Teknium
13c3958df5
Second leg of the masoria 20-minute 'Summarizing' stall (Aug 31 2026 bundle): after the Codex timeout, compression fell back to OpenRouter, which defaulted the omitted output cap to the model's full 65,536-token window and rejected with '402 ... can only afford 7117' — on an account whose balance easily covered a summary. Three fallbacks, three 402s, zero summaries. - _create_with_progress: when a 402 names an affordable budget, retry ONCE with that cap (minus 64-token margin, 512-token floor); plain exhaustion 402s and within-budget 402s re-raise unchanged. Single funnel covers primary, fallback, and retry call sites. - extends the #41055 OpenRouter max_tokens preservation onto the current _build_call_kwargs gate (explicit caps survive; None still omitted). Live A/B against a mock OpenRouter enforcing a 7117-token budget with the real SDK + real adapter: main fails with the exact bundle 402; fixed branch retries clamped and returns the summary.
73 lines
2.7 KiB
Python
73 lines
2.7 KiB
Python
"""Test that _build_call_kwargs preserves max_tokens for OpenRouter endpoints.
|
|
|
|
Regression test for #41035: OpenRouter free-tier credits exhausted when
|
|
max_tokens was stripped, causing HTTP 402 and fallback to text-only model.
|
|
"""
|
|
|
|
import pytest
|
|
|
|
from agent.auxiliary_client import _build_call_kwargs
|
|
|
|
|
|
class TestOpenRouterMaxTokens:
|
|
"""max_tokens must be included for OpenRouter to prevent free-tier 402."""
|
|
|
|
def test_openrouter_provider_includes_max_tokens(self):
|
|
"""Direct openrouter provider keeps max_tokens."""
|
|
kwargs = _build_call_kwargs(
|
|
provider="openrouter",
|
|
model="openai/gpt-4o-mini",
|
|
messages=[{"role": "user", "content": "test"}],
|
|
max_tokens=2000,
|
|
)
|
|
assert kwargs.get("max_tokens") == 2000 or kwargs.get("max_completion_tokens") == 2000
|
|
|
|
def test_openrouter_base_url_includes_max_tokens(self):
|
|
"""Custom endpoint with openrouter.ai base_url keeps max_tokens."""
|
|
kwargs = _build_call_kwargs(
|
|
provider="openai",
|
|
model="gpt-4o-mini",
|
|
messages=[{"role": "user", "content": "test"}],
|
|
max_tokens=2000,
|
|
base_url="https://openrouter.ai/api/v1",
|
|
)
|
|
assert kwargs.get("max_tokens") == 2000 or kwargs.get("max_completion_tokens") == 2000
|
|
|
|
def test_generic_provider_omits_max_tokens(self):
|
|
"""Generic OpenAI-compatible provider still omits max_tokens."""
|
|
kwargs = _build_call_kwargs(
|
|
provider="openai",
|
|
model="gpt-4o-mini",
|
|
messages=[{"role": "user", "content": "test"}],
|
|
max_tokens=2000,
|
|
)
|
|
assert "max_tokens" not in kwargs
|
|
|
|
def test_anthropic_compat_still_includes_max_tokens(self):
|
|
"""Anthropic-compatible endpoints still include max_tokens."""
|
|
kwargs = _build_call_kwargs(
|
|
provider="minimax",
|
|
model="MiniMax-Text-01",
|
|
messages=[{"role": "user", "content": "test"}],
|
|
max_tokens=4000,
|
|
)
|
|
assert kwargs["max_tokens"] == 4000
|
|
|
|
def test_none_max_tokens_never_included(self):
|
|
"""max_tokens=None is never added regardless of provider."""
|
|
for provider, base_url in [
|
|
("openrouter", None),
|
|
("openai", "https://openrouter.ai/api/v1"),
|
|
("minimax", None),
|
|
]:
|
|
kwargs = _build_call_kwargs(
|
|
provider=provider,
|
|
model="test-model",
|
|
messages=[{"role": "user", "content": "test"}],
|
|
max_tokens=None,
|
|
base_url=base_url,
|
|
)
|
|
assert "max_tokens" not in kwargs, (
|
|
f"max_tokens should not be set when None for {provider}"
|
|
)
|