teknium1
d10bb2ab6f
`scripts/run_tests.sh tests/<dir>/` is how a change gets its regression
coverage run, so a test filed under the wrong directory is a test nobody
runs when that code changes. Two kinds of drift had accumulated.
Parallel directories for one source package, folded into the mirror:
tests/acp -> tests/acp_adapter (its __init__/conftest move with it)
tests/cli -> tests/hermes_cli (prompt_toolkit fixture merged into
hermes_cli/conftest.py)
tests/run_agent -> tests/agent (backoff fixture becomes
agent/conftest.py)
tests/relay -> tests/gateway/relay
tests/state -> tests/hermes_state
246 loose files at tests/ root, routed by the package they import/patch:
hermes_cli, hermes_state, agent, gateway, tools, plugins, tui_gateway, cron.
Installer and desktop-update script tests go to tests/scripts/{install,
desktop_update}/. 43 tests of root-level modules (batch_runner, utils,
hermes_constants, packaging) stay at the root.
Filenames drop their issue numbers (95 files: test_89315_x.py -> test_x.py);
the number stays in the module docstring where it has context.
Collisions: test_cli_skin_integration.py existed in both tests/ and tests/cli
with different subsets — merged into one (10 tests, all kept);
run_agent/test_pre_compress_memory_context.py -> agent/..._handoff.py;
tests/test_account_usage.py -> agent/test_account_usage_fetch.py;
tests/test_web_server.py -> hermes_cli/test_web_server_ws_ping.py.
Deleted: test_minisweagent_path.py (empty since PR #2804),
test_model_picker_scroll.py (tested a private copy of the logic, imported
nothing), test_process_loop_event_loop_warning.py (asserted asyncio behaviour,
imported nothing from Hermes).
Repo-root path arithmetic (Path(__file__).parents[N], dirname chains) is
bumped for the 202 files that changed depth and verified by evaluating every
such expression against the new location. classify_changes' desktop-updater
lane prefix, tests-os.yml's ignore glob and every in-tree path comment follow
the moves. tests/test_tests_tree_layout.py keeps the tree from drifting back.
97 lines
3.3 KiB
Python
97 lines
3.3 KiB
Python
"""Native Gemini output-token reservation at agent init (#57275 claim 4).
|
||
|
||
When ``model.max_tokens`` is unset, the native generateContent adapter still
|
||
sends ``maxOutputTokens=65,535`` (GEMINI_DEFAULT_MAX_OUTPUT_TOKENS) — Gemini
|
||
treats an omitted cap as a low internal default, not "unlimited", so the
|
||
adapter always sends an explicit cap. The compressor's trigger is
|
||
``pct × (window − max_tokens)``; constructing it with ``max_tokens=None``
|
||
reserved 0 while the wire reserved 65,535: on a 128K window the trigger
|
||
landed at ~96K against a real safe input budget of ~65K, and the provider
|
||
400'd before compaction ever fired.
|
||
|
||
These tests assert the compressor's reservation mirrors the adapter default
|
||
on the native Gemini path, and ONLY there.
|
||
"""
|
||
|
||
|
||
|
||
|
||
from unittest.mock import patch
|
||
|
||
|
||
|
||
import agent.context_compressor as cc_mod
|
||
from agent.gemini_native_adapter import GEMINI_DEFAULT_MAX_OUTPUT_TOKENS
|
||
|
||
|
||
CFG = {"agent": {}}
|
||
|
||
|
||
def _build_agent(model, base_url, provider="", max_tokens=None, window=131072):
|
||
with (
|
||
patch("model_tools.get_tool_definitions", return_value=[]),
|
||
patch("model_tools.check_toolset_requirements", return_value={}),
|
||
patch("agent.process_bootstrap.OpenAI"),
|
||
patch("hermes_cli.config.load_config", return_value=CFG),
|
||
patch("hermes_cli.config.load_config_readonly", return_value=CFG),
|
||
patch(
|
||
"agent.model_metadata.get_model_context_length", return_value=window,
|
||
),
|
||
patch.object(cc_mod, "get_model_context_length", return_value=window),
|
||
):
|
||
from run_agent import AIAgent
|
||
|
||
return AIAgent(
|
||
model=model,
|
||
api_key="test-key-1234567890",
|
||
base_url=base_url,
|
||
provider=provider,
|
||
max_tokens=max_tokens,
|
||
quiet_mode=True,
|
||
skip_context_files=True,
|
||
skip_memory=True,
|
||
)
|
||
|
||
|
||
def test_native_gemini_unset_max_tokens_reserves_adapter_default():
|
||
agent = _build_agent(
|
||
"gemma-3-27b-it",
|
||
"https://generativelanguage.googleapis.com/v1beta",
|
||
)
|
||
cc = agent.context_compressor
|
||
assert cc.max_tokens == GEMINI_DEFAULT_MAX_OUTPUT_TOKENS
|
||
# Trigger must sit at/below the real safe input budget the wire leaves.
|
||
assert cc.threshold_tokens <= cc.context_length - GEMINI_DEFAULT_MAX_OUTPUT_TOKENS
|
||
|
||
|
||
def test_gemini_provider_name_also_reserves_default():
|
||
agent = _build_agent(
|
||
"gemini-3.7-flash", "https://example-proxy.invalid/v1", provider="google",
|
||
)
|
||
assert agent.context_compressor.max_tokens == GEMINI_DEFAULT_MAX_OUTPUT_TOKENS
|
||
|
||
|
||
def test_explicit_max_tokens_wins_over_adapter_default():
|
||
agent = _build_agent(
|
||
"gemma-3-27b-it",
|
||
"https://generativelanguage.googleapis.com/v1beta",
|
||
max_tokens=8192,
|
||
)
|
||
assert agent.context_compressor.max_tokens == 8192
|
||
|
||
|
||
def test_non_gemini_paths_keep_no_reservation():
|
||
agent = _build_agent(
|
||
"openai/gpt-4.1", "https://openrouter.ai/api/v1",
|
||
)
|
||
assert agent.context_compressor.max_tokens is None
|
||
|
||
|
||
def test_gemini_openai_compat_endpoint_not_treated_as_native():
|
||
# The /openai compatibility endpoint does not use the native adapter.
|
||
agent = _build_agent(
|
||
"gemma-3-27b-it",
|
||
"https://generativelanguage.googleapis.com/v1beta/openai",
|
||
)
|
||
assert agent.context_compressor.max_tokens is None
|