prompt to use 75% of max tokens

src/synthia/miner/BaseLLM.py

@@ -34,7 +34,7 @@ def get_context_prompt(self, max_tokens: int) -> str:
             "yourself clear and concisely, crystallizing thoughts and "
             "key concepts. You only respond with the explanations themselves, "
             "eliminating redundant conversational additions. "
-            f"Try to keep your answer below {max_tokens} tokens"
+            f"Keep your answer below {int(max_tokens * 0.75)} tokens"
         )
         return prompt
 

src/synthia/miner/anthropic.py

@@ -33,7 +33,7 @@ def __init__(self, settings: AnthropicSettings | None = None) -> None:
             "yourself clear and concisely, crystallizing thoughts and "
             "key concepts. You only respond with the explanations themselves, "
             "eliminating redundant conversational additions. "
-            f"Try to keep your answer below {self.settings.max_tokens} tokens"
+            f"Keep your answer below {int(self.settings.max_tokens * 0.75)} tokens"
         )
 
     def prompt(self, user_prompt: str, system_prompt: str | None | NotGiven = None):
@@ -140,9 +140,9 @@ def prompt(self, user_prompt: str, system_prompt: str | None = None):
     log(f"Running module with key {key.ss58_address}")
     claude = OpenrouterModule()
     refill_rate = 1/400
-    bucket = TokenBucketLimiter(2, refill_rate)
+    bucket = TokenBucketLimiter(15, refill_rate)
     server = ModuleServer(
-        claude, key, ip_limiter=bucket, subnets_whitelist=[3]
+        claude, key, ip_limiter=bucket, subnets_whitelist=None
         )
     app = server.get_fastapi_app()
     uvicorn.run(app, host="127.0.0.1", port=8000)

src/synthia/validator/generate_data.py

@@ -25,7 +25,7 @@ def gen_explanation(
             "yourself clear and concisely, crystallizing thoughts and "
             "key concepts. You only respond with the explanations themselves, "
             "eliminating redundant conversational additions. "
-            f"Try to keep your answer below {self.llm.max_tokens} tokens"
+            f"Keep your answer below {int(self.llm.max_tokens * 0.75)} tokens"
         )
 
         user_prompt, criteria = explanation_prompt()