Merge branch 'main' into fix/provider-registration-bugs

sources/agents/planner_agent.py

@@ -147,17 +147,23 @@ class PlannerAgent(Agent):
             pretty_print(f"{task['agent']} -> {task['task']}", color="info")
         pretty_print("▔▗ E N D ▖▔", color="status")
 
-    async def make_plan(self, prompt: str) -> str:
+    async def make_plan(self, prompt: str, max_retries: int = 4) -> str:
         """
         Asks the LLM to make a plan.
         Args:
             prompt (str): The prompt to be sent to the LLM.
+            max_retries (int): Maximum number of retries before giving up.
         Returns:
             str: The plan made by the LLM.
         """
         ok = False
         answer = None
+        retries = 0
         while not ok:
+            if retries >= max_retries:
+                pretty_print(f"Failed to make a plan after {max_retries} attempts. Giving up.", color="failure")
+                self.logger.warning(f"make_plan exceeded max retries ({max_retries}).")
+                return []
             animate_thinking("Thinking...", color="status")
             self.memory.push('user', prompt)
             answer, reasoning = await self.llm_request()
@@ -168,6 +174,7 @@ class PlannerAgent(Agent):
                 self.show_plan(agents_tasks, answer)
                 prompt = f"Failed to parse the tasks. Please write down your task followed by a json plan within ```json. Do not ask for clarification.\n"
                 pretty_print("Failed to make plan. Retrying...", color="warning")
+                retries += 1
                 continue
             self.show_plan(agents_tasks, answer)
             ok = True

sources/llm_provider.py

@@ -419,11 +419,13 @@ class Provider:
     def minimax_fn(self, history, verbose=False):
         """
         Use MiniMax API to generate text via OpenAI-compatible interface.
-        
+
         Supported models:
+        - MiniMax-M2.7: Latest flagship model with enhanced reasoning and coding
+        - MiniMax-M2.7-highspeed: High-speed version of M2.7 for low-latency scenarios
         - MiniMax-M2.5: Peak performance model (~60 tps), 204,800 context window
         - MiniMax-M2.5-highspeed: Same performance, faster (~100 tps)
-        
+
         Note: temperature must be in range (0.0, 1.0], default is 1.0
         """
         load_dotenv()