mirror of
https://github.com/razzant/ouroboros.git
synced 2026-10-03 04:07:04 +00:00
ffbd10f09c
NOT_REVIEWED integration checkpoint for READY 02-COGNITION. Contains linked knowledge, book composition, source retention, Nano view primitives and native multiwindow review. Full phase review, shared model/loop wiring, book migration and live acceptance remain pending. No version bump, tag or public delivery.
60 lines
2.6 KiB
Python
60 lines
2.6 KiB
Python
"""One physical request's input, headroom and actual output allowance."""
|
|
|
|
import pytest
|
|
|
|
from ouroboros.context_fit import resolve_call_context_fit
|
|
|
|
|
|
def _fit(**changes):
|
|
values = dict(input_tokens=60_826, input_is_exact=True, caller_max_tokens=65_536,
|
|
total_target_tokens=81_920, route_capacity_tokens=100_000,
|
|
minimum_free_tokens=8192, tokenizer_template_provenance={"source": "exact test tokenizer/template"},
|
|
output_limit_enforced=True, reasoning_included_in_limit=True, route_capacity_confirmed=True)
|
|
return resolve_call_context_fit(**{**values, **changes})
|
|
|
|
|
|
def test_nano_uses_available_output_not_fixed_floor_or_quarter_window():
|
|
fit = _fit()
|
|
assert fit.effective_max_tokens == 21_094
|
|
assert fit.strict_bound_proven and fit.fit_status == "fits"
|
|
assert _fit(caller_max_tokens=4096).effective_max_tokens == 4096
|
|
|
|
|
|
@pytest.mark.parametrize("input_tokens,status,proven", [
|
|
(73728, "fits", True), (73729, "insufficient_headroom", False),
|
|
(81921, "unfit", False),
|
|
])
|
|
def test_exact_headroom_boundary(input_tokens, status, proven):
|
|
fit = _fit(input_tokens=input_tokens)
|
|
assert fit.fit_status == status and fit.strict_bound_proven is proven
|
|
assert fit.effective_max_tokens == max(0, 81920 - input_tokens)
|
|
|
|
|
|
def test_known_smaller_route_wins_and_larger_route_does_not_expand_nano():
|
|
assert _fit(route_capacity_tokens=70000).effective_max_tokens == 9174
|
|
assert _fit(route_capacity_tokens=1_000_000).bound_tokens == 81920
|
|
|
|
|
|
@pytest.mark.parametrize("gap,changes", [
|
|
("exact_input_measurement", {"input_is_exact": False}),
|
|
("tokenizer_template_provenance", {"tokenizer_template_provenance": None}),
|
|
("confirmed_serving_capacity", {"route_capacity_confirmed": False}),
|
|
("enforced_output_limit", {"output_limit_enforced": False}),
|
|
("reasoning_within_measured_window", {"reasoning_included_in_limit": None}),
|
|
])
|
|
def test_unknown_evidence_is_disclosed_without_prohibiting_the_route(gap, changes):
|
|
fit = _fit(**changes)
|
|
assert fit.effective_max_tokens == 21094 and fit.fit_status == "fits"
|
|
assert not fit.strict_bound_proven and gap in fit.missing_evidence
|
|
|
|
|
|
def test_unknown_capacity_keeps_caller_allowance():
|
|
fit = _fit(total_target_tokens=None, route_capacity_tokens=None)
|
|
assert fit.effective_max_tokens == 65536
|
|
assert fit.fit_status == "unknown_capacity" and not fit.strict_bound_proven
|
|
|
|
|
|
def test_unfit_estimate_does_not_manufacture_a_zero_output_request():
|
|
fit = _fit(input_tokens=90000, input_is_exact=False)
|
|
assert fit.fit_status == "unfit" and fit.effective_max_tokens == 65536
|
|
assert not fit.strict_bound_proven
|