verifiers 0.3.2.dev199__py3-none-any.whl → 0.3.2.dev203__py3-none-any.whl

verifiers/v1/interception/tunnel/prime.py

@@ -6,8 +6,9 @@ pool scales with."""
 import asyncio
 import contextlib
 import logging
+import random
 from collections.abc import AsyncIterator
-from typing import Literal
+from typing import TYPE_CHECKING, Literal
 
 from verifiers.v1.interception.tunnel.base import BaseTunnelConfig, Tunnel
 from verifiers.v1.runtimes.limiters import CreationLimiter
@@ -15,6 +16,9 @@ from verifiers.v1.utils.aio import run_shielded
 from verifiers.v1.utils.prime import ensure_prime_auth
 from verifiers.v1.utils.scope import run_scope
 
+if TYPE_CHECKING:
+    from prime_tunnel import Tunnel as TunnelClient
+
 logger = logging.getLogger(__name__)
 
 
@@ -40,6 +44,45 @@ def tunnel_limiter() -> CreationLimiter:
     return CreationLimiter("prime-tunnel", run_scope(), _TUNNELS_PER_MIN / 60)
 
 
+# How often a held tunnel's state is checked with the tunnel service. The service marks a
+# tunnel disconnected about a minute after its frpc stops answering; two such checks in a
+# row restart frpc.
+CHECK_SECONDS = 20
+
+
+async def _watch(client: "TunnelClient", url: str) -> None:
+    """Keep a held tunnel up. A tunnel service hiccup can leave frpc running while the
+    service has it disconnected for good, and every request to the URL then hangs or
+    404s; frpc can also exit. Either way, restart it."""
+    strikes = 0
+    while True:
+        await asyncio.sleep(CHECK_SECONDS * random.uniform(0.75, 1.25))
+        if client.is_running:
+            try:
+                info = await client._client.get_tunnel(client.tunnel_id)
+            except Exception as e:  # noqa: BLE001 - the service is unreachable: no verdict
+                logger.debug("tunnel %s: state check failed: %s", url, e)
+                continue
+            if info is None:
+                logger.error("tunnel %s: registration gone; cannot repair", url)
+                return
+            strikes = strikes + 1 if info.status == "disconnected" else 0
+            if strikes < 2:
+                continue
+        logger.warning(
+            "tunnel %s: %s; restarting frpc",
+            url,
+            "disconnected" if client.is_running else "frpc exited",
+        )
+        try:
+            await client.restart()
+        except Exception as e:  # noqa: BLE001 - tried again on the next check
+            logger.warning("tunnel %s: frpc restart failed: %s", url, e)
+        else:
+            logger.warning("tunnel %s: back up", url)
+            strikes = 0
+
+
 class PrimeTunnelConfig(BaseTunnelConfig):
     """Expose the host interception port via `prime_tunnel` (frpc). No fields — the tunnel
     service mints a fresh public URL per exposed port."""
@@ -57,7 +100,8 @@ class PrimeTunnel(Tunnel[PrimeTunnelConfig]):
         """Bridge the host `port` to a public URL via prime_tunnel (frpc). Tunnel creation
         is network-bound and rate-capped (512/min, run-wide via the shared
         `tunnel_limiter`), so transient failures are retried; a terminal one raises
-        `TunnelError`. The tunnel is torn down on exit."""
+        `TunnelError`. While held, frpc is restarted if the tunnel goes down. The tunnel is
+        torn down on exit."""
         from prime_tunnel import Tunnel as TunnelClient
 
         from verifiers.v1.errors import TunnelError
@@ -72,10 +116,17 @@ class PrimeTunnel(Tunnel[PrimeTunnelConfig]):
                         url = str(await client.start()).rstrip("/")
         except Exception as e:
             raise TunnelError(f"{label} failed: {e}") from e
+        watch = asyncio.create_task(_watch(client, url))
+
+        async def close() -> None:
+            watch.cancel()  # first, so it can't start a new frpc after the stop
+            await asyncio.wait({watch})
+            await asyncio.to_thread(client.sync_stop)
+
         try:
             yield url
         finally:
-            # Run the synchronous stop to completion even under cancellation (`run_shielded`
-            # re-raises the cancellation after); tunnel-stop failures are best-effort.
+            # Run the stop to completion even under cancellation (`run_shielded` re-raises
+            # the cancellation after); tunnel-stop failures are best-effort.
             with contextlib.suppress(Exception):
-                await run_shielded(asyncio.to_thread(client.sync_stop))
+                await run_shielded(close())

{verifiers-0.3.2.dev199.dist-info → verifiers-0.3.2.dev203.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: verifiers
-Version: 0.3.2.dev199
+Version: 0.3.2.dev203
 Summary: Verifiers: Environments for LLM Reinforcement Learning
 Project-URL: Homepage, https://github.com/primeintellect-ai/verifiers
 Project-URL: Documentation, https://github.com/primeintellect-ai/verifiers
@@ -37,7 +37,7 @@ Requires-Dist: openai<3.0.0,>=2.54.0
 Requires-Dist: prime-pydantic-config[toml]>=0.4.3
 Requires-Dist: prime-runs>=0.1.4
 Requires-Dist: prime-sandboxes>=0.4.2
-Requires-Dist: prime-tunnel>=0.1.12
+Requires-Dist: prime-tunnel>=0.1.13
 Requires-Dist: pydantic>=2.12.3
 Requires-Dist: pyzmq>=27.1.0
 Requires-Dist: renderers[multimodal]>=0.1.12.dev2

{verifiers-0.3.2.dev199.dist-info → verifiers-0.3.2.dev203.dist-info}/RECORD

@@ -130,7 +130,7 @@ verifiers/v1/interception/server.py,sha256=yeKMo3gVZl4G_iXpY2PWkwnwGBlfeb3uOhnc7
 verifiers/v1/interception/tunnel/__init__.py,sha256=eVNZJszj6myrKhx9jmJz7VRd9qlfj0AUvUZwI2bXEj8,893
 verifiers/v1/interception/tunnel/base.py,sha256=DZB6uPLwM4Qy7n7m0vKeyg-Px87MhsmpYTpe2vpPEng,1998
 verifiers/v1/interception/tunnel/custom.py,sha256=yL4UbGf4bAuMsxk42mEP-t1wITQCT8BE9FHQIz0qRMA,1708
-verifiers/v1/interception/tunnel/prime.py,sha256=D1p0ecM0h6IVeYPD-1Xp605nx5OYcMwKqeUfqQtfq2s,3390
+verifiers/v1/interception/tunnel/prime.py,sha256=4RAn1eO9cyijfsgaE6cJmn_eFkReDMDiQAXZQ2UZBr8,5426
 verifiers/v1/judges/__init__.py,sha256=z-16awBqEZJrlVYIb9HD-rk12RaABh6jodJgwX4blnA,410
 verifiers/v1/judges/decision.py,sha256=U8BKuzJVMWfX76fa6jaG_rmcOhzMkZoXSnkUuHKpoV4,3991
 verifiers/v1/judges/reference.py,sha256=iEVHw-iJ6dbwLOOqrvWrEMJOcUEYJ0YspT04rZz7Sd4,3868
@@ -196,8 +196,8 @@ verifiers/v1/utils/scope.py,sha256=bzUEiWDOVdbj7dXOwGOlzsQ_-Xu0j9NR5TalQB-8TS8,8
 verifiers/v1/utils/score.py,sha256=493yJVMw8teCu9JxapxMFPFzI0hNUqdo0Y2nGW4kckk,6200
 verifiers/v1/utils/trace_store.py,sha256=RsCDonGR-bs-tvEstcppiJm3jFmAdIiG83fA5KtCbAw,2801
 verifiers/v1/utils/version.py,sha256=-obEo_-l9-D8FLef4hYxncOe-uJpxrM1g2Hig_37Sgs,1607
-verifiers-0.3.2.dev199.dist-info/METADATA,sha256=lgUJ7FxzxG4uAUS0yIAONykYcBi0cvlbZCZFouEgiqA,4228
-verifiers-0.3.2.dev199.dist-info/WHEEL,sha256=W3fkpkm7-wf9vBI5Z-7s0eWkeM-spu78I8Neb98DeEg,87
-verifiers-0.3.2.dev199.dist-info/entry_points.txt,sha256=iugElcdWPKbQM7uFF0lZ8iUpHsNr17-BwEAAjJWxV3U,259
-verifiers-0.3.2.dev199.dist-info/licenses/LICENSE,sha256=v0RrUsdV3IDoZhrRce297IXS3xMHNJ-_LdLpFAUWb9k,1072
-verifiers-0.3.2.dev199.dist-info/RECORD,,
+verifiers-0.3.2.dev203.dist-info/METADATA,sha256=l2QZRoOww5BGE-nOyicc5K8BnYwILZ8OlVcwbMPFsEM,4228
+verifiers-0.3.2.dev203.dist-info/WHEEL,sha256=W3fkpkm7-wf9vBI5Z-7s0eWkeM-spu78I8Neb98DeEg,87
+verifiers-0.3.2.dev203.dist-info/entry_points.txt,sha256=iugElcdWPKbQM7uFF0lZ8iUpHsNr17-BwEAAjJWxV3U,259
+verifiers-0.3.2.dev203.dist-info/licenses/LICENSE,sha256=v0RrUsdV3IDoZhrRce297IXS3xMHNJ-_LdLpFAUWb9k,1072
+verifiers-0.3.2.dev203.dist-info/RECORD,,