Fixed issues with benchmarking

.env.example

@@ -1,9 +1,11 @@
 # Your OpenAI API Key. If GPT-4 is available it will use that, otherwise will use 3.5-turbo
-OPENAI_API_KEY=MY-API-KEY
-
+OPENAI_API_KEY=abc
 # If you want to enable Helicone proxy and caching
-# HELICONE_KEY=MY-HELICONE-KEY
-# OPENAI_API_BASE=https://oai.hconeai.com/v1
+HELICONE_KEY=abc
+OPENAI_API_BASE=https://oai.hconeai.com/v1
 
 # Control log level
-LOG_LEVEL=INFO
+LOG_LEVEL=INFO
+DATABASE_NAME="./agent.db"
+PORT=8000
+AGENT_WORKSPACE="agbenchmark/workspace"

autogpt/__main__.py

@@ -11,9 +11,11 @@ if __name__ == "__main__":
     load_dotenv()
     database_name = os.getenv("DATABASE_NAME")
     port = os.getenv("PORT")
+    workspace = os.getenv("AGENT_WORKSPACE")
     auto_gpt = autogpt.agent.AutoGPT()
 
     database = autogpt.db.AgentDB(database_name)
     agent = Agent.setup_agent(auto_gpt.task_handler, auto_gpt.step_handler)
     agent.db = database
+    agent.workspace = workspace
     agent.start(port=port)

autogpt/agent.py

@@ -14,10 +14,12 @@ class AutoGPT:
         await Agent.db.create_step(task.task_id, task.input, is_last=True)
         time.sleep(2)
         autogpt.utils.run(task.input)
+        # print(f"Created Task id: {task.task_id}")
+        return task
         
 
     async def step_handler(self, step: Step) -> Step:
-        print(f"step: {step.input}")
+        # print(f"step: {step}")
         agent_step = await Agent.db.get_step(step.task_id, step.step_id)
         updated_step: Step = await Agent.db.update_step(agent_step.task_id, agent_step.step_id, status="completed")
         updated_step.output = agent_step.input

autogpt/db.py

@@ -137,7 +137,7 @@ class AgentDB(TaskDB):
                 for step in steps:
                     status = Status.created if step[3] == "created" else Status.completed
                     task.steps.append(Step(task_id=step[1], step_id=step[0], name=step[2], status=status, is_last=True if step[4] == 1 else False, additional_properties=step[5]))
-            print(f"Task: {task}")
+            # print(f"Getting task {task_id}.... Task details: {task}")
             return task
         else:
             raise DataNotFoundError("Task not found")
@@ -149,7 +149,14 @@ class AgentDB(TaskDB):
             "SELECT * FROM steps WHERE task_id=? AND step_id=?", (task_id, step_id)
         )
         if step := cursor.fetchone():
-            return Step(task_id=task_id, step_id=step_id, name=step[2], status=step[3])
+            return Step(
+                task_id=task_id,
+                step_id=step_id,
+                name=step[2],
+                status=step[3],
+                is_last=step[4] == 1,
+                additional_properties=step[5],
+            )
         else:
             raise DataNotFoundError("Step not found")
 

run

@@ -0,0 +1,6 @@
+#!/bin/bash
+
+export PYTHONPATH=$PYTHONPATH:$PWD
+
+python -m autogpt "$@"
+