Bug: reshapeArray silently corrupts data for 3+ dimensional tensors

[golldyck]

Problem

The reshapeArray function in src/utils.ts (line ~208) uses shape.slice(0, -1) when recursing into sub-dimensions. This removes the last dimension instead of the first (already consumed) dimension, causing silent data corruption for any tensor with 3 or more dimensions.

Steps to Reproduce

import { reshapeArray } from './utils';

const flat = [1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24];
const result = reshapeArray(flat, [2, 3, 4]);
// Expected: [[[1,2,3,4],[5,6,7,8],[9,10,11,12]],[[13,14,15,16],[17,18,19,20],[21,22,23,24]]]
// Actual: incorrectly shaped array due to wrong dimension slicing

Root Cause

Line 208: const subShape = shape.slice(0, -1);

For shape [2, 3, 4]:

• Current (broken): consumes dim 0 (size=2), recurses with [2, 3] — reuses the already-consumed dimension
• Correct: consumes dim 0 (size=2), recurses with [3, 4] — passes remaining dimensions

Fix: change to shape.slice(1)

Additional Issues

• src/client.ts line ~302: llmChat has gas estimation commented out, replaced with hardcoded 100000. The infer and llmCompletion methods properly use estimateGas()
• src/client.ts line ~142: uses var instead of const

Impact

Critical — any user working with 3D+ tensor model outputs through the TypeScript SDK gets silently corrupted data with no error or warning. This affects ML model inference results.