ci(docs): harden fabricated-docs checker + enforce --strict (QG v2 Fase 9 T9) (#4149)

Harden fabricated-docs checker (precision: env helpers/code-identifiers/api-prefixes/env-contract/tutorial-placeholders) + enforce --strict (QG v2 Fase 9 T9). Inclui testes anti-over-suppression + correção de refs stale de docs. Integrado em release/v3.8.29.
This commit is contained in:
Diego Rodrigues de Sa e Souza
2026-06-18 03:21:24 -03:00
committed by GitHub
parent 7fd4723f15
commit 0ca7a01c4b
12 changed files with 488 additions and 62 deletions

View File

@@ -80,7 +80,7 @@ claude.ai API
| File | Purpose |
| ----------------------------------------------------- | -------------------------------------------- |
| `src/shared/constants/providers.ts` | Provider registration (WEB_COOKIE_PROVIDERS) |
| `src/lib/providers/wrappers/claudeWeb.ts` | Type definitions + cookie utilities |
| `src/lib/providers/webCookieAuth.ts` | Cookie utilities (normalize/extract session cookies) |
| `open-sse/executors/claude-web.ts` | Executor implementation |
| `open-sse/executors/index.ts` | Executor registration |
| `open-sse/services/claudeTlsClient.ts` | TLS fingerprint spoofing via tls-client-node |

View File

@@ -292,7 +292,7 @@ Services (business logic):
- Thinking budget management: `open-sse/services/thinkingBudget.ts`
- Wildcard model routing: `open-sse/services/wildcardRouter.ts`
- Rate limit management: `open-sse/services/rateLimitManager.ts`
- Circuit breaker: `open-sse/services/circuitBreaker.ts`
- Circuit breaker: `src/shared/utils/circuitBreaker.ts`
- Context handoff: `open-sse/services/contextHandoff.ts` — handoff summary generation and injection for context-relay strategy
- Compression: `open-sse/services/compression/*` — proactive compression before provider translation;
includes Caveman rules, RTK filters, stacked pipelines, compression combos, stats, and validation
@@ -308,17 +308,17 @@ Services (business logic):
Domain layer modules:
- Cost rules/budgets: `src/lib/domain/costRules.ts`
- Fallback policy: `src/lib/domain/fallbackPolicy.ts`
- Combo resolver: `src/lib/domain/comboResolver.ts`
- Lockout policy: `src/lib/domain/lockoutPolicy.ts`
- Cost rules/budgets: `src/domain/costRules.ts`
- Fallback policy: `src/domain/fallbackPolicy.ts`
- Combo resolver: `src/domain/comboResolver.ts`
- Lockout policy: `src/domain/lockoutPolicy.ts`
- Policy engine: `src/domain/policyEngine.ts` — centralized lockout → budget → fallback evaluation
- Error codes catalog: `src/lib/domain/errorCodes.ts`
- Request ID: `src/lib/domain/requestId.ts`
- Fetch timeout: `src/lib/domain/fetchTimeout.ts`
- Request telemetry: `src/lib/domain/requestTelemetry.ts`
- Compliance/audit: `src/lib/domain/compliance/index.ts`
- Eval runner: `src/lib/domain/evalRunner.ts`
- Error codes catalog: `src/shared/constants/errorCodes.ts`
- Request ID: `src/shared/utils/requestId.ts`
- Fetch timeout: `src/shared/utils/fetchTimeout.ts`
- Request telemetry: `src/shared/utils/requestTelemetry.ts`
- Compliance/audit: `src/lib/compliance/index.ts`
- Eval runner: `src/lib/evals/evalRunner.ts`
- Domain state persistence: `src/lib/db/domainState.ts` — SQLite CRUD for fallback chains, budgets, cost history, lockout state, circuit breakers
OAuth provider modules (16 individual files under `src/lib/oauth/providers/`):

View File

@@ -143,4 +143,4 @@ Namespaces added by Group B:
| `activity.*` | All Activity page strings (title, verbs, filters, empty state) |
Source-of-truth locales: `pt-BR` and `en`. All other 39 locales fall back to
English via the fallback mechanism (`src/i18n/fallback.ts` or `next-intl` fallback).
English via the `next-intl` fallback mechanism (configured in `src/i18n/config.ts`).

View File

@@ -360,7 +360,6 @@ open-sse/
| `check-supported-node-runtime.ts` | Validate current Node version is supported |
| `check-pr-test-policy.mjs` | Enforce "tests required" rule on production code changes |
| **`gen-provider-reference.ts`** | NEW: auto-generate `docs/reference/PROVIDER_REFERENCE.md` from catalog |
| `generate-docs-index.mjs` | Build `src/app/docs/lib/docs-auto-generated.ts` from `docs/*.md` |
| `i18n/generate-multilang.mjs` | Translate UI strings + docs via Google Translate |
| `i18n_autotranslate.py` | LLM-based doc translation pipeline |
| `validate_translation.py` | Per-locale translation validation |

View File

@@ -129,7 +129,7 @@ from the current `PlaygroundState`. API key placeholder is always `$OMNIROUTE_AP
`PresetPicker.tsx``usePresets.ts``/api/playground/presets/*`:
- Stored in `playground_presets` SQLite table (migration `076_playground_presets.sql`).
- Stored in `playground_presets` SQLite table (migration `084_playground_presets.sql`).
- Each preset stores: `name`, `endpoint`, `model`, `system`, `params_json`, `created_at`.
- CRUD: `GET` list, `POST` create, `GET /:id`, `PUT /:id`, `DELETE /:id`.
@@ -186,7 +186,7 @@ Auth: optional (`REQUIRE_API_KEY`). Errors via `buildErrorBody()` (Hard Rule #12
| `src/app/api/playground/improve-prompt/route.ts` | Improve-prompt REST route |
| `src/app/api/playground/presets/route.ts` | Presets list + create |
| `src/app/api/playground/presets/[id]/route.ts` | Presets get/update/delete |
| `src/lib/db/migrations/076_playground_presets.sql` | DB migration |
| `src/lib/db/migrations/084_playground_presets.sql` | DB migration |
---

View File

@@ -571,15 +571,14 @@ omniroute setup --add-provider \
Recognized environment variables for non-interactive setup:
| Var | Purpose |
| ----------------------------- | -------------------------------------------- |
| `OMNIROUTE_SETUP_PASSWORD` | Admin password (>=8 chars) |
| `OMNIROUTE_PROVIDER` | Provider id (e.g. `openai`, `anthropic`) |
| `OMNIROUTE_PROVIDER_NAME` | Display name for the connection |
| `OMNIROUTE_PROVIDER_BASE_URL` | Optional OpenAI-compatible base URL override |
| `OMNIROUTE_API_KEY` | Provider API key |
| `OMNIROUTE_DEFAULT_MODEL` | Optional default model |
| `DATA_DIR` | Override the OmniRoute data directory |
| Var | Purpose |
| ------------------ | ----------------------------------------------------------------- |
| `OMNIROUTE_API_KEY` | Provider API key (bound to `--api-key` via Commander `.env()`) |
| `DATA_DIR` | Override the OmniRoute data directory |
All other non-interactive inputs are passed as flags, not environment variables:
`--password`, `--provider`, `--provider-name`, `--provider-base-url`, `--default-model`
(see the `omniroute setup` options above).
### Diagnostics

View File

@@ -13,7 +13,7 @@ lastUpdated: 2026-05-13
> Every variable documented here must also appear in `.env.example`, and
> every variable in `.env.example` must appear here. `npm run check:env-doc-sync`
> enforces this on commit and in CI. To omit a variable on purpose, add it to
> the allowlist inside `scripts/check-env-doc-sync.mjs`.
> the allowlist inside `scripts/check/check-env-doc-sync.mjs`.
---

View File

@@ -129,7 +129,7 @@ When adding a new route or executor, copy the assertion pattern from this file.
## Related controls
- `js/stack-trace-exposure` CodeQL alerts in `.github/security` should always be **either** fixed via these helpers **or** dismissed with a comment citing this doc.
- The `pino` redaction config (`src/lib/log/redaction.ts` — if present) handles structured log redaction separately. This doc covers only the response-message surface.
- The `pino` redaction config (`src/shared/utils/logRedaction.ts`) handles structured log redaction separately. This doc covers only the response-message surface.
- Upstream-header denylist (`src/shared/constants/upstreamHeaders.ts`) covers header leakage — keep both files aligned when adding a new exfiltration concern.
## Upstream details passthrough

View File

@@ -105,7 +105,7 @@
"check:docs-counts": "node scripts/check/check-docs-counts-sync.mjs",
"check:deprecated-versions": "node scripts/check/check-deprecated-versions.mjs",
"check:doc-links": "node scripts/check/check-doc-links.mjs",
"check:fabricated-docs": "node scripts/check/check-fabricated-docs.mjs",
"check:fabricated-docs": "node scripts/check/check-fabricated-docs.mjs --strict",
"check:docs-all": "npm run check:docs-sync && npm run check:docs-counts && npm run check:env-doc-sync && npm run check:deprecated-versions && npm run check:doc-links && npm run check:fabricated-docs",
"docs:render-diagrams": "node scripts/docs/render-diagrams.mjs",
"i18n:run": "node scripts/i18n/run-translation.mjs",

View File

@@ -60,6 +60,17 @@ const KNOWN_HOOKS = new Set([
"onActivate",
"onDeactivate",
"onUninstall",
// Real callbacks wired in code that docs reference (verified present in src/):
// onChunk/onFirstChunk — streaming callbacks (src/shared/utils/streamTracker.ts,
// playground ChatTab.tsx); onServerStatus/onPortChanged/onUpdateStatus — Electron
// IPC callbacks (src/shared/hooks/useElectron.ts, HomePageClient.tsx);
// onEmpty — model-metadata registry callback (src/lib/modelMetadataRegistry.ts).
"onChunk",
"onFirstChunk",
"onServerStatus",
"onPortChanged",
"onUpdateStatus",
"onEmpty",
]);
// Common false-positives the heuristic would otherwise flag. Add to this
@@ -85,6 +96,21 @@ const ENV_VAR_ALLOWLIST = new Set([
"OMNIROUTE_URL", // used by ad-hoc tooling, validated elsewhere
"OMNIROUTE_KEY", // ditto
"OPENCODE_API_KEY", // ditto
// ── External-tool / spawn-injected / ops env vars ────────────────────────
// Real environment variables, but they belong to an UPSTREAM CLI/tool, a
// docker-compose/electron-build pipeline, or are injected into a spawned
// subprocess — never read via `process.env.X` in OmniRoute's own source, so the
// code-read index can't see them. Documented (correctly) in the relevant guides.
"COPILOT_PROVIDER_BASE_URL", // GitHub Copilot CLI ≥v1.0.19's own env var (AGENTBRIDGE.md)
"OPENAI_BASE_URL", // env var OmniRoute passes to downstream CLIs (AGENT_PROTOCOLS_GUIDE.md)
"NINEROUTER_API_KEY", // injected into the 9router subprocess at spawn (EMBEDDED-SERVICES.md)
"CLAUDE_CODE_MAX_OUTPUT_TOKENS", // Claude Code CLI's own env var (CODEX-CLI-CONFIGURATION.md)
"CODEX_HOME", // Codex CLI's own config-home env var (CODEX-CLI-CONFIGURATION.md)
"REDIS_PORT", // docker-compose host-port override (DOCKER_GUIDE.md)
"AUTO_UPDATE_HOST_REPO_DIR", // docker-compose self-update mount (DOCKER_GUIDE.md)
"LINUX_GPG_KEY", // electron AppImage signing key, CI/build only (ELECTRON_GUIDE.md)
"BRANCH_LOCK_TOKEN", // release branch-protection ops token (QUALITY_GATE_PLAYBOOK.md)
"NEXT_LOCALE", // next-intl locale cookie name (I18N.md)
]);
// Common pluralized / column-header all-caps that aren't env vars
@@ -275,6 +301,18 @@ const ENV_VAR_DENYLIST = new Set([
"KNOWN_STALE_DOC_REFS", // export const in check-docs-symbols.mjs
"KNOWN_MISSING", // export const in check-fetch-targets.mjs
"KNOWN_RAW_SQL", // export const in check-db-rules.mjs
// ── Error / Node codes documented in prose (string-literal codes, not env vars) ──
"URL_GUARD_BLOCKED", // HTTP 422 guard-violation code (ARCHITECTURE.md)
"AUTHZ_NOT_INITIALIZED", // AuthzAssertionError code (AUTHZ_GUIDE.md)
"MODULE_NOT_FOUND", // Node runtime error code watched by service supervisor (ELECTRON_GUIDE.md)
"ERR_DLOPEN_FAILED", // Node native-module load error code (ELECTRON_GUIDE.md)
// ── Code-symbol / naming-convention examples documented in prose ─────────────
"UPPER_SNAKE", // the literal naming-convention token in the style guide (CODEBASE_DOCUMENTATION.md)
"DEFAULT_TIMEOUT", // example constant name in the UPPER_SNAKE convention row (AGENTS.md)
"SIDEBAR_DEFINITIONS", // code constant referenced in prose (MONITORING_SECTIONS.md)
"LOCAL_ONLY", // routeGuard classification label (AGENTBRIDGE.md)
"SPAWN_CAPABLE", // routeGuard classification label (AGENTBRIDGE.md)
"ZEROGRAVITY_SENSITIVE_WORDS", // cross-project constant named in a comparison (STEALTH_GUIDE.md)
]);
/** Endpoints that don't follow the standard route.ts pattern. */
@@ -306,6 +344,10 @@ const ENDPOINT_ALLOWLIST = new Set([
"/api/mcp/stream", // Streamable HTTP MCP transport
"/api/mcp/sse", // SSE MCP transport
"/api/health",
// Upstream/external provider endpoints documented in provider guides — these are
// paths on the UPSTREAM service (Claude.ai web, Blackbox), not OmniRoute routes.
"/api/organizations/{orgId}/chat_conversations/{convId}/completion", // claude-web upstream
"/api/chat", // Blackbox Web upstream (validated-token target)
]);
/** Doc files to skip (auto-generated, vendored, or third-party). */
@@ -316,12 +358,27 @@ const SKIP_DOC_FILES = new Set([
// Point-in-time documentation audit (v3.8.24): intentionally references drift,
// counts, and not-yet-existing files as part of documenting them — not living docs.
"docs/ops/DOCUMENTATION_AUDIT_REPORT.md",
// Design / research / plan docs: by definition describe not-yet-built files and
// proposed (not-yet-shipped) endpoints (each carries a `Status: Design`/`Active
// research`/`Plano` header). Same rationale as the audit report above — these are
// forward-looking specs, not living API docs, so their forward references are
// expected, not fabrications.
"docs/research", // DISCOVERY_TOOL_DESIGN.md, UNLIMITED_LLM_ACCESS.md, …
"docs/superpowers/plans", // dated implementation plans (files described before they exist)
// Release notes are historical, point-in-time records: they intentionally describe
// modules/paths as they were at that release (e.g. a module later moved or renamed).
// Rewriting them to today's layout would falsify history — out of scope for a
// living-docs accuracy gate.
"docs/releases",
// Forward-looking coverage plan: a `- [ ]` checklist of test targets and helper
// components to be created. Same rationale as the design/plan docs above.
"docs/ops/COVERAGE_PLAN.md",
]);
// ── File discovery ─────────────────────────────────────────────────────────
function walkMarkdown(dir, out = []) {
const abs = path.join(ROOT, dir);
function walkMarkdown(dir, out = [], root = ROOT) {
const abs = path.join(root, dir);
if (!fs.existsSync(abs)) return out;
const stat = fs.statSync(abs);
if (stat.isFile()) {
@@ -332,17 +389,17 @@ function walkMarkdown(dir, out = []) {
if (name === "node_modules" || name.startsWith(".")) continue;
const childAbs = path.join(abs, name);
const s = fs.statSync(childAbs);
if (s.isDirectory()) walkMarkdown(path.relative(ROOT, childAbs), out);
if (s.isDirectory()) walkMarkdown(path.relative(root, childAbs), out, root);
else if (childAbs.endsWith(".md") || childAbs.endsWith(".mdx")) out.push(childAbs);
}
return out;
}
function allScanFiles() {
function allScanFiles(root = ROOT) {
const files = [];
for (const p of SCAN_PATHS) walkMarkdown(p, files);
for (const p of SCAN_PATHS) walkMarkdown(p, files, root);
return files.filter((f) => {
const rel = path.relative(ROOT, f);
const rel = path.relative(root, f);
for (const skip of SKIP_DOC_FILES) {
if (rel === skip || rel.startsWith(skip + path.sep)) return false;
}
@@ -352,22 +409,39 @@ function allScanFiles() {
// ── Codebase index ─────────────────────────────────────────────────────────
function buildCodebaseIndex() {
// Env var helper wrappers used across OmniRoute — envInt(NAME, 5), envBool(NAME),
// envStr(NAME), … — read the named var from the environment, so a string-literal
// argument is a genuine env-var read, equivalent to a direct process.env member read.
// (Comment avoids a literal `process.env.<NAME>` token so the sibling env-doc-sync
// grep does not mistake this example for a real env-var read.)
const ENV_HELPER_CALL =
/\benv(?:Int|Bool|Str|Num|Float|String|Flag|Raw|List|Json)?\(\s*["'`]([A-Z][A-Z0-9_]+)["'`]/g;
export function buildCodebaseIndex(root = ROOT) {
// Set of /api/... paths that have a route.ts handler.
const apiRoutes = new Set();
// Set of /api/... prefixes that are an ancestor of (or equal to) a real route.ts.
// A documented prefix like /api/cloud/ is valid even without a route.ts at that
// exact level, as long as some src/app/api/cloud/**/route.ts exists.
const apiPrefixes = new Set();
// Map of /api/... → methods implemented in route.ts
const apiMethods = new Map();
function dynToBrace(seg) {
// [id] → {id}, [...path] → {path} — match the doc convention for dynamic segments.
return seg.replace(/^\[\.\.\.(.+)\]$/, "{$1}").replace(/^\[(.+)\]$/, "{$1}");
}
function walkApiRoutes(dir) {
const abs = path.join(ROOT, dir);
const abs = path.join(root, dir);
if (!fs.existsSync(abs)) return;
for (const name of fs.readdirSync(abs)) {
const child = path.join(abs, name);
const s = fs.statSync(child);
if (s.isDirectory()) walkApiRoutes(path.relative(ROOT, child));
if (s.isDirectory()) walkApiRoutes(path.relative(root, child));
else if (name === "route.ts" || name === "route.mjs") {
// Build the route path from the directory hierarchy
const rel = path.relative(ROOT, child).replace(/\\/g, "/");
const rel = path.relative(root, child).replace(/\\/g, "/");
const parts = rel.split("/");
// drop "src/app/api" and "route.ts"
parts.shift(); // src
@@ -378,6 +452,15 @@ function buildCodebaseIndex() {
apiRoutes.add(routePath);
apiRoutes.add(routePath + "/"); // trailing slash variant
// Register every ancestor prefix (and its {brace} dynamic-segment variant)
// so documented prefixes-with-subroutes resolve.
for (let i = 1; i <= parts.length; i++) {
const prefix = "/api/" + parts.slice(0, i).join("/");
const bracePrefix = "/api/" + parts.slice(0, i).map(dynToBrace).join("/");
apiPrefixes.add(prefix);
apiPrefixes.add(bracePrefix);
}
// Read the file to find exported HTTP methods
try {
const content = fs.readFileSync(child, "utf8");
@@ -397,29 +480,47 @@ function buildCodebaseIndex() {
}
walkApiRoutes("src/app/api");
// Set of env var names that are actually read in code.
// Set of env var names that are actually read in code, and the set of
// ALL_CAPS code identifiers (export const / enum / object-literal keys). A
// documented `UPPER_SNAKE` that resolves to a code identifier (e.g.
// LOCAL_ONLY_API_PREFIXES, HALF_OPEN) is NOT a fabricated env var.
const envVars = new Set();
const codeIdentifiers = new Set();
function walkForEnv(dir) {
const abs = path.join(ROOT, dir);
const abs = path.join(root, dir);
if (!fs.existsSync(abs)) return;
const skipDirs = new Set(["node_modules", ".next", "dist", ".build", "coverage"]);
for (const name of fs.readdirSync(abs)) {
if (skipDirs.has(name)) continue;
const child = path.join(abs, name);
const s = fs.statSync(child);
if (s.isDirectory()) walkForEnv(path.relative(ROOT, child));
if (s.isDirectory()) walkForEnv(path.relative(root, child));
else if (/\.(ts|tsx|js|mjs|cjs)$/.test(name)) {
try {
const content = fs.readFileSync(child, "utf8");
// process.env.X
const m1 = content.matchAll(/process\.env\.([A-Z][A-Z0-9_]+)/g);
for (const m of m1) envVars.add(m[1]);
for (const m of content.matchAll(/process\.env\.([A-Z][A-Z0-9_]+)/g)) envVars.add(m[1]);
// process.env["X"] / process.env['X'] (bracket notation)
for (const m of content.matchAll(/process\.env\[\s*["'`]([A-Z][A-Z0-9_]+)["'`]\s*\]/g))
envVars.add(m[1]);
// env.X (destructured in some handlers)
const m2 = content.matchAll(/\benv\.([A-Z][A-Z0-9_]+)\b/g);
for (const m of m2) envVars.add(m[1]);
for (const m of content.matchAll(/\benv\.([A-Z][A-Z0-9_]+)\b/g)) envVars.add(m[1]);
// env["X"] (bracket on a destructured env binding)
for (const m of content.matchAll(/\benv\[\s*["'`]([A-Z][A-Z0-9_]+)["'`]\s*\]/g))
envVars.add(m[1]);
// envInt("X", …) / envBool("X") / envStr("X") … helper wrappers
for (const m of content.matchAll(ENV_HELPER_CALL)) envVars.add(m[1]);
// import.meta.env.X (Vite-style, unlikely here but cheap)
const m3 = content.matchAll(/import\.meta\.env\.([A-Z][A-Z0-9_]+)/g);
for (const m of m3) envVars.add(m[1]);
for (const m of content.matchAll(/import\.meta\.env\.([A-Z][A-Z0-9_]+)/g))
envVars.add(m[1]);
// export const / const / let / var / enum NAME — JS identifiers, not env vars.
for (const m of content.matchAll(
/\b(?:export\s+)?(?:const|let|var|enum)\s+([A-Z][A-Z0-9_]{2,})\b/g
))
codeIdentifiers.add(m[1]);
// Object-literal / enum members on their own line: `HALF_OPEN: "HALF_OPEN"`.
for (const m of content.matchAll(/^[ \t]*([A-Z][A-Z0-9_]{2,})[ \t]*[:=]/gm))
codeIdentifiers.add(m[1]);
} catch {
/* ignore */
}
@@ -430,16 +531,42 @@ function buildCodebaseIndex() {
walkForEnv("open-sse");
walkForEnv("bin");
walkForEnv("scripts");
// Env vars that are only read by the test harness (e.g. RUN_CHAOS_INT) are still
// real env vars and must not be flagged as fabricated.
walkForEnv("tests");
// Env contract maintained by the sibling gate (check-env-doc-sync.mjs): a var
// listed in .env.example or docs/reference/ENVIRONMENT.md is, by definition, a
// documented OmniRoute env var (including external-CLI / docker / electron vars
// that are not read via process.env in our own source).
function readEnvContract() {
try {
const t = fs.readFileSync(path.join(root, ".env.example"), "utf8");
for (const line of t.split("\n")) {
const m = line.match(/^#?\s*([A-Z][A-Z0-9_]+)\s*=/);
if (m) envVars.add(m[1]);
}
} catch {
/* ignore */
}
try {
const t = fs.readFileSync(path.join(root, "docs", "reference", "ENVIRONMENT.md"), "utf8");
for (const m of t.matchAll(/`([A-Z][A-Z0-9_]{2,})`/g)) envVars.add(m[1]);
} catch {
/* ignore */
}
}
readEnvContract();
// Set of `omniroute <subcommand>` strings that exist in bin/
const cliCommands = new Set();
function walkCli(dir) {
const abs = path.join(ROOT, dir);
const abs = path.join(root, dir);
if (!fs.existsSync(abs)) return;
for (const name of fs.readdirSync(abs)) {
const child = path.join(abs, name);
const s = fs.statSync(child);
if (s.isDirectory()) walkCli(path.relative(ROOT, child));
if (s.isDirectory()) walkCli(path.relative(root, child));
else if (/\.(mjs|js|ts)$/.test(name)) {
try {
const content = fs.readFileSync(child, "utf8");
@@ -460,7 +587,7 @@ function buildCodebaseIndex() {
}
walkCli("bin");
return { apiRoutes, apiMethods, envVars, cliCommands };
return { apiRoutes, apiPrefixes, apiMethods, envVars, codeIdentifiers, cliCommands };
}
// ── Doc scanning ───────────────────────────────────────────────────────────
@@ -490,18 +617,44 @@ function lineOf(text, idx) {
return line;
}
function scanDocFile(absPath, index) {
const rel = path.relative(ROOT, absPath);
export function scanDocFile(absPath, index, root = ROOT) {
const rel = path.relative(root, absPath);
const text = fs.readFileSync(absPath, "utf8");
const textNoCode = stripCodeBlocksAndFences(text);
const findings = [];
// 1) API endpoints
for (const m of textNoCode.matchAll(COARSE_PATTERNS.apiPath)) {
const p = m[0].replace(/[\[\]\{\}]/g, ""); // strip wildcards for lookup
const candidate = p.replace(/\/$/, "");
if (ENDPOINT_ALLOWLIST.has(candidate) || ENDPOINT_ALLOWLIST.has(candidate + "/")) continue;
if (index.apiRoutes.has(candidate) || index.apiRoutes.has(candidate + "/")) continue;
// Normalize wildcard segments to {brace} form so [id] / [...path] / {id} all
// compare equally against the indexed routes/prefixes.
const normalized = m[0].replace(/\[\.\.\.(.+?)\]/g, "{$1}").replace(/\[(.+?)\]/g, "{$1}");
const candidate = normalized.replace(/\/$/, "");
const stripped = m[0].replace(/[\[\]\{\}]/g, "").replace(/\/$/, ""); // legacy lookup form
if (
ENDPOINT_ALLOWLIST.has(candidate) ||
ENDPOINT_ALLOWLIST.has(candidate + "/") ||
ENDPOINT_ALLOWLIST.has(stripped) ||
ENDPOINT_ALLOWLIST.has(stripped + "/")
)
continue;
if (
index.apiRoutes.has(stripped) ||
index.apiRoutes.has(stripped + "/") ||
index.apiRoutes.has(candidate) ||
index.apiRoutes.has(candidate + "/")
)
continue;
// Prefix-with-subroutes: a documented prefix like /api/cloud/ or /api/services/{name}/
// is valid when some src/app/api/<prefix>/**/route.ts exists. Accept when the
// documented path is (or is an ancestor of) a real route prefix, or vice-versa.
let isPrefix = false;
for (const pre of index.apiPrefixes) {
if (pre === candidate || pre.startsWith(candidate + "/") || candidate.startsWith(pre + "/")) {
isPrefix = true;
break;
}
}
if (isPrefix) continue;
// Allow docs that describe intended-but-not-yet-shipped routes by skipping lines that say "planned" / "TBD" / "future"
const ln = lineOf(text, m.index);
const lineText = text.split("\n")[ln - 1] || "";
@@ -525,11 +678,30 @@ function scanDocFile(absPath, index) {
if (ENV_VAR_ALLOWLIST.has(name)) continue;
if (!/_/.test(name)) continue; // real env vars have an underscore
if (index.envVars.has(name)) continue;
// Documented ALL_CAPS that resolves to a JS identifier (export const / enum /
// object-literal key) is a code symbol, not a fabricated env var.
// E.g. LOCAL_ONLY_API_PREFIXES, HALF_OPEN, A2A_SKILL_HANDLERS, MCP_SCOPE_PRESETS.
if (index.codeIdentifiers.has(name)) continue;
if (/^X-[A-Z]/.test(name)) continue;
if (ENV_VAR_DENYLIST.has(name)) continue;
const ln = lineOf(text, m.index);
const lineText = text.split("\n")[ln - 1] || "";
// Use the line as seen in the stripped text (where m.index is valid) for context
// checks — fenced-code removal shifts offsets, so text.split()[ln-1] can be wrong.
const lineStart = textNoCode.lastIndexOf("\n", m.index) + 1;
let lineEnd = textNoCode.indexOf("\n", m.index);
if (lineEnd === -1) lineEnd = textNoCode.length;
const lineText = textNoCode.slice(lineStart, lineEnd);
if (/example|placeholder|todo|tbd|\.\.\./i.test(lineText)) continue;
// A doc that explicitly states a var does NOT exist / is not implemented is
// documenting its absence, not fabricating it. Examples:
// "`MEMORY_RRF_VECTOR_WEIGHT` … do not exist"
// "a `ZED_CONFIG_PATH` environment variable override is not yet implemented"
if (
/\b(?:do(?:es)?\s+not\s+exist|no\s+such|not\s+a\s+real|isn't\s+a\s+real|never\s+(?:read|exists?)|not\s+(?:yet\s+)?(?:implemented|supported))\b/i.test(
lineText
)
)
continue;
findings.push({
kind: "env-var",
value: name,
@@ -582,10 +754,20 @@ function scanDocFile(absPath, index) {
// 5) File references
for (const m of textNoCode.matchAll(COARSE_PATTERNS.fileRef)) {
const ref = m[1].replace(/\\/g, "/");
const abs = path.join(ROOT, ref);
const abs = path.join(root, ref);
if (fs.existsSync(abs)) continue;
// Allow README/AGENTS to mention example files explicitly in a non-verified way
if (/\{\{|\.\.\./.test(ref)) continue; // templated / placeholder
// Tutorial placeholders in the "how to add a …" scenarios are intentional
// stand-ins, not real files: src/app/api/your-route/route.ts,
// src/lib/db/yourModule.ts, src/lib/guardrails/myGuardrail.ts, etc.
if (/(?:^|\/)(?:your-|your[A-Z]|my[A-Z])/.test(ref)) continue;
// Skip matches that are only the tail of a longer path, not a repo-root ref:
// the fileRef regex anchors on the `src`/`open-sse`/… token, so a leading `/`
// means the real reference is `<something>/src/...` — either a relative example
// path (`./src/index.ts`, a PII-pattern sample) or a workspace-package path
// (`@omniroute/opencode-provider/src/index.ts`). Neither resolves from repo root.
if (m.index > 0 && textNoCode[m.index - 1] === "/") continue;
const ln = lineOf(text, m.index);
findings.push({
kind: "file-ref",
@@ -601,12 +783,15 @@ function scanDocFile(absPath, index) {
// ── Main ───────────────────────────────────────────────────────────────────
export function runFabricatedDocsCheck(opts = {}) {
const index = buildCodebaseIndex();
const files = allScanFiles();
// `root` lets tests run the full pipeline against a fixture tree (with its own
// docs/, src/, .env.example) instead of the live repo.
const root = opts.root ?? ROOT;
const index = opts.index ?? buildCodebaseIndex(root);
const files = allScanFiles(root);
const allFindings = [];
for (const f of files) {
const result = scanDocFile(f, index);
const result = scanDocFile(f, index, root);
if (result.findings.length > 0) {
allFindings.push(result);
}
@@ -699,6 +884,10 @@ function main() {
console.log(formatHumanReport(result));
console.log();
if (totalFindings === 0) {
// Clean run — pass regardless of mode.
process.exit(0);
}
if (STRICT) {
console.error(`${totalFindings} claim(s) drift from source. Failing (--strict).`);
process.exit(1);

View File

@@ -69,7 +69,7 @@ For authoritative coverage requirements and test execution guidelines, see [`CON
### Adding a New Domain Module
1. Create `src/lib/db/[module].ts` with CRUD functions (create, read, update, delete, list)
2. Export from `src/lib/db/localDb.ts` (add re-export)
2. Export from `src/lib/localDb.ts` (add re-export)
3. If new tables required: create migration in `db/migrations/NNN_[description].sql`
4. Run migration via `migrationRunner.ts` at startup (automatic)
5. Add unit tests in `tests/unit/db/[module].test.mjs`

View File

@@ -9,6 +9,45 @@ import {
formatHumanReport,
} from "../../scripts/check/check-fabricated-docs.mjs";
// ── Fixture helpers ─────────────────────────────────────────────────────────
// The hardened checker accepts a `root` so we can run the full pipeline against a
// throwaway fixture repo (its own docs/, src/, .env.example) instead of the live
// tree. Each fixture is a minimal repo: a code surface to index + a single doc.
type Fixture = { docs?: Record<string, string>; files?: Record<string, string> };
function makeFixtureRoot(fx: Fixture): string {
const root = fs.mkdtempSync(path.join(os.tmpdir(), "fab-docs-"));
const write = (rel: string, content: string) => {
const abs = path.join(root, rel);
fs.mkdirSync(path.dirname(abs), { recursive: true });
fs.writeFileSync(abs, content);
};
// Always provide an AGENTS.md so allScanFiles() has at least one scan target.
write("AGENTS.md", fx.docs?.["AGENTS.md"] ?? "# Fixture\n");
for (const [rel, content] of Object.entries(fx.docs ?? {})) {
if (rel === "AGENTS.md") write(rel, content);
else write(path.join("docs", rel), content);
}
for (const [rel, content] of Object.entries(fx.files ?? {})) write(rel, content);
return root;
}
/** Returns the set of distinct `${kind}::${value}` findings for a fixture. */
function findingsFor(fx: Fixture): Set<string> {
const root = makeFixtureRoot(fx);
try {
const result = runFabricatedDocsCheck({ root });
const out = new Set<string>();
for (const f of result.files) {
for (const finding of f.findings) out.add(`${finding.kind}::${finding.value}`);
}
return out;
} finally {
fs.rmSync(root, { recursive: true, force: true });
}
}
// We can't easily mock the buildCodebaseIndex() which walks src/app/api etc.
// Instead, we test the report-formatting logic and the no-findings path on
// the real repo, which acts as a smoke test that the script runs end-to-end.
@@ -73,3 +112,203 @@ test("formatHumanReport: groups findings by kind", () => {
assert.match(out, /FAKE_VAR_X/);
assert.match(out, /onFake/);
});
// ─── Hardening: precision fixes (QG v2 Fase 9 T9) ───────────────────────────
//
// Each test runs the FULL pipeline (index + scan) over a fixture repo so it
// exercises the real heuristic, not a stubbed index.
test("env-var: an `export const` UPPER_SNAKE identifier in backticks is NOT flagged", () => {
const found = findingsFor({
files: {
"src/server/authz/routeGuard.ts":
"export const LOCAL_ONLY_API_PREFIXES: ReadonlyArray<string> = [];\n",
},
docs: {
"guide.md": "We expose `LOCAL_ONLY_API_PREFIXES` to gate loopback routes.\n",
},
});
assert.ok(
!found.has("env-var::LOCAL_ONLY_API_PREFIXES"),
"export const identifier must not be flagged as a fabricated env var"
);
});
test("env-var: an enum / object-literal member (HALF_OPEN) in backticks is NOT flagged", () => {
const found = findingsFor({
files: {
"src/shared/utils/circuitBreaker.ts":
'export const STATE = {\n CLOSED: "CLOSED",\n HALF_OPEN: "HALF_OPEN",\n} as const;\n',
},
docs: { "resilience.md": "After the reset window the breaker enters `HALF_OPEN`.\n" },
});
assert.ok(!found.has("env-var::HALF_OPEN"), "object-literal/enum key must not be flagged");
});
test('env-var: a var read via bracket notation process.env["X"] is NOT flagged', () => {
const found = findingsFor({
files: {
"src/lib/memory/vectorStore.ts": 'const k = Number(process.env["MEMORY_RRF_K"] ?? 60);\n',
},
docs: { "memory.md": "Tune ranking with `MEMORY_RRF_K`.\n" },
});
assert.ok(!found.has("env-var::MEMORY_RRF_K"), 'process.env["X"] read must be indexed');
});
test('env-var: a var read via an env helper (envInt("X")) is NOT flagged', () => {
const found = findingsFor({
files: {
"open-sse/config/constants.ts":
'const t = envInt("OMNIROUTE_CIRCUIT_BREAKER_OAUTH_THRESHOLD", 8);\n',
},
docs: { "cfg.md": "Override with `OMNIROUTE_CIRCUIT_BREAKER_OAUTH_THRESHOLD`.\n" },
});
assert.ok(
!found.has("env-var::OMNIROUTE_CIRCUIT_BREAKER_OAUTH_THRESHOLD"),
'envInt("X", …) helper read must be indexed'
);
});
test("env-var: a var read ONLY in tests/ (RUN_CHAOS_INT) is NOT flagged", () => {
const found = findingsFor({
files: {
"tests/integration/resilience-chaos.test.ts":
'const RUN = process.env.RUN_CHAOS_INT === "1";\n',
},
docs: { "testing.md": "Set `RUN_CHAOS_INT` to enable the chaos suite.\n" },
});
assert.ok(
!found.has("env-var::RUN_CHAOS_INT"),
"env vars read only in tests/ must be indexed, not flagged"
);
});
test("env-var: a var present only in .env.example is NOT flagged", () => {
const found = findingsFor({
files: { ".env.example": "# Contract\nSOME_DOCUMENTED_CONTRACT_VAR=value\n" },
docs: { "env.md": "Configure `SOME_DOCUMENTED_CONTRACT_VAR` in your environment.\n" },
});
assert.ok(
!found.has("env-var::SOME_DOCUMENTED_CONTRACT_VAR"),
".env.example is the env contract — its vars are documented, not fabricated"
);
});
test("api-path: a documented prefix with sub-routes (/api/cloud/) is NOT flagged", () => {
const found = findingsFor({
files: {
"src/app/api/cloud/auth/route.ts": "export async function GET() {}\n",
},
docs: { "cloud.md": "All cloud endpoints live under `/api/cloud/`.\n" },
});
assert.ok(
!found.has("api-path::/api/cloud/"),
"a prefix that is an ancestor of a real route.ts must resolve"
);
});
test("api-path: a dynamic-segment prefix (/api/services/{name}/) is NOT flagged", () => {
const found = findingsFor({
files: {
"src/app/api/services/[name]/status/route.ts": "export async function GET() {}\n",
},
docs: { "services.md": "Hit `/api/services/{name}/status` for health.\n" },
});
assert.ok(
!found.has("api-path::/api/services/{name}/status"),
"[name] dynamic segments must match the documented {name} convention"
);
});
test("hook: a real callback now in KNOWN_HOOKS (onChunk) is NOT flagged", () => {
const found = findingsFor({
docs: { "playground.md": "The stream invokes `onChunk` for each delta.\n" },
});
assert.ok(!found.has("hook::onChunk"), "onChunk is a real callback and must not be flagged");
});
test("file-ref: a tutorial placeholder (src/app/api/your-route/route.ts) is NOT flagged", () => {
const found = findingsFor({
docs: {
"scenario.md": "Create `src/app/api/your-route/route.ts` with GET/POST handlers.\n",
},
});
assert.ok(
!found.has("file-ref::src/app/api/your-route/route.ts"),
"your-* / my* placeholders in how-to scenarios must not be flagged"
);
});
// ─── ANTI-OVER-SUPPRESSION GUARD (the most important test) ──────────────────
//
// The hardening eliminates false positives by adding PRECISION, never by blinding
// the checker. A reference to something that genuinely does not exist MUST still be
// flagged — otherwise the gate is worthless.
test("ANTI-OVER-SUPPRESSION: a reference to a genuinely missing file IS still flagged", () => {
const found = findingsFor({
docs: { "bad.md": "See the handler in `src/nao/existe.ts` for details.\n" },
});
assert.ok(
found.has("file-ref::src/nao/existe.ts"),
"a non-existent file reference must remain flagged — precision must not blind detection"
);
});
test("ANTI-OVER-SUPPRESSION: a truly fabricated env var IS still flagged", () => {
const found = findingsFor({
// No code reads it, not a code identifier, not in .env.example.
docs: { "bad.md": "Set `TOTALLY_FABRICATED_ENV_VAR_XYZ` to enable nothing.\n" },
});
assert.ok(
found.has("env-var::TOTALLY_FABRICATED_ENV_VAR_XYZ"),
"a fabricated env var must remain flagged"
);
});
test("ANTI-OVER-SUPPRESSION: a fabricated API path with no backing route.ts IS still flagged", () => {
const found = findingsFor({
files: { "src/app/api/cloud/auth/route.ts": "export async function GET() {}\n" },
// /api/imaginary/* has no route.ts anywhere and is not an ancestor of one.
docs: { "bad.md": "Call `/api/imaginary/widget` to do magic.\n" },
});
assert.ok(
found.has("api-path::/api/imaginary/widget"),
"an API path with no backing route must remain flagged"
);
});
test("env-var: a doc explicitly stating a var does NOT exist is NOT flagged (documents absence)", () => {
const found = findingsFor({
docs: {
"memory.md":
"There are no env vars to tune weights (`MEMORY_RRF_VECTOR_WEIGHT` does not exist).\n",
},
});
assert.ok(
!found.has("env-var::MEMORY_RRF_VECTOR_WEIGHT"),
"documenting a var's absence is not fabricating it"
);
});
test("env-var: a doc saying an override is 'not yet implemented' is NOT flagged", () => {
const found = findingsFor({
docs: {
"zed.md": "A `ZED_CONFIG_PATH` environment variable override is not yet implemented.\n",
},
});
assert.ok(!found.has("env-var::ZED_CONFIG_PATH"), "not-yet-implemented is an absence statement");
});
test("ANTI-OVER-SUPPRESSION: a fabricated env var on a normal (non-negated) line IS still flagged", () => {
// Same shape as the negation cases above, but WITHOUT an absence statement —
// proves the negation skip is scoped to lines that disclaim the var, not a blanket
// suppression of every memory/config-looking name.
const found = findingsFor({
docs: { "bad.md": "Set `MADE_UP_TUNING_KNOB_FOR_NOTHING` to change behavior.\n" },
});
assert.ok(
found.has("env-var::MADE_UP_TUNING_KNOB_FOR_NOTHING"),
"a fabricated env var on a plain line must still be flagged"
);
});