Merge pull request #49 from vicsanity623/vicsanity623-api-patch

runs every 6 hours

Author: vicsanity623

.github/workflows/pyob_service.yml

@@ -6,7 +6,7 @@ concurrency:
 
 on:
   schedule:
-    - cron: '0 * * * *'
+    - cron: '0 */6 * * *'
   workflow_dispatch:
 
 jobs:

src/pyob/core_utils.py

@@ -446,15 +446,19 @@ def on_chunk():
 
     def get_valid_llm_response(self, prompt: str, validator, context: str = "") -> str:
         """
-        SOTA LLM Orchestrator:
-        - Rotates Gemini Keys
-        - Pivots to GitHub Models (Phi-4) in the cloud
-        - Enforces mandatory bucket refills (60s) on empty responses
-        - Prevents high-frequency API spam
+        Elite LLM Orchestrator (v0.3.2):
+        - Prevents high-frequency Machine Gun API spam.
+        - Immediate Cloud Pivot (Gemini -> GitHub Models).
+        - Enforces mandatory 60s bucket refills on cloud failure.
+        - Multi-tier wait logic for rate-limit protection.
         """
         attempts = 0
         is_cloud = os.environ.get("GITHUB_ACTIONS") == "true"
 
+        logger.info(
+            f"📊 Engine check: Found {len(self.key_cooldowns)} Gemini API keys."
+        )
+
         while True:
             key = None
             now = time.time()
@@ -465,53 +469,53 @@ def get_valid_llm_response(self, prompt: str, validator, context: str = "") -> s
                 logger.info(
                     f"Attempting Gemini API Key {attempts % len(available_keys) + 1}/{len(available_keys)}"
                 )
-                response_text = self._stream_single_llm(
-                    prompt, key=key, context=context
-                )
             elif is_cloud:
-                logger.warning(
-                    "⏳ Gemini keys limited. Pivoting to GitHub Models (Phi-4)..."
-                )
-                response_text = self._stream_single_llm(
-                    prompt, key=None, context=context
-                )
+                logger.warning("⏳ Gemini keys limited. Using GitHub Models (Phi-4)...")
             else:
                 logger.info("🏠 Using Local Ollama Engine...")
-                response_text = self._stream_single_llm(
-                    prompt, key=None, context=context
-                )
 
-            if response_text.startswith("ERROR_CODE_429"):
-                if key:
-                    self.key_cooldowns[key] = time.time() + 1200
-                    logger.warning(f"⚠️ Key {key[-4:]} rate-limited (429). Rotating...")
-                else:
-                    logger.warning(
-                        "🚫 GitHub Models rate-limited. Sleeping 2 minutes..."
-                    )
-                    time.sleep(120)
-                attempts += 1
-                continue
+            response_text = self._stream_single_llm(prompt, key=key, context=context)
 
             if is_cloud and (
                 not response_text or response_text.startswith("ERROR_CODE_")
             ):
                 if key is not None:
+                    if "429" in response_text:
+                        self.key_cooldowns[key] = time.time() + 1200
+                        logger.warning(f"⚠️ Key {key[-4:]} rate-limited. Pivoting...")
+
                     logger.warning(
-                        "☁️ Gemini failed/limited. Pivoting to GitHub Models (Phi-4) immediately..."
+                        "☁️ Gemini blipped/limited. Pivoting to GitHub Models (Phi-4) immediately..."
                     )
                     response_text = self._stream_single_llm(
                         prompt, key=None, context=context
                     )
+
                 if not response_text or response_text.startswith("ERROR_CODE_"):
                     wait_time = 60
                     logger.warning(
-                        f"⚠️ All Cloud Engines failed. Sleeping {wait_time}s to refill tokens..."
+                        f"⚠️ All Cloud Engines exhausted. Sleeping {wait_time}s to refill tokens..."
                     )
                     time.sleep(wait_time)
                     attempts += 1
                     continue
 
+            if response_text.startswith("ERROR_CODE_429"):
+                if key:
+                    self.key_cooldowns[key] = time.time() + 1200
+                    logger.warning(f"⚠️ Key {key[-4:]} rate-limited (429). Rotating...")
+                else:
+                    logger.warning(
+                        "🚫 GitHub Models rate-limited. Sleeping 2 minutes..."
+                    )
+                    time.sleep(120)
+                attempts += 1
+                continue
+
+            if is_cloud and key:
+                logger.info("⏳ Rotating keys... (10s anti-spam breather)")
+                time.sleep(10)
+
             if not response_text or response_text.startswith("ERROR_CODE_"):
                 logger.warning("⚠️ Generic LLM error. Retrying in 10s...")
                 time.sleep(10)