File size: 1,627 Bytes
c27193b | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 | """mindXtrain autoresearch — iterated search with verifiable per-attempt provenance.
A thin iteration shell over primitives that already pass their suite (`proof_loop`,
`feedback`, `mei`, `manifest`): the declarative `AttemptContract` decouples the
keep/reset loop from the benchmark, the git guard fences every attempt to one editable
file and dedups by commit, and the durable `Ledger` records the trail. See
`docs/mindXtrain_autoresearch_strategy.md`.
Import-light: stdlib + pydantic only, so `import mindxtrain.research` works on the base
install. Heavier eval metric sources are lazy-imported by the attempt runner.
"""
from __future__ import annotations
from mindxtrain.research.anchor import (
AutoresearchReceipt,
anchor_on_chain,
build_research_receipt,
)
from mindxtrain.research.attempt import AttemptResult, read_metric, run_attempt, run_logged
from mindxtrain.research.contract import AttemptContract, MetricSource
from mindxtrain.research.guard import (
ResearchAbort,
assert_clean_and_committed,
assert_new_commit,
repo_root,
)
from mindxtrain.research.ledger import AttemptMetric, Ledger, LedgerEntry
from mindxtrain.research.search import search, search_from_contract
__all__ = [
"AttemptContract",
"AttemptMetric",
"AttemptResult",
"AutoresearchReceipt",
"Ledger",
"LedgerEntry",
"MetricSource",
"ResearchAbort",
"anchor_on_chain",
"assert_clean_and_committed",
"assert_new_commit",
"build_research_receipt",
"read_metric",
"repo_root",
"run_attempt",
"run_logged",
"search",
"search_from_contract",
]
|