Add Neuralwatt as a compatible provider

[404oops]

Hello everyone! I love Puter and all other projects built with it. With that being said, I feel like self-hosted Puter instances could use

Neuralwatt is an open-source-exclusive LLM provider (with options to host your own infrastructure, more on that later) whose main premise is charging for how much power you use, with the option to switch to token-based pricing if you don't like power-based pricing.

Now for the technical part

Its API is OpenAI-compatible, however it includes an energy field which is utilized in the billing section. I cross-referenced what the console total said to the Neuralwatt dashboard's total and found the results to be accurately represented.

I wrote a unit test to verify if everything is as it should be, according to other unit tests (particularly the OpenRouter one), and I verified that all 17 of them pass.

I wrote the provider backend, which wasn't really complicated, but I followed a convention that fetches at "boot", since Neuralwatt doesn't change very often. I do feel like that behavior should be discussed with y'all, so if you have any objections to that, do let me know.

I also wired it into the provider driver which is necessary for it to even work.

I tested it with my own keys (inside the console, unfortunately), and it works fine. However, due to the nature of Puter's dev environment, I can't test it in apps, so I'm quite unsure on how I can test it with local apps and how models are handled. But it does work.

If you have any questions, or guidance on how I can improve this feature, please let me know. I'm still learning the codebase and how it works, and I think I got a hang of this part.

[CLAassistant]

All committers have signed the CLA.

[Copilot]

Pull request overview

This PR adds Neuralwatt as a new OpenAI-compatible backend provider for the ai-chat driver, including model catalog handling, metering/billing integration (cost + energy annotations), and automated tests to validate request shaping and usage accounting.

Changes:

• Added NeuralwattProvider implementation with model catalog + quota caching, vision-model gating, and OpenAI-compatible chat completions wiring.
• Added Neuralwatt model catalog mapping utilities and comprehensive offline unit tests (including streaming + metering behavior).
• Registered the Neuralwatt provider in ChatCompletionDriver and extended driver edge tests to cover provider registration.

Reviewed changes

Copilot reviewed 5 out of 5 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
src/backend/drivers/ai-chat/providers/neuralwatt/NeuralwattProvider.ts	Implements Neuralwatt chat provider with catalog/quota fetch, vision handling, and metering integration for cost/energy.
src/backend/drivers/ai-chat/providers/neuralwatt/NeuralwattProvider.test.ts	Adds unit tests covering catalog/quota caching, request shaping, streaming, and billing/metering behavior.
src/backend/drivers/ai-chat/providers/neuralwatt/models.ts	Defines Neuralwatt catalog types + mapping to IChatModel, plus helpers for vision detection and image-part detection.
src/backend/drivers/ai-chat/ChatCompletionDriver.ts	Registers Neuralwatt provider and updates model-map aggregator handling.
src/backend/drivers/ai-chat/ChatCompletionDriver.edges.test.ts	Adds Neuralwatt to edge-test provider config and stubs its model catalog for model-map building.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.