Handle new LangChain message content format in agents

launch/launch/agent/base_image.py

@@ -1,9 +1,13 @@
 """
 Base Docker image selection agent for repository environment setup.
 """
-from langchain.schema import HumanMessage
+try:  # LangChain >= 0.3.26
+    from langchain_core.messages import HumanMessage  # type: ignore[import-not-found]
+except ImportError:  # pragma: no cover - fallback for older LangChain
+    from langchain.schema import HumanMessage  # type: ignore
 
 from launch.agent.state import AgentState, auto_catch
+from launch.agent.utils import message_content_to_str
 from launch.utilities.language_handlers import get_language_handler
 
 
@@ -49,8 +53,9 @@ def select_base_image(state: AgentState) -> dict:
     while not base_image or trials < 5:
         trials += 1
         response = llm.invoke(messages)
-        if "<image>" in response.content:
-            image = response.content.split("<image>")[1].split("</image>")[0]
+        response_text = message_content_to_str(response.content)
+        if "<image>" in response_text:
+            image = response_text.split("<image>")[1].split("</image>")[0]
             if image in candidate_images:
                 base_image = image
                 break

launch/launch/agent/locate.py

@@ -3,10 +3,14 @@
 """
 import os
 
-from langchain.schema import HumanMessage
+try:  # LangChain >= 0.3.26
+    from langchain_core.messages import HumanMessage  # type: ignore[import-not-found]
+except ImportError:  # pragma: no cover
+    from langchain.schema import HumanMessage  # type: ignore
 
 from launch.agent.state import AgentState, auto_catch
 from launch.utilities.get_repo_structure import view_repo_structure
+from launch.agent.utils import message_content_to_str
 
 prompt = """Given this repository structure:
 ------ BEGIN REPOSITORY STRUCTURE ------
@@ -67,9 +71,10 @@ def locate_related_file(state: AgentState) -> dict:
         )
 
     response = llm.invoke([locate_prompt])
+    response_text = message_content_to_str(response.content)
     potential_files = [
         line.split("<file>")[1].split("</file>")[0].strip()
-        for line in response.content.split("\n")
+        for line in response_text.split("\n")
         if line.strip() and "<file>" in line
     ]
     potential_files = [
@@ -103,11 +108,12 @@ def locate_related_file(state: AgentState) -> dict:
         determine_input = HumanMessage(content=determine_prompt.format(file=file_info))
         try:
             response = llm.invoke([determine_input])
+            response_text = message_content_to_str(response.content)
         except Exception:
             logger.error(f"Error determining file: {file}")
             continue
-        logger.info(f"File: {file} - {response.content}")
-        if "<rel>Yes</rel>" in response.content:
+        logger.info(f"File: {file} - {response_text}")
+        if "<rel>Yes</rel>" in response_text:
             docs += f"File: {file}\n```\n"
             docs += content + "\n"
             docs += "```\n"

launch/launch/agent/setup.py

@@ -6,14 +6,21 @@
 import time
 from typing import Any, Literal
 
-from langchain_core.messages import HumanMessage, SystemMessage
+try:  # LangChain >= 0.3.26
+    from langchain_core.messages import (  # type: ignore[import-not-found]
+        HumanMessage,
+        SystemMessage,
+    )
+except ImportError:  # pragma: no cover
+    from langchain.schema import HumanMessage, SystemMessage  # type: ignore
 from pydantic import BaseModel, Field
 
 from launch.agent.action_parser import ActionParser
 from launch.agent.prompt import ReAct_prompt
 from launch.agent.state import AgentState, auto_catch
 from launch.runtime import start_session
 from launch.utilities.language_handlers import get_language_handler
+from launch.agent.utils import message_content_to_str
 
 system_msg = """You are a developer. Your task is to install dependencies and set up a environment that is able to run the tests of the project.
 
@@ -224,11 +231,10 @@ def setup(max_steps: int, state: AgentState) -> dict:
 
         response = llm.invoke(input_messages)
 
-
         # print(response.pretty_repr())
         logger.info("\n" + response.pretty_repr())
         messages.append(response)
-        action = parse_setup_action(response.content)
+        action = parse_setup_action(message_content_to_str(response.content))
         if action and action.action == "command":
             commands.append(action.args)
         observation = observation_for_setup_action(state, action)

launch/launch/agent/utils.py

@@ -0,0 +1,44 @@
+"""
+Utility helpers shared across agent modules.
+"""
+from __future__ import annotations
+
+from typing import Any
+
+
+def message_content_to_str(content: Any) -> str:
+    """
+    Normalize LangChain message content into a plain string.
+
+    Newer LangChain versions return message.content as either a raw string
+    or a list of structured blocks (e.g., [{"type": "text", "text": "..."}]).
+    This helper flattens those structures so legacy string-based parsing
+    logic keeps working.
+    """
+    if content is None:
+        return ""
+
+    if isinstance(content, str):
+        return content
+
+    if isinstance(content, list):
+        parts: list[str] = []
+        for block in content:
+            if isinstance(block, str):
+                parts.append(block)
+            elif isinstance(block, dict):
+                if block.get("type") == "text":
+                    parts.append(block.get("text", ""))
+                elif "text" in block:
+                    parts.append(str(block["text"]))
+                elif "content" in block:
+                    parts.append(str(block["content"]))
+                else:
+                    # Fallback to the full dict representation to preserve info
+                    parts.append(str(block))
+            else:
+                parts.append(str(block))
+        return "".join(parts)
+
+    return str(content)
+

launch/launch/agent/verify.py

@@ -3,13 +3,20 @@
 """
 from typing import Any, Literal
 
-from langchain.schema import HumanMessage, SystemMessage
+try:  # LangChain >= 0.3.26
+    from langchain_core.messages import (  # type: ignore[import-not-found]
+        HumanMessage,
+        SystemMessage,
+    )
+except ImportError:  # pragma: no cover
+    from langchain.schema import HumanMessage, SystemMessage  # type: ignore
 from pydantic import BaseModel, Field
 
 from launch.agent.action_parser import ActionParser
 from launch.agent.prompt import ReAct_prompt
 from launch.agent.state import AgentState, auto_catch
 from launch.runtime import SetupRuntime
+from launch.agent.utils import message_content_to_str
 
 system_msg = """You are a developer. Your task is to verify whether the environment for the given project is set up correctly. Your colleague has set up a Docker environment for the project. You need to verify if it can successfully run the tests of the project.
 - You interact with a Bash session inside this container.
@@ -170,7 +177,7 @@ def verify(max_steps: int, state: AgentState) -> dict:
         # print(response.pretty_repr())
         logger.info(response.pretty_repr())
         messages.append(response)
-        action = parse_verify_action(response.content)
+        action = parse_verify_action(message_content_to_str(response.content))
         if action.action == "command":
             commands.append(action.args)
         observation = observation_for_verify_action(action, session)