fabiotaroc/mike

fabiotaroc/mike - adds OpenAI as a third LLM provider alongside Claude and Gemini

Single topic recorded. The fork adds OpenAI provider support without restructuring anything upstream. Scope is narrow; there are no infrastructure changes, auth changes, or other divergences from the base.

View on GitHub →

fabiotaroc's fork has one recorded topic: wiring OpenAI into Mike's LLM abstraction. The change adds backend/src/lib/llm/openai.ts using the Responses API with streaming and tool calling, extends per-user key storage via a database migration, and updates the frontend model picker and account settings to include a third provider group.

The diff is additive - Claude and Gemini paths are unchanged. The registered model IDs (gpt-5.5, gpt-5.4, gpt-5.4-nano) are forward-dated placeholders that need replacing with real ones. A migration hygiene issue (the one-shot baseline 000_*.sql is patched alongside the additive 001_*.sql) is worth fixing before adoption.

What's in it

OpenAI backend Adds OpenAI as a third LLM provider alongside Mike's existing Claude and Gemini support, so GPT-family models can drive the assistant.
Streaming responses OpenAI replies stream in token-by-token, matching the responsiveness users already get from the other two backends.
Tool-calling support The OpenAI integration wires up tool use, so Mike's agentic features keep working when GPT is the model behind them.
Reasoning summaries Optionally surfaces reasoning summaries from OpenAI's newer reasoning models, giving users a peek at the model's thinking.

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.

📝 fabiotaroc puts OpenAI on Mike's menu, next to Claude and Gemini 1 commit 3mo ago integrationinfrastructure draft

A third AI engine option for Mike - bring your own OpenAI key and pick a model per task - though it won't run as-shipped just yet.

Threads of work (detailed view)

2 threads have been distilled into posts.

fabiotaroc puts OpenAI on Mike's menu, next to Claude and Gemini

A third AI engine option for Mike - bring your own OpenAI key and pick a model per task - though it won't run as-shipped just yet.

integrationinfrastructure

1 commit · latest 3mo ago

fabiotaroc adds OpenAI as a third LLM provider

fabiotaroc wired the OpenAI Responses API into Mike's provider abstraction, adding streaming, tool calling, and per-user key storage. The plumbing is additive and isolated - existing Claude and Gemini paths are untouched.

integrationinfrastructure

1 commit · latest 3mo ago