### Overview
jpbreda/mike
A fork exploring how to run Mike on AI you host yourself, not just the commercial cloud services.
This is jpbreda's take on Mike, and the through-line is control over where the model runs. Rather than being limited to the commercial cloud backends Mike ships with, the work here opens a path to point Mike at a model instance you host on your own hardware - an OpenAI-compatible local endpoint sitting alongside the existing options.
The why is easy to guess for anyone in legal tech: a self-hosted model means sensitive matter data never has to leave your infrastructure. jpbreda hasn't wrapped a rebrand or a niche around this - it's a focused, infrastructure-minded change to who Mike can talk to.
There's not much else diverging here yet. If the self-hosting angle matters to you, this is the fork to watch, and GitHub is the place to see how far along it actually is.
What's in it
- Self-hosted model support Point Mike at a privately-hosted AI instance instead of relying only on the commercial cloud providers.
- OpenAI-compatible endpoints Adds a third backend option that speaks the common local-inference protocol, so operators can bring their own hosted models.
- Keeps the existing providers The self-hosted route sits alongside the cloud backends Mike already supports rather than replacing them.
Direction
infrastructureintegration
Activity
### Overview
📝 jpbreda tries to let Mike run on your own hardware 3 commits 3mo ago integrationinfrastructure draft
A proposal to point Mike at self-hosted AI models, not just the big commercial clouds.
📝 jpbreda teaches Mike to run on a model you host yourself 3 commits 3mo ago infrastructureintegration draft
A new option lets Mike talk to a privately-hosted AI instead of only the commercial cloud services.
Threads of work (detailed view)
jpbreda tries to let Mike run on your own hardware
A proposal to point Mike at self-hosted AI models, not just the big commercial clouds.
jpbreda teaches Mike to run on a model you host yourself
A new option lets Mike talk to a privately-hosted AI instead of only the commercial cloud services.
jpbreda adds vLLM as a third LLM provider alongside Claude and Gemini
This fork wires a self-hosted vLLM endpoint into Mike's provider system, letting operators swap the model layer to a local inference server instead of routing all requests through Anthropic or Google. The PR closed without merging upstream on May 10.
Pull requests (detailed view)
⛔ Closed without merge (2)
by
jpbreda · opened 3mo ago · closed 2mo ago by
jpbreda · opened 3mo ago · closed 2mo ago