Auto-Heal

Auto-Heal

Auto-Heal is HydraSkill’s automatic failover system. When an IP gets blocked, banned, or rate-limited, HydraSkill switches to a new IP transparently — your agent never handles proxy errors.

How Detection Works

HydraSkill monitors responses for block signals:

Signal	Detection
HTTP 403 Forbidden	Direct block
HTTP 429 Too Many Requests	Rate limit
CAPTCHA pages	Content analysis
Connection reset	Network-level block
Empty response	Stealth block
Redirect to block page	URL pattern matching

Recovery Flow

Agent Request → Proxy → Target Site
                         ↓ (blocked!)
                    HydraSkill detects
                         ↓
                    New IP assigned
                         ↓
                    Request retried
                         ↓
Agent receives ← Success response

Your agent code stays simple:

# No try/except needed for proxy errors
response = requests.get(url, proxies=proxy.to_dict())
# HydraSkill already retried if the first IP was blocked

Configuration

proxy = client.get_proxy(
    target="amazon.com",
    auto_heal=True,          # enabled by default
    max_retries=3,           # retry up to 3 times
    retry_delay_ms=500,      # wait between retries
    rotate_on_block=True,    # get new IP on block
)

Retry Budget

Each request has a retry budget (default: 3 attempts). If all retries fail, HydraSkill raises a ProxyExhaustedError so your agent can handle the edge case:

from hydraskill.exceptions import ProxyExhaustedError

try:
    response = requests.get(url, proxies=proxy.to_dict())
except ProxyExhaustedError:
    # All retry attempts failed — target may be fully blocking
    log.warning(f"Could not reach {url} after 3 proxy rotations")

Metrics

Track Auto-Heal activity in your dashboard:

• Heal Rate — % of requests that needed IP rotation
• Avg Recovery Time — milliseconds to switch IP
• Block Sources — which targets trigger the most blocks

Disabling Auto-Heal

For debugging or when you need raw responses (including blocks):

proxy = client.get_proxy(
    target="example.com",
    auto_heal=False  # get raw responses, including 403s
)