bob_vector_db.embedder

Classes

EmbedderNode

Basic Embedder ROS Node.

Functions

main()

Module Contents

class bob_vector_db.embedder.EmbedderNode

Bases: rclpy.node.Node

Basic Embedder ROS Node. This ROS node subscribes to a std_msgs/msg/String topic to receive JSON data with embedding messages. Additionaly, when using a Qdrant vector Database, it’s possible to embed Images along with text and Payload data from a sensor_msgs/msg/Image or bob_msgs/msg/STTImage.

It embeds the delivered data into the configured Vector DB. A Qdrant Vector DB is the default DB. To embed into a Chroma DB set parameter ‘use_chroma’ to true. See ROS parameter for further configuration. The JSON data has to contain the following fields. (for Qdrant the ids are optional).

Example of a text embedding:

{
    "collection": "stories",
    "documents": [
        "some story text",
        "text abou something strange"
    ],
    "metadatas": [
        {"title":"The end"},
        {"title":"Dark star"}
    ],
    "ids":[
        "id1",
        "id2"
    ]
}

Example of a text + image embedding:

{
    "collection": "movie_cover",
    "documents": [
        "some story text",
        "text about something strange"
    ],
    "metadatas": [
        {"title":"The end"},
        {"title":"Dark star"}
    ],
    "images": [
        "/path/to/cover_image_id1.jpg",
        "/path/to/cover_image_id2.jpg"
    ], "ids":[
        "id1",
        "id2"
    ]
}

For the image embedding there exist following environment variables:

Enable / disable automatic storing of the image in Base64 format into payload data:

EMBED_IMAGES_BASE64 Default: 0

Used image format when embedding from sensor_msgs/msg/Image or bob_msgs/msg/STTImage:

EMBED_PIL_B64FORMAT Default: JPEG

See Package README.md for further information.

model

host

port

path

location

sub

sub_raw

sub_image

sub_ttimage

init_qdrant(): Init Qdrant embedder

init_chroma(): Init Chroma embedder

embed_mixed(j: str, media: list = None) → None

embed_callback(msg: std_msgs.msg.String) → None: Embeds the incoming JSON topic message into the Vector DB. Prints the incoming message also to the debug log. If it fails an error log entry will be written.

embed_raw_callback(msg: std_msgs.msg.String) → None: Embeds raw data of any type into default collection ‘embed_raw’. To use another collection set the env var or ROS parameter accordingly. If the collection comes with the input data that will be used instead.

embed_image(msg: sensor_msgs.msg.Image | bob_msgs.msg.TTImage) → None

embed_image_callback(msg: sensor_msgs.msg.Image) → None: Embeds Image only with simple payload into default collection ‘embed_raw’. To use another collection set the env var or ROS parameter accordingly.

embed_ttimage_callback(msg: bob_msgs.msg.TTImage) → None: Embeds TTImage with payload into default collection ‘embed_raw’. To use another collection set the env var or ROS parameter accordingly.

bob_vector_db.embedder.main()