Align HTTP locator matching for declared/discovered duplicates

domains/guardrailsai.com/integrations.json

@@ -22,13 +22,15 @@
       "authStatus": "unknown",
       "name": "Guardrails Server REST API",
       "slug": "guardrails-server-rest-api",
-      "type": "http"
+      "type": "http",
+      "url": "http://localhost:8000"
     },
     {
       "authStatus": "unknown",
       "name": "Guardrails OpenAI-compatible endpoint",
       "slug": "guardrails-openai-compatible-endpoint",
-      "type": "http"
+      "type": "http",
+      "url": "http://localhost:8000/guards/{guard_name}/openai/v1"
     }
   ]
 }

domains/lakera.ai/integrations.json

@@ -14,7 +14,8 @@
       "authStatus": "none",
       "name": "Lakera self-hosted Kubernetes probe endpoints",
       "slug": "lakera-self-hosted-kubernetes-probe-endpoints",
-      "type": "http"
+      "type": "http",
+      "url": "http://<self-hosted-lakera-instance>:8000/{startupz,readyz,livez}"
     },
     {
       "authStatus": "required",

domains/portkey.ai/integrations.json

@@ -15,7 +15,8 @@
       "authStatus": "unknown",
       "name": "Portkey Admin API",
       "slug": "portkey-admin-api",
-      "type": "http"
+      "type": "http",
+      "url": "https://api.portkey.ai/v1/admin"
     },
     {
       "authStatus": "required",

src/lib/discover.test.ts

@@ -1,6 +1,7 @@
 import { describe, expect, test } from "bun:test";
 import { PROBE_KEYS } from "./conventions.ts";
 import { discover, slackMcpAppManifestUrl, type ChatFn, type WebBackend } from "./discover.ts";
+import { locatorOf } from "../../worker/operations.ts";
 import type { DetectionResult } from "./detect.ts";
 
 const web: WebBackend = {
@@ -189,3 +190,75 @@ describe("discover MCP onboarding overrides", () => {
     expect(setup).not.toContain("api.slack.com/apps?new_app=1");
   });
 });
+
+describe("declared HTTP merge with discovered HTTP", () => {
+  test("merges declared and discovered HTTP surfaces when both point to the same base URL", async () => {
+    const chat: ChatFn = async () => ({
+      message: { role: "assistant", content: null },
+      toolCalls: [
+        {
+          id: "surface-1",
+          name: "record_surface",
+          arguments: {
+            name: "Buddy REST API",
+            type: "http",
+            url: "https://api.buddy.works",
+            spec: "https://api.buddy.works/openapi.json",
+            authStatus: "none",
+            publicEvidence: ["https://docs.buddy.works/reference"],
+            basis: { via: "discovered", evidence: ["https://docs.buddy.works/reference"] },
+          },
+        },
+        {
+          id: "finish-1",
+          name: "finish",
+          arguments: {
+            summary: "Buddy provides a REST API.",
+            description: "Buddy helps teams build software faster.",
+          },
+        },
+      ],
+    });
+
+    const detect: DetectionResult = {
+      domain: "buddy.works",
+      found: [],
+      probed: [],
+      mcp: [],
+      errors: [],
+      integrationsJson: {
+        url: "https://buddy.works/.well-known/integrations.json",
+        result: {
+          version: 3,
+          surfaces: [
+            {
+              name: "Buddy REST API",
+              type: "http",
+              url: "https://api.buddy.works",
+              auth: { status: "none", basis: { via: "discovered", evidence: ["https://docs.buddy.works/reference"] } },
+            },
+          ],
+        },
+      },
+    };
+
+    const result = await discover(detect.domain, detect, chat, web);
+    expect(result).not.toBeNull();
+    expect(result?.surfaces).toHaveLength(1);
+    expect(result?.surfaces[0]?.type).toBe("http");
+    expect(result?.surfaces[0]?.url).toBe("https://api.buddy.works");
+  });
+});
+
+describe("slug continuity locator strategy", () => {
+  test("uses URL when available for HTTP locator continuity", () => {
+    expect(
+      locatorOf({
+        slug: "buddy-rest-api",
+        type: "http",
+        url: "https://api.buddy.works",
+        spec: "https://api.buddy.works/openapi.json",
+      }),
+    ).toBe("http|https://api.buddy.works");
+  });
+});

src/lib/discover.ts

@@ -202,6 +202,15 @@ export interface DiscoveryResult {
   surfaces: Surface[];
 }
 
+type SurfaceLocatorInput = {
+  type: string;
+  name?: string;
+  url?: string;
+  spec?: string;
+  command?: string;
+  packages?: readonly { identifier?: string }[];
+};
+
 // ── streamed events (partials emitted as findings are confirmed) ──────────────
 
 export type DiscoverEvent =
@@ -420,7 +429,7 @@ export async function discover(
             return `auth references undefined credential id(s): ${missing.join(", ")}. Call record_credential for each first (or use authStatus "unknown" if the credential isn't a real user-minted one), then re-record this surface.`;
           }
         }
-        const key = `${s.type}|${(s.spec || s.url || s.name).toLowerCase()}`;
+        const key = surfaceLocator(s);
         if (surfaceKeys.has(key)) return `Already recorded ${s.name}.`;
         surfaceKeys.add(key);
         s.slug = assignSlug(s.name, surfaces);
@@ -736,9 +745,16 @@ function markDeclaredSurface(surface: Surface, source: string): Surface {
   return surface;
 }
 
-function surfaceLocator(s: Surface): string {
-  const packageId = s.packages?.[0]?.identifier;
-  return `${s.type}|${(s.spec || s.url || s.command || packageId || s.name).toLowerCase()}`;
+export function surfaceLocatorValue(s: SurfaceLocatorInput): string | undefined {
+  const packageId = s.packages?.[0]?.identifier ?? undefined;
+  if (s.type === "cli") return s.command ?? packageId;
+  if (s.type === "http" || s.type === "graphql") return s.url ?? s.spec;
+  return s.spec ?? s.url ?? s.command ?? packageId;
+}
+
+export function surfaceLocator(s: Surface): string {
+  const value = surfaceLocatorValue(s) ?? s.name ?? "unknown";
+  return `${s.type}|${value.toLowerCase()}`;
 }
 
 function entryKey(entry: AuthEntry): string {

worker/operations.ts

@@ -9,7 +9,7 @@
  */
 import { Effect, Schema } from "effect";
 import { detect } from "../src/lib/detect.ts";
-import { discover, type ChatFn, type DiscoverEvent, type WebBackend } from "../src/lib/discover.ts";
+import { discover, surfaceLocatorValue, type ChatFn, type DiscoverEvent, type WebBackend } from "../src/lib/discover.ts";
 import { naiveWeb } from "../src/lib/contextdev.ts";
 import { Credential, CredentialType, DISCOVERY_VERSION, Surface } from "../src/lib/discovery-schema.ts";
 import { canonicalDomain } from "../src/lib/domain-aliases.ts";
@@ -114,7 +114,7 @@ type SlugSurface = {
 };
 
 export const locatorOf = (s: SlugSurface): string | undefined => {
-  const loc = s.type === "cli" ? (s.command ?? s.packages?.[0]?.identifier) : (s.url ?? s.spec);
+  const loc = surfaceLocatorValue(s);
   return loc ? `${s.type}|${loc.toLowerCase()}` : undefined;
 };