kartAI · jathavaan · May 27, 2026 · May 27, 2026 · May 27, 2026 · May 27, 2026
diff --git a/.docker/Api.Dockerfile b/.docker/Api.Dockerfile
diff --git a/.github/workflows/publish-api.yml b/.github/workflows/publish-api.yml
diff --git a/.github/workflows/pull-request-tests.yml b/.github/workflows/pull-request-tests.yml
@@ -31,7 +31,6 @@ jobs:
       pull-requests: read
     outputs:
       orchestrator: ${{ steps.filter.outputs.orchestrator }}
-      api: ${{ steps.filter.outputs.api }}
       benchmarks: ${{ steps.filter.outputs.benchmarks }}
     steps:
       - uses: actions/checkout@v4
@@ -57,13 +56,6 @@ jobs:
               - '.docker/Setup.Dockerfile'
               - 'requirements.txt'
               - 'docker-compose.yml'
-            api:
-              - 'src/**'
-              - '!src/presentation/entrypoints/**'
-              - '!src/presentation/databricks/**'
-              - '.docker/Api.Dockerfile'
-              - 'requirements.txt'
-              - 'docker-compose.yml'
 
   compile:
     name: Check Python syntax
@@ -109,24 +101,6 @@ jobs:
       - name: Build Container Orchestrator from docker-compose
         run: docker compose build container-orchestrator
 
-  build-api-image:
-    name: Build VMT API Server
-    runs-on: ubuntu-latest
-    needs:
-      - compile
-      - detect-changes
-    if: needs.detect-changes.outputs.api == 'true'
-
-    steps:
-      - name: Checkout code
-        uses: actions/checkout@v4
-
-      - name: Set up Docker Buildx
-        uses: docker/setup-buildx-action@v3
-
-      - name: Build VMT API Server from docker-compose
-        run: docker compose build vmt-api-server
-
   build-benchmark-images:
     name: Build ${{ matrix.display_name }}
     runs-on: ubuntu-latest

diff --git a/CLAUDE.md b/CLAUDE.md
@@ -4,14 +4,14 @@ Reproducible benchmarking framework comparing cloud-native (DuckDB + GeoParquet,
 
 ## Stack
 
-Python, DuckDB (spatial), PostGIS on Azure Database for PostgreSQL, Apache Sedona on Databricks, Azure Blob Storage, Azure Container Instances, `dependency_injector`, FastAPI, PMTiles/MVT. See `requirements.txt` for versions.
+Python, DuckDB (spatial), PostGIS on Azure Database for PostgreSQL, Apache Sedona on Databricks, Azure Blob Storage, Azure Container Instances, `dependency_injector`. See `requirements.txt` for versions.
 
 ## Layout (Clean Architecture)
 
 - `src/domain/` — enums only; no dependencies on other layers.
 - `src/application/` — `contracts/` (service interfaces), `dtos/`, `common/` (logger, monitor).
 - `src/infra/` — `infrastructure/services/` (contract impls), `infrastructure/containers.py` (DI wiring), `persistence/context/` (DuckDB, Postgres, Blob clients).
-- `src/presentation/` — `entrypoints/` (one file per benchmark), `configuration/app_config.py` (`initialize_dependencies`), `databricks/` (notebook script), `endpoints/tile_server.py` (FastAPI VMT server).
+- `src/presentation/` — `entrypoints/` (one file per benchmark), `configuration/app_config.py` (`initialize_dependencies`), `databricks/` (notebook script).
 - `main.py` — outside-ACI orchestrator. Reads `benchmarks.yml`, launches one ACI per experiment.
 - `benchmark_runner.py` — in-container dispatcher. Matches `--script-id` to a function in `src/presentation/entrypoints/`.
 - `benchmarks.yml` — experiment manifest. Each entry: `id`, `image`, `cpu`, `memory_gb`, `related_script_ids`.

diff --git a/README.md b/README.md
@@ -2,7 +2,7 @@
 
 doppa is a reproducible benchmarking framework for evaluating traditional geospatial query stacks
 (PostGIS, shapefiles) against cloud-native geospatial (CNG) alternatives (DuckDB over GeoParquet in
-blob storage, PMTiles/MVT vector tiles, and Apache Sedona on Databricks) across a range of real-world
+blob storage and Apache Sedona on Databricks) across a range of real-world
 spatial query patterns: point-in-polygon lookups, k-nearest-neighbour search, bounding-box filtering,
 and a national-scale spatial join.
 
@@ -13,7 +13,7 @@ measurable and reproducible on identical datasets and hardware.
 
 <div align="center">
 
-[![Push containers to Azure Container Registry](https://github.com/kartAI/doppa-data/actions/workflows/push-containers-to-acr.yml/badge.svg)](https://github.com/kartAI/doppa-data/actions/workflows/push-containers-to-acr.yml) [![Publish APIs](https://github.com/kartAI/doppa-data/actions/workflows/publish-api.yml/badge.svg)](https://github.com/kartAI/doppa-data/actions/workflows/publish-api.yml)
+[![Push containers to Azure Container Registry](https://github.com/kartAI/doppa-data/actions/workflows/push-containers-to-acr.yml/badge.svg)](https://github.com/kartAI/doppa-data/actions/workflows/push-containers-to-acr.yml)
 
 </div>
 
@@ -60,7 +60,7 @@ format internals to client-observed cost is measured end to end.
 **Cloud-native vector formats vs. traditional formats on cloud storage.** Empirical comparisons in the literature
 (Holmes 2023; Flatgeobuf 2024) measure write times and file sizes on local disk and do not place cloud-native and
 traditional formats side by side on cloud storage. doppa benchmarks GeoParquet over Azure Blob Storage (via DuckDB)
-against PostGIS on Azure Database for PostgreSQL, and PMTiles against WMS-style vector tiles, across the active
+against PostGIS on Azure Database for PostgreSQL, across the active
 catalog of query patterns: point-in-polygon lookups, k-nearest-neighbour search, bounding-box filtering, and a
 national-scale spatial join. The local-Shapefile entrypoints sit on the side as a laptop-workflow reference, with the
 Shapefile downloaded ahead of the timed scope to emulate that workflow rather than to bench the format on cloud
@@ -167,8 +167,6 @@ to the elapsed-time distribution.
 | PostGIS                 | Single-node, managed service     | Azure Database for PostgreSQL Flexible Server                                          |
 | GeoPandas + Shapefile   | Single-node, local-disk baseline | Shapefile pre-downloaded to the container before the timed scope                       |
 | Apache Sedona           | Distributed                      | Azure Databricks, 2 / 4 / 8 / 12 / 16 `Standard_D4s_v3` workers, reading GeoParquet via ABFS |
-| PMTiles                 | Cloud-native vector tiles        | PMTiles archive in blob storage, accessed via HTTP range reads                         |
-| WMS-style vector tiles  | Traditional vector tiles         | `doppa-vmt` web app for containers, tiles assembled on demand                          |
 
 DuckDB and PostGIS each run inside an Azure Container Instance with 4 vCPU and 16 GB RAM, so CPU and memory baselines
 match between the single-node engines.
@@ -348,17 +346,16 @@ so.
 #### Resource naming
 
 The resource names used throughout this section (`doppa`, `doppabs`, `doppaacr`, `doppa-uami`,
-`doppa-db`, `doppa-vmt`, `doppa-databricks`) are baked into source and configuration. Keep them
+`doppa-db`, `doppa-databricks`) are baked into source and configuration. Keep them
 as-is for the simplest setup; this is also what the thesis deployment uses, so reproducing the
 published results requires these exact names.
 
 If you need to rename a resource, the following references must be updated together:
 
 | Location                            | What is hardcoded                                                              |
 |-------------------------------------|--------------------------------------------------------------------------------|
-| `src/config.py`                     | Default values for resource group, blob URL/account, VMT URL, STAC container   |
+| `src/config.py`                     | Default values for resource group, blob URL/account, STAC container            |
 | `benchmarks.yml`                    | ACR image references (`doppaacr.azurecr.io/<image>:latest`) for every benchmark |
-| `.github/workflows/publish-api.yml` | `webapp_name: doppa-vmt`                                                       |
 
 `src/config.py` defaults can also be overridden via the corresponding environment variables
 (see [Local development](#local-development) and [GitHub Actions](#github-actions)) without
@@ -464,34 +461,6 @@ same setting change the following:
 - `effective_cache_size`: `6291456`
 - `work_mem`: `65536`
 
-#### Web app for containers
-
-Create
-a [web app for containers](https://portal.azure.com/#view/Microsoft_Azure_Marketplace/GalleryItemDetailsBladeNopdl/id/Microsoft.AppSvcLinux/selectionMode~/false/resourceGroupId//resourceGroupLocation//dontDiscardJourney~/false/selectedMenuId/home/launchingContext~/%7B%22galleryItemId%22%3A%22Microsoft.AppSvcLinux%22%2C%22source%22%3A%5B%22GalleryFeaturedMenuItemPart%22%2C%22VirtualizedTileDetails%22%5D%2C%22menuItemId%22%3A%22home%22%2C%22subMenuItemId%22%3A%22Search%20results%22%2C%22telemetryId%22%3A%22135c4e97-6a92-446e-aa0a-3f2201ddfdb1%22%7D/searchTelemetryId/c154ee0a-06d6-49e4-a17f-3820937e6335)
-The process is the same for each of the following API servers:
-
-- `doppa-vmt`
-
-Under *Basics*:
-
-- Resource group: `doppa`
-- Name: `<name-from-list-above>`
-- Publish: `Container`
-- Operating system: `Linux`
-- Pricing plan: `Premium V4 P0V4`
-
-Under *Container*:
-
-- Image source: `Azure Container Registry`
-- Registry: `doppaacr`
-- Authentication: `Managed identity`
-- Identity: `doppa-uami`
-- Image: `<select the image that matches with the name>`
-- Tag: `latest`
-- Startup command `uvicorn src.presentation.endpoints.<API server script>:app --host 0.0.0.0 --port 8000`
-
-Navigate to *Review + create* and create the resource. Repeat this process for each name in the list.
-
 #### Databricks
 
 The national-scale spatial join benchmarks run on Azure Databricks using Apache Sedona. A separate Databricks

diff --git a/docker-compose.yml b/docker-compose.yml
@@ -217,14 +217,3 @@ services:
       dockerfile: .docker/Query.Dockerfile
     image: national-scale-spatial-join-databricks-partitioned-16-nodes:latest
     command: python benchmark_runner.py --script-id national-scale-spatial-join-databricks-partitioned-16-nodes --benchmark-run 1 --run-id ABCDEF
-
-  vmt-api-server:
-    env_file:
-      - .env
-    build:
-      context: .
-      dockerfile: .docker/Api.Dockerfile
-    ports:
-      - "8000:8000"
-    image: vmt-api-server:latest
-    command: uvicorn src.presentation.endpoints.tile_server:app --host 0.0.0.0 --port 8000
diff --git a/requirements.txt b/requirements.txt
@@ -44,7 +44,6 @@ dependency-injector==4.48.2
 dotenv==0.9.9
 duckdb==1.4.0
 executing==2.2.1
-fastapi==0.135.1
 fastjsonschema==2.21.2
 fiona==1.10.1
 folium==0.20.0
@@ -89,7 +88,6 @@ lark==1.3.0
 MarkupSafe==3.0.3
 matplotlib==3.10.6
 matplotlib-inline==0.1.7
-mercantile==1.2.1
 mistune==3.1.4
 msal==1.34.0
 msal-extensions==1.3.1
@@ -111,7 +109,6 @@ parso==0.8.5
 pexpect==4.9.0
 pillow==11.3.0
 platformdirs==4.4.0
-pmtiles==3.7.0
 prometheus_client==0.23.1
 prompt_toolkit==3.0.52
 propcache==0.4.1
@@ -152,7 +149,6 @@ sniffio==1.3.1
 soupsieve==2.8
 SQLAlchemy==2.0.47
 stack-data==0.6.3
-starlette==0.52.1
 terminado==0.18.1
 tinycss2==1.4.0
 tornado==6.5.2
@@ -164,7 +160,6 @@ typing_extensions==4.15.0
 tzdata==2025.2
 uri-template==1.3.0
 urllib3==2.5.0
-uvicorn==0.41.0
 viztracer==1.1.1
 watchfiles==1.1.1
 wcwidth==0.2.14

diff --git a/src/application/contracts/__init__.py b/src/application/contracts/__init__.py
@@ -12,13 +12,10 @@
 from .file_path_service_interface import IFilePathService
 from .fkb_service_interface import IFKBService
 from .monitoring_storage_service import IMonitoringStorageService
-from .mvt_service_interface import IMVTService
 from .open_street_map_file_service_interface import IOpenStreetMapFileService
 from .open_street_map_service_interface import IOpenStreetMapService
 from .release_service_interface import IReleaseService
 from .stac_io_service_interface import IStacIOService
 from .stac_service_interface import IStacService
 from .test_dataset_service_interface import ITestDatasetService
-from .tile_api_service_interface import ITileApiService
-from .tile_service_interface import ITileService
 from .vector_service_interface import IVectorService
diff --git a/src/application/contracts/bytes_service_interface.py b/src/application/contracts/bytes_service_interface.py
@@ -1,5 +1,4 @@
 from abc import ABC, abstractmethod
-from pathlib import Path
 
 import pandas as pd
 import geopandas as gpd
@@ -62,13 +61,3 @@ def convert_df_to_parquet_bytes(df: pd.DataFrame | gpd.GeoDataFrame) -> bytes:
         """
         raise NotImplementedError
 
-    @staticmethod
-    @abstractmethod
-    def convert_pmtiles_to_bytes(path: Path) -> bytes:
-        """
-        Converts a PMTiles file to a byte array.
-        :param path: Path to the PMTiles file.
-        :return: Byte array representation of the PMTiles file.
-        :rtype: bytes
-        """
-        raise NotImplementedError