sean-a-ward/mike

Automated upstream sync and OpenRouter/LiteLLM routing via env or per-user settings

Two topics of work. The fork-sync workflow is a single-file addition with no ongoing development cost. The OpenAI-compatible endpoint support spans three commits touching `openai.ts`, `userApiKeys.ts`, `schema.sql`, and several frontend settings files - it's an actively developed feature rather than a quick config tweak.

View on GitHub →

sean-a-ward's fork runs a daily GitHub Actions workflow that merges willchen96/mike's main branch automatically, keeping the fork current without manual intervention. On top of that, the fork adds first-class support for routing the OpenAI provider to any OpenAI-compatible Chat Completions endpoint - OpenRouter, LiteLLM, LM Studio - configurable at both the instance level (via env vars) and per user through the account settings UI.

The OpenRouter work is the more substantial change: new columns on user_api_keys, a model-ID remapping layer (OPENAI_MODEL_MAP), parallel streaming paths for the Responses API vs chat/completions, and a migration-tolerance fallback for rolling out the schema change without downtime.

What's in it

Bring-your-own model provider Point Mike at OpenRouter, LiteLLM, LM Studio, or anything else that speaks the OpenAI API - not just OpenAI itself.
Local and self-hosted inference With LM Studio support in the mix, you can run Mike against models on your own hardware instead of sending prompts to a third party.
Daily upstream sync Automation pulls in changes from the original Mike project every day, so the fork doesn't drift away from the parent.

Direction

integrationinfrastructure

Activity

Themed changes and pull requests touching this fork, newest first. Themed changes that haven't been turned into a public post yet still appear — they're real work even without a published writeup.

📝 sean-a-ward unhooks Mike from a single AI vendor 3 commits 2mo ago infrastructureintegration draft

The fork can now point its AI engine at whatever provider you choose - and let each user pick their own.

📝 Fork-sync GitHub Actions workflow 1 commit 2mo ago minor change

Threads of work (detailed view)

2 threads have been distilled into posts.

sean-a-ward unhooks Mike from a single AI vendor

The fork can now point its AI engine at whatever provider you choose - and let each user pick their own.

infrastructureintegration

3 commits · latest 2mo ago

sean-a-ward adds per-user OpenRouter/LiteLLM routing to the OpenAI provider

Three commits from sean-a-ward let Mike's OpenAI provider route to any OpenAI-compatible Chat Completions endpoint - OpenRouter, LiteLLM, LM Studio - both at the instance level via env vars and per user through the settings UI. Model IDs are remapped through a JSON map so Mike's internal names point at whatever the target provider expects.

infrastructureintegration

3 commits · latest 2mo ago