vision-agent 0.2.171__tar.gz → 0.2.173__tar.gz

{vision_agent-0.2.171 → vision_agent-0.2.173}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vision-agent
-Version: 0.2.171
+Version: 0.2.173
 Summary: Toolset for Vision Agent
 Author: Landing AI
 Author-email: dev@landing.ai

{vision_agent-0.2.171 → vision_agent-0.2.173}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry]
 name = "vision-agent"
-version = "0.2.171"
+version = "0.2.173"
 description = "Toolset for Vision Agent"
 authors = ["Landing AI <dev@landing.ai>"]
 readme = "README.md"

{vision_agent-0.2.171 → vision_agent-0.2.173}/vision_agent/agent/vision_agent.py

@@ -149,11 +149,11 @@ def execute_code_action(
     result = code_interpreter.exec_isolation(
         BoilerplateCode.add_boilerplate(code, remote_path=artifact_remote_path)
     )
-    extract_and_save_files_to_artifacts(artifacts, code)
 
     obs = str(result.logs)
     if result.error:
         obs += f"\n{result.error}"
+    extract_and_save_files_to_artifacts(artifacts, code, obs)
     return result, obs
 
 
@@ -182,6 +182,7 @@ def execute_user_code_action(
         )
         if user_result.error:
             user_obs += f"\n{user_result.error}"
+        extract_and_save_files_to_artifacts(artifacts, user_code_action, user_obs)
     return user_result, user_obs
 
 
@@ -491,19 +492,10 @@ class VisionAgent(Agent):
                     }
                 )
 
+                code_action = response.get("execute_python", None)
                 # sometimes it gets stuck in a loop, so we force it to exit
                 if last_response == response:
                     response["let_user_respond"] = True
-
-                finished = response.get("let_user_respond", False)
-
-                code_action = response.get("execute_python", None)
-                if code_action is not None:
-                    code_action = use_extra_vision_agent_args(
-                        code_action, test_multi_plan, custom_tool_names
-                    )
-
-                if last_response == response:
                     self.streaming_message(
                         {
                             "role": "assistant",
@@ -513,7 +505,7 @@ class VisionAgent(Agent):
                                 "value": "Agent is stuck in conversation loop, exited",
                                 "traceback_raw": [],
                             },
-                            "finished": finished and code_action is None,
+                            "finished": code_action is None,
                         }
                     )
                 else:
@@ -523,10 +515,18 @@ class VisionAgent(Agent):
                             "content": new_format_to_old_format(
                                 add_step_descriptions(response)
                             ),
-                            "finished": finished and code_action is None,
+                            "finished": response.get("let_user_respond", False)
+                            and code_action is None,
                         }
                     )
 
+                finished = response.get("let_user_respond", False)
+
+                if code_action is not None:
+                    code_action = use_extra_vision_agent_args(
+                        code_action, test_multi_plan, custom_tool_names
+                    )
+
                 if code_action is not None:
                     result, obs = execute_code_action(
                         artifacts,

{vision_agent-0.2.171 → vision_agent-0.2.173}/vision_agent/tools/meta_tools.py

@@ -451,11 +451,6 @@ def generate_vision_code(
         custom_tool_names=custom_tool_names,
     )
 
-    # capture and save any files that were saved in the code to the artifacts
-    extract_and_save_files_to_artifacts(
-        artifacts, response["code"] + "\n" + response["test"]
-    )
-
     redisplay_results(response["test_result"])
     code = response["code"]
     artifacts[name] = code
@@ -536,10 +531,6 @@ def edit_vision_code(
         test_multi_plan=False,
         custom_tool_names=custom_tool_names,
     )
-    # capture and save any files that were saved in the code to the artifacts
-    extract_and_save_files_to_artifacts(
-        artifacts, response["code"] + "\n" + response["test"]
-    )
 
     redisplay_results(response["test_result"])
     code = response["code"]
@@ -766,7 +757,9 @@ def use_object_detection_fine_tuning(
     return diff
 
 
-def extract_and_save_files_to_artifacts(artifacts: Artifacts, code: str) -> None:
+def extract_and_save_files_to_artifacts(
+    artifacts: Artifacts, code: str, obs: str
+) -> None:
     """Extracts and saves files used in the code to the artifacts object.
 
     Parameters:
@@ -776,12 +769,16 @@ def extract_and_save_files_to_artifacts(artifacts: Artifacts, code: str) -> None
     try:
         response = extract_json(
             AnthropicLMM()(  # type: ignore
-                f"""You are a helpful AI assistant. Your job is to look at a snippet of code and return the file paths that are being saved in the file. Below is the code snippet:
+                f"""You are a helpful AI assistant. Your job is to look at a snippet of code and the output of running that code and return the file paths that are being saved in the file. Below is the code snippet:
 
 ```python
 {code}
 ```
 
+```output
+{obs}
+```
+
 Return the file paths in the following JSON format:
 {{"file_paths": ["/path/to/image1.jpg", "/other/path/to/data.json"]}}"""
             )