LMM

Class: LMMBlockV1

Source: inference.core.workflows.core_steps.models.foundation.lmm.v1.LMMBlockV1

Ask a question to a Large Multimodal Model (LMM) with an image and text.

You can specify arbitrary text prompts to an LMMBlock.

The LLMBlock supports two LMMs:

• OpenAI's GPT-4 with Vision;

You need to provide your OpenAI API key to use the GPT-4 with Vision model.

If you want to classify an image into one or more categories, we recommend using the dedicated LMMForClassificationBlock.

Type identifier

Use the following identifier in step "type" field: roboflow_core/lmm@v1to add the block as as step in your workflow.

Properties

Name	Type	Description	Refs
name	str	Enter a unique identifier for this step..	❌
prompt	str	Holds unconstrained text prompt to LMM mode.	✅
lmm_type	str	Type of LMM to be used.	✅
lmm_config	LMMConfig	Configuration of LMM.	❌
remote_api_key	str	Holds API key required to call LMM model - in current state of development, we require OpenAI key when lmm_type=gpt_4v..	✅
json_output	Dict[str, str]	Holds dictionary that maps name of requested output field into its description.	❌

The Refs column marks possibility to parametrise the property with dynamic values available in workflow runtime. See Bindings for more info.

Available Connections

Compatible Blocks

Check what blocks you can connect to LMM in version v1.

• inputs: Camera Focus, Polygon Zone Visualization, Slack Notification, Local File Sink, Grid Visualization, Image Convert Grayscale, Trace Visualization, Instance Segmentation Model, Absolute Static Crop, Roboflow Custom Metadata, Perspective Correction, OpenAI, Circle Visualization, Clip Comparison, Image Slicer, OpenAI, Triangle Visualization, Halo Visualization, Corner Visualization, Email Notification, Object Detection Model, LMM, Roboflow Dataset Upload, Dynamic Crop, Depth Estimation, Model Monitoring Inference Aggregator, Stitch Images, Llama 3.2 Vision, Anthropic Claude, Model Comparison Visualization, Crop Visualization, Blur Visualization, CogVLM, Image Threshold, Stability AI Inpainting, VLM as Detector, Relative Static Crop, Image Preprocessing, Keypoint Visualization, Background Color Visualization, Color Visualization, Twilio SMS Notification, OCR Model, Multi-Label Classification Model, Classification Label Visualization, Google Vision OCR, Camera Calibration, Pixelate Visualization, Stitch OCR Detections, Label Visualization, Image Slicer, Reference Path Visualization, Single-Label Classification Model, Webhook Sink, Google Gemini, Roboflow Dataset Upload, Line Counter Visualization, Image Blur, Florence-2 Model, SIFT Comparison, LMM For Classification, Image Contours, Polygon Visualization, CSV Formatter, SIFT, Florence-2 Model, Ellipse Visualization, Mask Visualization, Keypoint Detection Model, Bounding Box Visualization, Dot Visualization, Stability AI Image Generation, VLM as Classifier
• outputs: Line Counter, Single-Label Classification Model, Slack Notification, YOLO-World Model, Image Convert Grayscale, Absolute Static Crop, Perspective Correction, OpenAI, Distance Measurement, Image Slicer, OpenAI, Cosine Similarity, Halo Visualization, Byte Tracker, Corner Visualization, Email Notification, Object Detection Model, Detections Classes Replacement, Template Matching, Roboflow Dataset Upload, Overlap Filter, Dynamic Crop, Cache Set, VLM as Classifier, Depth Estimation, Dynamic Zone, Model Monitoring Inference Aggregator, Cache Get, Llama 3.2 Vision, Anthropic Claude, Crop Visualization, Blur Visualization, Dominant Color, Image Threshold, Stability AI Inpainting, Relative Static Crop, Path Deviation, Clip Comparison, Moondream2, Twilio SMS Notification, OCR Model, Data Aggregator, Google Vision OCR, Pixelate Visualization, Property Definition, Stitch OCR Detections, Image Slicer, Time in Zone, Webhook Sink, JSON Parser, Line Counter Visualization, Byte Tracker, Image Blur, Florence-2 Model, Delta Filter, Detections Stabilizer, LMM For Classification, Instance Segmentation Model, Keypoint Detection Model, Detections Stitch, Bounding Rectangle, Stability AI Image Generation, VLM as Classifier, Identify Changes, Detection Offset, Camera Focus, First Non Empty Or Default, Polygon Zone Visualization, Detections Filter, Time in Zone, Local File Sink, Grid Visualization, Instance Segmentation Model, Trace Visualization, Roboflow Custom Metadata, Continue If, Circle Visualization, Dimension Collapse, Clip Comparison, Triangle Visualization, QR Code Detection, Gaze Detection, Line Counter, Size Measurement, Byte Tracker, LMM, Detections Consensus, Velocity, Buffer, Stitch Images, Segment Anything 2 Model, Object Detection Model, Model Comparison Visualization, Keypoint Detection Model, SIFT Comparison, CLIP Embedding Model, CogVLM, Expression, SmolVLM2, VLM as Detector, Image Preprocessing, Keypoint Visualization, Background Color Visualization, Pixel Color Count, Barcode Detection, Color Visualization, Multi-Label Classification Model, Classification Label Visualization, Camera Calibration, Label Visualization, Reference Path Visualization, Single-Label Classification Model, Google Gemini, Roboflow Dataset Upload, Qwen2.5-VL, Identify Outliers, Multi-Label Classification Model, Detections Transformation, SIFT Comparison, Image Contours, Polygon Visualization, CSV Formatter, SIFT, Florence-2 Model, Detections Merge, Mask Visualization, Ellipse Visualization, Rate Limiter, Bounding Box Visualization, Dot Visualization, VLM as Detector, Path Deviation

Input and Output Bindings

The available connections depend on its binding kinds. Check what binding kinds LMM in version v1 has.

Bindings

• input

  • images (image): The image to infer on..
  • prompt (string): Holds unconstrained text prompt to LMM mode.
  • lmm_type (string): Type of LMM to be used.
  • remote_api_key (Union[string, secret]): Holds API key required to call LMM model - in current state of development, we require OpenAI key when lmm_type=gpt_4v..

• output

  • parent_id (parent_id): Identifier of parent for step output.
  • root_parent_id (parent_id): Identifier of parent for step output.
  • image (image_metadata): Dictionary with image metadata required by supervision.
  • structured_output (dictionary): Dictionary.
  • raw_output (string): String value.
  • * (*): Equivalent of any element.

Example JSON definition of step LMM in version v1

{
    "name": "<your_step_name_here>",
    "type": "roboflow_core/lmm@v1",
    "images": "$inputs.image",
    "prompt": "my prompt",
    "lmm_type": "gpt_4v",
    "lmm_config": {
        "gpt_image_detail": "low",
        "gpt_model_version": "gpt-4o",
        "max_tokens": 200
    },
    "remote_api_key": "xxx-xxx",
    "json_output": {
        "count": "number of cats in the picture"
    }
}