@tiens.nguyen/gonext-local-worker 1.0.132 → 1.0.134

package/gonext_agent_chat.py

@@ -66,6 +66,22 @@ def _http_request_impl(method, url, headers=None, body=None, timeout=25):
         return f"Error: {e}"
 
 
+def _normalize_openai_base(url: str) -> str:
+    """Normalize a model-server URL to its OpenAI-compatible /v1 root.
+
+    Users paste whatever their server docs show: an MLX root (http://host:8089), an
+    Ollama native endpoint (http://host:11434/api/generate), or a full /v1 URL. Ollama
+    serves the OpenAI-compatible API at /v1 on the same host, so strip any native
+    /api/... path (including a mistakenly appended /v1 after it) and ensure /v1.
+    """
+    u = (url or "").strip().rstrip("/")
+    if not u:
+        return ""
+    u = re.sub(r"/api(/(generate|chat|tags|embeddings|embed))?(/v1)?$", "", u,
+               flags=re.IGNORECASE)
+    return u if u.lower().endswith("/v1") else u + "/v1"
+
+
 def _html_to_text(html_text, limit=3000):
     """Strip HTML to readable plain text (zero-dep). Used by fetch_url so the weak
     model receives prose instead of raw tags it cannot parse."""
@@ -934,7 +950,10 @@ 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.
-    raw_coding_base = (cfg.get("codingBaseURL") or "").strip()
+    # The URL may be a pasted Ollama endpoint (http://host:11434/api/generate) or a
+    # bare host — normalize to the OpenAI-compatible /v1 root either way (defensive:
+    # also fixes an older API layer that appended /v1 after the native /api path).
+    raw_coding_base = _normalize_openai_base(cfg.get("codingBaseURL") or "")
     raw_coding_model = (cfg.get("codingModelId") or "").strip()
     if raw_coding_base:
         same_server = raw_coding_base.rstrip("/") == (agent_base_url or "").rstrip("/")

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@tiens.nguyen/gonext-local-worker",
-  "version": "1.0.132",
+  "version": "1.0.134",
   "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",