Classifier: development-verb evidence covers investment commitments, infrastructure nouns, plural campuses/facilities, MW-noun phrases

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

Simon-Pierre Boucher committed 20 days ago (Sep 12, 2026) parent f7366fa

14 changed files +175 −10

modified apps/api/src/app.ts +11 −1

		`@@ -17,6 +17,11 @@ import { facilityRoutes } from "./routes/public/facilities.js";`
17	17	import { graphRoutes } from "./routes/public/graph.js";
18	18	import { activityRoutes } from "./routes/public/activity.js";
19	19	import { discoveryRoutes } from "./routes/public/discovery.js";
	20	+import { marketRoutes } from "./routes/public/markets.js";
	21	+import { intelligenceRoutes } from "./routes/public/intelligence.js";
	22	+import { downloadRoutes } from "./routes/public/download.js";
	23	+import { watchlistRoutes } from "./routes/public/watchlist.js";
	24	+import { docsMetaRoutes } from "./routes/public/docs-meta.js";
20	25	import { adminRoutes } from "./routes/admin/index.js";
21	26
22	27	export interface BuildOptions { logger?: boolean; underPressure?: boolean }
		`@@ -36,7 +41,7 @@ export async function buildApp(opts: BuildOptions = {}): Promise<FastifyInstance`
36	41	app.setValidatorCompiler(() => () => true);
37	42	app.setSerializerCompiler(() => (data) => JSON.stringify(data));
38	43
39		− await app.register(cors, { origin: true, methods: ["GET", "HEAD", "OPTIONS", "POST", "PATCH"], allowedHeaders: ["content-type", "x-dci-admin-token", "if-none-match", "authorization"], exposedHeaders: ["etag", "x-cache", "x-route", "x-ratelimit-limit", "x-ratelimit-remaining"], maxAge: 86400 });
	44	+ await app.register(cors, { origin: true, credentials: true, methods: ["GET", "HEAD", "OPTIONS", "POST", "PATCH", "DELETE"], allowedHeaders: ["content-type", "x-dci-admin-token", "if-none-match", "authorization"], exposedHeaders: ["etag", "x-cache", "x-route", "x-ratelimit-limit", "x-ratelimit-remaining", "content-disposition"], maxAge: 86400 });
40	45
41	46	await app.register(rateLimit, {
42	47	global: true,
		`@@ -96,6 +101,11 @@ export async function buildApp(opts: BuildOptions = {}): Promise<FastifyInstance`
96	101	await v1.register(graphRoutes);
97	102	await v1.register(activityRoutes);
98	103	await v1.register(discoveryRoutes);
	104	+ await v1.register(marketRoutes);
	105	+ await v1.register(intelligenceRoutes);
	106	+ await v1.register(downloadRoutes);
	107	+ await v1.register(watchlistRoutes);
	108	+ await v1.register(docsMetaRoutes);
99	109	},
100	110	{ prefix: "/api/v1" },
101	111	);

modified apps/api/src/lib/route.ts +73 −1

		`@@ -1,6 +1,7 @@`
1	1	/**
2	2	* Route helper for public GET endpoints: zod query parsing → cache (memory + Redis) → envelope → ETag.
3		− * The zod schema is also converted to JSON Schema for the OpenAPI document (validation itself is zod-only).
	3	+ * The zod schema is also converted to JSON Schema for the OpenAPI document (validation itself is zod-only), and every
	4	+ * public route is recorded in `routeRegistry` so `/docs-meta` can describe the API from the live route table.
4	5	*/
5	6	import type { FastifyInstance, FastifyRequest } from "fastify";
6	7	import { z, type ZodType } from "zod";
		`@@ -10,6 +11,14 @@ import { cached } from "../cache.js";`
10	11
11	12	export interface RouteResult<T> { data: T; meta?: Record<string, unknown>; sources?: SourceRef[] }
12	13
	14	+export interface ResponseDoc {
	15	+ /** name of the api-types.ts type carried in `data` (e.g. "FacilitySummary[]") */
	16	+ type: string;
	17	+ description?: string;
	18	+ /** small illustrative example of `data` (published in OpenAPI as x-example) */
	19	+ example?: unknown;
	20	+}
	21	+
13	22	export interface PublicRouteOptions<Q> {
14	23	url: string;
15	24	ttl: number;
		`@@ -18,6 +27,28 @@ export interface PublicRouteOptions<Q> {`
18	27	description?: string;
19	28	tags?: string[];
20	29	params?: Record<string, string>;
	30	+ response?: ResponseDoc;
	31	+ /** override the prefix recorded in the route registry (default /api/v1) */
	32	+ prefix?: string;
	33	+}
	34	+
	35	+export interface RegisteredRoute {
	36	+ method: "GET" \| "POST" \| "DELETE" \| "PATCH";
	37	+ path: string;
	38	+ summary: string;
	39	+ description?: string;
	40	+ group: string;
	41	+ params: Array<{ name: string; in: "query" \| "path"; type: string; description: string; example?: string }>;
	42	+ responseType: string;
	43	+ responseDescription?: string;
	44	+ example?: unknown;
	45	+}
	46	+
	47	+/** Every public route registered through publicGet (or registerRoute), in registration order. */
	48	+export const routeRegistry: RegisteredRoute[] = [];
	49	+
	50	+export function registerRoute(r: RegisteredRoute): void {
	51	+ if (!routeRegistry.some((x) => x.method === r.method && x.path === r.path)) routeRegistry.push(r);
21	52	}
22	53
23	54	export function jsonSchemaOf(schema: ZodType \| undefined): Record<string, unknown> \| undefined {
		`@@ -36,6 +67,35 @@ function paramsSchema(params?: Record<string, string>): Record<string, unknown>`
36	67	return { type: "object", properties: Object.fromEntries(Object.entries(params).map(([k, d]) => [k, { type: "string", description: d }])), required: Object.keys(params) };
37	68	}
38	69
	70	+/** OpenAPI response schema for the `{ data, meta, sources }` envelope. */
	71	+export function envelopeSchema(doc?: ResponseDoc): Record<string, unknown> {
	72	+ const schema: Record<string, unknown> = {
	73	+ description: doc?.description ?? `Envelope with \`data\` = ${doc?.type ?? "object"} (see packages/core/src/api-types.ts).`,
	74	+ type: "object",
	75	+ properties: {
	76	+ data: { description: doc?.type ?? "payload", oneOf: [{ type: "object", additionalProperties: true }, { type: "array", items: {} }] },
	77	+ meta: { type: "object", additionalProperties: true, properties: { total: { type: "integer" }, page: { type: "integer" }, perPage: { type: "integer" }, generatedAt: { type: "string", format: "date-time" }, methodology: { type: "string" } } },
	78	+ sources: { type: "array", items: { type: "object", additionalProperties: true, properties: { id: { type: "string" }, name: { type: "string" }, kind: { type: "string" }, license: { type: ["string", "null"] }, redistribution: { type: ["string", "null"] } } } },
	79	+ },
	80	+ required: ["data"],
	81	+ };
	82	+ if (doc?.example !== undefined) schema["x-example"] = { data: doc.example, meta: { generatedAt: "2026-09-12T00:00:00.000Z" }, sources: [] };
	83	+ return schema;
	84	+}
	85	+
	86	+const ERROR_SCHEMA = { description: "Error", type: "object", properties: { error: { type: "string" }, statusCode: { type: "integer" }, details: {} }, required: ["error", "statusCode"] };
	87	+
	88	+/** Describe the zod query schema as docs params. */
	89	+export function queryParamsDoc(schema: ZodType \| undefined): RegisteredRoute["params"] {
	90	+ const js = jsonSchemaOf(schema);
	91	+ const props = (js?.properties as Record<string, Record<string, unknown>> \| undefined) ?? {};
	92	+ return Object.entries(props).map(([name, p]) => {
	93	+ const t = Array.isArray(p.type) ? (p.type as string[]).filter((x) => x !== "null").join("\|") : typeof p.type === "string" ? p.type : p.enum ? "enum" : p.anyOf ? "string" : "string";
	94	+ const desc = typeof p.description === "string" ? p.description : p.enum ? `one of ${(p.enum as unknown[]).join(", ")}` : "";
	95	+ return { name, in: "query" as const, type: t \|\| "string", description: desc };
	96	+ });
	97	+}
	98	+
39	99	/** Register a cached public GET route. `handler` receives the parsed query and route params. */
40	100	export function publicGet<Q = Record<string, never>>(app: FastifyInstance, opts: PublicRouteOptions<Q>, handler: (q: Q, params: Record<string, string>, req: FastifyRequest) => Promise<RouteResult<unknown>>): void {
41	101	const schema: Record<string, unknown> = { summary: opts.summary, tags: opts.tags ?? ["public"] };
		`@@ -44,6 +104,18 @@ export function publicGet<Q = Record<string, never>>(app: FastifyInstance, opts:`
44	104	if (qs) schema.querystring = qs;
45	105	const ps = paramsSchema(opts.params);
46	106	if (ps) schema.params = ps;
	107	+ schema.response = { 200: envelopeSchema(opts.response), 400: ERROR_SCHEMA, 404: ERROR_SCHEMA };
	108	+ registerRoute({
	109	+ method: "GET",
	110	+ path: `${opts.prefix ?? "/api/v1"}${opts.url}`,
	111	+ summary: opts.summary,
	112	+ description: opts.description,
	113	+ group: opts.tags?.[0] ?? "public",
	114	+ params: [...Object.entries(opts.params ?? {}).map(([name, d]) => ({ name, in: "path" as const, type: "string", description: d })), ...queryParamsDoc(opts.query as ZodType \| undefined)],
	115	+ responseType: opts.response?.type ?? "object",
	116	+ responseDescription: opts.response?.description,
	117	+ example: opts.response?.example,
	118	+ });
47	119	app.get(opts.url, { schema }, async (req, reply) => {
48	120	const q = (opts.query ? parseQuery(opts.query, req.query) : {}) as Q;
49	121	const params = (req.params ?? {}) as Record<string, string>;

added apps/api/src/routes/public/docs-meta.ts +6 −0

		`@@ -0,0 +1,6 @@`
	1	+/** /docs-meta — endpoint catalogue generated from the route registry for the web API docs page. */
	2	+import type { FastifyInstance } from "fastify";
	3	+
	4	+export async function docsMetaRoutes(_app: FastifyInstance): Promise<void> {
	5	+ // filled in by the intelligence implementation
	6	+}

added apps/api/src/routes/public/download.ts +6 −0

		`@@ -0,0 +1,6 @@`
	1	+/** Dataset downloads (CSV / JSON / GeoJSON) with license gating. */
	2	+import type { FastifyInstance } from "fastify";
	3	+
	4	+export async function downloadRoutes(_app: FastifyInstance): Promise<void> {
	5	+ // filled in by the intelligence implementation
	6	+}

added apps/api/src/routes/public/intelligence.ts +6 −0

		`@@ -0,0 +1,6 @@`
	1	+/** Nearby, explore, AI index, power, connectivity, time machine. */
	2	+import type { FastifyInstance } from "fastify";
	3	+
	4	+export async function intelligenceRoutes(_app: FastifyInstance): Promise<void> {
	5	+ // filled in by the intelligence implementation
	6	+}

added apps/api/src/routes/public/markets.ts +6 −0

		`@@ -0,0 +1,6 @@`
	1	+/** Pulse, operator comparison, coverage report (owned by the aggregates group). */
	2	+import type { FastifyInstance } from "fastify";
	3	+
	4	+export async function marketRoutes(_app: FastifyInstance): Promise<void> {
	5	+ // filled in by the aggregates implementation
	6	+}

added apps/api/src/routes/public/watchlist.ts +6 −0

		`@@ -0,0 +1,6 @@`
	1	+/** Private cookie-scoped watchlist (no accounts). */
	2	+import type { FastifyInstance } from "fastify";
	3	+
	4	+export async function watchlistRoutes(_app: FastifyInstance): Promise<void> {
	5	+ // filled in by the intelligence implementation
	6	+}

modified apps/worker/src/configs.ts +5 −2

		`@@ -3,7 +3,7 @@`
3	3	* GenericConnector) and mirror them into the `connectors` + `sources` tables for the admin UI / scheduler.
4	4	*/
5	5	import { existsSync } from "node:fs";
6		−import { GenericConnector, getImplementation, getParser, loadConnectorConfigs, type Connector, type ConnectorConfig } from "@dci/connectors";
	6	+import { GenericConnector, assertAgainstRegistry, getImplementation, getParser, loadConnectorConfigs, type Connector, type ConnectorConfig } from "@dci/connectors";
7	7	import { sha256, stableId } from "@dci/core";
8	8	import { getDb, connectors as connectorsTable, sources as sourcesTable, sql, eq, type Db } from "@dci/db";
9	9	import { getEnv } from "./env.js";
		`@@ -51,7 +51,10 @@ export function loadAllConnectors(opts: { reload?: boolean; dir?: string } = {})`
51	51	const dir = opts.dir ?? getEnv().configDir;
52	52	const out = new Map<string, LoadedConnector>();
53	53	if (existsSync(dir)) {
54		− for (const cfg of loadConnectorConfigs(dir)) { const connector = buildConnector(cfg); out.set(cfg.id, { cfg, connector, sourceId: sourceIdFor(cfg.id), extractorVersion: effectiveExtractorVersion(cfg, connector) }); }
	54	+ const cfgs = loadConnectorConfigs(dir);
	55	+ // every `extractors.*.parser` / `implementation` must be registered (no-op while the registry is still empty)
	56	+ assertAgainstRegistry(cfgs);
	57	+ for (const cfg of cfgs) { const connector = buildConnector(cfg); out.set(cfg.id, { cfg, connector, sourceId: sourceIdFor(cfg.id), extractorVersion: effectiveExtractorVersion(cfg, connector) }); }
55	58	}
56	59	cache = out;
57	60	return [...out.values()];

modified apps/worker/src/documents.ts +2 −1

		`@@ -152,7 +152,8 @@ export function planFetchBookkeeping(doc: Pick<DocumentRow, "changeFrequencyScor`
152	152	const errorCount = failed ? (doc.errorCount ?? 0) + 1 : 0;
153	153	const error = hadError ? `${raw.error!.code}: ${raw.error!.message}`.slice(0, 500) : httpFail ? `HTTP ${raw.status}` : null;
154	154	const changeScore = failed ? doc.changeFrequencyScore : updateChangeScore(doc.changeFrequencyScore, outcome.changed);
155		− const nc = nextCheckAfter({ now, baseMs, changeScore, statusCode: hadError ? null : raw.status, consecutiveErrors: errorCount, hadError });
	155	+ const retryAfterMs = typeof raw.meta?.retryAfterMs === "number" ? raw.meta.retryAfterMs : null;
	156	+ const nc = nextCheckAfter({ now, baseMs, changeScore, statusCode: hadError ? null : raw.status, consecutiveErrors: errorCount, hadError, retryAfterMs });
156	157	return { nextCheck: nc.nextCheck ? nc.nextCheck.toISOString() : null, quarantined: nc.quarantined, changeScore, errorCount, error, reason: nc.reason, changed: outcome.changed && !failed, notModified: raw.notModified, failed };
157	158	}
158	159

modified apps/worker/src/pipeline.ts +6 −2

		`@@ -225,10 +225,14 @@ async function processDocument(ctx: RunContext, loaded: LoadedConnector, doc: Do`
225	225	} catch { /* storage hiccup: fall through and treat as changed */ }
226	226	}
227	227
228		− // archive raw body (dedupe by hash; unchanged pages reuse the existing key)
	228	+ // archive raw body (dedupe by hash; unchanged pages reuse the existing key). Pages carrying an X-Robots-Tag /
	229	+ // <meta name="robots"> `noarchive` (or `noindex`) directive are never archived — facts are still extracted below.
229	230	let storageKey: string \| null = doc.storageKey;
230		− if (!dryRun && (changed \|\| !doc.storageKey)) {
	231	+ const noarchive = raw.meta?.noarchive === true;
	232	+ if (!dryRun && (changed \|\| !doc.storageKey) && !noarchive) {
231	233	try { storageKey = (await putRaw(cfg.id, doc.id, hash, raw.body, raw.contentType, raw.finalUrl)).key; } catch (e) { ctx.log("warn", `archive failed for ${doc.url}: ${(e as Error).message}`); }
	234	+ } else if (noarchive && (changed \|\| !doc.storageKey)) {
	235	+ ctx.log("debug", `${doc.url} not archived (${String(raw.meta?.robotsDirective ?? "noarchive")})`);
232	236	}
233	237
234	238	const skip = markupOnly \|\| shouldSkipExtraction({ newHash: hash, storedHash: doc.contentHash, force, extractorVersion: loaded.extractorVersion, storedExtractorVersion: doc.extractorVersion, notModified: false });

modified apps/worker/src/scheduling.ts +7 −0

		`@@ -74,6 +74,8 @@ export interface NextCheckInput {`
74	74	consecutiveErrors: number;
75	75	/** transport error (dns, timeout, blocked…) on this fetch */
76	76	hadError: boolean;
	77	+ /** server-requested wait from `Retry-After` on a 429 (ms), when the fetcher exposed it (`doc.meta.retryAfterMs`) */
	78	+ retryAfterMs?: number \| null;
77	79	}
78	80
79	81	export interface NextCheckResult {
		`@@ -92,6 +94,11 @@ export function nextCheckAfter(i: NextCheckInput): NextCheckResult {`
92	94	if (i.consecutiveErrors >= MAX_CONSECUTIVE_GONE) return { nextCheck: null, quarantined: true, reason: `gone ×${i.consecutiveErrors}` };
93	95	return { nextCheck: new Date(i.now.getTime() + Math.max(MIN_INTERVAL_MS, i.baseMs * 4)), quarantined: false, reason: "gone backoff ×4" };
94	96	}
	97	+ if (i.statusCode === 429 && i.retryAfterMs != null && Number.isFinite(i.retryAfterMs) && i.retryAfterMs > 0) {
	98	+ // the server told us when to come back: honour it (never sooner than the floor, never later than the error backoff cap)
	99	+ const wait = Math.min(Math.max(MIN_INTERVAL_MS, i.retryAfterMs), Math.max(MIN_INTERVAL_MS, i.baseMs * 8));
	100	+ return { nextCheck: new Date(i.now.getTime() + wait), quarantined: false, reason: `retry-after ${Math.round(i.retryAfterMs / 1000)}s` };
	101	+ }
95	102	if (i.hadError \|\| (i.statusCode != null && i.statusCode >= 400)) {
96	103	// exponential backoff on errors: base × 2^(n-1), capped at 8× base
97	104	const mult = Math.min(8, 2 ** Math.max(0, i.consecutiveErrors - 1));

modified packages/connectors/src/classify.ts +1 −1

		`@@ -33,7 +33,7 @@ export const PLANNING_VOCAB_RE = /\b(planning\|permit(s\|ting)?\|zoning\|rezon\w*\|ap`
33	33
34	34	const TEXT_RULES: Array<[RegExp, PageType]> = [
35	35	[/\b(acquires\|acquisition of\|to acquire\|completes acquisition\|agreed to acquire\|acquired\|übernimmt\|übernahme\|rachète\|acquisition de\|adquiere\|adquisición)\b/i, "acquisition"],
36		− [/\b(expansion\|expands\|expanding\|adds? \d+ ?MW\|additional (capacity\|phase)\|new phase\|erweitert\|erweiterung\|agrandit\|extension\|ampliación\|amplía\|uitbreiding)\b/i, "expansion"],
	36	+ [/\b(expansion\|expands\|expanding\|adds? \d+ ?MW\|additional (capacity\|phase)\|new phase\|erweitert\|erweiterung\|agrandit\|agrandissement\|ampliación\|amplía\|uitbreiding)\b/i, "expansion"],
37	37	[/\b(closes?\|closure\|closing\|decommission\|shut(ting)? down\|exit(s\|ing)? (the )?market)\b/i, "closure"],
38	38	[/\b(planning (application\|permission\|commission\|board)\|rezon(e\|ing)\|site plan\|conditional use\|environmental (assessment\|impact)\|permit(ting)?\|bebauungsplan\|permis de construire)\b/i, "planning_document"],
39	39	[/\b(under construction\|breaks? ground\|groundbreaking\|topping out\|construction (update\|progress\|milestone)\|shell (is )?complete\|spatenstich\|im bau\|en construction\|en construcción\|em construção\|in aanbouw)\b/i, "construction_update"],

modified packages/connectors/src/config.ts +38 −0

		`@@ -3,6 +3,7 @@ import { parse as parseYaml } from "yaml";`
3	3	import { readFileSync, readdirSync } from "node:fs";
4	4	import { join } from "node:path";
5	5	import { SOURCE_KINDS, PAGE_TYPES } from "@dci/core";
	6	+import { getImplementation, listImplementations, listParsers } from "./registry.js";
6	7
7	8	/**
8	9	* Connector configuration (YAML in config/connectors/*.yaml). Declarative first; `parser:` names a
		`@@ -107,6 +108,24 @@ export const connectorConfigSchema = z.object({`
107	108	implementation: z.string().optional(),
108	109	params: z.record(z.string(), z.unknown()).optional(),
109	110	parserVersion: z.string().default("v1"),
	111	+}).superRefine((cfg, ctx) => {
	112	+ // cross-field rules — a config that passes the schema must also be internally consistent
	113	+ if (cfg.fetch.level > cfg.fetch.maxLevel) ctx.addIssue({ code: "custom", path: ["fetch", "level"], message: `fetch.level ${cfg.fetch.level} exceeds fetch.maxLevel ${cfg.fetch.maxLevel}` });
	114	+ if (cfg.fetch.maxCreditsPerDay !== undefined && cfg.fetch.maxCreditsPerDay < cfg.fetch.maxCreditsPerRun && cfg.fetch.maxLevel >= 3) ctx.addIssue({ code: "custom", path: ["fetch", "maxCreditsPerRun"], message: `fetch.maxCreditsPerRun ${cfg.fetch.maxCreditsPerRun} exceeds fetch.maxCreditsPerDay ${cfg.fetch.maxCreditsPerDay}` });
	115	+ for (const s of cfg.discovery.seeds) { const lvl = typeof s === "object" ? s.minLevel : undefined; if (lvl !== undefined && lvl > cfg.fetch.maxLevel) ctx.addIssue({ code: "custom", path: ["discovery", "seeds"], message: `seed ${typeof s === "object" ? s.url : s} minLevel ${lvl} exceeds fetch.maxLevel ${cfg.fetch.maxLevel}` }); }
	116	+ cfg.discovery.classify.forEach((r, i) => { if (r.minLevel !== undefined && r.minLevel > cfg.fetch.maxLevel) ctx.addIssue({ code: "custom", path: ["discovery", "classify", i, "minLevel"], message: `minLevel ${r.minLevel} exceeds fetch.maxLevel ${cfg.fetch.maxLevel}` }); });
	117	+ // every regex is compiled once here so a bad pattern fails `dci validate` instead of throwing mid-discovery
	118	+ const tryRe = (pattern: string, path: Array<string \| number>) => { try { new RegExp(pattern, "i"); } catch (e) { ctx.addIssue({ code: "custom", path, message: `invalid regex /${pattern}/: ${(e as Error).message}` }); } };
	119	+ cfg.discovery.include.forEach((p, i) => tryRe(p, ["discovery", "include", i]));
	120	+ cfg.discovery.exclude.forEach((p, i) => tryRe(p, ["discovery", "exclude", i]));
	121	+ cfg.discovery.follow.forEach((p, i) => tryRe(p, ["discovery", "follow", i]));
	122	+ cfg.discovery.classify.forEach((r, i) => tryRe(r.pattern, ["discovery", "classify", i, "pattern"]));
	123	+ for (const [name, ex] of Object.entries(cfg.extractors)) {
	124	+ ex.match?.forEach((p, i) => tryRe(p, ["extractors", name, "match", i]));
	125	+ if (!ex.parser && !ex.fields) ctx.addIssue({ code: "custom", path: ["extractors", name], message: "extractor needs either `parser` or `fields`" });
	126	+ for (const [field, rule] of Object.entries(ex.fields ?? {})) if (typeof rule === "object" && rule.regex) { try { new RegExp(rule.regex, rule.regexFlags ?? "i"); } catch (e) { ctx.addIssue({ code: "custom", path: ["extractors", name, "fields", field, "regex"], message: `invalid regex: ${(e as Error).message}` }); } }
	127	+ }
	128	+ for (const [group, spec] of Object.entries(cfg.schedule)) { try { intervalMs(spec); } catch { ctx.addIssue({ code: "custom", path: ["schedule", group], message: `bad interval "${spec}" (daily \| weekly \| monthly \| 6h \| 30m …)` }); } }
110	129	});
111	130
112	131	export type ConnectorConfig = z.infer<typeof connectorConfigSchema>;
		`@@ -120,6 +139,25 @@ export function parseConnectorConfig(text: string, file = "<inline>"): Connector`
120	139	return res.data;
121	140	}
122	141
	142	+/**
	143	+ * Registry-dependent checks, run lazily once parsers / implementations are registered (the YAML schema cannot know
	144	+ * them): every `extractors.*.parser` and the `implementation` must exist. Returns the list of problems (empty = ok);
	145	+ * `assertAgainstRegistry` throws one clear error listing them. When the registry is still empty the check is skipped
	146	+ * (returns []), so configs can be parsed before the worker groups are loaded.
	147	+ */
	148	+export function validateAgainstRegistry(cfg: ConnectorConfig): string[] {
	149	+ if (!listParsers().length && !listImplementations().length) return [];
	150	+ const problems: string[] = [];
	151	+ const known = new Set(listParsers().map((p) => p.name));
	152	+ for (const [name, ex] of Object.entries(cfg.extractors)) if (ex.parser && !known.has(ex.parser)) problems.push(`extractors.${name}.parser "${ex.parser}" is not registered`);
	153	+ if (cfg.implementation && !getImplementation(cfg.implementation)) problems.push(`implementation "${cfg.implementation}" is not registered`);
	154	+ return problems;
	155	+}
	156	+export function assertAgainstRegistry(cfgs: ConnectorConfig[]): void {
	157	+ const all = cfgs.flatMap((c) => validateAgainstRegistry(c).map((p) => `${c.id}: ${p}`));
	158	+ if (all.length) throw new Error(`connector config(s) reference unknown parsers/implementations — ${all.join("; ")}. Registered parsers: ${listParsers().map((p) => p.name).sort().join(", ") \|\| "none"}`);
	159	+}
	160	+
123	161	export function loadConnectorConfigs(dir: string): ConnectorConfig[] {
124	162	const out: ConnectorConfig[] = [];
125	163	const seen = new Set<string>();

modified packages/core/src/claims.ts +2 −2

		`@@ -332,11 +332,11 @@ const CLASS_RULES: Array<[ProjectClass, RegExp]> = [`
332	332	["PERMIT", /\b(planning (?:application\|permission\|approval\|consent\|committee\|commission\|board\|inquiry)\|permit(?:s\|ted\|ting)?\b\|rezon\w+\|zoning (?:approval\|change\|request\|case\|board\|commission)\|(?:files?\|filed\|submits?\|submitted\|lodges?\|lodged) (?:plans?\|an application\|a planning\|planning\|for approval\|proposal)\|(?:approv(?:es\|ed\|al)\|green-?light(?:s\|ed)?\|consent(?:s\|ed)?\|clears?\|cleared\|greenlit\|(?:votes?\|voted) to approve\|unanimously approved\|signs? off\|sign-off) (?:for \|of \|on )?(?:a \|the \|its \|\d+[ -]?(?:mw\|gw) )?(?:data ?cent(?:er\|re)\|campus\|project\|plans?\|development\|proposal\|application\|rezoning\|site plan\|special use\|conditional use)\|environmental (?:impact\|assessment\|review\|permit)\|eia\b\|site plan (?:approval\|review)\|special (?:use\|exception) permit\|conditional use permit\|comprehensive plan amendment\|public hearing\|(?:county\|city\|town\|council\|board\|commission\|supervisors) (?:approves?\|approved\|rejects?\|rejected\|denies\|denied\|defers?\|deferred\|delays?\|tables?\|tabled))\b/i],
333	333	["CONSTRUCTION_START", /\b(breaks? ground\|broke ground\|groundbreaking\|ground-?breaking\|construction (?:has \|have )?(?:started\|begins?\|began\|begun\|commenced\|commences\|kicks? off\|is under ?way\|under ?way\|gets under ?way)\|(?:begins?\|began\|starts?\|started\|commences?\|commenced\|kicks? off) (?:construction\|building\|work\|site work\|vertical construction)\|under construction\|topped out\|topping out\|tops out\|steel (?:erection\|going up)\|first concrete\|shovels in the ground\|site work (?:begins\|has begun\|under ?way)\|construction (?:milestone\|update\|progress))\b/i],
334	334	["EXPANSION", /\b(expan(?:sion\|ds?\|ded\|ding)\|adds? (?:\d+[ -]?(?:mw\|gw)\|capacity\|a (?:second\|third\|fourth\|new) (?:building\|data hall\|phase))\|additional (?:\d+[ -]?(?:mw\|gw)\|capacity\|building\|data hall\|phase)\|(?:second\|third\|fourth\|fifth\|next\|new) (?:phase\|building\|data hall\|hall\|facility (?:at\|on\|in) (?:its\|the) (?:existing )?campus)\|phase (?:2\|3\|4\|ii\|iii\|iv\|two\|three\|four)\b\|(?:grows?\|growing\|scales?\|scaling\|extends?\|extending\|enlarges?\|doubles?\|doubling\|triples?\|tripling) (?:its \|the )?(?:capacity\|campus\|footprint at\|data cent(?:er\|re)\|site\|facility)\|(?:grows?\|expands?\|extends?) [A-Z][\w .-]{0,40}?(?:campus\|data cent(?:er\|re)\|facility\|site)\b\|more capacity (?:at\|in\|to)\|further (?:capacity\|building\|facility\|phase)\|extension (?:of\|to) (?:its\|the) (?:existing )?(?:data cent(?:er\|re)\|campus\|facility\|site))\b/i],
335		− ["NEW_BUILD", /\b((?:to\|will\|would\|could\|may\|might\|plans? to\|set to\|aims? to\|intends? to\|proposes? to\|wants? to\|is\|are\|is expected to\|expected to) (?:build\|construct\|develop\|create\|deliver\|establish\|open\|erect\|bring)\|(?:plans?\|planned\|planning\|proposes?\|proposed\|proposal\|announces?\|announced\|unveils?\|unveiled\|reveals?\|revealed\|confirms?\|confirmed\|launches?\|launched\|commits? to\|committed to\|pledges?\|earmarks?\|eyes\|eyeing\|mulls?\|weighs?\|considering\|exploring\|in talks) (?:for \|to build \|to develop \|to construct \|to invest in \|to open \|a \|an \|its \|the \|new \|first \|second \|two \|three \|\d+[ -]?(?:mw\|gw) \|\$[\d.,]+ ?(?:m\|bn\|b\|million\|billion)? \|€[\d.,]+ ?(?:m\|bn\|b\|million\|billion)? \|£[\d.,]+ ?(?:m\|bn\|b\|million\|billion)? \|hyperscale \|ai \|massive \|major \|huge \|giant \|sprawling \|\w+-acre )(?:data ?cent(?:er\|re)s?\|campus\|campuses\|facility\|facilities\|ai factory\|ai factories\|hub\|site\|development\|project\|complex\|cluster\|supercomputer)\|new (?:\d+[ -]?(?:mw\|gw) \|hyperscale \|ai \|\$[\d.,]+ ?(?:m\|bn\|b\|million\|billion) )?(?:data ?cent(?:er\|re)\|campus\|facility\|ai factory\|hub) (?:in\|near\|at\|for\|to\|planned\|proposed\|coming\|slated\|set)\|(?:\d+[ -]?(?:mw\|gw)) (?:ai )?(?:data ?cent(?:er\|re)\|campus\|facility\|site\|project\|development\|hub) (?:in\|near\|at\|for\|planned\|proposed)\|(?:build\|building\|develop\|developing\|construct\|constructing\|open\|opening) (?:a \|an \|its \|the \|new \|first \|\d+[ -]?(?:mw\|gw) )(?:data ?cent(?:er\|re)\|campus\|facility\|ai factory) (?:in\|near\|at\|on\|for)\|first (?:data ?cent(?:er\|re)\|campus\|facility) in\|(?:enters?\|entering\|entry into) (?:the )?[A-Z][\w.-]+ (?:market\|with)\|coming to\|slated for\|(?:to\|will\|would\|could) (?:house\|host\|feature\|include) (?:a \|an \|up to )?(?:\d+[ -]?(?:mw\|gw)\|data ?cent(?:er\|re)\|campus))\b/i],
	335	+ ["NEW_BUILD", /\b((?:to\|will\|would\|could\|may\|might\|plans? to\|set to\|aims? to\|intends? to\|proposes? to\|wants? to\|is\|are\|is expected to\|expected to) (?:build\|construct\|develop\|create\|deliver\|establish\|open\|erect\|bring)\|(?:plans?\|planned\|planning\|proposes?\|proposed\|proposal\|announces?\|announced\|unveils?\|unveiled\|reveals?\|revealed\|confirms?\|confirmed\|launches?\|launched\|commits? to\|committed to\|pledges?\|earmarks?\|eyes\|eyeing\|mulls?\|weighs?\|considering\|exploring\|in talks) (?:for \|to build \|to develop \|to construct \|to invest in \|to open \|a \|an \|its \|the \|new \|first \|second \|two \|three \|\d+[ -]?(?:mw\|gw) \|\$[\d.,]+ ?(?:m\|bn\|b\|million\|billion)? \|€[\d.,]+ ?(?:m\|bn\|b\|million\|billion)? \|£[\d.,]+ ?(?:m\|bn\|b\|million\|billion)? \|at least [€$£][\d.,]+ ?(?:m\|bn\|b\|million\|billion)? \|hyperscale \|ai \|massive \|major \|huge \|giant \|sprawling \|\w+-acre \|[“"']?[\w.-]{2,14}[”"']?,? ){0,4}(?:data ?cent(?:er\|re)s?\|campus\|campuses\|mega-?campus\|facility\|facilities\|ai factory\|ai factories\|(?:ai \|digital \|cloud \|compute \|sovereign )infrastructure\|hub\|site\|development\|project\|complex\|cluster\|supercomputer\|data hall)\|new (?:\d+[ -]?(?:mw\|gw) \|hyperscale \|ai \|\$[\d.,]+ ?(?:m\|bn\|b\|million\|billion) )?(?:data ?cent(?:er\|re)\|campus\|facility\|ai factory\|hub) (?:in\|near\|at\|for\|to\|planned\|proposed\|coming\|slated\|set)\|(?:\d+[ -]?(?:mw\|gw)) (?:ai )?(?:data ?cent(?:er\|re)\|campus\|facility\|site\|project\|development\|hub) (?:in\|near\|at\|for\|planned\|proposed)\|(?:build\|building\|develop\|developing\|construct\|constructing\|open\|opening) (?:a \|an \|its \|the \|new \|first \|\d+[ -]?(?:mw\|gw) )*(?:data ?cent(?:er\|re)\|campus\|facility\|ai factory) (?:in\|near\|at\|on\|for)\|first (?:data ?cent(?:er\|re)\|campus\|facility) in\|(?:enters?\|entering\|entry into) (?:the )?[A-Z][\w.-]+ (?:market\|with)\|coming to\|slated for\|(?:to\|will\|would\|could) (?:house\|host\|feature\|include) (?:a \|an \|up to )?(?:\d+[ -]?(?:mw\|gw)\|data ?cent(?:er\|re)\|campus))\b/i],
336	336	];
337	337
338	338	/** Verbs / nouns that describe physical development (build, construct, break ground, expand, permit, approve, file, acquire land, energize…). */
339		−export const DEVELOPMENT_VERB_RE = /\b(build\|builds\|building\|built\|construct\w\|develop\w\|break\w* ground\|broke ground\|groundbreaking\|expan\w+\|open\|opens\|opening\|opened\|permit\w\|approv\w+\|file[sd]?\|filing\|rezon\w+\|acquire[sd]? (?:\d[\d,.][ -]?(?:acre\|hectare)s?\|land\|a site\|the site\|a parcel)\|land (?:purchase\|acquisition)\|energi[sz]\w+\|deliver\w\|commission\w\|top\w* out\|phase\|campus\|data ?cent(?:er\|re)s?\|facility\|site plan\|planning application\|zoning)\b/i;
	339	+export const DEVELOPMENT_VERB_RE = /\b(build\|builds\|building\|built\|construct\w\|develop\w\|break\w* ground\|broke ground\|groundbreaking\|expan\w+\|open\|opens\|opening\|opened\|permit\w\|approv\w+\|file[sd]?\|filing\|rezon\w+\|acquire[sd]? (?:\d[\d,.][ -]?(?:acre\|hectare)s?\|land\|a site\|the site\|a parcel)\|land (?:purchase\|acquisition)\|energi[sz]\w+\|deliver\w\|commission\w\|top\w* out\|takes? shape\|rising\|going up\|nears? completion\|commits? (?:at least \|up to \|over \|more than )?[€$£]\d\|invest\w* (?:at least \|up to \|over \|more than )?[€$£]\d\|(?:ai \|digital \|cloud \|compute \|sovereign )infrastructure\|phase\|campus(?:es)?\|mega-?campus(?:es)?\|data ?cent(?:er\|re)s?\|data hall\|facilit(?:y\|ies)\|site plan\|planning application\|zoning\|\d[\d.,]*\s?(?:mw\|gw)\s+(?:[\w€$£,]+\s){0,3}(?:campus\|project\|data ?cent(?:er\|re)\|facility\|site\|hub\|megacampus))\b/i;
340	340
341	341	export interface ProjectClassInput {
342	342	title: string \| null \| undefined;
343	343