Skip to main content
Review a proposed ML plan against knowledge base best practices. Returns approvals, risks, and improvement suggestions.

​
When to Use

  • Before executing a plan: Validate your approach before writing code
  • Catching assumptions: Find incorrect assumptions about how a framework works
  • Risk assessment: Know what pitfalls or edge cases to watch out for

​
Parameters

ParameterRequiredTypeDescription
proposalYesstringThe plan or proposal to review
goalYesstringThe intended goal of the plan

​
Example

Tool: review_plan
Proposal: "1. Load model in 4-bit NF4. 2. Apply LoRA rank 64 to all linear layers. 
           3. Train with lr=2e-5 for 3 epochs. 4. Merge and save."
Goal: "QLoRA fine-tuning of Llama-3.1-8B for instruction following"
Returns: Approvals (what looks good), risks (e.g., learning rate too low for QLoRA), and improvement suggestions.