@tiens.nguyen/gonext-local-worker 1.0.82 → 1.0.84

package/gonext_agent_chat.py

@@ -64,6 +64,32 @@ def _http_request_impl(method, url, headers=None, body=None, timeout=25):
         return f"Error: {e}"
 
 
+def _detect_model_id(base_url, api_key=""):
+    """Ask an OpenAI-compatible server which model it serves.
+
+    Queries GET {base_url}/models and returns the first reported model id.
+    `base_url` already ends with /v1. Returns "" on any failure so callers can
+    fall back. Used when the user supplies a coding-model URL but no model name.
+    """
+    url = base_url.rstrip("/") + "/models"
+    headers = {"Accept": "application/json"}
+    if api_key and api_key != "local":
+        headers["Authorization"] = f"Bearer {api_key}"
+    req = urllib.request.Request(url, method="GET", headers=headers)
+    try:
+        with urllib.request.urlopen(req, timeout=10, context=_ssl_context()) as resp:
+            payload = json.loads(resp.read().decode("utf-8", errors="replace"))
+    except Exception as e:  # noqa: BLE001
+        _log(f"model detect failed {url}: {e}")
+        return ""
+    data = payload.get("data") if isinstance(payload, dict) else None
+    if isinstance(data, list) and data:
+        first = data[0]
+        if isinstance(first, dict) and isinstance(first.get("id"), str):
+            return first["id"].strip()
+    return ""
+
+
 def _summarise_step(step_log):
     """Return a short human-readable description of an agent step."""
     tool_calls = getattr(step_log, "tool_calls", None) or []
@@ -249,8 +275,26 @@ def run_agent_chat(cfg):
     # Optional dedicated coding/reasoning model for the CodeAgent's tool-use loop.
     # Routing, plain replies and summarization stay on the chat model (better at
     # natural language); the code model only drives http_request reasoning.
-    coding_base_url = (cfg.get("codingBaseURL") or "").strip() or agent_base_url
-    coding_model_id = (cfg.get("codingModelId") or "").strip() or agent_model_id
+    raw_coding_base = (cfg.get("codingBaseURL") or "").strip()
+    raw_coding_model = (cfg.get("codingModelId") or "").strip()
+    if raw_coding_base:
+        # A dedicated coding server is configured. If no model name was given,
+        # ask the server which model it serves (mlx_lm.server otherwise tries to
+        # download a mismatched name from HF and 404s).
+        detected = raw_coding_model or _detect_model_id(raw_coding_base, agent_api_key)
+        if detected:
+            coding_base_url = raw_coding_base
+            coding_model_id = detected
+        else:
+            _log(
+                f"coding model id unresolved for {raw_coding_base!r}; "
+                "falling back to chat model"
+            )
+            coding_base_url = agent_base_url
+            coding_model_id = agent_model_id
+    else:
+        coding_base_url = agent_base_url
+        coding_model_id = agent_model_id
     max_steps = int(cfg.get("maxSteps") or 5)
 
     _log(

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tiens.nguyen/gonext-local-worker",
-  "version": "1.0.82",
+  "version": "1.0.84",
   "description": "Polls GoNext cloud API for async local LLM jobs and runs them against Ollama/OpenAI-compatible servers on this Mac",
   "type": "module",
   "license": "MIT",