Fast local LLMs on the AMD GPU you already own.

Qwen3.8-Flash on Strix Halo, native MTP, greedy, thinking off. This run: 482 output tokens at 67.6 tok/s server-measured decode; 183 ms client time to first token. The displayed command is the native CLI equivalent of the recorded streaming request.

~ $ hipfire run qwen3.8:flash-next "Write a minimal Python LRU cache with per-key TTL and two pytest tests"

```python
import time
from collections import OrderedDict
from typing import Any, Optional


class LRUCache:
    """A minimal LRU cache with per-key TTL support."""

    def __init__(self, capacity: int):
        self.capacity = capacity
        self.cache: OrderedDict[str, tuple[Any, Optional[float]]] = OrderedDict()

    def get(self, key: str) -> Optional[Any]:
        """Get value by key. Returns None if not found or expired."""
        if key not in self.cache:
            return None
        value, expires_at = self.cache[key]
        if expires_at is not None and time.time() > expires_at:
            del self.cache[key]
            return None
        # Move to end (most recently used)
        self.cache.move_to_end(key)
        return value

    def put(self, key: str, value: Any, ttl: Optional[float] = None) -> None:
        """Put value with optional TTL in seconds."""
        expires_at = time.time() + ttl if ttl is not None else None
        if key in self.cache:
            self.cache.move_to_end(key)
        self.cache[key] = (value, expires_at)
        if len(self.cache) > self.capacity:
            self.cache.popitem(last=False)


# Tests
import pytest


def test_lru_eviction():
    cache = LRUCache(capacity=2)
    cache.put("a", 1)
    cache.put("b", 2)
    # Access "a" to make it most recently used
    assert cache.get("a") == 1
    # Add "c", should evict "b" (least recently used)
    cache.put("c", 3)
    assert cache.get("b") is None
    assert cache.get("a") == 1
    assert cache.get("c") == 3


def test_ttl_expiry():
    cache = LRUCache(capacity=10)
    cache.put("x", 42, ttl=0.1)
    assert cache.get("x") == 42
    time.sleep(0.15)
    assert cache.get("x") is None
```