This repository acts as the central brain and configuration storage for your local developer environment, defining your PowerShell profile rules, LiteLLM API proxies, and the personalized OpenCode multi-agent framework.

You have a highly optimized, multi-agent AI system deployed in the ./opencode directory which operates to save API tokens and avoid re-reading context.

Instead of using generic models, you summon these specific personas directly in OpenCode (e.g., @orchestrator):

• @orchestrator: The project manager. Start complex tasks here. Conducts an interview, plans the architecture, and delegates all work down the chain. Configured to explicitly use Gemini logic.
• @coder: The builder. Listens to the orchestrator and exclusively executes logic and implementation using your local Ollama model (qwen3-coder:30b).
• @qa: The tester. Fuses code-review and test-engineering into one agent. Validates code without duplicating token reads.
• @docs: The memory bank. Constantly updates CONTEXT.md in your project to maintain active awareness of decisions, stack, and rules.

The global engine limits tokens by actively routing tasks:

• Easy Tasking: Natively routes to ollama/qwen3.5:0.8b for basic OS filesystem commands.
• Default Plan: Uses Gemini.
• Default Build: Uses local 30b model.

Because top-tier reasoning engines (like Gemini 3.1 Flash Lite constraint) hit API Quota bottlenecks due to OpenCode's fast sequential execution, we use a LiteLLM Proxy.

The litellm_config.yaml explicitly enforces an rpm: 14 (Requests Per Minute) rule. Instead of the Google Cloud API outright rejecting your requests with "Resource Exhausted", the LiteLLM proxy intercepts the requests and smoothly trickles them into the API at exactly 14 requests per minute, preventing hard crashes.

• To connect OpenCode to LiteLLM: Configure a custom OpenAI provider format in opencode.json pointing to http://127.0.0.1:4000/v1.

Your $PROFILE automatically loads the directories for Functions and Aliases ensuring you have Unix-like commands seamlessly mapped in Windows PowerShell.

You have natively mapped classic Unix operations directly to your PowerShell session so muscle memory works smoothly:

• touch (Creates empty files)
• grep (Regex search matching)
• ls (Directory listing aliases)
• kill-port (Wraps Stop-Process-On-Port)
• which (Locates executables on PATH)

Maintained actively to ensure efficient API routing and streamlined CLI flows.