Vector Strategies

This folder contains the default vector pipeline used by AIKPs that publish vector data through OCLI:

• makegeojson.py → convert vector input into a styled GeoJSON + create a DB-facing metadata document

• upload.py → upload the GeoJSON payload + DB metadata to Cloud Object Storage (COS)

• publish.py → register (publish) the metadata document into the docs database

These strategies are typically called by an AIKP template (for example ocli.aikp.vector_basic.template.Template).

High-level flow

1. ai makegeojson (vector)

   • Reads the vector source.

   • Writes a processed GeoJSON for the UI.

   • Writes a companion *_db.geojson document that points at the COS object.

2. ai upload

   • Uploads both the processed GeoJSON and the *_db.geojson document to COS.

3. ai publish post

   • Publishes the *_db.geojson document to the docs DB so it becomes visible in the UI.

The three scripts rely on consistent file naming and task/recipe fields (described below).

makegeojson.py (VectorGeoJsonStrategy)

Purpose: turn a vector dataset into:

• a styled GeoJSON file (the “payload”)

• a metadata GeoJSON document (the “DB doc”) that references the payload in COS

Inputs

From task.config:

• file_path (optional): path to the vector input

• output_file_name (optional): output GeoJSON filename; defaults to <input_stem>.geojson

• group_by_column (optional): attribute to color by; defaults to index

• colormap (optional): matplotlib colormap name; defaults to viridis

• alpha (optional): alpha channel used in rgba(...); defaults to 128

From task.get_recipe():

• OUTDIR: where outputs are written (this should typically be the task AI results folder)

• COS.bucket, COS.ResultKey, COS.endpoint: used to populate the DB doc

• GeoJSON.static.url: a string template used to build properties.source.url

From Click context (ai makegeojson populates these in ctx.meta):

• --cos-key: overrides the recipe COS key; if it starts with +, it is treated as a suffix

• --friendly-name: overrides the friendly name; if it starts with +, it is treated as a suffix

• --print / --less: pretty-print the generated DB doc

What it produces

1. Processed GeoJSON written to OUTDIR:

• File name: output_file_name or <input_stem>.geojson

• Content: original features plus two styling fields:

  • fill: rgba(r, g, b, a)

  • stroke: same as fill

  Note: upload.py/publish.py currently derive expected filenames from task.config.file_path (the <input_stem>). If you set output_file_name to a different stem, the built-in upload/publish steps will not find the files unless you also adjust those steps or keep the default naming.

Color assignment is computed by:

• selecting group_by_column (default index)

• assigning each unique group a color from the chosen matplotlib colormap

2. DB metadata document written to OUTDIR:

• File name: <stem>_db.geojson

• Content: a GeoJSON Feature with:

  • properties.ResultKey, properties.Bucket, properties.Endpoint

  • properties.friendly_name

  • properties.source.url built from GeoJSON.static.url and the bucket + result key

  • geometry set to the dataset bounding box

Source discovery behavior

If task.config.file_path is not set, the strategy searches the task AI results directory for the first match in:

• **/*.shp

• **/*.kml

• **/*.geojson

upload.py (VectorUploadStrategy)

Purpose: upload the processed GeoJSON and the DB doc to COS.

Inputs

From task.config or recipe:

• file_path: used only to determine the output stem/name

From Click context (ai upload populates these in ctx.meta):

• --dry-run: print what would upload

• --cos-key: overrides upload key

Files it expects

The upload step looks in the task AI results folder for files derived from task.config.file_path:

• <input_stem>.geojson

• <input_stem>_db.geojson

It does not currently consider task.config.output_file_name when locating the files to upload.

It then uploads them as:

• COS_KEY

• COS_KEY_db.geojson

Where COS_KEY is determined by:

• --cos-key (if provided), otherwise

• properties.ResultKey read from <input_stem>_db.geojson

--cos-key is passed through string.Template(...).substitute(task.config), so keys can reference task config values like ${id}.

publish.py (VectorPublishStrategy)

Purpose: publish the DB doc (*_db.geojson) to the configured docs database.

Inputs

From Click context (set by ai publish post):

• --dry-run

• --print-json

From recipe:

• APP_DOCS_DB: target docs database name/URL (consumed by ocli.cli.publish.publish_json)

File it publishes

The publish step reads from the task AI results folder, derived from task.config.file_path:

• <input_stem>_db.geojson

It does not currently consider task.config.output_file_name when locating the DB doc to publish.

…and sends it to the docs DB.

Using the strategies in an AIKP template

An AIKP template can call the strategy classes directly:

from ocli.classes.task_template import TaskTemplate
from ocli.ai.vector_strategies.makegeojson import VectorGeoJsonStrategy
from ocli.ai.vector_strategies.upload import VectorUploadStrategy
from ocli.ai.vector_strategies.publish import VectorPublishStrategy

class Template(TaskTemplate):
    @classmethod
    def makegeojson(cls, task, cos_key=None, friendly_name=None):
        VectorGeoJsonStrategy(task).make_geojson()

    @classmethod
    def upload(cls, task, dry_run=False, cos_key=None):
        VectorUploadStrategy(task).upload()

    @classmethod
    def publish(cls, task, dry_run=False, cos_key=None):
        VectorPublishStrategy(task).publish()