Langchain V0.2 and Sync API deprecated (#116)

* upgrade langchain and remove sync apis

* remove return prompt from agenerate

* remove prompt from agenerate

* add sync apis

examples/generate_specific_envs.py

@@ -12,7 +12,7 @@
 import numpy as np
 from datasets import DatasetDict, load_dataset
 
-from sotopia.generation_utils.generate import StrOutputParser, generate
+from sotopia.generation_utils.generate import StrOutputParser, agenerate
 
 
 async def generate_mutual_friend_envs() -> tuple[str, list[str]]:
@@ -78,7 +78,7 @@ async def generate_craigslist_bargains_envs() -> tuple[str, list[str]]:
     all_data = craigslist_bargains_dataset["train"]
     # sample one datum from all data
     datum = np.random.choice(all_data)
-    scenario = generate(
+    scenario = await agenerate(
         model_name="gpt-4",
         template="The following sentence is automatically generated with the following"
         'template: "One person is selling <item> for <price>, another person is'
@@ -100,7 +100,7 @@ async def generate_craigslist_bargains_envs() -> tuple[str, list[str]]:
             datum["agent_info"]["Target"][i] = datum["items"]["Price"][0] / (
                 1 + markup_ratio
             )
-        goal = generate(
+        goal = await agenerate(
             model_name="gpt-4",
             template="The following sentence is automatically generated with the following"
             'template: "You want to <role> this item. Your target price '

pyproject.toml

@@ -11,7 +11,7 @@ packages = [{include = "sotopia"}]
 python = ">=3.10, <3.13"
 lxml = ">=4.9.3,<6.0.0"
 openai = "^1.11.0"
-langchain = "0.1.5"
+langchain = "~0.2.5"
 rich = "^13.6.0"
 PettingZoo = "1.24.3"
 redis-om = "^0.2.1"
@@ -21,7 +21,7 @@ absl-py = "^2.0.0"
 together = "^0.2.4"
 pydantic = "1.10.12"
 beartype = "^0.14.0"
-langchain-openai = ">=0.0.5,<0.0.7"
+langchain-openai = "~0.1.8"
 litellm = "~1.23.12"
 
 # dependency versions for extras

sotopia/agents/__init__.py

@@ -8,7 +8,6 @@
     HumanAgent,
     LLMAgent,
     ScriptWritingAgent,
-    SpeakAgent,
 )
 from .redis_agent import RedisAgent
 
@@ -17,7 +16,6 @@
     "LLMAgent",
     "Agents",
     "HumanAgent",
-    "SpeakAgent",
     "generate_background",
     "generate_background_conversation",
     "RedisAgent",

sotopia/agents/generate_agent_background.py

@@ -2,14 +2,11 @@
 import os
 from typing import Callable
 
-from sotopia.generation_utils.generate import (
-    convert_narratives,
-    generate_init_profile,
-)
+from sotopia.generation_utils.generate import convert_narratives, agenerate_init_profile
 from sotopia.messages import Message, ScriptBackground
 
 
-def generate_background(
+async def generate_background(
     info_json_file: str, basic_info: dict[str, str]
 ) -> tuple[str, str, str, str, list[dict[str, str]]]:
     if os.path.isfile(info_json_file):
@@ -22,7 +19,7 @@ def generate_background(
             previous_messages = info_dict["messages"]
     else:
         initial_profile = str(basic_info)
-        profile = generate_init_profile(
+        profile = await agenerate_init_profile(
             model_name="gpt-3.5-turbo", basic_info=basic_info
         )
         first_narrative = convert_narratives(

sotopia/agents/llm_agent.py

@@ -1,16 +1,14 @@
 import asyncio
 from concurrent.futures import ThreadPoolExecutor
-from typing import Callable, cast
+from typing import cast
 
 from sotopia.agents import BaseAgent
 from sotopia.database import AgentProfile
 from sotopia.generation_utils.generate import (
     LLM_Name,
     agenerate_action,
+    agenerate_goal,
     agenerate_script,
-    generate_action,
-    generate_action_speak,
-    generate_goal,
 )
 from sotopia.messages import AgentAction, Observation
 from sotopia.messages.message_classes import ScriptBackground
@@ -44,48 +42,34 @@ def __init__(
     def goal(self) -> str:
         if self._goal is not None:
             return self._goal
-        assert (
-            len(self.inbox) > 0
-        ), "attribute goal has to be called after at least one step"
-        goal = generate_goal(
-            self.model_name,
-            background=self.inbox[0][
-                1
-            ].to_natural_language(),  # Only consider the first message for now
-        )
-        return goal
+        else:
+            raise Exception("Goal is not set.")
 
     @goal.setter
     def goal(self, goal: str) -> None:
         self._goal = goal
 
     def act(
         self,
-        obs: Observation,
-        gen_func: Callable[..., AgentAction] = generate_action,
+        _obs: Observation,
     ) -> AgentAction:
+        raise Exception("Sync act method is deprecated. Use aact instead.")
+
+    async def aact(self, obs: Observation) -> AgentAction:
         self.recv_message("Environment", obs)
 
-        if len(obs.available_actions) == 1 and "none" in obs.available_actions:
-            return AgentAction(action_type="none", argument="")
-        else:
-            action = gen_func(
+        if self._goal is None:
+            self._goal = await agenerate_goal(
                 self.model_name,
-                history="\n".join(f"{y.to_natural_language()}" for x, y in self.inbox),
-                turn_number=obs.turn_number,
-                action_types=obs.available_actions,
-                agent=self.agent_name,
-                goal=self.goal,
+                background=self.inbox[0][
+                    1
+                ].to_natural_language(),  # Only consider the first message for now
             )
-            return action
-
-    async def aact(self, obs: Observation) -> AgentAction:
-        self.recv_message("Environment", obs)
 
         if len(obs.available_actions) == 1 and "none" in obs.available_actions:
             return AgentAction(action_type="none", argument="")
         else:
-            action, prompt = await agenerate_action(
+            action = await agenerate_action(
                 self.model_name,
                 history="\n".join(f"{y.to_natural_language()}" for x, y in self.inbox),
                 turn_number=obs.turn_number,
@@ -147,15 +131,6 @@ async def aact(self, obs: Observation) -> AgentAction:
         return returned_action
 
 
-class SpeakAgent(LLMAgent):
-    def act(
-        self,
-        obs: Observation,
-        gen_func: Callable[..., AgentAction] = generate_action_speak,
-    ) -> AgentAction:
-        return super().act(obs, gen_func=gen_func)
-
-
 class HumanAgent(BaseAgent[Observation, AgentAction]):
     """
     A human agent that takes input from the command line.

sotopia/envs/evaluators.py

@@ -273,7 +273,7 @@ async def __acall__(
             response: (
                 EnvResponsePlus | EnvResponse
             )  # fix type error from langchain 0.0.264. we don't need this line for langchain 0.0.263
-            response, prompt = await agenerate(
+            response = await agenerate(
                 model_name=self.model_name,
                 template="""{history},
                     Based on previous interactions, evaluate how well participants achieve their goals.
@@ -286,7 +286,6 @@ async def __acall__(
                 ),
                 temperature=temperature,
             )
-            self.prompt = prompt
             response_list = []
             # TODO: multiple agents
             for dimension in response.agent_1_evaluation.dict().keys():

sotopia/generation_utils/__init__.py

@@ -2,14 +2,21 @@
     EnvResponse,
     LLM_Name,
     agenerate_env_profile,
-    fill_in_background,
-    generate_goal,
+    agenerate,
+    agenerate_action,
+)
+
+from .sync import (
+    generate,
+    generate_action,
 )
 
 __all__ = [
     "EnvResponse",
     "agenerate_env_profile",
     "LLM_Name",
-    "fill_in_background",
-    "generate_goal",
+    "agenerate",
+    "agenerate_action",
+    "generate",
+    "generate_action",
 ]