Protocol, Evals + Draft App

README.md

@@ -127,7 +127,7 @@ This project was developed and tested on the following system configuration:
 
 1. **Start the MCP server**
    ```bash
-   python mcp_server.py
+   python run_call_centre_server.py
    ```
    Server will be available at `http://localhost:8001/mcp`
 

agent_planning_workflow.py

@@ -438,12 +438,13 @@ async def run_plan(
 
                 prompt = prompt_map[step_name]
                 # Most prompts need schema, user input
-                schema = memory.get("get_experiment_parameter_schema")
+                schema = memory.get("get_schema")
                 result = await client.get_prompt(
                     prompt.name, {"schema": schema, "user_input": user_input}
                 )
                 llm_prompt_text = result.messages[0].content.text
-                # Actually run LLM (synchronously!)
+
+                # run the prompt
                 llm_result = llm.invoke(llm_prompt_text)
                 parameters = json.loads(clean_llm_response(llm_result))
                 memory["parameters"] = parameters
@@ -551,7 +552,7 @@ async def main(
         transient=True,  # Removes progress bar after completion
     ) as progress:
         task = progress.add_task("executing", total=None)
-        memory = await run_plan(plan_steps, features, llm, user_input)
+        memory = await run_plan(plan_steps, features, llm, user_input, debug_mode)
         progress.remove_task(task)
 
     # 4. Results

agent_self_reflection.py

@@ -34,7 +34,7 @@
 Hard constraints
 -------------
 MAX_RETRIES : int
-    Maximum number of parameter generation attempts before bailout (default: 4)
+    Maximum number of parameter generation attempts before bailout (default: mam4)
 
 Examples
 --------
@@ -445,20 +445,6 @@ def parse_arguments():
 
 if __name__ == "__main__":
 
-    # TM notes: mistral:7b can fail and successed with reflection (or 2nd go)
-    # qwen2:math:7b is not suitable
-    # gemma3:27b both variants work well
-    # deepseek-r1:32b is too slow on 4080
-
-    #model_name = "gemma3n:e4b"
-    #model_name = "deepseek-r1:32b"
-    #model_name = "llama3:latest"
-    #model_name = "llama3.1:8b"
-    #model_name = "gemma3:27b"
-    # model_name = "gemma3:27b-it-qat"
-    #model_name = "qwen2-math:7b"
-    #model_name = "mistral:7b"
-
     # Parse command line arguments
     args = parse_arguments()