Skip to main content

API Development

This section of the guide provides guidance on how to work with the Meltano API, which serves as the backend of Meltano and is built with the Python framework: Flask.

Getting Set Up

After all of your dependencies installed, we recommend opening a new window/tab in your terminal so you can run the following commands:

# Activate your poetry created virtual environment if needed.
# If you manage your virtualenv activation through other means you can omit this command.
poetry shell

# Create a new Meltano project

# Change directory into your newly created project

Debugging Your Code

To debug your code, we recommend using the Python debugger. It can be invoked by adding breakpoint() in your Python code at the location you want to drop into the debugger. From there you can drop into a Python interactive console by entering interact.

API V2 Design Guidelines

For V2 of the Meltano API, we generally aim to implement and adhere to our variation of a Resource Oriented Architecture.

The key characteristic of a resource-oriented API is that it emphasizes resources (data model) over the methods performed on the resources (functionality). A typical resource-oriented API exposes a large number of resources with a small number of methods. The methods can be either the standard methods or custom methods.

Where API functionality naturally maps to one of the standard methods, that method should be used in the API design. For functionality that does not naturally map to one of the standard methods,custom methods may be used. Custom methods offer the same design freedom as traditional RPC APIs, which can be used to implement common programming patterns, such as database transactions or data analysis.

Concrete methods/verbs

We stick to traditional methods/verbs List, Get, Create, Update, and Delete, with the following HTTP equivalents:

  • List
    • HTTP GET <collection>, No request body, Resource list as a response
  • Get
    • HTTP GET <resource>, No request body, Resource as a response
  • Create
    • HTTP POST <collection>, Resource as request body, Resource as response body
  • Update
    • HTTP PUT, PATCH <resource>, Resource as request body, Resource as response body
  • Delete
    • HTTP DELETE <resource>, No request body, No response body
    • HTTP DELETE <resource>, No request body, Resource as response body with a state indicating deletion (if a soft delete is being performed)

Custom methods/verbs

Occasionally, during development we may also have a need for custom verbs (e.g. for things like copying state or bulk operations). In those case, we'll map to an appropriate HTTP verb, but the MR/issue will require justification for the use of a custom verb, and be documented in this guide. Where a custom verb is used, the verb will be appended to the resource as a new path segment:

  • Move
    • HTTP POST <collection>/<resource>/move (we follow a more conventional style, rather than using Google's <resource>:<verb> style)

Collections and resources

API service umbrella and namespace:

  • meltano/core/v2beta (during development)
  • meltano/core/v2

Our top-level collection is projects/* with an intermediate of /envs/* as we've opted to include support for operating on multiple projects. The V2 API spec is in flux, we don't yet actually have strong ties between a lot of the topics beyond this level, but as we start specing v2 we expect a grouping around the below features to evolve.

  • envs/*/plugins/*
  • envs/*/state/*
  • envs/*/schedules/*
  • envs/*/run/*
  • envs/*/test/*
  • envs/*/service/*
  • envs/*/jobs/*

As a hypothetical on how the spec might evolve, we may also end up organize around a job as resource or the like. Nesting under a jobs collection might look more like:

  • envs/*/jobs/*
    • envs/*/jobs/run (potentially a custom verb - need to support ad hoc meltano run like invocations, where the submitting client doesn't know what the state IDs will be)
    • envs/*/jobs/*/runs/*
    • envs/*/jobs/*/schedule/*
    • envs/*/jobs/*/state/*? (up for discussion whether state should be stand-alone or part of the jobs' collection.)
    • envs/*/jobs/*/tests/*? (would meltano test ever become a "job")

API Design Examples

These are for illustration only, and subject to change.

# list all runs
GET meltano/core/v2/envs/prod/jobs/tap-gitlab-target-jsonl/runs
Response: {[the runs]}

# get a specific run
GET meltano/core/v2/envs/prod/jobs/tap-gitlab-target-jsonl/runs/ff643ba2
Response: {A run}

# meltano run equivalent (custom google style verb: run)
POST meltano/core/v2/envs/prod/jobs:run
{"command": "tap mapper target dbt:run", "args"...}
Reponse: {The run, with run_id , and "inprogress" status field}

# meltano state set equivalent
PUT meltano/core/v2/envs/stage/jobs/tap-gh-to-target-sql/state
{new state message}
Reponse: {"The new state after update"}

# meltano state copy equivalent (custom verb: copy)
POST meltano/core/v2/envs/stage/jobs/tap-gh-to-target-sql/state/copy
{destination: "different-state-id"}
Response: {"ref to the dest")

Guidelines for FastAPI pydantic models and auto generated documention