewable AI around sensing, EW, PNT, or mission-system workflows.Partner / integration
Download PDFThree send-ready versions
Choose the version for the conversation.
Cyber & Networks is the clearest direct route. C3BM is enabling; Electronic Systems is partner-led.
Strongest direct fit
Cyber & Networks
For AI/ML, DevSecOps, cyber, and enterprise IT workflows where context, memory, and evidence must stay inside the controlled network.Cyber / IT
Download PDFEnabling / adjacent fit
C3BM
For software factories and distributed decision workflows where Perseus can sit beside—not replace—the mission platform.Software factory / decisions
Download PDFPartner-led / conditional fit
Electronic Systems
For a prime, integrator, or technical partner exploring local, reviewable AI around sensing, EW, PNT, or mission-system workflows.Partner / integration
Download PDFWhat is inside
A reader can scan it in two minutes.
Page 01 / fit
What Perseus builds
The problem with lost state, the local context–memory–evidence system, the audience fit, and the smallest useful first conversation.
Page 02 / proof
Why the boundary matters
Measured evidence, procurement identifiers, licensing and SBOM posture, deployment options, and a direct contact path.
At a glance
The facts a buyer needs first.
Deployment
Local-firstAir-gapped, on-premises, private VPC, or controlled network.
Assessment
110 / 110SPRS score; CMMC Level 2 self-assessment posture is scoped explicitly.
Identity
CAGE 22JC5Perseus Computing LLC · UEI PJS2LW7HAK35.
License
MITSBOM materials published; no mandatory proprietary runtime.
Need the right front door?
Bring the workflow. We will bring the smallest credible path.
Send the mission workflow, its data boundary, and the integration surface. Thomas can point you to the right version or technical conversation.
Government · primes · technical partners
Capability statement.
Infrastructure for AI agents that need their context, memory, and evidence to stay inside the mission boundary.
Perseus Computing builds local-first infrastructure for controlled, disconnected, private-VPC, and regulated environments. Start with the document; choose a tailored version only when the conversation calls for it.
Three send-ready versions
Choose the version for the conversation.
Cyber & Networks is the clearest direct route. C3BM is enabling; Electronic Systems is partner-led.
Strongest direct fit
Cyber & Networks
For AI/ML, DevSecOps, cyber, and enterprise IT workflows where context, memory, and evidence must stay inside the controlled network.Cyber / IT
Download PDFEnabling / adjacent fit
C3BM
For software factories and distributed decision workflows where Perseus can sit beside—not replace—the mission platform.Software factory / decisions
Download PDFPartner-led / conditional fit
Electronic Systems
For a prime, integrator, or technical partner exploring local, reviewable AI around sensing, EW, PNT, or mission-system workflows.Partner / integration
Download PDFGovernment · primes · technical partners
Mission fit for controlled AI.
Infrastructure for AI agents that need their context, memory, and evidence to stay inside the mission boundary.
Perseus Computing builds local-first infrastructure for controlled, disconnected, private-VPC, and regulated environments. Start with the document; choose a tailored version only when the conversation calls for it.
Three audience routes
Choose the version for the conversation.
Cyber & Networks is the clearest direct route. C3BM is enabling; Electronic Systems is partner-led.
Strongest direct fit
Cyber & Networks
For AI/ML, DevSecOps, cyber, and enterprise IT workflows where context, memory, and evidence must stay inside the controlled network.Cyber / IT
Download PDFEnabling / adjacent fit
C3BM
For software factories and distributed decision workflows where Perseus can sit beside—not replace—the mission platform.Software factory / decisions
Download PDFPartner-led / conditional fit
Electronic Systems
For a prime, integrator, or technical partner exploring local, reviewable AI around sensing, EW, PNT, or mission-system workflows.Partner / integration
Download PDFWhat is inside
A reader can scan it in two minutes.
Page 01 / fit
What Perseus builds
The problem with lost state, the local context–memory–evidence system, the audience fit, and the smallest useful first conversation.
Page 02 / proof
Why the boundary matters
Measured evidence, procurement identifiers, licensing and SBOM posture, deployment options, and a direct contact path.
At a glance
The facts a buyer needs first.
Deployment
Local-firstAir-gapped, on-premises, private VPC, or controlled network.
Assessment
110 / 110SPRS score; CMMC Level 2 self-assessment posture is scoped explicitly.
Identity
CAGE 22JC5Perseus Computing LLC · UEI PJS2LW7HAK35.
License
MITSBOM materials published; no mandatory proprietary runtime.
Need the right front door?
Bring the workflow. We will bring the smallest credible path.
Send the mission workflow, its data boundary, and the integration surface. Thomas can point you to the right version or technical conversation.
Need the right front door?
Bring the workflow. We will bring the smallest credible path.
Send the mission workflow, its data boundary, and the integration surface. We can point you to the right version or technical conversation.
Run the context, memory, and evidence layers where the program keeps its working data. Air-gapped, on-premises, private VPC, and controlled-network deployment are supported postures.", st["body_tiny"]), - rich("CONTACT
Thomas Connally, Founder
perseus@perseus.observer", st["body_tiny"]), + rich("CONTACT
perseus@perseus.observer", st["body_tiny"]), ]], colWidths=[4.72 * inch, 2.63 * inch]) boundary.setStyle(TableStyle([ ("BACKGROUND", (0, 0), (0, 0), PALE_BLUE), diff --git a/tests/test_capability_download.py b/tests/test_capability_download.py index 492e6e8a..0288ece7 100644 --- a/tests/test_capability_download.py +++ b/tests/test_capability_download.py @@ -47,7 +47,9 @@ def test_tailored_capability_artifacts_exist_and_public_copy_is_not_internal_wor def test_government_landing_is_single_purpose_and_exposes_all_three_custom_routes(): landing = (ROOT / "government" / "index.html").read_text(encoding="utf-8") - assert "Capability" in landing and "statement." in landing + assert "Mission fit for" in landing and "controlled AI" in landing + assert "Email Thomas" not in landing + assert "Thomas Connally" not in landing assert "Choose the version for the conversation." in landing assert "Cyber-Networks.pdf" in landing assert "C3BM.pdf" in landing From 36266069af5a165ea235a720a4e34d87affce268 Mon Sep 17 00:00:00 2001 From: Perseus Computing <51974392+tcconnally@users.noreply.github.com> Date: Tue, 18 Aug 2026 00:27:34 +0000 Subject: [PATCH 3/3] feat(government): publish flat capability brief layout --- ...omputing-LLC-Capability-Statement-C3BM.pdf | Bin 7769 -> 7562 bytes ...LC-Capability-Statement-Cyber-Networks.pdf | Bin 7771 -> 7545 bytes ...apability-Statement-Electronic-Systems.pdf | Bin 7792 -> 7567 bytes ...eus-Computing-LLC-Capability-Statement.pdf | Bin 7914 -> 7704 bytes scripts/build_capability_statement.py | 523 ++++++------------ 5 files changed, 175 insertions(+), 348 deletions(-) diff --git a/government/assets/Perseus-Computing-LLC-Capability-Statement-C3BM.pdf b/government/assets/Perseus-Computing-LLC-Capability-Statement-C3BM.pdf index c53945c9eb6098d0736fcd4be441cd2f4c72bb18..1013ca60773e1beb3d1057a10740b6e80dd6ccaa 100644 GIT binary patch literal 7562 zcmdUU+4izZvgiHx6iz6J$RL7>IDq1a0xF6)fPji1qjp#IMZZAzUElY!Hve6FSM5`E zR-d!_u30W2gUHB?UqmDs8Gtkr5WG(Q0{++k{2%{2r#MlVe3s@dH*o?dF+Yx-%Q$j& zXno&K@X3K;
"), style) -def rich(text: str, style: ParagraphStyle) -> Paragraph: +def markup(text: str, style: ParagraphStyle) -> Paragraph: return Paragraph(text, style) -def section_heading(title: str, st: dict[str, ParagraphStyle], color=BLUE) -> list: - return [ - rich(escape(title.upper()), st["section"]), - HRFlowable(width="100%", thickness=0.7, color=color, spaceBefore=1, spaceAfter=7), - ] - +def rule_section(title: str, st: dict[str, ParagraphStyle]) -> list: + return [markup(escape(title.upper()), st["section"]), HRFlowable(width="100%", thickness=0.65, color=RULE, spaceBefore=3, spaceAfter=8)] -def bullet(label: str, text: str, st: dict[str, ParagraphStyle]) -> Paragraph: - return rich(f"• {escape(label)} {escape(text)}", st["body_small"]) - -def header(st: dict[str, ParagraphStyle], page_label: str) -> Table: - left = [rich("PERSEUS COMPUTING LLC", st["brand"]), rich("CONTEXT · MEMORY · EVIDENCE", st["brand_sub"])] - right = [rich(escape(page_label.upper()), st["eyebrow_white"]), rich("CAPABILITY STATEMENT · 2026", st["brand_sub"])] - table = Table([[left, right]], colWidths=[4.65 * inch, 2.75 * inch]) - table.setStyle(TableStyle([ +def header(page_title: str, st: dict[str, ParagraphStyle]) -> Table: + left = [markup("PERSEUS COMPUTING LLC", st["header_brand"]), markup("CONTEXT · MEMORY · EVIDENCE", st["header_meta"])] + right = [markup(escape(page_title.upper()), st["header_meta"]), markup("DEFENSE CAPABILITY BRIEF · 2026", st["header_meta"])] + t = Table([[left, right]], colWidths=[4.65 * inch, 2.70 * inch]) + t.setStyle(TableStyle([ ("BACKGROUND", (0, 0), (-1, -1), NAVY), ("VALIGN", (0, 0), (-1, -1), "MIDDLE"), ("ALIGN", (1, 0), (1, 0), "RIGHT"), - ("LEFTPADDING", (0, 0), (0, 0), 13), - ("RIGHTPADDING", (0, 0), (0, 0), 5), - ("LEFTPADDING", (1, 0), (1, 0), 5), - ("RIGHTPADDING", (1, 0), (1, 0), 13), - ("TOPPADDING", (0, 0), (-1, -1), 10), - ("BOTTOMPADDING", (0, 0), (-1, -1), 10), + ("LEFTPADDING", (0, 0), (0, 0), 12), ("RIGHTPADDING", (0, 0), (0, 0), 4), + ("LEFTPADDING", (1, 0), (1, 0), 4), ("RIGHTPADDING", (1, 0), (1, 0), 12), + ("TOPPADDING", (0, 0), (-1, -1), 9), ("BOTTOMPADDING", (0, 0), (-1, -1), 9), ])) - return table + return t -def at_a_glance(profile: dict, st: dict[str, ParagraphStyle]) -> Table: +def info_band(profile: dict, st: dict[str, ParagraphStyle]) -> Table: + fit = "Three audience routes" if profile["audience_label"].startswith("MASTER") else str(profile.get("fit_tag") or "") rows = [ - [p("PRIMARY FIT", st["at_glance_label"]), p(profile["fit_tag"], st["at_glance_value"])], - [p("DEPLOYMENT", st["at_glance_label"]), p("Local / air-gapped / private VPC", st["at_glance_value"])], - [p("PROCUREMENT", st["at_glance_label"]), p("SAM active · CAGE 22JC5", st["at_glance_value"])], - [p("CONTACT", st["at_glance_label"]), p("perseus@perseus.observer", st["at_glance_value"])], + [markup("FOCUS
" + escape(fit), st["body_small"]), markup("DEPLOYMENT
Local · air-gapped · private VPC", st["body_small"]), markup("IDENTITY
UEI PJS2LW7HAK35 · CAGE 22JC5", st["body_small"]), markup("LICENSE
MIT · SBOM published", st["body_small"])], ] - table = Table(rows, colWidths=[0.88 * inch, 2.13 * inch]) - table.setStyle(TableStyle([ - ("BACKGROUND", (0, 0), (-1, -1), PALE_BLUE), - ("BOX", (0, 0), (-1, -1), 0.6, RULE), - ("INNERGRID", (0, 0), (-1, -1), 0.35, RULE), + t = Table(rows, colWidths=[1.84 * inch] * 4) + t.setStyle(TableStyle([ + ("BACKGROUND", (0, 0), (-1, -1), PALE), + ("LINEABOVE", (0, 0), (-1, 0), 0.55, RULE), + ("LINEBELOW", (0, 0), (-1, 0), 0.55, RULE), + ("LINEBEFORE", (1, 0), (-1, 0), 0.35, RULE), ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (-1, -1), 7), - ("RIGHTPADDING", (0, 0), (-1, -1), 7), - ("TOPPADDING", (0, 0), (-1, -1), 7), - ("BOTTOMPADDING", (0, 0), (-1, -1), 7), + ("LEFTPADDING", (0, 0), (-1, -1), 9), ("RIGHTPADDING", (0, 0), (-1, -1), 9), + ("TOPPADDING", (0, 0), (-1, -1), 8), ("BOTTOMPADDING", (0, 0), (-1, -1), 8), ])) - return table + return t -def module_card(number: str, title: str, text: str, background, st: dict[str, ParagraphStyle]) -> Table: - cell = [ - rich(escape(number), st["label"]), - Spacer(1, 9), - rich(escape(title), st["card_title"]), - Spacer(1, 5), - p(text, st["card_text"]), - ] - table = Table([[cell]], colWidths=[2.36 * inch]) - table.setStyle(TableStyle([ - ("BACKGROUND", (0, 0), (-1, -1), background), - ("BOX", (0, 0), (-1, -1), 0.45, RULE), +def two_column_copy(left_label: str, left_text: str, right_label: str, right_text: str, st: dict[str, ParagraphStyle]) -> Table: + left = [markup(escape(left_label.upper()), st["label_blue"]), Spacer(1, 6), plain(left_text, st["body"])] + right = [markup(escape(right_label.upper()), st["label"]), Spacer(1, 6), plain(right_text, st["body"])] + t = Table([[left, right]], colWidths=[3.65 * inch, 3.70 * inch]) + t.setStyle(TableStyle([ + ("LINEBEFORE", (1, 0), (1, 0), 0.7, RULE), ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (-1, -1), 10), - ("RIGHTPADDING", (0, 0), (-1, -1), 10), - ("TOPPADDING", (0, 0), (-1, -1), 10), - ("BOTTOMPADDING", (0, 0), (-1, -1), 11), + ("LEFTPADDING", (0, 0), (0, 0), 0), ("RIGHTPADDING", (0, 0), (0, 0), 14), + ("LEFTPADDING", (1, 0), (1, 0), 14), ("RIGHTPADDING", (1, 0), (1, 0), 0), + ("TOPPADDING", (0, 0), (-1, -1), 0), ("BOTTOMPADDING", (0, 0), (-1, -1), 0), ])) - return table + return t -def fit_table(profile: dict, st: dict[str, ParagraphStyle]) -> Table: - data = [] - for label, text, tag in profile["fit_rows"]: - tag_style = st["label_gold"] if tag == "PRIMARY FIT" else st["label_blue"] if tag == "ENABLING FIT" else st["label"] - data.append([rich(f"{escape(label)}
{escape(tag)}", tag_style), p(text, st["body_small"])]) - table = Table(data, colWidths=[1.62 * inch, 5.58 * inch]) - table.setStyle(TableStyle([ - ("BACKGROUND", (0, 0), (0, -1), PALE_GREEN), - ("GRID", (0, 0), (-1, -1), 0.4, RULE), +def system_table(st: dict[str, ParagraphStyle]) -> Table: + rows = [ + [markup("CONTEXT
Before action", st["module_title"]), plain("Resolves repository, ticket, deployment, and decision state into bounded context before an agent uses it.", st["body_small"])], + [markup("MEMORY
Across sessions", st["module_title"]), plain("Retains governed, bitemporal memory with durable journaling and explicit authority boundaries.", st["body_small"])], + [markup("EVIDENCE
After action", st["module_title"]), plain("Links actions and decisions to evidence, authority, and time so a reviewer can reconstruct what happened.", st["body_small"])], + ] + t = Table(rows, colWidths=[1.85 * inch, 5.50 * inch]) + t.setStyle(TableStyle([ + ("BACKGROUND", (0, 0), (-1, -1), PALE), + ("LINEABOVE", (0, 0), (-1, 0), 0.55, RULE), + ("LINEBELOW", (0, -1), (-1, -1), 0.55, RULE), + ("LINEBELOW", (0, 0), (-1, -2), 0.35, RULE), ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (-1, -1), 8), - ("RIGHTPADDING", (0, 0), (-1, -1), 8), - ("TOPPADDING", (0, 0), (-1, -1), 7), - ("BOTTOMPADDING", (0, 0), (-1, -1), 7), + ("LEFTPADDING", (0, 0), (-1, -1), 9), ("RIGHTPADDING", (0, 0), (-1, -1), 9), + ("TOPPADDING", (0, 0), (-1, -1), 8), ("BOTTOMPADDING", (0, 0), (-1, -1), 8), ])) - return table + return t -def metric_card(value: str, label: str, detail: str, st: dict[str, ParagraphStyle]) -> Table: - cell = [rich(escape(value), st["metric_value"]), Spacer(1, 5), rich(escape(label), st["metric_label"]), Spacer(1, 4), p(detail, st["metric_detail"])] - table = Table([[cell]], colWidths=[3.54 * inch]) - table.setStyle(TableStyle([ - ("BACKGROUND", (0, 0), (-1, -1), PALE_GREEN), - ("BOX", (0, 0), (-1, -1), 0.5, RULE), +def fit_table(profile: dict, st: dict[str, ParagraphStyle]) -> Table: + data = [[markup("ROUTE", st["label_muted"]), markup("USE", st["label_muted"]), markup("ROLE", st["label_muted"])]] + for label, text, tag in profile["fit_rows"]: + role = {"PRIMARY FIT": "Direct fit", "ENABLING FIT": "Enabling fit", "PARTNER-LED": "Partner-led"}.get(tag, tag.title()) + data.append([markup(f"{escape(label)}", st["body_small"]), plain(text, st["body_small"]), markup(escape(role), st["label"])]) + t = Table(data, colWidths=[1.62 * inch, 4.53 * inch, 1.20 * inch]) + t.setStyle(TableStyle([ + ("BACKGROUND", (0, 0), (-1, 0), PALE_BLUE), + ("LINEABOVE", (0, 0), (-1, 0), 0.55, RULE), + ("LINEBELOW", (0, -1), (-1, -1), 0.55, RULE), + ("LINEBELOW", (0, 1), (-1, -2), 0.35, RULE), ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (-1, -1), 11), - ("RIGHTPADDING", (0, 0), (-1, -1), 11), - ("TOPPADDING", (0, 0), (-1, -1), 11), - ("BOTTOMPADDING", (0, 0), (-1, -1), 11), + ("LEFTPADDING", (0, 0), (-1, -1), 8), ("RIGHTPADDING", (0, 0), (-1, -1), 8), + ("TOPPADDING", (0, 0), (-1, -1), 7), ("BOTTOMPADDING", (0, 0), (-1, -1), 7), ])) - return table + return t -def fact_block(title: str, rows: list[tuple[str, str]], background, st: dict[str, ParagraphStyle]) -> Table: - body = [rich(escape(title.upper()), st["eyebrow"])] - for label, value in rows: - body.extend([rich(f"{escape(label)}
{escape(value)}", st["body_small"]), Spacer(1, 5)]) - table = Table([[body]], colWidths=[3.54 * inch]) - table.setStyle(TableStyle([ - ("BACKGROUND", (0, 0), (-1, -1), background), - ("BOX", (0, 0), (-1, -1), 0.5, RULE), +def evidence_table(claims: dict, st: dict[str, ParagraphStyle]) -> Table: + rows = [[markup("MEASURE", st["label_muted"]), markup("RESULT", st["label_muted"]), markup("WHAT IT SHOWS", st["label_muted"]), markup("SCOPE", st["label_muted"])]] + values = [ + ("LongMemEval QA", claims["longmemeval_cot"]["value"], "Answer quality when the agent receives the official context-of-thought protocol.", "Official-CoT mean · 3 signed runs"), + ("Session retrieval", claims["longmemeval_retrieval_recall10"]["value"], "Relevant session memories remain available to the later task.", "Retrieval-only · recall@10"), + ("BEAM correctness", claims["beam_correctness"]["value"], "The deterministic gauntlet stays correct across token tiers.", "128K–10M token tiers"), + ("Durable write", claims["vault_durable_write_100k"]["value"], "Sustained write behavior at a 100K-entity scale.", "Signed report · not bulk insert"), + ] + for name, value, meaning, scope in values: + rows.append([markup(f"{escape(name)}", st["metric_label"]), markup(escape(value), st["metric_value"]), plain(meaning, st["body_small"]), plain(scope, st["metric_detail"])]) + t = Table(rows, colWidths=[1.45 * inch, .82 * inch, 3.12 * inch, 1.96 * inch]) + t.setStyle(TableStyle([ + ("BACKGROUND", (0, 0), (-1, 0), PALE_BLUE), + ("LINEABOVE", (0, 0), (-1, 0), 0.55, RULE), + ("LINEBELOW", (0, -1), (-1, -1), 0.55, RULE), + ("LINEBELOW", (0, 1), (-1, -2), 0.35, RULE), + ("VALIGN", (0, 0), (-1, -1), "MIDDLE"), + ("LEFTPADDING", (0, 0), (-1, -1), 7), ("RIGHTPADDING", (0, 0), (-1, -1), 7), + ("TOPPADDING", (0, 0), (-1, -1), 8), ("BOTTOMPADDING", (0, 0), (-1, -1), 8), + ])) + return t + + +def flat_fact_table(title: str, rows: list[tuple[str, str]], st: dict[str, ParagraphStyle]) -> Table: + data = [[markup(escape(title.upper()), st["eyebrow_blue"]), ""]] + data.extend([[markup(f"{escape(k)}", st["body_small"]), plain(v, st["body_small"])] for k, v in rows]) + t = Table(data, colWidths=[1.18 * inch, 2.50 * inch]) + t.setStyle(TableStyle([ + ("SPAN", (0, 0), (1, 0)), + ("LINEABOVE", (0, 0), (-1, 0), 0.55, RULE), + ("LINEBELOW", (0, -1), (-1, -1), 0.55, RULE), + ("LINEBELOW", (0, 1), (-1, -2), 0.35, RULE), ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (-1, -1), 11), - ("RIGHTPADDING", (0, 0), (-1, -1), 11), - ("TOPPADDING", (0, 0), (-1, -1), 11), - ("BOTTOMPADDING", (0, 0), (-1, -1), 7), + ("LEFTPADDING", (0, 0), (-1, -1), 0), ("RIGHTPADDING", (0, 0), (-1, -1), 8), + ("TOPPADDING", (0, 0), (-1, -1), 6), ("BOTTOMPADDING", (0, 0), (-1, -1), 6), ])) - return table + return t -def page_chrome(canvas, doc) -> None: +def footer(canvas, doc) -> None: canvas.saveState() canvas.setStrokeColor(RULE) - canvas.setLineWidth(0.45) - canvas.line(doc.leftMargin, 0.34 * inch, PAGE_W - doc.rightMargin, 0.34 * inch) + canvas.setLineWidth(.45) + canvas.line(doc.leftMargin, .32 * inch, PAGE_W - doc.rightMargin, .32 * inch) canvas.setFillColor(FAINT) - canvas.setFont("Helvetica", 6.7) - canvas.drawString(doc.leftMargin, 0.20 * inch, "Perseus Computing LLC · perseus.observer · perseus@perseus.observer") - canvas.drawRightString(PAGE_W - doc.rightMargin, 0.20 * inch, f"{doc.page} / 2") + canvas.setFont("Helvetica", 6.6) + canvas.drawString(doc.leftMargin, .18 * inch, "Perseus Computing LLC · perseus.observer · perseus@perseus.observer") + canvas.drawRightString(PAGE_W - doc.rightMargin, .18 * inch, f"{doc.page} / 2") canvas.restoreState() -def build(profile_name: str, profile: dict, claims: dict) -> Path: - st = make_styles() - output = OUTPUT_DIR / profile["filename"] - output.parent.mkdir(parents=True, exist_ok=True) - doc = BaseDocTemplate( - str(output), - pagesize=letter, - leftMargin=0.56 * inch, - rightMargin=0.56 * inch, - topMargin=0.42 * inch, - bottomMargin=0.52 * inch, - title=f"Perseus Computing LLC — {profile['audience_label']}", - author="Perseus Computing LLC", - subject="Capability statement", - ) - frame = Frame(doc.leftMargin, doc.bottomMargin, doc.width, doc.height, id="capability") - doc.addPageTemplates([PageTemplate(id="capability", frames=[frame], onPage=page_chrome)]) - - c = claims +def build(name: str, profile: dict, claims: dict) -> Path: + st = styles() + OUT.mkdir(parents=True, exist_ok=True) + path = OUT / profile["filename"] + doc = BaseDocTemplate(str(path), pagesize=letter, leftMargin=.60*inch, rightMargin=.60*inch, topMargin=.42*inch, bottomMargin=.50*inch, title="Perseus Computing Defense Capability Brief", author="Perseus Computing LLC") + frame = Frame(doc.leftMargin, doc.bottomMargin, doc.width, doc.height, id="brief") + doc.addPageTemplates([PageTemplate(id="brief", frames=[frame], onPage=footer)]) story: list = [] - # PAGE 1: orient the reader before listing proof. - story.append(header(st, profile["audience_label"])) - story.append(Spacer(1, 13)) - story.append(rich("CAPABILITY STATEMENT", st["eyebrow"])) - hero_title_style = st["hero_title_compact"] if len(profile["headline"]) > 62 else st["hero_title"] - hero = [ - rich(escape(profile["headline"]), hero_title_style), - p(profile["lead"], st["lead"]), - Spacer(1, 8), - rich("UEI PJS2LW7HAK35 CAGE 22JC5 SAM Active HQ Austin, Texas", st["identity"]), - ] - hero_table = Table([[hero, at_a_glance(profile, st)]], colWidths=[4.25 * inch, 3.10 * inch]) - hero_table.setStyle(TableStyle([ - ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (-1, -1), 0), - ("RIGHTPADDING", (0, 0), (0, 0), 13), - ("RIGHTPADDING", (1, 0), (1, 0), 0), - ("TOPPADDING", (0, 0), (-1, -1), 0), - ("BOTTOMPADDING", (0, 0), (-1, -1), 0), - ])) - story.append(hero_table) - story.append(Spacer(1, 16)) - - story.extend(section_heading("The problem we solve", st)) - problem = Table([[ - p(profile["problem"], st["body"]), - rich(f"WHAT CHANGES
{escape(profile['change'])}", st["callout"]), - ]], colWidths=[4.25 * inch, 3.10 * inch]) - problem.setStyle(TableStyle([ - ("BACKGROUND", (1, 0), (1, 0), PALE_GOLD), - ("BOX", (1, 0), (1, 0), 0.5, HexColor("#E5C58C")), - ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (0, 0), 0), - ("RIGHTPADDING", (0, 0), (0, 0), 13), - ("LEFTPADDING", (1, 0), (1, 0), 11), - ("RIGHTPADDING", (1, 0), (1, 0), 11), - ("TOPPADDING", (0, 0), (-1, -1), 10), - ("BOTTOMPADDING", (0, 0), (-1, -1), 10), - ])) - story.append(problem) - story.append(Spacer(1, 16)) - - story.extend(section_heading("The Perseus system", st)) - modules = Table([[ - module_card("01 / CONTEXT", "Before the agent acts", "Resolves repository, ticket, deployment, and decision state into bounded context before a model uses it.", PALE_BLUE, st), - module_card("02 / MEMORY", "After the session ends", "Vault retains encrypted, bitemporal memory with durable journaling and explicit authority boundaries.", PALE_GREEN, st), - module_card("03 / EVIDENCE", "When the work is reviewed", "Ledger links actions and decisions to evidence, authority, and time instead of leaving the record in an opaque dashboard.", PALE_GOLD, st), - ]], colWidths=[2.36 * inch, 2.36 * inch, 2.36 * inch]) - modules.setStyle(TableStyle([ - ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (-1, -1), 0), - ("RIGHTPADDING", (0, 0), (0, 0), 7), - ("RIGHTPADDING", (1, 0), (1, 0), 7), - ("RIGHTPADDING", (2, 0), (2, 0), 0), - ("TOPPADDING", (0, 0), (-1, -1), 0), - ("BOTTOMPADDING", (0, 0), (-1, -1), 0), - ])) - story.append(modules) - story.append(Spacer(1, 16)) - - story.extend(section_heading("Where this fits", st)) - story.append(fit_table(profile, st)) - story.append(Spacer(1, 15)) - first = Table([[rich(f"FIRST CONVERSATION {escape(profile['discussion'])}", st["callout"])]], colWidths=[7.35 * inch]) - first.setStyle(TableStyle([ - ("BACKGROUND", (0, 0), (-1, -1), NAVY), - ("TEXTCOLOR", (0, 0), (-1, -1), WHITE), - ("BOX", (0, 0), (-1, -1), 0.5, NAVY), - ("LEFTPADDING", (0, 0), (-1, -1), 11), - ("RIGHTPADDING", (0, 0), (-1, -1), 11), - ("TOPPADDING", (0, 0), (-1, -1), 9), - ("BOTTOMPADDING", (0, 0), (-1, -1), 9), - ])) - story.append(first) - - # PAGE 2: evidence and procurement details with more room to breathe. - story.append(PageBreak()) - story.append(header(st, "Evidence and procurement")) - story.append(Spacer(1, 14)) - story.append(rich("EVIDENCE AND PROCUREMENT", st["eyebrow"])) - story.append(rich("Proof first. Claims stay attached to their method.", st["hero_title_compact"])) - story.append(p("The figures below are drawn from the public claims registry. They are presented with the measurement family and condition that give each number meaning.", st["lead"])) - story.append(Spacer(1, 16)) - - story.extend(section_heading("Measured evidence", st)) - metrics = [ - metric_card(c["longmemeval_cot"]["value"], "LongMemEval QA", "Official-CoT mean of three signed runs.", st), - metric_card(c["longmemeval_retrieval_recall10"]["value"], "Session retrieval recall@10", "Retrieval-only session-level result.", st), - metric_card(c["beam_correctness"]["value"], "BEAM correctness", "Every 128K–10M token tier; deterministic gauntlet.", st), - metric_card(c["vault_durable_write_100k"]["value"], "Durable sustained write", "Signed scale report at 100K entities; not an in-memory insert rate.", st), - ] - metric_grid = Table([[metrics[0], metrics[1]], [metrics[2], metrics[3]]], colWidths=[3.68 * inch, 3.68 * inch]) - metric_grid.setStyle(TableStyle([ - ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (0, -1), 0), - ("RIGHTPADDING", (0, 0), (0, -1), 7), - ("LEFTPADDING", (1, 0), (1, -1), 0), - ("RIGHTPADDING", (1, 0), (1, -1), 0), - ("TOPPADDING", (0, 0), (-1, -1), 0), - ("BOTTOMPADDING", (0, 0), (-1, 0), 10), - ("BOTTOMPADDING", (0, 1), (-1, 1), 0), - ])) - story.append(metric_grid) - story.append(Spacer(1, 7)) - story.append(p("Measured on named hardware with signed or committed reports and rerunnable methods. Methodology variants remain separately labeled. No customer ROI, customer cost savings, or compliance certification is implied by these figures.", st["body_tiny"])) - story.append(Spacer(1, 17)) - - story.extend(section_heading("Procurement facts", st)) - facts = Table([[ - fact_block("Company", [ - ("Legal name", "Perseus Computing LLC"), - ("Business", "U.S. small business · technical access"), - ("UEI / CAGE", "PJS2LW7HAK35 / 22JC5"), - ("NAICS", "541715 · 541511 · 541512"), - ("SAM.gov", "Active — All Awards"), - ], PALE_BLUE, st), - fact_block("Security and deployment", [ - ("Boundary", "On-premises, air-gapped, private VPC, or controlled network"), - ("Readiness", "SPRS 110/110 · CMMC Level 2 final self-assessment, enclave scope"), - ("Software", "MIT license · published SBOM · NIST AI RMF-aligned architecture"), - ("Data posture", "No required cloud service, API key, or vendor runtime"), - ("Integration", "Partner-led where mission hardware or sensors are involved"), - ], PALE_GREEN, st), - ]], colWidths=[3.68 * inch, 3.68 * inch]) - facts.setStyle(TableStyle([ - ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (0, 0), 0), - ("RIGHTPADDING", (0, 0), (0, 0), 7), - ("LEFTPADDING", (1, 0), (1, 0), 0), - ("RIGHTPADDING", (1, 0), (1, 0), 0), - ("TOPPADDING", (0, 0), (-1, -1), 0), - ("BOTTOMPADDING", (0, 0), (-1, -1), 0), - ])) - story.append(facts) - story.append(Spacer(1, 17)) - - display_label = { - "MASTER CAPABILITY STATEMENT": "Master", - "CYBER & NETWORKS": "Cyber & Networks", - "C3BM": "C3BM", - "ELECTRONIC SYSTEMS": "Electronic Systems", - }[profile["audience_label"]] - story.append(Spacer(1, 12)) - boundary = Table([[ - rich("DEPLOYMENT BOUNDARY
Run the context, memory, and evidence layers where the program keeps its working data. Air-gapped, on-premises, private VPC, and controlled-network deployment are supported postures.", st["body_tiny"]), - rich("CONTACT
perseus@perseus.observer", st["body_tiny"]), - ]], colWidths=[4.72 * inch, 2.63 * inch]) - boundary.setStyle(TableStyle([ - ("BACKGROUND", (0, 0), (0, 0), PALE_BLUE), - ("BACKGROUND", (1, 0), (1, 0), PALE_GREEN), - ("BOX", (0, 0), (-1, -1), 0.5, RULE), - ("INNERGRID", (0, 0), (-1, -1), 0.35, RULE), - ("VALIGN", (0, 0), (-1, -1), "TOP"), - ("LEFTPADDING", (0, 0), (-1, -1), 8), - ("RIGHTPADDING", (0, 0), (-1, -1), 8), - ("TOPPADDING", (0, 0), (-1, -1), 7), - ("BOTTOMPADDING", (0, 0), (-1, -1), 7), - ])) - story.append(boundary) - story.append(Spacer(1, 5)) - story.append(rich(f"Updated {escape(str(c['_meta']['updated']))} · {escape(display_label)} version · Source-linked public claims", st["footer"])) - + # Page 1: orientation and fit. + story += [header("Mission fit", st), Spacer(1, 14), markup("PERSEUS COMPUTING LLC", st["eyebrow"])] + headline = "Infrastructure for AI agents in controlled environments." if name == "master" else profile["headline"] + story += [markup(escape(headline), st["title"] if len(headline) < 60 else st["title_compact"]), plain(profile["lead"], st["lead"]), Spacer(1, 12), info_band(profile, st), Spacer(1, 13)] + story += rule_section("The useful difference", st) + story += [two_column_copy("The problem", profile["problem"], "The layer", profile["change"], st), Spacer(1, 12)] + story += rule_section("What Perseus provides", st) + [system_table(st), Spacer(1, 12)] + story += rule_section("Where it fits", st) + [fit_table(profile, st), Spacer(1, 8)] + call = Table([[markup(f"FIRST CONVERSATION {escape(profile['discussion'])}", st["callout"])]], colWidths=[7.35*inch]) + call.setStyle(TableStyle([("BACKGROUND", (0, 0), (-1, -1), NAVY), ("LEFTPADDING", (0, 0), (-1, -1), 10), ("RIGHTPADDING", (0, 0), (-1, -1), 10), ("TOPPADDING", (0, 0), (-1, -1), 9), ("BOTTOMPADDING", (0, 0), (-1, -1), 9)])) + story.append(call) + + # Page 2: evidence and procurement without unequal boxes. + story += [PageBreak(), header("Evidence and procurement", st), Spacer(1, 14), markup("EVIDENCE AND PROCUREMENT", st["eyebrow"]), markup("Proof with the condition attached.", st["title_compact"]), plain("The figures below are kept separate by measurement family so a reviewer can see what each result does—and does not—establish.", st["lead"]), Spacer(1, 16)] + story += rule_section("Measured evidence", st) + [evidence_table(claims, st), Spacer(1, 7), plain("Measured on named hardware with signed or committed reports and rerunnable methods. No customer ROI, customer cost savings, or compliance certification is implied by these figures.", st["body_tiny"]), Spacer(1, 18)] + story += rule_section("Procurement and deployment", st) + left = flat_fact_table("Company", [("Legal name", "Perseus Computing LLC"), ("Business", "U.S. small business · technical access"), ("UEI / CAGE", "PJS2LW7HAK35 / 22JC5"), ("NAICS", "541715 · 541511 · 541512"), ("SAM.gov", "Active — All Awards")], st) + right = flat_fact_table("Security and boundary", [("Deployment", "On-premises, air-gapped, private VPC, or controlled network"), ("Readiness", "SPRS 110/110 · CMMC Level 2 self-assessment, enclave scope"), ("Software", "MIT license · published SBOM"), ("Data posture", "No required cloud service, API key, or vendor runtime"), ("Integration", "Partner-led where mission hardware or sensors are involved")], st) + facts = Table([[left, right]], colWidths=[3.68*inch, 3.67*inch]) + facts.setStyle(TableStyle([("VALIGN", (0, 0), (-1, -1), "TOP"), ("LINEBEFORE", (1, 0), (1, 0), .7, RULE), ("LEFTPADDING", (0, 0), (0, 0), 0), ("RIGHTPADDING", (0, 0), (0, 0), 14), ("LEFTPADDING", (1, 0), (1, 0), 14), ("RIGHTPADDING", (1, 0), (1, 0), 0), ("TOPPADDING", (0, 0), (-1, -1), 0), ("BOTTOMPADDING", (0, 0), (-1, -1), 0)])) + story += [facts, Spacer(1, 17)] + story += rule_section("Scope in one sentence", st) + story += [plain("Perseus is the infrastructure around the model: context before action, governed memory across sessions, and evidence after action. It is not a claim to replace a mission platform, prime integrator, or sensor system.", st["body"]), Spacer(1, 14)] + contact = Table([[markup("CONTACT perseus@perseus.observer · perseus.observer", st["body_small"]), markup("UPDATED " + escape(str(claims["_meta"]["updated"])), st["body_small"])]], colWidths=[5.75*inch, 1.60*inch]) + contact.setStyle(TableStyle([("BACKGROUND", (0, 0), (-1, -1), PALE_BLUE), ("LINEABOVE", (0, 0), (-1, 0), .55, RULE), ("LINEBELOW", (0, 0), (-1, 0), .55, RULE), ("VALIGN", (0, 0), (-1, -1), "MIDDLE"), ("LEFTPADDING", (0, 0), (-1, -1), 9), ("RIGHTPADDING", (0, 0), (-1, -1), 9), ("TOPPADDING", (0, 0), (-1, -1), 8), ("BOTTOMPADDING", (0, 0), (-1, -1), 8)])) + story.append(contact) doc.build(story) - print(f"wrote {output}") - return output + return path def main() -> None: parser = argparse.ArgumentParser() - parser.add_argument("--profile", choices=["master", "cyber-networks", "c3bm", "electronic-systems"], default="master") - parser.add_argument("--all", action="store_true", help="build the master and all tailored variants") + parser.add_argument("--all", action="store_true") + parser.add_argument("--profile", default="master") args = parser.parse_args() registry = json.loads((ROOT / "claims.json").read_text(encoding="utf-8")) profiles = json.loads(PROFILES.read_text(encoding="utf-8")) - selected = list(profiles) if args.all else [args.profile] - for name in selected: - build(name, profiles[name], registry["claims"] | {"_meta": registry["_meta"]}) + names = list(profiles) if args.all else [args.profile] + for name in names: + print(build(name, profiles[name], registry["claims"] | {"_meta": registry["_meta"]})) if __name__ == "__main__":