Skip to content

Codebase intelligence for AI and humans: code health scores, auto-generated docs, git analytics, dead code detection, and architectural decisions via MCP.

refactor(answer): grade answers through an ordered table of gate functions (#4059)

Raghav Chamadiya111ad90bcf51
diff --git a/packages/server/src/repowise/server/mcp_server/tool_answer/confidence.py b/packages/server/src/repowise/server/mcp_server/tool_answer/confidence.py
index 8264367ac..dd42327a0 100644
--- a/packages/server/src/repowise/server/mcp_server/tool_answer/confidence.py
+++ b/packages/server/src/repowise/server/mcp_server/tool_answer/confidence.py
@@ -14,6 +14,7 @@ gate reads live beside it.
from __future__ import annotations
import re
+from dataclasses import dataclass, field
from typing import NamedTuple
from repowise.server.mcp_server._answer_context import (
@@ -710,6 +711,198 @@ class _Grade(NamedTuple):
named_body_cut: dict | None
+@dataclass
+class _GradeState:
+ """One :func:`_grade_answer` pass: its inputs, running grade and findings."""
+
+ question: str
+ question_ids: set[str]
+ answer_text: str
+ hits: list[dict]
+ citations: list[str]
+ symbol_bodies: list[dict]
+ served_named_body: bool
+ dominant: bool
+ earn_high: bool
+ uncovered: list[str] | None
+ confidence: str
+ hedged: bool = False
+ ungrounded_values: list[str] = field(default_factory=list)
+ frame_unsupported: list[str] = field(default_factory=list)
+ exclusivity_over_truncated: bool = False
+ withheld_implicated: list[str] = field(default_factory=list)
+ lookup_body_truncated: bool = False
+ named_body_cut: dict | None = None
+ gap: list[str] = field(default_factory=list)
+
+
+def _earned_high(
+ question: str,
+ answer_text: str,
+ hits: list[dict],
+ citations: list[str],
+ served_named_body: bool,
+ top_score: float,
+) -> tuple[bool, bool]:
+ """``(earned_body, earned_grounding)``: the two routes to "high" on a weak retrieval.
+
+ ``earned_body``: the question-named symbol's live body is inlined, resolved
+ by exact name rather than ranking. ``earned_grounding``: every mechanism term
+ the answer names was in the material shown; that shows no fabrication, not
+ that retrieval found the right page, so it is opt-in. Demotion gates still
+ apply to an earned high.
+ """
+ if not (_flag_on(_EARN_HIGH_GROUNDING_ENV) and top_score >= _HIGH_CONFIDENCE_SCORE_FLOOR):
+ return False, False
+ earned_grounding = False
+ if _opt_in(_EARN_HIGH_ON_WEAK_RETRIEVAL_ENV):
+ cited = set(citations)
+ cited_has_body = any(h.get("symbols") for h in hits if h.get("target_path") in cited)
+ unsupported, grounded = _frame_term_grounding(answer_text, question, hits)
+ earned_grounding = cited_has_body and grounded >= 1 and not unsupported
+ return served_named_body, earned_grounding
+
+
+def _hedge_gate(s: _GradeState) -> None:
+ """Dominance says the right file was indexed, not that the prose is usable.
+
+ With the question-named body served the ground truth is in hand, so the
+ demotion stops at "medium" rather than sending the agent to Read.
+ """
+ s.hedged = _answer_is_hedged(s.answer_text)
+ if s.hedged:
+ s.confidence = "medium" if s.served_named_body else "low"
+
+
+def _identifier_gate(s: _GradeState) -> None:
+ """The question names identifiers but no top hit hydrates any of them."""
+ if s.confidence != "high" or not s.question_ids:
+ return
+ top_n = [h for h in s.hits[:_ENRICH_TOP_N_HITS] if h.get("symbols")]
+ if not any(sym.get("_matched") for h in top_n for sym in (h.get("symbols") or [])):
+ s.confidence = "medium"
+
+
+def _value_gate(s: _GradeState) -> None:
+ """On a value question every asserted number must appear in the retrieved material.
+
+ A value derived from grounded operands softens one notch, but only on a
+ dominant high with another asserted number grounded; a lone invented value
+ goes "low".
+ """
+ if s.hedged or not _is_value_question(s.question):
+ return
+ s.ungrounded_values = _ungrounded_numbers(s.answer_text, s.hits)
+ if not s.ungrounded_values:
+ return
+ grounded_sibling = len(_asserted_numbers(s.answer_text)) > len(s.ungrounded_values)
+ if grounded_sibling and s.confidence == "high" and s.dominant:
+ s.confidence = "medium"
+ else:
+ s.confidence = "low"
+
+
+def _citation_gate(s: _GradeState) -> None:
+ """A high answer must cite a page that contributed hydrated symbols."""
+ if s.confidence != "high":
+ return
+ cited = set(s.citations)
+ if not any(h.get("symbols") for h in s.hits if h.get("target_path") in cited):
+ s.confidence = "medium"
+
+
+def _claim_gate(s: _GradeState) -> None:
+ """A why/how answer naming a code-like mechanism term retrieval never showed.
+
+ That is the tell of a conflated mechanism (right file, wrong function).
+ Lookups are out of scope: they legitimately echo the named symbol.
+ """
+ in_scope = _is_why_question(s.question) or (
+ _flag_on(_CLAIM_SUPPORT_GATE_ENV) and _is_mechanism_question(s.question)
+ )
+ if s.confidence != "high" or s.hedged or not in_scope:
+ return
+ unsupported, grounded = _frame_term_grounding(s.answer_text, s.question, s.hits)
+ if unsupported and len(unsupported) >= grounded:
+ s.frame_unsupported = unsupported
+ s.confidence = "medium"
+
+
+def _exclusivity_gate(s: _GradeState) -> None:
+ """"The only" / "entirely" asserted over a truncated body."""
+ if s.confidence != "high" or s.hedged:
+ return
+ s.exclusivity_over_truncated = _has_unqualified_exclusivity_over_truncated(
+ s.answer_text, s.symbol_bodies
+ )
+ if s.exclusivity_over_truncated:
+ s.confidence = "medium"
+
+
+def _withheld_gate(s: _GradeState) -> None:
+ """A withheld symbol the question names or the answer references as code."""
+ s.withheld_implicated = implicated_withheld_symbols(
+ s.question, s.answer_text, s.symbol_bodies
+ )
+ if s.confidence == "high" and s.withheld_implicated:
+ s.confidence = "medium"
+
+
+def _lookup_gate(s: _GradeState) -> None:
+ """A bare-name lookup whose own single-definition body we cut.
+
+ The question names the served symbol, so the withheld gate cannot fire.
+ The entry is kept so the note can quote the real range.
+ """
+ s.named_body_cut = next(
+ (b for b in s.symbol_bodies if _is_question_named_body_cut_by_us(b, s.question_ids)),
+ None,
+ )
+ s.lookup_body_truncated = (
+ s.confidence == "high"
+ and s.named_body_cut is not None
+ and is_symbol_lookup_question(s.question, s.question_ids)
+ )
+ if s.lookup_body_truncated:
+ s.confidence = "medium"
+
+
+def _coverage_gate(s: _GradeState) -> None:
+ """The top hit leaves out question words that outweigh what it covers.
+
+ Dominance says the top hit beat the others, not that it is about the
+ question; a served body does not exempt it.
+ """
+ if s.confidence != "high":
+ return
+ s.gap = uncovered_terms(s.question, s.hits) if s.uncovered is None else s.uncovered
+ if s.gap:
+ s.confidence = "medium"
+
+
+def _non_dominant_ceiling(s: _GradeState) -> None:
+ """With no dominant page the answer stays below "high" unless it earned high."""
+ if not s.dominant and not s.earn_high and s.confidence == "high":
+ s.confidence = "medium"
+
+
+#: The demotion gates, in order. Several fire only while the answer is still
+#: high, so one problem never pushes a response down two levels. Coverage runs
+#: late so a more specific gate writes the note.
+_GRADE_GATES = (
+ _hedge_gate,
+ _identifier_gate,
+ _value_gate,
+ _citation_gate,
+ _claim_gate,
+ _exclusivity_gate,
+ _withheld_gate,
+ _lookup_gate,
+ _coverage_gate,
+ _non_dominant_ceiling,
+)
+
+
def _grade_answer(
*,
question: str,
@@ -733,157 +926,47 @@ def _grade_answer(
``uncovered`` is :func:`uncovered_terms`, when the caller already has it.
"""
dominant = dominance is not None
- _ratio = _top_two_score_ratio(hits)
- _top_score = _confidence_score(hits[0]) if hits else 0.0
- _second_score = _confidence_score(hits[1]) if len(hits) >= 2 else 0.0
-
- # Two routes EARN "high" on a non-dominant (weak) retrieval, tracked apart:
- # * `earned_body`: the question-named symbol's live body is inlined, resolved
- # by exact name rather than ranking, so the source is in the response.
- # * `earned_grounding`: every mechanism term the answer names was in the
- # material shown. Shows no fabrication, not that retrieval found the right
- # page, so it is off by default (see the env flag's comment).
- # The demotion gates below still pull an earned high back down.
- earned_body = False
- earned_grounding = False
- if _flag_on(_EARN_HIGH_GROUNDING_ENV) and _top_score >= _HIGH_CONFIDENCE_SCORE_FLOOR:
- earned_body = served_named_body
- if _opt_in(_EARN_HIGH_ON_WEAK_RETRIEVAL_ENV):
- _cited = set(citations)
- _cited_has_body = any(h.get("symbols") for h in hits if h.get("target_path") in _cited)
- _fu, _fg = _frame_term_grounding(answer_text, question, hits)
- earned_grounding = _cited_has_body and _fg >= 1 and not _fu
- earn_high = earned_body or earned_grounding
-
- if (dominant or earn_high) and _top_score >= _HIGH_CONFIDENCE_SCORE_FLOOR:
- confidence = "high"
- elif dominant:
- # Dominant but weak: right file relative to its siblings, but not
- # strong enough to trust the synthesis without verification.
- confidence = "medium"
- else:
- confidence = "medium"
-
- # Hedge gate: dominance says we indexed the right file, not that the
- # synthesised text is usable.
- hedged = _answer_is_hedged(answer_text)
- if hedged:
- # When the question-named symbol's body is served, the ground truth is
- # in hand; "low" would send the agent to Read for nothing.
- confidence = "medium" if served_named_body else "low"
-
- # Identifier-citation gate: the question names identifiers but no top hit
- # hydrates any of them, so retrieval may be on plausible-but-wrong files.
- if confidence == "high" and question_ids:
- top_n = [h for h in hits[:_ENRICH_TOP_N_HITS] if h.get("symbols")]
- has_match = any(s.get("_matched") for h in top_n for s in (h.get("symbols") or []))
- if not has_match:
- confidence = "medium"
-
- # Value gate: on value-shaped questions every asserted number must appear in
- # the retrieved material. An invented number delivered as "high" is the
- # worst calibration failure, because the consumer was told not to verify.
- ungrounded_values: list[str] = []
- if not hedged and _is_value_question(question):
- ungrounded_values = _ungrounded_numbers(answer_text, hits)
- if ungrounded_values:
- # A value derived from grounded operands softens one notch, but only
- # on a dominant high and only when another asserted number is
- # grounded; a lone invented value is what this gate exists to catch.
- grounded_sibling = len(_asserted_numbers(answer_text)) > len(ungrounded_values)
- if grounded_sibling and confidence == "high" and dominant:
- confidence = "medium"
- else:
- confidence = "low"
-
- # Citation-source gate: a high answer must cite a page that contributed
- # hydrated symbols; summary-only grounding lets plausible-wrong prose through.
- if confidence == "high":
- cited = set(citations)
- if not any(h.get("symbols") for h in hits if h.get("target_path") in cited):
- confidence = "medium"
-
- # Claim-support gate: a synthesis can conflate mechanisms (right file, wrong
- # function) and still clear dominance. The tell is a code-like term the
- # answer names as the mechanism that retrieval never showed. Covers why and
- # how questions; lookups legitimately echo the named symbol.
- frame_unsupported: list[str] = []
- _claim_scope = _is_why_question(question) or (
- _flag_on(_CLAIM_SUPPORT_GATE_ENV) and _is_mechanism_question(question)
- )
- if confidence == "high" and not hedged and _claim_scope:
- frame_unsupported, _grounded_terms = _frame_term_grounding(answer_text, question, hits)
- if frame_unsupported and len(frame_unsupported) >= _grounded_terms:
- confidence = "medium"
- else:
- frame_unsupported = []
-
- # Exclusivity gate: "the only" / "entirely" asserted over a truncated body.
- exclusivity_over_truncated = False
- if confidence == "high" and not hedged:
- exclusivity_over_truncated = _has_unqualified_exclusivity_over_truncated(
- answer_text, symbol_bodies
- )
- if exclusivity_over_truncated:
- confidence = "medium"
-
- # Withheld-dependency gate: a withheld symbol is named by the question or
- # referenced as code by the answer. Unlike the exclusivity gate it needs no
- # prose token; truncation alone is deliberately not enough.
- withheld_implicated = implicated_withheld_symbols(question, answer_text, symbol_bodies)
- if confidence == "high" and withheld_implicated:
- confidence = "medium"
-
- # Lookup gate: on a bare-name lookup the question names the SERVED symbol,
- # so the withheld-dependency gate can never fire. A multi-def name caps on
- # truncation in the union path; a single-def name never reaches it, so it
- # is capped here. Narrow on purpose: a symbol-lookup question whose own
- # symbol was cut. Kept as the entry so the note can quote the real range.
- named_body_cut = next(
- (b for b in symbol_bodies if _is_question_named_body_cut_by_us(b, question_ids)),
- None,
+ top_score = _confidence_score(hits[0]) if hits else 0.0
+ earned_body, earned_grounding = _earned_high(
+ question, answer_text, hits, citations, served_named_body, top_score
)
- lookup_body_truncated = (
- confidence == "high"
- and named_body_cut is not None
- and is_symbol_lookup_question(question, question_ids)
+ earn_high = earned_body or earned_grounding
+ # Dominant but under the score floor is "medium": the right file relative to
+ # its siblings, not strong enough to trust unverified.
+ strong = top_score >= _HIGH_CONFIDENCE_SCORE_FLOOR
+ start = "high" if (dominant or earn_high) and strong else "medium"
+ s = _GradeState(
+ question=question,
+ question_ids=question_ids,
+ answer_text=answer_text,
+ hits=hits,
+ citations=citations,
+ symbol_bodies=symbol_bodies,
+ served_named_body=served_named_body,
+ dominant=dominant,
+ earn_high=earn_high,
+ uncovered=uncovered,
+ confidence=start,
)
- if lookup_body_truncated:
- confidence = "medium"
-
- # Coverage gate, last so a more specific gate writes the note: dominance
- # says the top hit beat the others, not that it is about the question. A
- # served body does not exempt it: a question can name a symbol in passing.
- gap: list[str] = []
- if confidence == "high":
- gap = uncovered_terms(question, hits) if uncovered is None else uncovered
- if gap:
- confidence = "medium"
-
- # Non-dominant ceiling: with no single dominant page the answer must not
- # read "high" (cite without verifying) unless it EARNED high. Keyed on
- # `earn_high`, not `earned_body`, so the grounding opt-in really restores
- # the old behaviour. Usually a no-op; explicit so the contract is visible.
- if not dominant and not earn_high and confidence == "high":
- confidence = "medium"
-
+ for gate in _GRADE_GATES:
+ gate(s)
# The specific dominance tier, not "dominant": gap and agreement fire near a
# 1.0 ratio, so quoting the ratio would print a near tie as justification.
high_reason = None
- if confidence == "high":
+ if s.confidence == "high":
high_reason = dominance or ("symbol_body" if earned_body else "grounding")
return _Grade(
- confidence=confidence,
+ confidence=s.confidence,
high_reason=high_reason,
- second_score=_second_score,
- hedged=hedged,
- ratio=_ratio,
- top_score=_top_score,
- ungrounded_values=ungrounded_values,
- uncovered=gap,
- frame_unsupported=frame_unsupported,
- exclusivity_over_truncated=exclusivity_over_truncated,
- withheld_implicated=withheld_implicated,
- lookup_body_truncated=lookup_body_truncated,
- named_body_cut=named_body_cut,
+ second_score=_confidence_score(hits[1]) if len(hits) >= 2 else 0.0,
+ hedged=s.hedged,
+ ratio=_top_two_score_ratio(hits),
+ top_score=top_score,
+ ungrounded_values=s.ungrounded_values,
+ uncovered=s.gap,
+ frame_unsupported=s.frame_unsupported,
+ exclusivity_over_truncated=s.exclusivity_over_truncated,
+ withheld_implicated=s.withheld_implicated,
+ lookup_body_truncated=s.lookup_body_truncated,
+ named_body_cut=s.named_body_cut,
)
diff --git a/tests/fixtures/mcp/grade_answer_cases.json b/tests/fixtures/mcp/grade_answer_cases.json
new file mode 100644
index 000000000..ead41e2aa
--- /dev/null
+++ b/tests/fixtures/mcp/grade_answer_cases.json
@@ -0,0 +1,38 @@
+[
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 2.5, "second_score": 1.0, "top_score": 2.5, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The default timeout is 30 seconds (pkg/a.py:12).", "citations": ["pkg/c.py"], "dominance": "abs_gap", "hits": [{"score": 2.5, "snippet": "Scheduler", "symbols": [{"_matched": false, "name": "_validate", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"score": 1.0, "snippet": "timeout = 30", "symbols": [{"_matched": false, "name": "retry_loop", "signature": "class Scheduler"}], "target_path": "pkg/b.py"}, {"score": 2.5, "snippet": "", "symbols": [{"_matched": true, "name": "retry_loop", "signature": "class Scheduler"}], "target_path": "pkg/c.py"}], "question": "Why is MAX_RETRIES 5?", "question_ids": ["MAX_RETRIES"], "served_named_body": true, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [1, 120], "name": "Scheduler", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "_validate"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": true, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.0, "second_score": 0.2, "top_score": 0.2, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The context provided does not contain the answer.", "citations": [], "dominance": null, "hits": [{"score": 0.2, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"_confidence_score_factor": -1, "score": 0.2, "snippet": "retry_loop runs", "symbols": [{"_matched": false, "name": "retry_loop", "signature": "def retry_loop(x)"}], "target_path": "pkg/b.py"}, {"_confidence_score_factor": -1, "score": 1.0, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/c.py"}], "question": "How does retry_loop work?", "question_ids": ["retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": null, "lines": [10, 160], "name": "retry_loop", "truncated": false, "withheld_symbols": [{"body_continues": true, "name": "on"}]}, {"continuation": "pkg/a.py:200", "lines": [10, 30], "name": "parse_config", "truncated": false, "withheld_symbols": []}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "1"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [1, 120], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "_validate"}]}, "ratio": 1.25, "second_score": 1.0, "top_score": 1.25, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The default is 30 and the cap is 7.", "citations": ["pkg/d.py"], "dominance": "agreement", "hits": [{"_confidence_score_factor": 0.5, "score": 2.5, "snippet": "Scheduler", "symbols": [{"_matched": false, "name": "retry_loop", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"score": 1.0, "snippet": "", "symbols": [{"_matched": true, "name": "_validate", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/b.py"}, {"_confidence_score_factor": -1, "score": 0.2, "snippet": "retry_loop runs", "symbols": [{"_matched": true, "name": "retry_loop", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/c.py"}], "question": "parse_config", "question_ids": ["parse_config"], "served_named_body": true, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [1, 120], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "_validate"}]}], "uncovered": []}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.0, "second_score": 1.0, "top_score": 1.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "The default timeout is 30 seconds (pkg/a.py:12).", "citations": ["pkg/d.py", "pkg/b.py"], "dominance": null, "hits": [{"_confidence_score_factor": -1, "score": 1.0, "snippet": "timeout = 30", "symbols": [{"_matched": false, "name": "retry_loop", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"score": 1.0, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": false, "name": "retry_loop", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/b.py"}], "question": "Why does FooBar call retry_loop?", "question_ids": ["FooBar", "retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": null, "lines": [1, 151], "name": "Scheduler", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}, {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": false, "withheld_symbols": [{"body_continues": true, "name": "on"}]}], "uncovered": []}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "low", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": true, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": Infinity, "second_score": 0.0, "top_score": 0.2, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The context provided does not contain the answer.", "citations": ["pkg/a.py"], "dominance": "ratio", "hits": [{"score": 0.2, "snippet": "", "target_path": "pkg/a.py"}], "question": "Where is the cache written?", "question_ids": [], "served_named_body": false, "symbol_bodies": [{"continuation": null, "lines": [10, 129], "name": "retry_loop", "truncated": false, "withheld_symbols": [{"body_continues": true, "name": "_validate"}]}], "uncovered": ["cache"]}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "low", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": true, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 5.0, "second_score": 0.8, "top_score": 4.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The context provided does not contain the answer.", "citations": [], "dominance": "ratio", "hits": [{"score": 4.0, "snippet": "Scheduler", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "def retry_loop(x)"}], "target_path": "pkg/a.py"}, {"_confidence_score_factor": 0.5, "score": 1.6, "snippet": "retry_loop runs", "target_path": "pkg/b.py"}, {"_confidence_score_factor": 0.5, "score": 1.0, "snippet": "retry_loop runs", "symbols": [{"_matched": false, "name": "retry_loop", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/c.py"}, {"score": 2.5, "snippet": "Scheduler", "symbols": [{"_matched": false, "name": "_validate", "signature": "class Scheduler"}], "target_path": "pkg/d.py"}, {"_confidence_score_factor": -1, "score": 0.2, "snippet": "timeout = 30", "symbols": [{"_matched": false, "name": "Scheduler", "signature": "def retry_loop(x)"}], "target_path": "pkg/a.py"}], "question": "parse_config", "question_ids": ["parse_config"], "served_named_body": false, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [1, 21], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "_validate"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "1"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.25, "second_score": 1.6, "top_score": 2.0, "uncovered": ["cache"], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The default is 30 and the cap is 7.", "citations": ["pkg/a.py", "pkg/b.py"], "dominance": null, "hits": [{"_confidence_score_factor": 0.5, "score": 4.0, "snippet": "timeout = 30", "target_path": "pkg/a.py"}, {"score": 1.6, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": true, "name": "parse_config", "signature": "class Scheduler"}], "target_path": "pkg/b.py"}, {"_confidence_score_factor": 1.0, "score": 0.2, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": true, "name": "_validate", "signature": "class Scheduler"}], "target_path": "pkg/c.py"}], "question": "parse_config", "question_ids": ["parse_config"], "served_named_body": true, "symbol_bodies": [], "uncovered": ["cache"]}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.5625, "second_score": 1.6, "top_score": 2.5, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "FooBar uses retry_loop because HiddenThing_v2 requires it.", "citations": [], "dominance": "agreement", "hits": [{"score": 2.5, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"score": 1.6, "snippet": "Scheduler", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "def retry_loop(x)"}], "target_path": "pkg/b.py"}, {"_confidence_score_factor": 0.5, "score": 1.0, "snippet": "Scheduler", "symbols": [{"_matched": false, "name": "_validate", "signature": "def retry_loop(x)"}], "target_path": "pkg/c.py"}], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": false, "symbol_bodies": [], "uncovered": []}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": ["RateLimiter"], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 2.5, "second_score": 1.0, "top_score": 2.5, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "MAX_RETRIES is 5 because of RateLimiter.", "citations": ["pkg/d.py", "pkg/b.py"], "dominance": null, "hits": [{"score": 2.5, "snippet": "retry_loop runs", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/a.py"}, {"score": 1.0, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": true, "name": "_validate", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/b.py"}], "question": "Why is MAX_RETRIES 5?", "question_ids": ["MAX_RETRIES"], "served_named_body": true, "symbol_bodies": [], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "0", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": true, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.0, "second_score": 1.6, "top_score": 1.6, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The context provided does not contain the answer.", "citations": ["pkg/b.py"], "dominance": "ratio", "hits": [{"_confidence_score_factor": -1, "score": 1.6, "snippet": "retry_loop runs", "target_path": "pkg/a.py"}, {"score": 1.6, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/b.py"}], "question": "retry_loop _validate", "question_ids": ["_validate", "retry_loop"], "served_named_body": true, "symbol_bodies": [], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "high", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": "ratio", "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.6, "second_score": 2.5, "top_score": 4.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "Scheduler starts 4 workers via spawn_pool in pkg/a.py.", "citations": ["pkg/d.py", "pkg/a.py"], "dominance": "ratio", "hits": [{"_confidence_score_factor": 1.0, "score": 4.0, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": true, "name": "_validate", "signature": "def retry_loop(x)"}], "target_path": "pkg/a.py"}, {"score": 2.5, "snippet": "", "symbols": [{"_matched": false, "name": "parse_config", "signature": "def retry_loop(x)"}], "target_path": "pkg/b.py"}, {"score": 2.5, "snippet": "", "target_path": "pkg/c.py"}], "question": "parse_config", "question_ids": [], "served_named_body": false, "symbol_bodies": [{"continuation": null, "lines": [1, 151], "name": "retry_loop", "truncated": false, "withheld_symbols": [{"body_continues": true, "name": "on"}]}], "uncovered": []}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "0", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "low", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 0.0, "second_score": 0.0, "top_score": 0.0, "uncovered": [], "ungrounded_values": ["4"], "withheld_implicated": []}, "kwargs": {"answer_text": "Scheduler starts 4 workers via spawn_pool in pkg/a.py.", "citations": ["pkg/a.py", "pkg/b.py"], "dominance": null, "hits": [], "question": "How many workers does Scheduler start?", "question_ids": [], "served_named_body": true, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "Scheduler", "truncated": true, "withheld_symbols": []}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "1"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [1, 120], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}, "ratio": 4.0, "second_score": 1.0, "top_score": 4.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "retry_loop calls _validate( before writing; see pkg/a.py.", "citations": ["pkg/d.py"], "dominance": "ratio", "hits": [{"score": 4.0, "snippet": "Scheduler", "target_path": "pkg/a.py"}, {"score": 1.0, "snippet": "retry_loop runs", "symbols": [{"_matched": true, "name": "_validate", "signature": "def retry_loop(x)"}], "target_path": "pkg/b.py"}, {"score": 1.0, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": false, "name": "Scheduler", "signature": "def retry_loop(x)"}], "target_path": "pkg/c.py"}], "question": "How does retry_loop work?", "question_ids": ["retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [1, 120], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}, {"continuation": null, "lines": [10, 160], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "on"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "0", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 20.0, "second_score": 0.2, "top_score": 4.0, "uncovered": [], "ungrounded_values": ["7"], "withheld_implicated": []}, "kwargs": {"answer_text": "The default is 30 and the cap is 7.", "citations": ["pkg/d.py"], "dominance": "agreement", "hits": [{"score": 4.0, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"_confidence_score_factor": -1, "score": 0.2, "snippet": "", "symbols": [{"_matched": true, "name": "retry_loop", "signature": "class Scheduler"}], "target_path": "pkg/b.py"}], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": false, "symbol_bodies": [], "uncovered": []}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": true, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.6, "second_score": 1.0, "top_score": 1.6, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "It is the only place that writes the cache, in pkg/b.py.", "citations": ["pkg/b.py", "pkg/d.py"], "dominance": "ratio", "hits": [{"score": 1.6, "snippet": "timeout = 30", "target_path": "pkg/a.py"}, {"score": 1.0, "snippet": "Scheduler", "symbols": [{"_matched": true, "name": "retry_loop", "signature": "class Scheduler"}], "target_path": "pkg/b.py"}, {"score": 1.6, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "class Scheduler"}], "target_path": "pkg/c.py"}], "question": "Why does FooBar call retry_loop?", "question_ids": ["FooBar", "retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": null, "lines": [10, 160], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "on"}]}, {"continuation": "pkg/a.py:200", "lines": [1, 151], "name": "Scheduler", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": true, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.25, "second_score": 1.6, "top_score": 2.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "It is the only place that writes the cache, in pkg/b.py.", "citations": ["pkg/b.py", "pkg/d.py"], "dominance": "agreement", "hits": [{"_confidence_score_factor": 0.5, "score": 4.0, "snippet": "Scheduler", "symbols": [{"_matched": false, "name": "Scheduler", "signature": "def retry_loop(x)"}], "target_path": "pkg/a.py"}, {"score": 1.6, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "_validate", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/b.py"}, {"score": 4.0, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "parse_config", "signature": "def retry_loop(x)"}], "target_path": "pkg/c.py"}, {"score": 0.2, "snippet": "retry_loop runs", "symbols": [{"_matched": false, "name": "_validate", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/d.py"}, {"score": 1.6, "snippet": "Scheduler", "symbols": [{"_matched": false, "name": "parse_config", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/a.py"}], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": null, "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": []}], "uncovered": []}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": true, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [1, 151], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "on"}]}, "ratio": 0.0, "second_score": 0.0, "top_score": 0.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The context provided does not contain the answer.", "citations": ["pkg/d.py"], "dominance": null, "hits": [], "question": "How does retry_loop work?", "question_ids": ["retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [10, 30], "name": "Scheduler", "truncated": false, "withheld_symbols": []}, {"continuation": "pkg/a.py:200", "lines": [1, 151], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "on"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "low", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": true, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": []}, "ratio": 0.0, "second_score": 0.0, "top_score": 0.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The context provided does not contain the answer.", "citations": [], "dominance": "agreement", "hits": [], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": []}, {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": false, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}], "uncovered": []}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "low", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": true, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.5625, "second_score": 1.6, "top_score": 2.5, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "The context provided does not contain the answer.", "citations": ["pkg/c.py", "pkg/d.py"], "dominance": "ratio", "hits": [{"_confidence_score_factor": -1, "score": 2.5, "snippet": "", "target_path": "pkg/a.py"}, {"score": 1.6, "snippet": "retry_loop runs", "target_path": "pkg/b.py"}, {"_confidence_score_factor": -1, "score": 0.2, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "_validate", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/c.py"}, {"score": 1.6, "snippet": "Scheduler", "target_path": "pkg/d.py"}, {"score": 4.0, "snippet": "", "target_path": "pkg/a.py"}], "question": "How does retry_loop work?", "question_ids": ["retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": null, "lines": [1, 21], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": []}, "ratio": 2.5, "second_score": 1.6, "top_score": 4.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "MAX_RETRIES is 5 because of RateLimiter.", "citations": ["pkg/d.py", "pkg/b.py"], "dominance": "agreement", "hits": [{"_confidence_score_factor": -1, "score": 4.0, "snippet": "", "symbols": [{"_matched": true, "name": "retry_loop", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"score": 1.6, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": false, "name": "parse_config", "signature": "def retry_loop(x)"}], "target_path": "pkg/b.py"}, {"_confidence_score_factor": -1, "score": 0.2, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "parse_config", "signature": "class Scheduler"}], "target_path": "pkg/c.py"}], "question": "retry_loop _validate", "question_ids": ["_validate", "retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": null, "lines": [10, 30], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "retry_loop"}]}, {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": []}], "uncovered": []}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "1"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": true, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [10, 160], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}, "ratio": 1.6, "second_score": 1.0, "top_score": 1.6, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The default is 30 and the cap is 7.", "citations": ["pkg/b.py"], "dominance": null, "hits": [{"score": 1.6, "snippet": "timeout = 30", "target_path": "pkg/a.py"}, {"score": 1.0, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "class Scheduler"}], "target_path": "pkg/b.py"}], "question": "parse_config", "question_ids": ["parse_config"], "served_named_body": true, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [10, 160], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}], "uncovered": []}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "low", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": Infinity, "second_score": 0.0, "top_score": 4.0, "uncovered": [], "ungrounded_values": ["30", "7"], "withheld_implicated": []}, "kwargs": {"answer_text": "The default is 30 and the cap is 7.", "citations": ["pkg/b.py"], "dominance": "abs_gap", "hits": [{"score": 4.0, "snippet": "retry_loop runs", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [1, 21], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "on"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [1, 120], "name": "retry_loop", "truncated": true, "withheld_symbols": []}, "ratio": 2.5, "second_score": 1.0, "top_score": 2.5, "uncovered": ["default", "retry", "loop"], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "retry_loop calls _validate( before writing; see pkg/a.py.", "citations": ["pkg/b.py", "pkg/d.py"], "dominance": "ratio", "hits": [{"score": 2.5, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "parse_config", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"score": 1.0, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/b.py"}, {"score": 2.5, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "retry_loop", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/c.py"}], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [10, 160], "name": "parse_config", "truncated": true, "withheld_symbols": []}, {"continuation": "pkg/a.py:200", "lines": [1, 120], "name": "retry_loop", "truncated": true, "withheld_symbols": []}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": true, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": Infinity, "second_score": 0.0, "top_score": 2.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "The context provided does not contain the answer.", "citations": ["pkg/c.py"], "dominance": "ratio", "hits": [{"_confidence_score_factor": 0.5, "score": 4.0, "snippet": "workers = 4 spawn_pool", "target_path": "pkg/a.py"}], "question": "retry_loop _validate", "question_ids": [], "served_named_body": true, "symbol_bodies": [{"continuation": null, "lines": [10, 160], "name": "retry_loop", "truncated": false, "withheld_symbols": [{"body_continues": false, "name": "on"}]}, {"continuation": null, "lines": [1, 120], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "1"}, "expected": {"confidence": "high", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": "agreement", "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "on"}]}, "ratio": 0.625, "second_score": 4.0, "top_score": 2.5, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "The default is 30 and the cap is 7.", "citations": ["pkg/b.py", "pkg/a.py"], "dominance": "agreement", "hits": [{"score": 2.5, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": true, "name": "_validate", "signature": "def retry_loop(x)"}], "target_path": "pkg/a.py"}, {"score": 4.0, "snippet": "", "target_path": "pkg/b.py"}, {"score": 1.0, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/c.py"}, {"score": 1.0, "snippet": "workers = 4 spawn_pool", "target_path": "pkg/d.py"}, {"score": 4.0, "snippet": "retry_loop runs", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "def retry_loop(x)"}], "target_path": "pkg/a.py"}], "question": "How does retry_loop work?", "question_ids": ["retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "on"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "0", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "low", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [10, 160], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "on"}]}, "ratio": 0.05, "second_score": 4.0, "top_score": 0.2, "uncovered": [], "ungrounded_values": ["4"], "withheld_implicated": []}, "kwargs": {"answer_text": "Scheduler starts 4 workers via spawn_pool in pkg/a.py.", "citations": ["pkg/a.py", "pkg/d.py"], "dominance": null, "hits": [{"score": 0.2, "snippet": "Scheduler", "symbols": [{"_matched": false, "name": "retry_loop", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/a.py"}, {"_confidence_score_factor": -1, "score": 4.0, "snippet": "retry_loop runs", "symbols": [{"_matched": true, "name": "parse_config", "signature": "class Scheduler"}], "target_path": "pkg/b.py"}, {"score": 2.5, "snippet": "MAX_RETRIES = 5", "target_path": "pkg/c.py"}, {"score": 0.2, "snippet": "Scheduler", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "class Scheduler"}], "target_path": "pkg/d.py"}, {"score": 1.0, "snippet": "MAX_RETRIES = 5", "target_path": "pkg/a.py"}], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [10, 160], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "on"}]}], "uncovered": []}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "0", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "low", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [1, 120], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}, "ratio": 0.8, "second_score": 2.0, "top_score": 1.6, "uncovered": [], "ungrounded_values": ["30"], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "The default timeout is 30 seconds (pkg/a.py:12).", "citations": ["pkg/a.py"], "dominance": "ratio", "hits": [{"score": 1.6, "snippet": "retry_loop runs", "target_path": "pkg/a.py"}, {"_confidence_score_factor": 0.5, "score": 4.0, "snippet": "Scheduler", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/b.py"}, {"score": 1.6, "snippet": "Scheduler", "target_path": "pkg/c.py"}], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [1, 120], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "0", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": ["MAX_RETRIES", "RateLimiter"], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "on"}]}, "ratio": 0.8, "second_score": 2.0, "top_score": 1.6, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "MAX_RETRIES is 5 because of RateLimiter.", "citations": ["pkg/d.py", "pkg/b.py"], "dominance": "agreement", "hits": [{"_confidence_score_factor": -1, "score": 1.6, "snippet": "retry_loop runs", "symbols": [{"_matched": true, "name": "parse_config", "signature": "def retry_loop(x)"}], "target_path": "pkg/a.py"}, {"_confidence_score_factor": 0.5, "score": 4.0, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "class Scheduler"}], "target_path": "pkg/b.py"}, {"score": 0.2, "snippet": "timeout = 30", "target_path": "pkg/c.py"}], "question": "Why does FooBar call retry_loop?", "question_ids": ["FooBar", "retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "on"}]}], "uncovered": ["cache"]}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "1"}, "expected": {"confidence": "low", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 0.0, "second_score": 0.0, "top_score": 0.0, "uncovered": [], "ungrounded_values": ["30", "7"], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "The default is 30 and the cap is 7.", "citations": [], "dominance": null, "hits": [], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": null, "lines": [1, 120], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "retry_loop"}]}, {"continuation": "pkg/a.py:200", "lines": [10, 160], "name": "retry_loop", "truncated": false, "withheld_symbols": []}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "low", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": true, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [1, 151], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "_validate"}]}, "ratio": 0.0, "second_score": 0.0, "top_score": 0.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "The context provided does not contain the answer.", "citations": [], "dominance": null, "hits": [], "question": "How does retry_loop work?", "question_ids": ["retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [1, 151], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "_validate"}]}, {"continuation": null, "lines": [1, 151], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "0"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": ["RateLimiter"], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}, "ratio": 1.0, "second_score": 2.5, "top_score": 2.5, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "MAX_RETRIES is 5 because of RateLimiter.", "citations": ["pkg/b.py"], "dominance": "abs_gap", "hits": [{"_confidence_score_factor": -1, "score": 2.5, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": true, "name": "parse_config", "signature": "def retry_loop(x)"}], "target_path": "pkg/a.py"}, {"score": 2.5, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/b.py"}, {"score": 2.5, "snippet": "retry_loop runs", "symbols": [{"_matched": false, "name": "parse_config", "signature": "def retry_loop(x)"}], "target_path": "pkg/c.py"}], "question": "Why does FooBar call retry_loop?", "question_ids": ["FooBar", "retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": null, "lines": [10, 160], "name": "Scheduler", "truncated": false, "withheld_symbols": []}, {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "1"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": true, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "retry_loop"}]}, "ratio": Infinity, "second_score": 0.0, "top_score": 2.5, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "The context provided does not contain the answer.", "citations": ["pkg/a.py"], "dominance": "ratio", "hits": [{"score": 2.5, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": false, "name": "parse_config", "signature": "def retry_loop(x)"}], "target_path": "pkg/a.py"}], "question": "retry_loop _validate", "question_ids": ["_validate", "retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": null, "lines": [1, 120], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "on"}]}, {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "retry_loop"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "1", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": "1"}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": ["_validate"], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.6, "second_score": 2.5, "top_score": 4.0, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "retry_loop calls _validate( before writing; see pkg/a.py.", "citations": ["pkg/a.py", "pkg/d.py"], "dominance": null, "hits": [{"score": 4.0, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "def retry_loop(x)"}], "target_path": "pkg/a.py"}, {"score": 2.5, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": true, "name": "retry_loop", "signature": "def retry_loop(x)"}], "target_path": "pkg/b.py"}, {"_confidence_score_factor": -1, "score": 1.6, "snippet": "", "symbols": [{"_matched": true, "name": "parse_config", "signature": "MAX_RETRIES = 5"}], "target_path": "pkg/c.py"}], "question": "Why does FooBar call retry_loop?", "question_ids": ["FooBar", "retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": null, "lines": [10, 30], "name": "Scheduler", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "retry_loop"}]}, {"continuation": null, "lines": [10, 30], "name": "Scheduler", "truncated": true, "withheld_symbols": []}], "uncovered": ["cache"]}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "1", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "medium", "exclusivity_over_truncated": false, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": null, "ratio": 1.0, "second_score": 1.6, "top_score": 1.6, "uncovered": [], "ungrounded_values": ["7"], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "The default is 30 and the cap is 7.", "citations": ["pkg/d.py"], "dominance": "abs_gap", "hits": [{"score": 1.6, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "parse_config", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"score": 1.6, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": true, "name": "retry_loop", "signature": "def retry_loop(x)"}], "target_path": "pkg/b.py"}], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": true, "symbol_bodies": [{"continuation": null, "lines": [10, 160], "name": "Scheduler", "truncated": true, "withheld_symbols": [{"body_continues": true, "name": "retry_loop"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "0", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "medium", "exclusivity_over_truncated": true, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": []}, "ratio": 1.5625, "second_score": 1.6, "top_score": 2.5, "uncovered": [], "ungrounded_values": [], "withheld_implicated": ["retry_loop"]}, "kwargs": {"answer_text": "It is the only place that writes the cache, in pkg/b.py.", "citations": ["pkg/a.py"], "dominance": "ratio", "hits": [{"_confidence_score_factor": 1.0, "score": 2.5, "snippet": "MAX_RETRIES = 5", "symbols": [{"_matched": false, "name": "Scheduler", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"_confidence_score_factor": -1, "score": 1.6, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "class Scheduler"}], "target_path": "pkg/b.py"}], "question": "retry_loop _validate", "question_ids": ["_validate", "retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": "pkg/a.py:200", "lines": [10, 129], "name": "retry_loop", "truncated": true, "withheld_symbols": []}, {"continuation": null, "lines": [10, 129], "name": "parse_config", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "retry_loop"}]}], "uncovered": null}},
+{"env": {"REPOWISE_ANSWER_CLAIM_SUPPORT_GATE": "0", "REPOWISE_ANSWER_EARN_HIGH_GROUNDING": "0", "REPOWISE_ANSWER_EARN_HIGH_ON_WEAK_RETRIEVAL": ""}, "expected": {"confidence": "medium", "exclusivity_over_truncated": true, "frame_unsupported": [], "hedged": false, "high_reason": null, "lookup_body_truncated": false, "named_body_cut": {"continuation": "pkg/a.py:200", "lines": [10, 160], "name": "retry_loop", "truncated": true, "withheld_symbols": []}, "ratio": 1.28, "second_score": 1.25, "top_score": 1.6, "uncovered": [], "ungrounded_values": [], "withheld_implicated": []}, "kwargs": {"answer_text": "It is the only place that writes the cache, in pkg/b.py.", "citations": ["pkg/b.py"], "dominance": "ratio", "hits": [{"score": 1.6, "snippet": "retry_loop runs", "symbols": [{"_matched": true, "name": "parse_config", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}, {"_confidence_score_factor": 0.5, "score": 2.5, "snippet": "timeout = 30", "symbols": [{"_matched": true, "name": "Scheduler", "signature": "def retry_loop(x)"}], "target_path": "pkg/b.py"}, {"score": 2.5, "snippet": "MAX_RETRIES = 5", "target_path": "pkg/c.py"}, {"score": 0.2, "snippet": "timeout = 30", "target_path": "pkg/d.py"}, {"score": 1.6, "snippet": "workers = 4 spawn_pool", "symbols": [{"_matched": true, "name": "_validate", "signature": "class Scheduler"}], "target_path": "pkg/a.py"}], "question": "What is the default timeout for retry_loop?", "question_ids": ["retry_loop"], "served_named_body": false, "symbol_bodies": [{"continuation": null, "lines": [1, 151], "name": "Scheduler", "truncated": true, "withheld_symbols": [{"body_continues": false, "name": "on"}]}, {"continuation": "pkg/a.py:200", "lines": [10, 160], "name": "retry_loop", "truncated": true, "withheld_symbols": []}], "uncovered": ["cache"]}}
+]
diff --git a/tests/unit/server/mcp/test_grade_answer_table.py b/tests/unit/server/mcp/test_grade_answer_table.py
new file mode 100644
index 000000000..7e3d93c1d
--- /dev/null
+++ b/tests/unit/server/mcp/test_grade_answer_table.py
@@ -0,0 +1,34 @@
+"""``_grade_answer`` pinned over a table of inputs.
+
+Each row was graded once and its full ``_Grade`` recorded. The rows cover every
+outcome shape and every gate that demotes a grade, so a reordered or rewritten
+gate changes a recorded verdict.
+"""
+
+from __future__ import annotations
+
+import json
+from pathlib import Path
+
+import pytest
+
+from repowise.server.mcp_server.tool_answer.confidence import _grade_answer
+
+_CASES = json.loads(
+ (Path(__file__).parents[3] / "fixtures" / "mcp" / "grade_answer_cases.json").read_text(
+ encoding="utf-8"
+ )
+)
+
+
[email protected]("row", _CASES, ids=[f"case{i}" for i in range(len(_CASES))])
+def test_grade_matches_recorded_verdict(row: dict, monkeypatch: pytest.MonkeyPatch) -> None:
+ for name, value in row["env"].items():
+ monkeypatch.setenv(name, value)
+ kwargs = dict(row["kwargs"], question_ids=set(row["kwargs"]["question_ids"]))
+ got = json.loads(json.dumps(_grade_answer(**kwargs)._asdict()))
+ assert got == row["expected"]
+
+
+def test_table_reaches_every_grade() -> None:
+ assert {row["expected"]["confidence"] for row in _CASES} == {"high", "medium", "low"}