Lower MAX_SUBTASKS

README.md

@@ -17,7 +17,7 @@ tags:
 ---
 
 [Demo](https://huggingface.co/spaces/MCP-1st-Birthday/MedLLM-Agent)  
-
+[LinkedIn Post](https://www.linkedin.com/posts/dang-khoa-le-96a6332a8_medllm-agent-a-hugging-face-space-by-mcp-activity-7396162709618225152-WSt4?utm_source=social_share_send&utm_medium=member_desktop_web&rcm=ACoAAEoh1KIB2Np4-PEynHrT7uchKFL5ByBIpio)
 
 # 🩺 MedLLM Agent
 
@@ -79,6 +79,7 @@ Sample [conversation](sample.md) between MAC-system and patient are recorded.
 - **MedSwin Specialist**: Executes tasks sequentially, providing concise clinical answers
 - **Enhanced Synthesis**: Supervisor synthesizes all MedSwin responses with clear context into comprehensive final answers
 - **Iterative Improvement**: Supervisor challenges and enhances answers until confirmed optimal (up to 2 iterations)
+- **History-Aware Follow-Ups**: In agentic modes, Gemini uses the last assistant answer when breaking down follow-up queries (e.g., "clarify your answer") so subtasks stay grounded in prior responses
 - **Search Mode**: Gemini creates 1-4 search strategies → executes ddgs searches (1-2 sources each) → summarizes briefly
 - **Conditional Search Trigger**: When search mode enabled, supervisor can trigger additional searches if answer is unclear or has gaps
 - **RAG Mode**: Gemini brainstorms retrieved documents into 1-4 short contexts for MedSwin decision-making

supervisor.py

@@ -9,7 +9,7 @@ from config import GEMINI_MODEL, GEMINI_MODEL_LITE
 from utils import format_prompt_manually
 
 # Maximum number of subtasks for query breakdown
-MAX_SUBTASKS = 5
+MAX_SUBTASKS = 3
 # Maximum number of search strategies
 MAX_SEARCH_STRATEGIES = 3