Skip to main content

🤖 vlm_robot_agent

Author: Edison Bejarano

Library designed to use VLMs in the context of robotics actions for planning and interactions

Un agente robótico inteligente basado en modelos de lenguaje visual (VLM), que puede percibir el entorno desde una imagen, planificar acciones y decidir entre navegar o interactuar con personas para alcanzar un objetivo dado (como entrar en una habitación, buscar un baño, etc).


🚀 Características

  • 📷 Percepción visual usando un VLM de OpenAI
  • 🧠 Razonamiento basado en objetivos (macro y micro-goals)
  • 🧭 Acciones de navegación:
    • forward, left, right, forward_left, forward_right
  • 🙋 Acciones de interacción:
    • Conversar con una persona que bloquea el paso
    • Hacer gestos para pedir que se mueva
  • 💾 Memoria de interacciones y lectura/ejecución de prompts desde un folder

📦 Instalación

pip install vlm_robot_agent

🛠 Uso básico

from vlm_robot_agent import VLMRobotAgent

agent = VLMRobotAgent(prompt_folder="./prompts")

image = obtener_imagen_de_tu_robot()
goal = "entrar a la oficina 3"

# Loop de ejecución
while True:
    action = agent.step(image, goal)
    ejecutar_action_en_robot(action)
    if objetivo_cumplido():
        break

📁 Estructura de prompts

Los prompts se almacenan como archivos .json dentro del folder configurado, y puedes cargarlos con:

prompts = agent.load_prompts()

🧩 Integración con robots

  • Puede usarse en sistemas ROS, simuladores como Gazebo, o cualquier entorno de robots.
  • El agente necesita:
    • Imagen actual del entorno (image)
    • Objetivo a cumplir (goal)
    • Una función que ejecute la acción devuelta (Navigate, Interact)

📚 Ejemplo de acciones

from vlm_robot_agent import Navigate, Interact

# Navegar hacia adelante
Navigate(direction="forward")

# Pedir a una persona que se mueva
Interact(strategy="ask_to_move")

📄 Licencia

MIT


🧠 Futuras mejoras

  • Seguimiento de progreso con StateTracker
  • Manejo de múltiples agentes o flujos conversacionales
  • Soporte para entrada multimodal (texto + imagen)

Edison Bejarano

Metadata

Release files for vlm-robot-agent 0.3.2

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for vlm-robot-agent 0.3.2
File Size Uploaded Provenance Actions
vlm_robot_agent-0.3.2.tar.gz 22.0 kB Without provenance Details for vlm_robot_agent-0.3.2.tar.gz

Built distribution (wheel)

Table of built distributions (wheels) for vlm-robot-agent 0.3.2
File Interpreter ABI Platform Actions
vlm_robot_agent-0.3.2-py3-none-any.whl Python 3 none any Details for vlm_robot_agent-0.3.2-py3-none-any.whl

Total release size: 48.2 kB

Release files / vlm_robot_agent-0.3.2.tar.gz

Download URL vlm_robot_agent-0.3.2.tar.gz
Size 22.0 kB
Tags Source
SHA-256 checksum
How to use checksums
ce26b344f6a18ed59b036ab9104b47c1ad19ca0c25c1f396e065eb4fc91ce64a
BLAKE2b-256 checksum
How to use checksums
e07940e9a784c095090150e4929e70b8c4b34e340b8fb64228da63ab02d60007
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.1.0 CPython/3.12.2

Release files / vlm_robot_agent-0.3.2-py3-none-any.whl

Download URL vlm_robot_agent-0.3.2-py3-none-any.whl
Size 26.1 kB
Tags Python 3
SHA-256 checksum
How to use checksums
df1faf2070cc79d7de7eca911db353ebe6bb496c30e5cf6544dbf4ae347fc79f
BLAKE2b-256 checksum
How to use checksums
219807044997cec46b24e98c150b9f9f0598fb71bef80b9f1297127574ab16de
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via twine/6.1.0 CPython/3.12.2

Security for vlm-robot-agent 0.3.2

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

No provenance found

None of the files in this release were uploaded with provenance metadata.

A note on package security

PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.

To learn more about protecting yourself from malicious actors, visit our package security documentation.

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page