fix(providers): register llm7 + route Cohere via compatibility layer (#2361 #2360)

Two related provider-registry fixes:

#2361: LLM7.io was visible in the dashboard provider catalog (entry in
src/shared/constants/providers.ts) but the executor registry in
open-sse/config/providerRegistry.ts had no llm7 entry. Every connection
test fell through with a credential error because there was no baseUrl
or authType configured. Added the standard OpenAI-compatible v1
endpoint with a small seed model catalogue.

#2360: Cohere was pointed at the native /v2/chat upstream which returns
the proprietary { message: { content: [{type:'text', text:...}] } }
shape. The combo test validator (extractComboTestResponseText) only
reads the OpenAI choices[] envelope, so a successful Cohere call
surfaced as 'Provider returned HTTP 200 but no text content.' Cohere
publishes an OpenAI-compatible compatibility layer at
/compatibility/v1 that returns { choices: [{ message: { content }}] }
so we route there instead of needing a Cohere-specific translator.

Regression test asserts both registry entries match the expected
shape so a future refactor that reverts the URLs fails loudly.

open-sse/config/providerRegistry.ts

@@ -2288,7 +2288,16 @@ export const REGISTRY: Record<string, RegistryEntry> = {
     alias: "cohere",
     format: "openai",
     executor: "default",
-    baseUrl: "https://api.cohere.com/v2/chat",
+    // Issue #2360: Cohere's native /v2/chat endpoint returns the upstream
+    // proprietary shape ({ message: { content: [{type:"text", text:...}] } })
+    // which the combo test validator (extractComboTestResponseText) does not
+    // know how to read, surfacing as "Provider returned HTTP 200 but no text
+    // content." Cohere also publishes an OpenAI-compatible compatibility
+    // layer at /compatibility/v1 that returns the canonical
+    // { choices: [{ message: { content: "..." } }] } shape, so we route
+    // through it instead of needing a Cohere-specific response translator.
+    baseUrl: "https://api.cohere.com/compatibility/v1/chat/completions",
+    modelsUrl: "https://api.cohere.com/compatibility/v1/models",
     authType: "apikey",
     authHeader: "bearer",
     models: [
@@ -3120,6 +3129,30 @@ export const REGISTRY: Record<string, RegistryEntry> = {
     models: CHAT_OPENAI_COMPAT_MODELS.bytez,
   },
 
+  // Issue #2361: LLM7.io was visible in the dashboard provider list
+  // (entry in `src/shared/constants/providers.ts`) but missing from the
+  // executor registry, so test-connection and chat requests had no
+  // baseUrl / authType to route to and returned a credential error.
+  // The provider exposes a standard OpenAI-compatible v1 endpoint with
+  // an optional bearer token (set the literal string "unused" when no
+  // key is configured, per upstream docs).
+  llm7: {
+    id: "llm7",
+    alias: "llm7",
+    format: "openai",
+    executor: "default",
+    baseUrl: "https://api.llm7.io/v1/chat/completions",
+    modelsUrl: "https://api.llm7.io/v1/models",
+    authType: "apikey",
+    authHeader: "bearer",
+    models: [
+      { id: "gpt-4o-mini-2024-07-18", name: "GPT-4o mini (LLM7)" },
+      { id: "gpt-4.1-nano-2025-04-14", name: "GPT-4.1 nano (LLM7)" },
+      { id: "deepseek-r1-0528", name: "DeepSeek R1 (LLM7)" },
+      { id: "qwen2.5-coder-32b-instruct", name: "Qwen2.5 Coder 32B (LLM7)" },
+    ],
+  },
+
   aimlapi: {
     id: "aimlapi",
     alias: "aiml",

tests/unit/provider-registry-llm7-cohere.test.ts

@@ -0,0 +1,46 @@
+/**
+ * Issues #2361 + #2360 — Registry entries for LLM7.io and Cohere.
+ *
+ * #2361: `llm7` was visible in the dashboard provider catalog (entry in
+ * `src/shared/constants/providers.ts`) but missing from the executor
+ * registry, so every connection attempt failed at the test step with a
+ * credential error. Verify the executor entry now exists with the public
+ * v1 base URL.
+ *
+ * #2360: Cohere was pointed at the native `/v2/chat` endpoint which
+ * returns the Cohere-proprietary shape — the combo validator could not
+ * extract any text and surfaced "Provider returned HTTP 200 but no text
+ * content." Verify the registry now uses the OpenAI-compatible
+ * compatibility layer.
+ */
+import test from "node:test";
+import assert from "node:assert/strict";
+
+const { REGISTRY } = await import("../../open-sse/config/providerRegistry.ts");
+
+test("#2361 llm7 is registered with the OpenAI-compatible v1 endpoint", () => {
+  const entry = (REGISTRY as Record<string, Record<string, unknown>>).llm7;
+  assert.ok(entry, "llm7 should be present in the executor registry");
+  assert.equal(entry.format, "openai");
+  assert.equal(entry.baseUrl, "https://api.llm7.io/v1/chat/completions");
+  assert.equal(entry.authType, "apikey");
+  assert.equal(entry.authHeader, "bearer");
+  assert.ok(Array.isArray(entry.models), "llm7 must expose a model catalogue");
+});
+
+test("#2360 cohere routes via the OpenAI-compatible compatibility layer", () => {
+  const entry = (REGISTRY as Record<string, Record<string, unknown>>).cohere;
+  assert.ok(entry, "cohere should be present in the executor registry");
+  assert.equal(entry.format, "openai");
+  // Must be the compatibility endpoint, NOT the native /v2/chat one
+  // (which returns the proprietary shape the combo validator cannot read).
+  assert.ok(
+    typeof entry.baseUrl === "string" &&
+      entry.baseUrl.includes("/compatibility/v1/chat/completions"),
+    `cohere baseUrl must use the OpenAI-compatible compatibility layer, got: ${entry.baseUrl}`
+  );
+  assert.ok(
+    typeof entry.modelsUrl === "string" && entry.modelsUrl.includes("/compatibility/v1/models"),
+    "cohere modelsUrl must use the compatibility endpoint so /v1/models import works"
+  );
+});