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

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"
    }
}