Files
hermes-agent/agent/bounded_response.py
Teknium 6a88ec4eb3 refactor(agent/adapters): simplify azure identity, response guards, error surface, retry utils (-603 LOC)
Drop dead read_error_body_or_default / LAYER_RUNTIME; inline single-use
_build_default_credential/_safe_close; compact incident narratives to their
invariants in the guards. Byte-cap and deadline semantics of
read_streaming_error_body verified identical.
2026-09-02 13:29:47 -07:00

95 lines
3.4 KiB
Python

"""Bounded reads of HTTP error response bodies.
On a non-OK *streaming* response Hermes reads the body for a diagnostic. A
bare ``response.read()`` is unbounded two ways: a server can stream an
arbitrarily large body (memory), or open the body and stall forever (hang).
The diagnostic is only ever shown truncated to a few hundred chars, so
``read_streaming_error_body`` caps bytes and enforces a hard wall-clock
deadline; callers feed the returned text to their error builders instead of
touching ``response.text`` (unbounded / raises after a partial stream read).
Subtlety: ``httpx.iter_bytes()`` blocks *inside* the socket read, so a
deadline checked only between chunks can't interrupt a mid-chunk stall until
httpx's own (30s+) read timeout fires. The read therefore runs on a daemon
thread; on timeout we close the response (unblocking the read) and return the
partial bytes collected so far.
Covers the three streaming error-body sites: native Gemini, Gemini Cloud
Code, Antigravity Cloud Code.
"""
from __future__ import annotations
import logging
import threading
from typing import List
import httpx
logger = logging.getLogger(__name__)
# Comfortably holds any real provider error envelope (Google RPC / Anthropic
# error JSON) while rejecting pathological bodies.
DEFAULT_ERROR_BODY_MAX_BYTES = 64 * 1024
# Hard deadline for the whole read; past it the connection is closed and the
# partial bytes are kept.
DEFAULT_ERROR_BODY_TIMEOUT_S = 10.0
def read_streaming_error_body(
response: httpx.Response,
*,
max_bytes: int = DEFAULT_ERROR_BODY_MAX_BYTES,
timeout_s: float = DEFAULT_ERROR_BODY_TIMEOUT_S,
) -> str:
"""Read a non-OK streaming body with a byte cap and a hard deadline.
Returns UTF-8 text (errors replaced) truncated to ``max_bytes``. Never
raises: transport errors, stalls and oversize bodies yield best-effort
partial text (or ""), so a read error can't mask the original failure.
"""
chunks: List[bytes] = []
state = {"truncated": False}
done = threading.Event()
def _drain() -> None:
total = 0
try:
for chunk in response.iter_bytes():
if not chunk:
continue
remaining = max_bytes - total
if len(chunk) > remaining:
if remaining > 0:
chunks.append(chunk[:remaining])
state["truncated"] = True
break
chunks.append(chunk)
total += len(chunk)
except Exception as exc: # noqa: BLE001 - error path must not raise
logger.debug("bounded error-body read failed: %s", exc)
finally:
done.set()
threading.Thread(target=_drain, name="bounded-error-body-read", daemon=True).start()
if not done.wait(timeout=timeout_s):
logger.debug(
"bounded error-body read: hard timeout after %.1fs (%d bytes so far)",
timeout_s,
sum(len(c) for c in chunks),
)
# Closing cancels any in-flight socket read so the worker unwinds. We do
# not join (it is a daemon and may be blocked in C).
try:
response.close()
except Exception: # noqa: BLE001
pass
if state["truncated"]:
logger.debug(
"bounded error-body read: capped at %d bytes (max=%d)",
sum(len(c) for c in chunks),
max_bytes,
)
return b"".join(chunks).decode("utf-8", errors="replace")