Evanfeenstra ↗ analysis ↗ GitHub stakwork/mikeoss
stakwork is turning Mike into an assistant you can run on your own hardware, with local inference and voice-driven chat.
stakwork/mikeoss is a fork of Mike pointed squarely at self-hosted deployment. The headline move is bringing Ollama into the mix as an LLM provider, so the assistant can talk to a model running on your own machine instead of routing every prompt out to a hosted API. Alongside that, the fork is fleshing out the day-to-day chat experience and the operational story for actually running it.
The work is coming from stakwork, with contributions landing under @Evanfeenstra and @pitoi. Recent threads touch the chat surface itself (a microphone button for dictating prompts, polish on the Ollama model picker and a state for when account settings tools aren't available) and the deployment side (a Docker-based setup that targets both local development and production).
Taken together, the direction reads as: keep the chat UI usable and hands-free, give people a real path to run Mike on their own hardware, and make standing it up - locally or in production - a known quantity. If that overlap interests you, the GitHub repo is where the detail lives.
What's in it
- Local inference via Ollama A provider adapter lets the fork route chats to a model running on your own hardware instead of a hosted API, with model listing and routing wired in.
- Voice input in chat A mic button on the chat box uses speech recognition so users can dictate prompts instead of typing them.
- Ollama model picker polish Targeted UI work on the model picker that appears once Ollama is in play, alongside an unavailable-state for account settings tools.
- Docker setup for dev and prod A containerised setup intended to cover both local development and production deployment of the fork.
Direction
infrastructurechat-uiintegration
Activity
pitoi ↗ analysis ↗ GitHub pitoi ↗ analysis ↗ GitHub Evanfeenstra ↗ analysis ↗ GitHub Threads of work (detailed view)
stakwork lets you run Mike on your own hardware
The fork wires in Ollama so the assistant can talk to a local model instead of Claude or Gemini.
Evan Feenstra gives stakwork's Mike a microphone
A mic button on the chat box lets users dictate prompts instead of typing them.
stakwork brings local inference to mikeoss with Ollama
A new provider adapter lets the fork run models on your own hardware instead of shipping prompts to a hosted API.
Pull requests (detailed view)
🟢 Open (1)
#2 Generated with Hive: Add docker directory with local dev compose, setup scripts, and production Dockerfiles Evanfeenstra · opened 22d ago ✅ Merged (3)
#3 Generated with Hive: Add voice input with mic button and speech recognition to ChatInput Evanfeenstra · opened 22d ago · merged 15d ago by pitoi #5 Generated with Hive: Add Ollama model picker, account settings section, and tool-unavailability notice to UI pitoi · opened 18d ago · merged 18d ago by pitoi #4 Generated with Hive: Add Ollama LLM provider adapter, routing, and model list endpoint pitoi · opened 18d ago · merged 18d ago by pitoi ⛔ Closed without merge (1)
fayekelmith · opened 24d ago · closed 24d ago