NimbRo API

NimbRo API is a robust and flexible framework for API clients in Python.

---

✨ Features

🚢 Ships with clients for the following APIs:

OpenAI:

• Chat Completions (nimbro_api.openai.ChatCompletions)
• Embeddings (nimbro_api.openai.Embeddings)
• Images (nimbro_api.openai.Images)
• Speech (nimbro_api.openai.Speech)
• Transcriptions (nimbro_api.openai.Transcriptions)
• Translations (nimbro_api.openai.Translations)

NimbRo Vision Servers:

• Describe Anything Model (nimbro_api.nimbro_vision_servers.Dam)
• Florence-2 (nimbro_api.nimbro_vision_servers.Florence2)
• Kosmos-2 (nimbro_api.nimbro_vision_servers.Kosmos2)
• MM-Grounding-DINO and LLMDet (nimbro_api.nimbro_vision_servers.MmGroundingDino)
• Segment Anything Model 2 real-time (nimbro_api.nimbro_vision_servers.Sam2Realtime)

Miscellaneous:

• No-as-a-Service (nimbro_api.misc.No)
• VLM-GIST (nimbro_api.misc.VlmGist)

🪧 All clients feature:

• Standardized interfaces
• Sensible default settings
• Monitoring of operation success
• Configurable retry behavior and timeouts
• Extensive documentation, logging, and tests

🪧 In addition, some clients feature:

• Response caching
• Response healing
• Batching of large requests
• API generalization across providers

⚙️ Global settings and utilities are exposed at the package level.

🛠️ Custom clients can be implemented using the nimbro_api.Client and nimbro_api.ClientBase classes.

---

🚀 Quick Start

Install this package:

pip install nimbro-api

Set the API key for the provider you want to use (e.g. OPENROUTER_API_KEY, OPENAI_API_KEY, VLLM_API_KEY):

export OPENROUTER_API_KEY='supersecretkey'

Create and use a ChatCompletions client:

import nimbro_api
# nimbro_api.set_api_key(name="OPENROUTER_API_KEY", key="supersecretkey") # alternatively, set API key here
client = nimbro_api.openai.ChatCompletions(endpoint="OpenRouter", model="google/gemini-3-flash-preview")
success, message, response = client.prompt(text="Tell me a joke!")

---

📖 Citation

If you utilize this package in your research, please cite one of our relevant publications.

• Leveraging Vision-Language Models for Open-Vocabulary Instance Segmentation and Tracking
  [arXiv:2503.16538]

  @article{paetzold25vlmgist,
      author={Bastian P{\"a}tzold and Jan Nogga and Sven Behnke},
      title={Leveraging Vision-Language Models for Open-Vocabulary Instance Segmentation and Tracking},
      journal={IEEE Robotics and Automation Letters (RA-L)},
      volume={10},
      number={11},
      pages={11578-11585},
      year={2025}
  }

• A Comparison of Prompt Engineering Techniques for Task Planning and Execution in Service Robotics
  [arXiv:2410.22997]

  @article{bode24prompt,
      author={Jonas Bode and Bastian P{\"a}tzold and Raphael Memmesheimer and Sven Behnke},
      title={A Comparison of Prompt Engineering Techniques for Task Planning and Execution in Service Robotics},
      journal={IEEE-RAS International Conference on Humanoid Robots (Humanoids)},
      pages={309-314},
      year={2024}
  }

• RoboCup@Home 2024 OPL Winner NimbRo: Anthropomorphic Service Robots using Foundation Models for Perception and Planning
  [arXiv:2412.14989]

  @article{memmesheimer25robocup,
      author={Raphael Memmesheimer and Jan Nogga and Bastian P{\"a}tzold and Evgenii Kruzhkov and Simon Bultmann and Michael Schreiber and Jonas Bode and Bertan Karacora and Juhui Park and Alena Savinykh and Sven Behnke},
      title={{RoboCup@Home 2024 OPL Winner NimbRo}: Anthropomorphic Service Robots using Foundation Models for Perception and Planning},
      journal={RoboCup 2024: RoboCup World Cup XXVII},
      volume={15570},
      pages={515-527},
      year={2025}
  }

---

📄 License

NimbRo API is licensed under the BSD-3-Clause License.

---

👤 Author

Bastian Pätzold – paetzoldbastian@gmail.com