feat: task cancellation (wip)

backend/src/agents/expert.py

@@ -2,10 +2,8 @@ import inference
 from agent import Agent, AgentField
 
 CLASS_DESCRIPTION = """
-Expert AI in any field, powered by Llama 3 70b.
+Expert AI with customizable system prompt.
 Will get guidelines, documents, and other advanced features in the future.
-
-Available models are: [llama/qwen]-[big/small] (replicate), ollama:MODEL_NAME
 """.strip()
 
 class ExpertAgent(Agent):
@@ -22,7 +20,6 @@ class ExpertAgent(Agent):
         Agent.__init__(self, app, data)
 
     async def run(self, run):
-        inp = run.input or {}
         opts = {}
 
         if self.config.get("temperature") is not None:
@@ -30,26 +27,20 @@ class ExpertAgent(Agent):
         if self.config.get("repetition_penalty") is not None:
             opts["repetition_penalty"] = self.config["repetition_penalty"]
 
-        # 1. Build message list based on input mode
-        if "messages" in inp:
-            messages = inp["messages"]
-        else:
-            instruction = inp.get("instruction")
-            input_text = inp.get("input")
-            messages = inference.build_messages(
-                system_prompt=self.config.get("system_prompt", ""),
-                instruction=instruction,
-                input=input_text,
-                messages=inp.get("messages"),
-            )
+        # 1. Build message list based on input
+        messages = inference.build_messages(
+            system_prompt=self.config.get("system_prompt", ""),
+            instruction=run.input.get("instruction"),
+            input=run.input.get("input"),
+            messages=run.input.get("messages"),
+        )
 
-        # 2. Call inference, passing the run so it feeds run.stream
-        #    and can emit tool_call events through the run.
+        # 2. Call inference, passing the run to feed run.stream and tool calls
         await inference.chat_completion(
             self.app,
             model_id=self.config.get("model"),
             messages=messages,
-            tools=inp.get("tools", []),
+            tools=run.input.get("tools", []),
             opts=opts,
             agent_run=run,
         )