SeoBuildIntelligencePort has no implementation and no caller, and its request types cannot compile against the live SEO-Bot contract

[cryptoxdog]

Summary

src/intelligence/SeoBuildIntelligencePort.ts defines the SEO build-intelligence seam, but nothing implements it and nothing calls it. It is a contract surface with no runtime behind it. Separately, four of its request fields do not match the live SEO-Bot endpoint schemas, so a naive adapter written against it would be rejected at the boundary.

Observed at Website-Bot 70f3f9a4, SEO-Bot 519879e6.

The seam is inert

$ grep -rn "implements SeoBuildIntelligencePort\|SeoBuildIntelligencePort" src/ scripts/ tests/
src/intelligence/SeoBuildIntelligencePort.ts:47:export interface SeoBuildIntelligencePort {
tests/unit/seo-build-intelligence-port.test.ts:12:import { assertWebsiteBlueprintLandscape } from ...

$ grep -rn "build-intelligence" src/ scripts/
(no matches)

Zero implementations, zero call sites, and no code anywhere references the three endpoints. compilePageContentContract (src/intelligence/compile-page-content-contract.ts) is correctly deterministic and zero-LLM — and also has no caller.

Contract mismatches vs. live SEO-Bot

SEO-Bot's request bodies are z.object({...}).strict() (SEO-Bot/src/api/build-intelligence.ts:60-101), so extra or wrongly-typed fields are a hard 400, not a coercion.

Field	Website-Bot port	SEO-Bot schema	Result
seed_queries	string[]	Array<{query, intent, weight?}>	incompatible — 400
desired_donor_count	required	optional (default 10)	over-constrained
market.device	required	optional	over-constrained
operator_exclusions	absent	supported	capability lost
seo_config	absent	supported	capability lost

The intent discriminator is not decorative: competitive-landscape.ts uses it to build the query portfolio, and weight feeds visibilityContribution(weight, rank). Flattening to string[] discards the ranking-truth inputs.

Auth

Build-intelligence routes are not auth-exempt (SEO-Bot/src/api/security.ts — exempt list is /health and /api/clients/register only). Operator auth accepts Bearer <token> or HTTP Basic (password portion), compared constant-time against OPERATOR_API_KEY. Fail-closed: if OPERATOR_API_KEY is unset, every protected route returns 401.

Expected

A single canonical real adapter (SeoBuildIntelligenceHttpAdapter or repo-native equivalent) performing authenticated HTTP against:

• POST /api/build-intelligence/competitive-landscape
• POST /api/build-intelligence/seo-content-blueprint
• POST /api/build-intelligence/structured-content

with typed requests/responses, timeout, bounded retries only where safe, fail-closed auth and malformed-response handling, artifact schema validation, ArtifactRef validation, payload-digest validation, and exact input-lineage validation. No provider/model/temperature/raw-prompt fields in application requests (SEO-Bot's .strict() bodies already reject these at the boundary — the adapter must not attempt them).

Port request types must be corrected to the five points above first, or the adapter cannot typecheck against reality.

Explicitly not an SEO-Bot defect

SEO-Bot's side of this seam is correct and needs no change. The mismatch is entirely in the Website-Bot port declaration.