Add Lynkr router submission

model_cost/model_cost.json

@@ -326,5 +326,9 @@
   "gpt-5.4": {
     "input_token_price_per_million": 2.5,
     "output_token_price_per_million": 15.0
+  },
+  "z-ai/glm-4.7": {
+    "input_token_price_per_million": 0.4,
+    "output_token_price_per_million": 1.5
   }
 }

router_inference/config/lynkr.json

@@ -0,0 +1,19 @@
+{
+  "pipeline_params": {
+    "router_name": "lynkr",
+    "router_cls_name": "LynkrRouter",
+    "models": [
+      "openai/gpt-oss-120b",
+      "qwen/qwen3-235b-a22b-2507",
+      "z-ai/glm-4.7"
+    ],
+    "default_model": "openai/gpt-oss-120b",
+    "tier_map": {
+      "SIMPLE": "openai/gpt-oss-120b",
+      "MEDIUM": "qwen/qwen3-235b-a22b-2507",
+      "COMPLEX": "z-ai/glm-4.7",
+      "REASONING": "z-ai/glm-4.7"
+    },
+    "description": "Lynkr (github.com/Fast-Editor/Lynkr) — self-hosted LLM gateway; complexity-tier routing via WS7 anchor intent scoring (local embeddings), queried live through Lynkr's /routing/analyze endpoint."
+  }
+}

router_inference/router/__init__.py

@@ -9,6 +9,7 @@
 from router_inference.router.auto_router import auto_router
 from router_inference.router.chuzom_solo_v32 import ChuzomSoloV32Router
 from router_inference.router.llm_router import LLMRouter
+from router_inference.router.lynkr_router import LynkrRouter
 
 __all__ = [
     "BaseRouter",
@@ -17,4 +18,5 @@
     "auto_router",
     "LLMRouter",
     "ChuzomSoloV32Router",
+    "LynkrRouter",
 ]

router_inference/router/lynkr_router.py

@@ -0,0 +1,69 @@
+# SPDX-FileCopyrightText: Copyright contributors to the RouterArena project
+# SPDX-License-Identifier: Apache-2.0
+
+"""
+Lynkr router adapter.
+
+Queries a running Lynkr gateway (github.com/Fast-Editor/Lynkr) for its
+routing decision on each prompt and maps Lynkr's complexity tier
+(SIMPLE / MEDIUM / COMPLEX / REASONING) to a model from the config's
+tier_map. Lynkr's /routing/analyze?mode=intent endpoint runs the same
+WS7 anchor intent scorer used by its live proxy routing, without
+performing any LLM inference.
+
+Requires a Lynkr instance (default http://localhost:8081, override with
+LYNKR_URL). Lynkr's embedding backend (Ollama nomic-embed-text) should be
+up so scoring runs in anchor mode; if embeddings are unavailable Lynkr
+falls back to lexical scoring, which is also its live-serving behavior.
+"""
+
+import json
+import os
+import time
+import urllib.request
+
+from router_inference.router.base_router import BaseRouter
+
+
+class LynkrRouter(BaseRouter):
+    """Adapter that proxies routing decisions to a live Lynkr gateway."""
+
+    def __init__(self, router_name: str):
+        super().__init__(router_name)
+        params = self.config["pipeline_params"]
+        self.tier_map = params["tier_map"]
+        self.default_model = params.get("default_model", self.models[0])
+        self.base_url = os.environ.get("LYNKR_URL", "http://localhost:8081")
+        self.endpoint = f"{self.base_url}/routing/analyze?mode=intent"
+        for tier, model in self.tier_map.items():
+            if model not in self.models:
+                raise ValueError(
+                    f"tier_map[{tier}] = '{model}' is not in the config model pool"
+                )
+
+    def _get_prediction(self, query: str) -> str:
+        payload = json.dumps({"messages": [{"role": "user", "content": query}]}).encode(
+            "utf-8"
+        )
+        last_err = None
+        for attempt in range(3):
+            try:
+                req = urllib.request.Request(
+                    self.endpoint,
+                    data=payload,
+                    headers={"Content-Type": "application/json"},
+                    method="POST",
+                )
+                with urllib.request.urlopen(req, timeout=30) as resp:
+                    body = json.loads(resp.read().decode("utf-8"))
+                tier = body.get("tier")
+                if tier in self.tier_map:
+                    return self.tier_map[tier]
+                raise ValueError(f"Lynkr returned unknown tier: {tier!r}")
+            except (urllib.error.URLError, TimeoutError, ValueError) as err:
+                last_err = err
+                time.sleep(1 + attempt)
+        raise RuntimeError(
+            f"Lynkr routing failed after 3 attempts ({last_err}). "
+            f"Is Lynkr running at {self.base_url}?"
+        )

universal_model_names.py

@@ -121,6 +121,9 @@
     "gemini-2.5-flash-lite",
     "qwen3-235b-a22b-instruct-2507",
     "qwen3-30b-a3b-instruct-2507",
+    # Lynkr pool additions (all served via OpenRouter)
+    "openai/gpt-oss-120b",
+    "z-ai/glm-4.7",
 ]