Process documents
Extract fields, classify text, normalize records, summarize content, and validate model output in a later phase.
Open source · Node.js · CLI + library
Model work you can inspect.
Ploinky Workers runs small, explicit task flows. Give it a request or a reusable task file, choose models by tier, and let one CLI and its local proxy handle execution, status, limits, and optional batching.
The open source part of a broader family of Axiologic internal tools.
pworker / terminal
# Connect a provider and assign a model to a tier $ pworker # Start work and get a task ID $ pworker run ./extract.json \ --input '{"input":"report.txt"}' \ --cwd ./project --async { "id": "9f3…", "status": "queued" } # Check the current phase or final result $ pworker --status 9f3…
Explicit phasesReviewable JSON task files
Model tiersChoose the provider separately
Background workIDs, phase status, persisted results
Optional batchingCombine eligible requests
Use cases
For repeatable work with changing inputs.
Describe the processing steps once, then run them against many requests or files.
Work with project files
Let a task read and write files within the exact working directory supplied by its caller.
Run many tasks
Queue work, flush it explicitly, return IDs immediately, and fetch progress or results later.
The idea in one example
Input → model → code → result.
A task can ask a model on the tiny tier to extract a title, then use a model-free JavaScript phase to save that title. The task controls its next phase explicitly. The tier chooses the configured model; the file describes the work.
Open source and evolving
Built from practical experience.
Ploinky Workers is an open source part of more sophisticated internal Axiologic tooling. This edition provides task phases, provider routing, local model support, monitoring, background execution, and prompt batching. We will add features gradually. For particular deployments and more sophisticated use cases, Axiologic offers consulting and customization around task flows, providers, batching, and local inference.