modify config files,schema based on new node repo

babyagi_task_finalizer/.env.example

@@ -0,0 +1,2 @@
+OPENAI_API_KEY=
+NODE_URL=http://localhost:7001

babyagi_task_finalizer/config/agent_deployments.json

@@ -0,0 +1,17 @@
+[
+    {
+        "name": "agent_deployment_1",
+        "module": {"name": "babyagi_task_finalizer"},
+        "worker_node_url": "http://localhost:7001",
+        "agent_config": {
+            "config_name": "TaskFinalizerAgentConfig",
+            "llm_config": {"config_name": "model_2"},
+            "persona_module" : null,
+            "system_prompt": {
+                "role": "You are a helpful AI assistant.",
+                "persona": ""
+            },
+            "user_message_template": "You are given the following task: {{task}}. The task is to accomplish the following objective: {{objective}}."
+        }
+    }
+]

babyagi_task_finalizer/config/llm_configs.json

@@ -0,0 +1,18 @@
+[
+    {
+        "config_name": "model_1",
+        "client": "ollama",
+        "model": "ollama/phi",
+        "temperature": 0.7,
+        "max_tokens": 1000,
+        "api_base": "http://localhost:11434"
+    },
+    {
+        "config_name": "model_2",
+        "client": "openai",
+        "model": "gpt-4o-mini",
+        "temperature": 0.7,
+        "max_tokens": 1000,
+        "api_base": "https://api.openai.com/v1"
+    }
+]

babyagi_task_finalizer/run.py

@@ -1,73 +1,97 @@
+#!/usr/bin/env python
+from dotenv import load_dotenv
+from babyagi_task_finalizer.schemas import InputSchema, TaskExecutorPromptSchema, TaskFinalizerAgentConfig, TaskFinalizer
+import json
+from litellm import completion
 import os
-import yaml
-import instructor
-from litellm import Router
-from babyagi_task_finalizer.schemas import InputSchema, TaskFinalizer
-from babyagi_task_finalizer.utils import get_logger
-
+from naptha_sdk.schemas import AgentDeployment, AgentRunInput
+from naptha_sdk.utils import get_logger
 
+load_dotenv()
 logger = get_logger(__name__)
 
-client = instructor.patch(
-    Router(
-        model_list=
-        [
-            {
-                "model_name": "gpt-3.5-turbo",
-                "litellm_params": {
-                    "model": "openai/gpt-3.5-turbo",
-                    "api_key": os.getenv("OPENAI_API_KEY"),
-                },
-            }
-        ],
-        # default_litellm_params={"acompletion": True},
-    )
-)
-
-def llm_call(messages, response_model=None):
-    if response_model:
-        response = client.chat.completions.create(
-            model="gpt-3.5-turbo",
-            response_model=response_model,
+class TaskFinalizerAgent:
+    def __init__(self, agent_deployment: AgentDeployment):
+        self.agent_deployment = agent_deployment
+
+    def execute_task(self, inputs: InputSchema):
+        if isinstance(self.agent_deployment.agent_config, dict):
+            self.agent_deployment.agent_config = TaskFinalizerAgentConfig(**self.agent_deployment.agent_config)
+
+        user_prompt = self.agent_deployment.agent_config.user_message_template.replace(
+            "{{task}}", inputs.tool_input_data.task
+        ).replace(
+            "{{objective}}", inputs.tool_input_data.objective
+        )
+
+        messages = [
+            {"role": "system", "content": json.dumps(self.agent_deployment.agent_config.system_prompt)},
+            {"role": "user", "content": user_prompt}
+        ]
+
+        api_key = None if self.agent_deployment.agent_config.llm_config.client == "ollama" else (
+            "EMPTY" if self.agent_deployment.agent_config.llm_config.client == "vllm" else os.getenv("OPENAI_API_KEY")
+        )
+
+        response = completion(
+            model=self.agent_deployment.agent_config.llm_config.model,
             messages=messages,
-            temperature=0.0,
-            max_tokens=1000,
+            temperature=self.agent_deployment.agent_config.llm_config.temperature,
+            max_tokens=self.agent_deployment.agent_config.llm_config.max_tokens,
+            api_base=self.agent_deployment.agent_config.llm_config.api_base,
+            api_key=api_key
         )
-    return response
-
-def run(inputs: InputSchema, *args, **kwargs):
-    logger.info(f"Running with inputs {inputs.objective}")
-    logger.info(f"Running with inputs {inputs.task}")
-    cfg = kwargs["cfg"]
-
-    user_prompt = cfg["inputs"]["user_message_template"].replace("{{task}}", inputs.task).replace("{{objective}}", inputs.objective)
-
-    messages = [
-        {"role": "system", "content": cfg["inputs"]["system_message"]},
-        {"role": "user", "content": user_prompt}
-    ]
-
-    response = llm_call(messages, response_model=TaskFinalizer)
-
-    logger.info(f"Result: {response}")
-
-    return response.model_dump_json()
-
+
+        # Parse the response into the TaskFinalizer model
+        response_content = response.choices[0].message.content
+
+        try:
+            # Attempt to parse the response as JSON
+            parsed_response = json.loads(response_content)
+            task_finalizer = TaskFinalizer(**parsed_response)
+        except (json.JSONDecodeError, TypeError):
+            # If parsing fails, create a TaskFinalizer with the raw content
+            task_finalizer = TaskFinalizer(
+                final_report=response_content,
+                new_tasks=[],
+                objective_met=False
+            )
+
+        logger.info(f"Response: {task_finalizer}")
+        return task_finalizer.model_dump_json()
+
+def run(agent_run: AgentRunInput, *args, **kwargs):
+    logger.info(f"Running with inputs {agent_run.inputs.tool_input_data}")
+    task_finalizer_agent = TaskFinalizerAgent(agent_run.agent_deployment)
+    method = getattr(task_finalizer_agent, agent_run.inputs.tool_name, None)
+    return method(agent_run.inputs)
 
 if __name__ == "__main__":
-    with open("babyagi_task_finalizer/component.yaml", "r") as f:
-        cfg = yaml.safe_load(f)
-
-    inputs = InputSchema(
-        task="Weather pattern between year 1900 and 2000 was cool andry", 
-        objective="Write a blog post about the weather in London."
+    from naptha_sdk.client.naptha import Naptha
+    from naptha_sdk.configs import load_agent_deployments
+
+    naptha = Naptha()
+
+    # Configs
+    agent_deployments = load_agent_deployments(
+        "babyagi_task_finalizer/configs/agent_deployments.json", 
+        load_persona_data=False, 
+        load_persona_schema=False
     )
-
-    r = run(inputs, cfg=cfg)
-    logger.info(f"Result: {type(r)}")
-
-
-    import json
-    t = TaskFinalizer(**json.loads(r))
-    logger.info(f"Final report: {type(t)}")
-
+
+    input_params = InputSchema(
+        tool_name="execute_task",
+        tool_input_data=TaskExecutorPromptSchema(
+            task="Weather pattern between year 1900 and 2000", 
+            objective="Write a blog post about the weather in London."
+        ),
+    )
+
+    agent_run = AgentRunInput(
+        inputs=input_params,
+        agent_deployment=agent_deployments[0],
+        consumer_id=naptha.user.id,
+    )
+
+    response = run(agent_run)
+    print(response)

babyagi_task_finalizer/schemas.py

@@ -1,10 +1,15 @@
+from naptha_sdk.schemas import AgentConfig
 from pydantic import BaseModel, Field
 from typing import List
 
-class InputSchema(BaseModel):
+class TaskExecutorPromptSchema(BaseModel):
     task: str
     objective: str
 
+class InputSchema(BaseModel):
+    tool_name: str
+    tool_input_data: TaskExecutorPromptSchema
+
 class Task(BaseModel):
     """Class for defining a task to be performed."""
     name: str = Field(..., description="The name of the task to be performed.")
@@ -16,4 +21,7 @@ class TaskFinalizer(BaseModel):
     """Class for finalizing the tasks."""
     final_report: str = Field("", description="The final report of the tasks.")
     new_tasks: List[Task] = Field([], description="A list of new tasks to be performed.")
-    objective_met: bool = Field(False, description="The status of the objective. True if the objective have been met, False otherwise.")
+    objective_met: bool = Field(False, description="The status of the objective. True if the objective have been met, False otherwise.")
+
+class TaskFinalizerAgentConfig(AgentConfig):
+    user_message_template: str

pyproject.toml

@@ -9,10 +9,12 @@ packages = [
 ]
 
 [tool.poetry.dependencies]
-python = "^3.10"
+python = ">=3.10,<=3.13"
 pydantic = "^2.7.0"
 litellm = "^1.35.15"
-instructor = "^1.3.2"
+naptha-sdk = {git = "https://github.com/NapthaAI/naptha-sdk.git"}
+python-dotenv = "^1.0.1"
+instructor="^1.3.2"
 
 
 [build-system]