Goblin.nvim is a Neovim plugin that allows you to create custom, automed workflows via "steps". Using steps allows for taking modular actions and stringing them together to form complex workflows. These steps can be reused in other workflows, shared with others, etc. Most of my work has gone into workflows that are trying to update the buffers of a current project, similar to Github Copilot Workflow, where the ai agents need guidance and context beyond a single buffer.
Goblin works by following a series of steps that you define in a configuration file (.goblin.json) or in your Neovim setup. Each step takes the input from the previous step and returns the output for the next step. Steps can be anything. The default steps are: data, plan, code, and push.
-
Data: Define input sources. The input source is something like a Github Issue or a Jira ticket that should provide the context for what needs to be updated in the codebase. The data step also has you select one of the input sources as the output of the step. You can also define a manual entry step that lets you enter the data when the workflow runs. Currently Goblin supports the following input sources: • Jira • Linear Roadmap: • Github • Manual Entry
Input: None Output: Buffer with selected data from the input source.
-
Convert: Take the instructions and format them in the Goob language. Goob allows us to take command problems and express them as
Input: Buffer with instructions. Output: Buffer with plan formatted in markdown.
-
Plan: Plan a step by step guide on updating a codebase. This takes in a general instruction and determines the best strategy to take to update the current project. This is accomplished by
Input: Buffer with instructions formatted in Goob. Output: Buffer with plan formatted in markdown.
-
Code:
Input: Buffer with plan. Output: Updates files with code changes.
-
Push: Performs the following steps Input: Output: Creates Github PR and pushes changes
Goblin requires a few dependencies to work. Install with your plugin manager.
- Determine where your input is coming from.
- Define workflow
- Add .goblin file to your project directory.
- Add configuration to .goblin file (see Workflows)
The .goblin file is a configuration file that allows you to specify the workflow for your project. It should be placed in the root of your project directory. The following values can be set in the .goblin file:
workflow: This is the workflow that your goblin will follow. It is a list of steps that your goblin will follow. Each step has a name, description, type, and extra field. The name is the name of the step, the description is a description of the step, the type is the type of step (data, plan, code, accept), and the extra field is any extra information needed for the step.
credentials: This is the credentials that your goblin will use to access the input source. It is a list of credentials that your goblin will use to access the input source. Each credential has a name and a value. The name is the name of the credential, and the value is the value of the credential.
connections = [{ "name": "jira", "username": "username", "password": "password" # do not store in plain text, load from a secure location }, { "name": "github", "username": "username", "password": "password" # do not store in plain text, load from a secure location }]
A workflow is a list of steps your Goblin will delicately follow. Workflows are defined in your .goblin file. An example workflow looks like this.
workflow = [{ name: "Select Jira Ticket", description: "Grab current jira tickets and list them. Select one to start.", type: "data", extra: { "source": "jira", "project_id": "1234" }, ui: "select", }, { name: "Plan", description: "Plan the work that needs to be done.", type: "plan", extra: { "" }, ui: "textbox", }, { name: "Code", description: "", type: "code", ui: "code", }, { name: "Create Github PR", description: "Review the changes and create a github PR.", type: "create" ui: "confirm" }]
data: input: none (calls data source or manual entry) ouptut: data
plan: creates a plan of action from a buffer input: buffer output: table of plan steps
code: uses a plan or a buffer to generate code changes input: buffer or table of plan steps ouptut: table of file changes
push: pushes a tables of file changes to a remote git repository input: table of file changes output: success or error
aider: takes a buffer and an instruction and runs it through aider input: buffer ouput: aider will update files and git commit the changed files
You can define your own steps either via your goblin configuration (less configurable) or via a lua file in your plugin directory. If you create a lua file for a step that you think is particularly useful, please submit a pull request to add it to the available steps. Keep in mind that any step should clean up its own messes - make sure to close any open popups or other UI elements when the step is complete.
Other workflows are available in the workflows folder in this repository. If you have a workflow you think is useful, please submit a pull request to add it to the available workflows.
select textbox code confirm
{ "name": this is the name of the workflow step, it can be anything "description": if you want to add some more info "type": this is the type of step, it can be "data", "plan", "code", "accept" "extra": this is used for any extra info needed for each workflow step. For example, for data steps, you might need to specify the source of the data. }
I map :Goblin to ai which initiates the workflow. Browse/Search through issues and select one to start. From there, your default
workflow will dictate how the algorithm proceeds.
The purpose of this plugin is to speed up development time, and this includes the training of the
These are sources that I used to create this plugin. Some might be flawed, use incorrect assumptions, or be flat out wrong. If you see mistakes, let me know :)
How Copilot Works Under The Hood https://thakkarparth007.github.io/copilot-explorer/posts/copilot-internals.html
CrewAI: Autonomous AI Agents https://www.crewai.com/
Q: why is it called Goblin? A: the menace that will take your (my) job
Q: what does it do? A: it allows you to create workflows via easy configuration that you can use to develop code. i imagine as editors evolve, they will continue to abstract the complexity of the problems we work on. goblin is an attempt at that abstraction.
data = {
order = 1,
source = "jira",
source_options = {
domain = "cityflavor.atlassian.net",
user = "weston@cityflavor.com",
token = "ATATT4xFfGF006Q8hXZFnRD430aTWKWLnHGy0TVNiUb5R5mPElQ2APpj5IGMpUOTQLKiHrShWKT1TAObdEjilbYg_3B3IrLX3jjFuv1_GFdXEXpQrw8hXQJowSR1h6S8rJ6aULHmi2CX0ZzSvJaUpJXt19dtkb2S0TuPfstEBZrIYDm6nOTyH_Y=061A49EB",
params = {
project = "CFDEV",
-- sprint = "current", -- TODO
}
},
},