logicsrc/packages/opencontext/src/doctor.ts
Anthony Ettinger ec3ed64f20 Add the LogicSRC OpenContext specification
OpenContext is an open specification for durable, portable, permissioned,
provenance-aware context shared between humans and AI agents. It defines how
organizational knowledge is described, authorized, versioned, resolved,
audited, and handed between replaceable workers without losing institutional
state.

Follows the OpenPRD/OpenOntology pattern already in the repo: self-contained
JSON Schemas in @logicsrc/schemas, a reference implementation package, CLI
subcommands, docs, examples, and an OpenPRD record.

Schemas (8, all self-contained so a third party can fetch one file and
validate against it with no further resolution):
  manifest, object, bundle, role, provenance, decision, diagnostic,
  audit-event — registered in @logicsrc/validators and schemas:validate.

Reference implementation (@logicsrc/opencontext):
  loader with upward manifest discovery, the full resolution pipeline,
  authority/supersession, permissions, redaction, lifecycle, provenance,
  deterministic digests, doctor, search, graph, history/diff, guarded writes,
  audit events, and file/http/git/sqlite adapters.

CLI: all 15 specified commands, as a standalone `opencontext` binary and as
`logicsrc context`, sharing one implementation so the two cannot drift.

Design decisions worth noting:

- Supersession is declared, never inferred from version numbers. Inferring it
  would hide the governance failure it represents and make
  multiple-active-versions and duplicate-canonical impossible to detect.

- The bundle digest identifies the resolved context, not the moment it was
  computed, so generated_at/bundle_id/digest/as_of are excluded while objects,
  lifecycle states, exclusions and warnings are covered. That is what lets a
  decision record cite exactly the context that produced it.

- A role's own max_classification beats an inherited one, so a ceiling on a
  shared base role cannot silently cap a role deliberately granted more;
  requesting several roles at once still takes the lowest, so combining roles
  never escalates.

- Scope wildcards match whole dotted segments only. A trailing .* covers a
  subtree; an interior * matches exactly one segment. Substring matching here
  would be an access-control bug.

- --include narrows an existing scope and is applied after it, never merged
  into it, so a request can never widen what a role holds.

Verified: 226 tests across core primitives, permissions/redaction, the
resolution pipeline, security, the published conformance fixtures (13 valid,
35 invalid, 8 resolution scenarios), project behaviour, and the five shipped
examples — which are held to --strict and a 100% health score. Benchmarks meet
every published budget (resolve 1,000 objects in ~33ms against a 2s target).

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-08-09 18:22:42 +00:00

247 lines
8.8 KiB
TypeScript
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/**
* Context health.
*
* `doctor` is validation plus the questions that only make sense against a
* clock: what has gone stale, what expired, what is overdue for review, whose
* sources have broken. Context rot is quiet — nothing fails, agents just start
* answering from last year's pricing — so the score exists to make the rot
* visible on a dashboard and to give CI something to fail on.
*
* The formula is documented and configurable, because an undocumented score is
* a number people learn to ignore:
*
* deduction = Σ (weight[code] × affected objects) / max(objects, 1)
* score = clamp(100 deduction × 100, 0, 100)
*
* A weight is "how much of the repository's health one instance of this problem
* costs". A single broken canonical conflict in a ten-object repository costs
* far more than one in a thousand-object repository, which is the intent.
*/
import type { ContextStore, Diagnostic, DiagnosticCode, DiagnosticReport } from "./types.js";
import { computeLifecycle, isReviewOverdue } from "./lifecycle.js";
import { isSuperseded, resolveSupersession } from "./authority.js";
import { sortDiagnostics, validateStore, type ValidateOptions } from "./validate.js";
import { resolveAsOf } from "./time.js";
import { SPEC_VERSION } from "./manifest.js";
/**
* Default weights.
*
* Anything that makes the resolver produce a *wrong* answer costs the most:
* duplicate canonical objects, unsettleable conflicts, broken supersession.
* Anything that merely makes it produce an *incomplete* answer costs less.
*/
export const DEFAULT_WEIGHTS: Partial<Record<DiagnosticCode, number>> = {
"schema-invalid": 1.0,
"manifest-invalid": 1.0,
"duplicate-id": 1.0,
"duplicate-canonical": 1.0,
"conflict-ambiguous": 1.0,
"broken-supersession": 0.8,
"supersession-cycle": 0.8,
"secret-detected": 1.0,
"untrusted-canonical": 0.8,
"path-traversal": 1.0,
"unknown-scheme": 0.6,
"source-unavailable": 0.6,
"broken-reference": 0.5,
"multiple-active-versions": 0.4,
"conflict-declared": 0.3,
"expired": 0.4,
"missing-provenance": 0.3,
"missing-digest": 0.2,
"invalid-permission": 0.3,
"unknown-role": 0.5,
"role-cycle": 0.8,
"missing-owner": 0.2,
"stale": 0.15,
"review-overdue": 0.15,
"orphaned": 0.1,
"unapproved": 0.1,
"empty-scope": 0.1,
"not-yet-valid": 0.05,
"unknown-authority": 0.5,
"unknown-extension": 0.1
};
export interface DoctorOptions extends ValidateOptions {
at?: string | Date;
}
export function doctor(store: ContextStore, options: DoctorOptions = {}): DiagnosticReport {
const asOf = resolveAsOf(options.at);
const findings: Diagnostic[] = [...validateStore(store, options)];
const supersession = resolveSupersession(store.objects, store.byId);
for (const entry of store.objects) {
const object = entry.object;
if (isSuperseded(object, supersession.superseded)) continue;
const lifecycle = computeLifecycle(object, { asOf, manifest: store.manifest });
if (lifecycle === "stale") {
findings.push({
code: "stale",
severity: store.manifest.freshness?.stale_is_error ? "error" : "warning",
message: `${object.id} is stale — last updated ${object.updated ?? "never"}.`,
id: object.id,
file: entry.file,
field: "updated",
remediation: `Review it and bump updated, extend its ttl, or supersede it.`
});
}
if (lifecycle === "expired") {
findings.push({
code: "expired",
severity: "error",
message: `${object.id} expired on ${object.expires}.`,
id: object.id,
file: entry.file,
field: "expires",
remediation: `Renew it, supersede it, or remove the expiry if it should be durable.`
});
}
if (lifecycle === "future") {
findings.push({
code: "not-yet-valid",
severity: "info",
message: `${object.id} is not valid until ${object.valid_from}.`,
id: object.id,
file: entry.file,
field: "valid_from"
});
}
if (isReviewOverdue(object, asOf, store.manifest)) {
findings.push({
code: "review-overdue",
severity: "warning",
message: `${object.id} is overdue for review${object.review?.next_review ? ` (due ${object.review.next_review})` : ""}.`,
id: object.id,
file: entry.file,
field: "review",
remediation: `Review it and set review.last_review, or push review.next_review out deliberately.`
});
}
}
const sorted = sortDiagnostics(findings);
const score = computeScore(sorted, store);
const failOn = store.manifest.health?.fail_on ?? "error";
const minimum = store.manifest.health?.minimum_score;
const ok =
!sorted.some((finding) => severityAtLeast(finding.severity, failOn)) &&
(minimum === undefined || score >= minimum);
return {
opencontext: SPEC_VERSION,
ok,
generated_at: new Date().toISOString(),
namespace: store.manifest.id,
score,
counts: countBy(sorted, store),
findings: sorted
};
}
export function computeScore(findings: Diagnostic[], store: ContextStore): number {
const weights = { ...DEFAULT_WEIGHTS, ...(store.manifest.health?.weights ?? {}) } as Record<string, number>;
const denominator = Math.max(store.objects.length, 1);
let deduction = 0;
for (const finding of findings) {
deduction += weights[finding.code] ?? 0.1;
}
const score = 100 - (deduction / denominator) * 100;
return Math.round(Math.min(100, Math.max(0, score)) * 10) / 10;
}
function severityAtLeast(severity: string, threshold: string): boolean {
const order = { error: 0, warning: 1, info: 2 } as Record<string, number>;
return (order[severity] ?? 3) <= (order[threshold] ?? 0);
}
function countBy(findings: Diagnostic[], store: ContextStore): Record<string, number> {
const count = (code: DiagnosticCode): number => findings.filter((finding) => finding.code === code).length;
return {
objects: store.objects.length,
errors: findings.filter((finding) => finding.severity === "error").length,
warnings: findings.filter((finding) => finding.severity === "warning").length,
info: findings.filter((finding) => finding.severity === "info").length,
stale: count("stale"),
expired: count("expired"),
conflicting: count("conflict-ambiguous") + count("conflict-declared") + count("duplicate-canonical"),
orphaned: count("orphaned"),
missing_owner: count("missing-owner"),
broken_sources: count("source-unavailable") + count("unknown-scheme")
};
}
/** The human-readable health report. */
export function renderHealth(report: DiagnosticReport, store: ContextStore): string {
const lines: string[] = [];
const counts = report.counts ?? {};
lines.push("OpenContext Health");
lines.push("────────────────────────────────");
// Named entries first: these are the objects a reader looks for by name.
const named = store.objects
.filter((entry) => !entry.collection)
.slice(0, 8);
for (const entry of named) {
const issues = report.findings.filter((finding) => finding.id === entry.object.id);
const worst = issues.find((finding) => finding.severity === "error") ?? issues[0];
const status = worst
? `${worst.severity === "error" ? "✗" : "⚠"} ${worst.code}`
: `${entry.object.authority ?? "current"}`;
lines.push(`${column(entry.object.title ?? entry.object.id)}${status}`);
}
if (named.length > 0) lines.push("");
const rows: Array<[string, number]> = [
["Orphaned context", counts.orphaned ?? 0],
["Conflicting context", counts.conflicting ?? 0],
["Expired context", counts.expired ?? 0],
["Stale context", counts.stale ?? 0],
["Missing owners", counts.missing_owner ?? 0],
["Broken sources", counts.broken_sources ?? 0]
];
for (const [label, value] of rows) {
lines.push(`${column(label)}${value}`);
}
lines.push("");
lines.push(`Context health: ${report.score ?? 100}%`);
if (report.findings.length > 0) {
lines.push("");
for (const finding of report.findings.slice(0, 40)) {
const marker = finding.severity === "error" ? "✗" : finding.severity === "warning" ? "⚠" : "·";
const where = finding.file ? ` (${finding.file}${finding.line ? `:${finding.line}` : ""})` : "";
lines.push(` ${marker} ${finding.code}: ${finding.message}${where}`);
if (finding.remediation) lines.push(`${finding.remediation}`);
}
if (report.findings.length > 40) {
lines.push(` … and ${report.findings.length - 40} more. Use --format json for the full report.`);
}
}
return `${lines.join("\n")}\n`;
}
const COLUMN_WIDTH = 24;
/** Pad to the report column, truncating long titles so the second column stays aligned. */
function column(label: string): string {
if (label.length >= COLUMN_WIDTH) return `${label.slice(0, COLUMN_WIDTH - 2)}`;
return label.padEnd(COLUMN_WIDTH);
}