I Built a Free 0/1 Knapsack Solver API (with a New Heuristic)

Hi everyone 👋

Over the past few months I’ve been working on a small optimization engine for the classic 0/1 Knapsack Problem. It’s now live and completely free to use.

You can try it here:

https://psi-knapsack-api.onrender.com/solve

What it supports

The API has three modes:

• exact → classic dynamic programming
• fast → a new heuristic I developed
• auto → automatically chooses between them based on estimated runtime and the time budget you provide

Quick Python example

import requests

url = "https://psi-knapsack-api.onrender.com/solve"

payload = {
    "capacity": 50,
    "items": [
        {"id": "laptop", "weight": 10, "value": 60},
        {"id": "book", "weight": 20, "value": 100},
        {"id": "camera", "weight": 30, "value": 120}
    ],
    "mode": "auto",          # "auto", "fast" or "exact"
    "time_budget_sec": 5.0
}

response = requests.post(url, json=payload)
print(response.json())

Example response:

{
  "selected_ids": ["camera", "book"],
  "metrics": {
    "total_value": 220,
    "total_weight": 50,
    "latency_ms": 1.24,
    "fractional_bound_gap": 0.0
  },
  "decision": {
    "candidate_chosen": "dp",
    "mode_requested": "auto",
    "time_budget_sec": 5.0
  },
  "reason": "AUTO selected DP due to low ETA (0.0001s ± 0.0s)."
}

Limits

Max items per request: 20,000
Rate limit: 20 requests per minute per IP
Time budget: 0.1s – 30s per request

Note: It’s running on a free-tier instance on Render, so the first request may take ~30 seconds (cold start). Subsequent requests respond in milliseconds.

I’d love to hear any feedback, especially about:

The quality of the new heuristic
Whether the auto mode makes good decisions
The overall API design

Feel free to test it with your own instances and let me know what you think!
Thanks for reading 🙌