vision-agent 0.2.244__tar.gz → 0.2.246__tar.gz

{vision_agent-0.2.244 → vision_agent-0.2.246}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: vision-agent
-Version: 0.2.244
+Version: 0.2.246
 Summary: Toolset for Vision Agent
 Author: Landing AI
 Author-email: dev@landing.ai

{vision_agent-0.2.244 → vision_agent-0.2.246}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "poetry.core.masonry.api"
 
 [tool.poetry]
 name = "vision-agent"
-version = "0.2.244"
+version = "0.2.246"
 description = "Toolset for Vision Agent"
 authors = ["Landing AI <dev@landing.ai>"]
 readme = "README.md"

{vision_agent-0.2.244 → vision_agent-0.2.246}/vision_agent/agent/vision_agent_planner.py

@@ -410,7 +410,7 @@ class VisionAgentPlanner(Agent):
             plans = write_plans(
                 chat,
                 get_tool_descriptions_by_names(
-                    custom_tool_names, T.FUNCTION_TOOLS, T.UTIL_TOOLS  # type: ignore
+                    custom_tool_names, T.tools.FUNCTION_TOOLS, T.tools.UTIL_TOOLS  # type: ignore
                 ),
                 format_feedback(working_memory),
                 self.planner,

{vision_agent-0.2.244 → vision_agent-0.2.246}/vision_agent/agent/vision_agent_planner_v2.py

@@ -317,7 +317,8 @@ def maybe_run_code(
         media_data = capture_media_from_exec(execution)
         int_chat_elt = AgentMessage(role="observation", content=obs, media=None)
         if media_list:
-            int_chat_elt.media = cast(List[Union[str, Path]], media_data)
+            # Don't sned more than 3 media files to the planner or it can get overloaded
+            int_chat_elt.media = cast(List[Union[str, Path]], media_data[:3])
         return_chat.append(int_chat_elt)
     else:
         return_chat.append(AgentMessage(role="planner", content=response, media=None))

{vision_agent-0.2.244 → vision_agent-0.2.246}/vision_agent/agent/vision_agent_prompts_v2.py

@@ -54,12 +54,22 @@ AGENT: <response>I am VisionAgent, an agent built by LandingAI, to help users wr
 - Understanding documents
 - Pose estimation
 - Visual question answering for both images and videos
-- Action recognition in videos
+- Activity recognition in videos
 - Image inpainting
 
 How can I help you?</response>
 --- END EXAMPLE2 ---
 
+**Helpful Context**:
+- You are helping the user write vision code that utilizes the `vision-agent` library to solve different vision tasks.
+- The goal is to write code the user can use to solve a visual task and then deploy the code so they can run it repeatedly on that task.
+- If the user wants to run the code, they can `pip install vision-agent` to install the library, copy the code, and run it locally themselves.
+- They can find the vision-agent source code at https://github.com/landing-ai/vision-agent. They can find the documentation for vision-agent at https://landing-ai.github.io/vision-agent/.
+- The code utilizes LandingAI's API to run the different tools, the API is free to use and does not require an API key.
+- The user can run VisionAgent (you) from two places, either on our website at https://va.landing.ai or locally.
+- If the user is running VisionAgent on the website, they can only upload images or videos, nothing else at the moment. They can view files created by VisionAgent by clicking the "Files" button in the upper right part of the screen.
+- If the user is running VisionAgent locally they can ask it to do more things like run over files in a folder. They can view files created by VisionAgent by checking their local file system.
+
 **Conversation**:
 Here is the current conversation so far:
 --- START CONVERSATION ---