mirror of
https://github.com/morluto/rea.git
synced 2026-11-16 19:53:50 +00:00
* feat(evidence)!: name the Evidence subject by its target identity (format revision 2) The Evidence subject carries subject.identity (kind, sha256, format, architecture, selection) and the Evidence ID commits to it, so raw/DOL entry points and ESP flash offsets of the same bytes are different subjects. Every record declares evidence_format: 2; revision 1 bundles, standalone records and snapshots are refused with the recreate guidance. binaryTargetIdentity is the one projection of an open target, used by Evidence and snapshot comparison. BREAKING CHANGE: Evidence format revision 2. Evidence IDs commit to the subject's target identity (subject.identity, including the raw/DOL entry point or ESP flash offsets), and every record declares evidence_format: 2. Evidence bundles, records and analysis snapshots from earlier releases are refused; recreate the observations with this release. * refactor(analysis-view): read parent subjects through the Evidence identity accessors (#11423) Analysis views compare their parent's file identity (evidenceFileTarget) and take path and digest from evidenceSubjectTarget instead of reading the subject's digest, format and architecture members, which duplicate subject.identity until they are removed. A counted guard (evidence-subject-declaration-read) keeps new direct reads out and records the 20 remaining files. * fix(evidence): carry format refusals through Evidence readers and bind target selections - safeParseEvidence/safeParseImmutableEvidence return the format refusal as an expected failure; every consumer maps it to the one integrity refusal with the recreate guidance. An unmarked object is an earlier record only when it has every member records have always carried. - ArtifactProvider inventory roots and provider subjects that name the target's own bytes carry the target's selection (ESP flash offsets). - Snapshot rows bind only Evidence whose identity, selection included, matches the snapshot target. - Entry-point selections use the raw-binary entry point form. * fix(evidence): refuse unauthenticated or earlier-format records instead of reading them as absent - recognizeEvidence (evidenceValidation) replaces tryParseEvidence: input without an evidence_id is not Evidence (ok(undefined)); a record that names one but fails authentication or is from another format revision is an EvidenceIntegrityError carrying the recreate guidance, never undefined. - Direct analysis retains that refusal as a secondary failure instead of silently skipping source capture; cleanup projection, source-closure annotation and the flutter, debugger and release-exposure recoveries use the same owner (the last three previously read the unauthenticated schema). - A desktop capture.json from an earlier format is refused with the recreate guidance instead of "not desktop application capture Evidence". * fix(evidence): advertise revision 2 ledger examples and rank delivery omissions by serialized size - The obligation ledger and reconstruction coverage examples named evidence_format 1, which revision 2 refuses; they now use EVIDENCE_FORMAT_REVISION. - MCP delivery chose which unretained recovery to omit by a parsed-storage lower bound clipped at the response budget. Once revision 2 subjects made the source record's estimate cross the budget, both candidates tied at the clip and the smaller source was omitted ahead of the larger completed observation. Candidates are now ranked by serialized size, which is what the budget counts. * refactor(native): compare function snapshots by their Evidence target identity (#11450) Function snapshots carry the subject's file target identity instead of the duplicated subject declaration. Call paths and function comparisons compare targets through sameTargetIdentity, so the same bytes opened at a different raw entry point or ESP flash offset no longer count as one coordinate space: build_call_path refuses the mix, and compare_functions no longer matches entries by address across selections. Structural candidates and native function matching read the architecture and digest through evidenceFileTarget. Refs #11417 * refactor(artifacts): compare artifact and WASM summaries by their Evidence target identity (#11461) Artifact comparison authenticates a summary against its retained inventory through sameTargetIdentity and rebinds the alias with the whole target, selection included, so a flash-selected inventory summary taken under an alias path is no longer refused. WASM comparison, the analysis view artifact projection and the inventory root check read the digest, format and display path through the Evidence identity accessors. Refs #11417 * refactor(evidence): read record subjects through the identity accessors in JS, managed, ledger and bundle readers (#11470) JavaScript application graphs, runtime reconciliation, managed inspection and native verification sources, the record summary filter, the ledger's residual-unknown target and bundle scope selection read the subject digest, format and display path through evidenceFileTarget and evidenceSubjectDisplay. No comparison changes: each reader matches one record's facts against its own payload or a caller filter. Refs #11417 * refactor(evidence)!: name the Evidence subject only by its target identity (#11497) * refactor(evidence)!: name the Evidence subject only by its target identity The subject keeps `identity` and the display facts `name` and `local_path`. The duplicated `digest`, `format` and `architecture` members are removed; the format-owned declaration now validates the identity itself. Readers the type checker found move to the owner accessors; untyped reads in verify scripts, the efficiency harness and test matchers move to `subject.identity`. The `evidence-subject-declaration-read` text guard and its baseline are deleted. Fixes #11473 Refs #11417 BREAKING CHANGE: the Evidence subject no longer has `digest`, `format` or `architecture`; read `subject.identity.sha256`, `subject.identity.format` and `subject.identity.architecture`. A record that still carries them is refused. * test(evidence): give the Flutter verifier fixture the identity subject and check the DOL declaration without the ID * test(evidence): drop the removed subject members from the declaration matrix and tamper fixtures * feat(evidence)!: give every Evidence record an explicit subject identity (#11536) * feat(evidence)!: give every Evidence record an explicit subject identity A producer that cannot name a file passes an `unknown` target identity with a closed reason (derived-from-linked-evidence, not-observed, unsupported-subject) instead of no subject. The subject is non-null; unknown subjects have no display facts. The generic missing-subject limitation and `subjectUnavailableReason` are removed; producer-specific explanations are ordinary limitations. Comparison identity requires the derived unknown, and null-subject consumers ask the owner accessors. Refs #11417 BREAKING CHANGE: every Evidence record has a subject. A record without a file target names `subject.identity` `{ "kind": "unknown", "reason": … }` with `name` and `local_path` null, instead of `subject: null` and the limitation "Artifact identity is unavailable for this observation.". Their Evidence IDs change. * fix(evidence): name explicit unknown subjects in JavaScript callers and keep a managed graph's file subject The verification scripts are not type-checked, so five createEvidence callers still passed undefined (or a missing first target) after the subject became required. They now pass unknownTarget("not-observed"). A managed application graph chose the first present input subject. Every subject is now present, so an unknown artifact subject hid a file subject on the member or boundary Evidence. The graph now takes the first input subject that names a file. * test(evidence): assert the subject kind instead of a presence the type now guarantees Every Evidence record has a subject after this change, so test guards and assertions that checked it for null always passed. Each now checks that the fixture subject names a file, or is removed where a later fileIdentity, fileSubjectTarget or fileSubjectDisplay call already requires that. Non-null assertions on Evidence subjects are removed; analysis-view inputs, whose subject is optional, keep an undefined check. * test(evidence): expect the Evidence subject format under its identity Two boundary tests still matched subject.format, which slice 2e moved into subject.identity. toMatchObject is not type-checked against the Evidence shape, so they failed only when run (the signature test on macOS).
560 lines
18 KiB
TypeScript
560 lines
18 KiB
TypeScript
import { visit } from "jsonc-parser";
|
|
import { z } from "zod";
|
|
|
|
import {
|
|
parseEvidence,
|
|
evidenceAuthoritySchema,
|
|
evidenceSchema,
|
|
type Evidence,
|
|
} from "../../../src/domain/evidence.js";
|
|
import {
|
|
jsonValueSchema,
|
|
type JsonValue,
|
|
} from "../../../src/domain/jsonValue.js";
|
|
import { canonicalJson } from "../../../src/domain/comparisonSemantics.js";
|
|
|
|
/** Only authenticated, successful, non-conflicting transcript deliveries. */
|
|
export interface FixtureClaimContext {
|
|
readonly evidence: Evidence;
|
|
readonly deliveredEvidence: readonly Evidence[];
|
|
}
|
|
|
|
/** One closed fixture assertion and the producer facts that must support it. */
|
|
export interface FixtureClaimExpectation {
|
|
readonly id: string;
|
|
readonly expectedValue: JsonValue;
|
|
readonly source: {
|
|
readonly operation: string;
|
|
readonly authority: Evidence["authority"];
|
|
readonly confidence: Evidence["confidence"];
|
|
readonly subject?: { readonly path: string; readonly sha256?: string };
|
|
readonly parameters?: Readonly<Record<string, JsonValue>>;
|
|
readonly linkedSubjects?: readonly {
|
|
readonly operation: string;
|
|
readonly path: string;
|
|
readonly sha256?: string;
|
|
}[];
|
|
readonly select: (
|
|
normalizedResult: JsonValue,
|
|
context: FixtureClaimContext,
|
|
) => JsonValue | undefined;
|
|
};
|
|
}
|
|
|
|
/** Failed fixture constraints, distinct from unrestricted natural-language correctness. */
|
|
export type FixtureClaimFailure =
|
|
| "claim_missing"
|
|
| "value_mismatch"
|
|
| "claim_confidence_mismatch"
|
|
| "claim_authority_mismatch"
|
|
| "unknown_evidence"
|
|
| "conflicting_evidence"
|
|
| "source_operation_mismatch"
|
|
| "source_confidence_mismatch"
|
|
| "source_authority_mismatch"
|
|
| "source_subject_mismatch"
|
|
| "source_parameters_mismatch"
|
|
| "source_links_mismatch"
|
|
| "source_assertion_mismatch"
|
|
| "selector_failed";
|
|
|
|
/** Diagnostics for a single configured fixture assertion. */
|
|
export interface FixtureClaimAssessment {
|
|
readonly id: string;
|
|
readonly status: "passed" | "failed";
|
|
readonly failures: readonly FixtureClaimFailure[];
|
|
}
|
|
|
|
/** Only the configured closed fixture claims are assessed. */
|
|
export interface KnownAnswerAssessment {
|
|
readonly scope: "configured_fixture_claims";
|
|
readonly status: "passed" | "failed" | "not_assessed";
|
|
readonly manifestFailures: readonly string[];
|
|
readonly claims: readonly FixtureClaimAssessment[];
|
|
}
|
|
|
|
const subjectSchema = z.strictObject({
|
|
path: z.string().min(1),
|
|
sha256: z
|
|
.string()
|
|
.regex(/^[a-f0-9]{64}$/u)
|
|
.optional(),
|
|
});
|
|
const expectationSchema = z.strictObject({
|
|
id: z.string().min(1),
|
|
expectedValue: jsonValueSchema,
|
|
source: z.strictObject({
|
|
operation: z.string().min(1),
|
|
authority: evidenceAuthoritySchema,
|
|
confidence: evidenceSchema.shape.confidence,
|
|
subject: subjectSchema.optional(),
|
|
parameters: z.record(z.string(), jsonValueSchema).optional(),
|
|
linkedSubjects: z
|
|
.array(subjectSchema.extend({ operation: z.string().min(1) }))
|
|
.optional(),
|
|
select: z.custom<FixtureClaimExpectation["source"]["select"]>(
|
|
(value) => typeof value === "function",
|
|
),
|
|
}),
|
|
});
|
|
const finalAnswerSchema = z.strictObject({
|
|
claims: z.array(
|
|
z.strictObject({
|
|
id: z.string().min(1),
|
|
value: jsonValueSchema,
|
|
evidence_id: z.string().regex(/^ev_[a-f0-9]{64}$/u),
|
|
confidence: evidenceSchema.shape.confidence,
|
|
authority: evidenceAuthoritySchema,
|
|
}),
|
|
),
|
|
});
|
|
type FinalClaim = z.infer<typeof finalAnswerSchema>["claims"][number];
|
|
const isRecord = (value: unknown): value is Record<string, unknown> =>
|
|
typeof value === "object" && value !== null && !Array.isArray(value);
|
|
const record = (value: unknown): Record<string, unknown> | undefined =>
|
|
isRecord(value) ? value : undefined;
|
|
const sameJson = (left: unknown, right: unknown): boolean => {
|
|
const parsedLeft = jsonValueSchema.safeParse(left);
|
|
const parsedRight = jsonValueSchema.safeParse(right);
|
|
return (
|
|
parsedLeft.success &&
|
|
parsedRight.success &&
|
|
canonicalJson(parsedLeft.data) === canonicalJson(parsedRight.data)
|
|
);
|
|
};
|
|
|
|
interface TranscriptJson {
|
|
readonly value: unknown;
|
|
readonly hasDuplicateMembers: boolean;
|
|
}
|
|
|
|
const parseTranscriptJson = (text: string): TranscriptJson | undefined => {
|
|
const objects: Set<string>[] = [];
|
|
let invalid = false;
|
|
let hasDuplicateMembers = false;
|
|
try {
|
|
visit(
|
|
text,
|
|
{
|
|
onObjectBegin: () => {
|
|
objects.push(new Set());
|
|
},
|
|
onObjectProperty: (key) => {
|
|
const keys = objects.at(-1);
|
|
if (keys === undefined) invalid = true;
|
|
else if (keys.has(key)) hasDuplicateMembers = true;
|
|
keys?.add(key);
|
|
},
|
|
onObjectEnd: () => {
|
|
objects.pop();
|
|
},
|
|
onError: () => {
|
|
invalid = true;
|
|
},
|
|
},
|
|
{
|
|
disallowComments: true,
|
|
allowTrailingComma: false,
|
|
allowEmptyContent: false,
|
|
},
|
|
);
|
|
if (invalid) return undefined;
|
|
// Keep ambiguity separate from syntax: a failure can be observed without
|
|
// choosing one contradictory code or authenticating ambiguous Evidence.
|
|
return { value: JSON.parse(text) as unknown, hasDuplicateMembers };
|
|
} catch (error) {
|
|
// Syntax and nesting failures are malformed transcript input, not an evaluation crash.
|
|
if (error instanceof SyntaxError || error instanceof RangeError)
|
|
return undefined;
|
|
throw error;
|
|
}
|
|
};
|
|
|
|
interface DeliveredEvidence {
|
|
readonly records: Map<string, Evidence>;
|
|
readonly conflicts: Set<string>;
|
|
}
|
|
|
|
/** One completed tool call, with aliases normalized and successful Evidence authenticated. */
|
|
export interface ParsedCodexToolCall {
|
|
readonly error: boolean;
|
|
readonly errorCode: string | null;
|
|
readonly id: string | null;
|
|
readonly server: string | null;
|
|
readonly tool: string;
|
|
readonly arguments: JsonValue;
|
|
readonly evidence: readonly Evidence[];
|
|
}
|
|
|
|
const consistentAliases = (
|
|
value: Record<string, unknown>,
|
|
left: string,
|
|
right: string,
|
|
): boolean =>
|
|
value[left] === undefined ||
|
|
value[right] === undefined ||
|
|
value[left] === value[right];
|
|
|
|
const consistentJsonFields = (
|
|
value: Record<string, unknown>,
|
|
left: string,
|
|
right: string,
|
|
): boolean =>
|
|
value[left] === undefined ||
|
|
value[right] === undefined ||
|
|
sameJson(value[left], value[right]);
|
|
|
|
const consistentJsonAliases = (
|
|
value: Record<string, unknown>,
|
|
left: string,
|
|
right: string,
|
|
): boolean =>
|
|
value[left] === undefined ||
|
|
value[left] === null ||
|
|
value[right] === undefined ||
|
|
value[right] === null ||
|
|
sameJson(value[left], value[right]);
|
|
|
|
/** Parse result representations once; authentication and diagnostic certainty stay distinct. */
|
|
const parseToolResult = (
|
|
item: Record<string, unknown>,
|
|
server: string | null,
|
|
): {
|
|
readonly failed: boolean;
|
|
readonly errorCode: string | null;
|
|
readonly evidenceValue: unknown;
|
|
} => {
|
|
const result = record(item.result ?? item.output);
|
|
const structured = record(
|
|
result?.structured_content ?? result?.structuredContent,
|
|
);
|
|
const direct = record(structured?.error ?? result?.error ?? item.error);
|
|
let text: TranscriptJson | undefined;
|
|
if (direct === undefined && structured === undefined) {
|
|
// Only a complete single JSON block can supply a missing structured result.
|
|
const content = result?.content;
|
|
const block =
|
|
Array.isArray(content) && content.length === 1
|
|
? record(content[0])
|
|
: undefined;
|
|
if (block?.type === "text" && typeof block.text === "string")
|
|
text = parseTranscriptJson(block.text);
|
|
}
|
|
const error = direct ?? record(record(text?.value)?.error);
|
|
const failed =
|
|
item.status === "failed" ||
|
|
(item.error !== undefined && item.error !== null) ||
|
|
result?.isError === true ||
|
|
structured?.error !== undefined ||
|
|
error !== undefined;
|
|
const errorCode =
|
|
(direct !== undefined || text?.hasDuplicateMembers === false) &&
|
|
typeof error?.code === "string" &&
|
|
error.code.length > 0
|
|
? error.code
|
|
: null;
|
|
const evidenceValue =
|
|
server?.toLowerCase() === "rea" &&
|
|
item.status !== "failed" &&
|
|
(item.error === undefined || item.error === null) &&
|
|
consistentJsonAliases(item, "result", "output") &&
|
|
result !== undefined &&
|
|
result.isError !== true &&
|
|
(result.error === undefined || result.error === null) &&
|
|
(result.structuredContent === undefined ||
|
|
result.structured_content === undefined ||
|
|
sameJson(result.structuredContent, result.structured_content))
|
|
? (result.structuredContent ??
|
|
result.structured_content ??
|
|
(text?.hasDuplicateMembers === false ? text.value : undefined))
|
|
: undefined;
|
|
return { failed, errorCode, evidenceValue };
|
|
};
|
|
|
|
const evidenceFromCall = (
|
|
value: unknown,
|
|
tool: string,
|
|
): Evidence | undefined => {
|
|
const structured = record(value);
|
|
if (
|
|
structured === undefined ||
|
|
(structured.error !== undefined && structured.error !== null)
|
|
)
|
|
return undefined;
|
|
let evidence: Evidence;
|
|
try {
|
|
evidence = parseEvidence(structured);
|
|
} catch {
|
|
// Transcript output is untrusted; invalid Evidence cannot support a claim.
|
|
return undefined;
|
|
}
|
|
if (tool !== evidence.operation) return undefined;
|
|
return evidence;
|
|
};
|
|
|
|
const completedToolItem = (
|
|
event: unknown,
|
|
): Record<string, unknown> | undefined => {
|
|
const value = record(event);
|
|
const wrapped = value?.type === "item.completed";
|
|
const item = wrapped
|
|
? record(value.item)
|
|
: value?.type === "mcp_tool_call"
|
|
? value
|
|
: undefined;
|
|
if (item?.type !== "mcp_tool_call") return undefined;
|
|
const terminalStatus =
|
|
item.status === "completed" || item.status === "failed";
|
|
if (
|
|
(wrapped
|
|
? item.status !== undefined && !terminalStatus
|
|
: !terminalStatus) ||
|
|
!consistentAliases(item, "server", "server_name") ||
|
|
!consistentAliases(item, "tool", "name") ||
|
|
!consistentJsonFields(item, "arguments", "input")
|
|
)
|
|
return undefined;
|
|
return item;
|
|
};
|
|
|
|
const artifactDelivery = (evidence: Evidence): readonly Evidence[] => {
|
|
const records = [evidence];
|
|
if (evidence.operation !== "inspect_artifact") return records;
|
|
const substeps = record(evidence.normalized_result)?.substeps;
|
|
if (!Array.isArray(substeps)) return records;
|
|
// Artifact inspection delivers linked inventory Evidence inline; bare IDs in
|
|
// normalized facts are references, not authenticated deliveries.
|
|
for (const value of substeps) {
|
|
const substep = record(value);
|
|
if (
|
|
substep?.status !== "completed" ||
|
|
substep.operation !== "inventory_artifact" ||
|
|
typeof substep.evidence_id !== "string" ||
|
|
!evidence.evidence_links.includes(substep.evidence_id)
|
|
)
|
|
continue;
|
|
try {
|
|
const nested = parseEvidence(substep.evidence);
|
|
if (
|
|
nested.evidence_id === substep.evidence_id &&
|
|
nested.operation === substep.operation
|
|
)
|
|
records.push(nested);
|
|
} catch {
|
|
// An authenticated parent does not authenticate an embedded record's ID.
|
|
}
|
|
}
|
|
return records;
|
|
};
|
|
|
|
/** Parse one completed transcript call and authenticate any successful REA Evidence. */
|
|
export const parseCodexToolCall = (
|
|
event: unknown,
|
|
): ParsedCodexToolCall | undefined => {
|
|
const item = completedToolItem(event);
|
|
if (item === undefined) return undefined;
|
|
const server = item.server ?? item.server_name;
|
|
const tool = item.tool ?? item.name;
|
|
if (
|
|
((item.server !== undefined || item.server_name !== undefined) &&
|
|
typeof server !== "string") ||
|
|
typeof tool !== "string" ||
|
|
tool.length === 0
|
|
)
|
|
return undefined;
|
|
const argumentsValue =
|
|
item.arguments !== undefined
|
|
? item.arguments
|
|
: item.input !== undefined
|
|
? item.input
|
|
: {};
|
|
const parsedArguments = jsonValueSchema.safeParse(argumentsValue);
|
|
if (!parsedArguments.success) return undefined;
|
|
const serverName = typeof server === "string" ? server : null;
|
|
const result = parseToolResult(item, serverName);
|
|
const parsed: ParsedCodexToolCall = {
|
|
error: result.failed,
|
|
errorCode: result.errorCode,
|
|
id: typeof item.id === "string" && item.id.length > 0 ? item.id : null,
|
|
server: serverName,
|
|
tool,
|
|
arguments: parsedArguments.data,
|
|
evidence: [],
|
|
};
|
|
const evidence = evidenceFromCall(result.evidenceValue, tool);
|
|
if (evidence === undefined) return parsed;
|
|
return { ...parsed, evidence: artifactDelivery(evidence) };
|
|
};
|
|
|
|
const collectDeliveredEvidence = (
|
|
calls: readonly (ParsedCodexToolCall | undefined)[],
|
|
): DeliveredEvidence => {
|
|
const records = new Map<string, Evidence>();
|
|
const conflicts = new Set<string>();
|
|
for (const call of calls) {
|
|
for (const evidence of call?.evidence ?? []) {
|
|
const previous = records.get(evidence.evidence_id);
|
|
if (previous !== undefined && !sameJson(previous, evidence))
|
|
conflicts.add(evidence.evidence_id);
|
|
else records.set(evidence.evidence_id, evidence);
|
|
}
|
|
}
|
|
return { records, conflicts };
|
|
};
|
|
|
|
const matchesSubject = (
|
|
evidence: Evidence,
|
|
expected: { readonly path: string; readonly sha256?: string },
|
|
): boolean =>
|
|
evidence.subject?.local_path === expected.path &&
|
|
(expected.sha256 === undefined ||
|
|
evidence.subject.identity.sha256 === expected.sha256);
|
|
|
|
const matchesParameters = (
|
|
evidence: Evidence,
|
|
expected: Readonly<Record<string, JsonValue>> | undefined,
|
|
): boolean =>
|
|
expected === undefined ||
|
|
Object.entries(expected).every(
|
|
([key, value]) =>
|
|
Object.hasOwn(evidence.parameters, key) &&
|
|
sameJson(evidence.parameters[key], value),
|
|
);
|
|
|
|
const matchesLinks = (
|
|
evidence: Evidence,
|
|
source: FixtureClaimExpectation["source"],
|
|
delivered: DeliveredEvidence,
|
|
): boolean =>
|
|
(source.linkedSubjects ?? []).every((expected) =>
|
|
evidence.evidence_links.some((id) => {
|
|
const linked = delivered.records.get(id);
|
|
return (
|
|
linked !== undefined &&
|
|
!delivered.conflicts.has(id) &&
|
|
linked.operation === expected.operation &&
|
|
matchesSubject(linked, expected)
|
|
);
|
|
}),
|
|
);
|
|
|
|
const assessClaim = (
|
|
expected: FixtureClaimExpectation,
|
|
claim: FinalClaim | undefined,
|
|
delivered: DeliveredEvidence,
|
|
): FixtureClaimAssessment => {
|
|
const failures: FixtureClaimFailure[] = [];
|
|
if (claim === undefined)
|
|
return { id: expected.id, status: "failed", failures: ["claim_missing"] };
|
|
if (!sameJson(claim.value, expected.expectedValue))
|
|
failures.push("value_mismatch");
|
|
if (claim.confidence !== expected.source.confidence)
|
|
failures.push("claim_confidence_mismatch");
|
|
if (claim.authority !== expected.source.authority)
|
|
failures.push("claim_authority_mismatch");
|
|
const evidence = delivered.records.get(claim.evidence_id);
|
|
if (evidence === undefined) failures.push("unknown_evidence");
|
|
else if (delivered.conflicts.has(claim.evidence_id))
|
|
failures.push("conflicting_evidence");
|
|
else {
|
|
if (evidence.operation !== expected.source.operation)
|
|
failures.push("source_operation_mismatch");
|
|
if (evidence.confidence !== expected.source.confidence)
|
|
failures.push("source_confidence_mismatch");
|
|
if (evidence.authority !== expected.source.authority)
|
|
failures.push("source_authority_mismatch");
|
|
if (
|
|
expected.source.subject !== undefined &&
|
|
!matchesSubject(evidence, expected.source.subject)
|
|
)
|
|
failures.push("source_subject_mismatch");
|
|
if (!matchesParameters(evidence, expected.source.parameters))
|
|
failures.push("source_parameters_mismatch");
|
|
if (!matchesLinks(evidence, expected.source, delivered))
|
|
failures.push("source_links_mismatch");
|
|
try {
|
|
if (
|
|
!sameJson(
|
|
expected.source.select(evidence.normalized_result, {
|
|
evidence,
|
|
deliveredEvidence: [...delivered.records.values()].filter(
|
|
(record) => !delivered.conflicts.has(record.evidence_id),
|
|
),
|
|
}),
|
|
expected.expectedValue,
|
|
)
|
|
)
|
|
failures.push("source_assertion_mismatch");
|
|
} catch {
|
|
// Producer-specific selectors may reject malformed representations; preserve a failed assessment.
|
|
failures.push("selector_failed");
|
|
}
|
|
}
|
|
return {
|
|
id: expected.id,
|
|
status: failures.length === 0 ? "passed" : "failed",
|
|
failures,
|
|
};
|
|
};
|
|
|
|
/** Grade closed answers only against configured fixture truth and delivered authenticated Evidence. */
|
|
export const evaluateKnownAnswers = (
|
|
calls: readonly (ParsedCodexToolCall | undefined)[],
|
|
finalMessage: string,
|
|
expectations: readonly FixtureClaimExpectation[] | undefined,
|
|
): KnownAnswerAssessment => {
|
|
const scope = "configured_fixture_claims";
|
|
if (expectations === undefined)
|
|
return { scope, status: "not_assessed", manifestFailures: [], claims: [] };
|
|
const rubric = z.array(expectationSchema).min(1).safeParse(expectations);
|
|
if (!rubric.success)
|
|
return {
|
|
scope,
|
|
status: "failed",
|
|
manifestFailures: ["invalid_fixture_rubric"],
|
|
claims: [],
|
|
};
|
|
const expectedIds = new Set(expectations.map(({ id }) => id));
|
|
if (expectedIds.size !== expectations.length)
|
|
return {
|
|
scope,
|
|
status: "failed",
|
|
manifestFailures: ["duplicate_fixture_claim_ids"],
|
|
claims: [],
|
|
};
|
|
const answer = parseTranscriptJson(finalMessage);
|
|
const manifest = finalAnswerSchema.safeParse(
|
|
answer?.hasDuplicateMembers === false ? answer.value : undefined,
|
|
);
|
|
if (!manifest.success)
|
|
return {
|
|
scope,
|
|
status: "failed",
|
|
manifestFailures: ["invalid_final_claim_manifest"],
|
|
claims: [],
|
|
};
|
|
const manifestFailures: string[] = [];
|
|
const submitted = new Map<string, FinalClaim>();
|
|
for (const claim of manifest.data.claims) {
|
|
if (submitted.has(claim.id))
|
|
manifestFailures.push(`duplicate_claim:${claim.id}`);
|
|
if (!expectedIds.has(claim.id))
|
|
manifestFailures.push(`unknown_claim:${claim.id}`);
|
|
submitted.set(claim.id, claim);
|
|
}
|
|
const delivered = collectDeliveredEvidence(calls);
|
|
const claims = expectations.map((expected) =>
|
|
assessClaim(expected, submitted.get(expected.id), delivered),
|
|
);
|
|
return {
|
|
scope,
|
|
status:
|
|
manifestFailures.length === 0 &&
|
|
claims.every(({ status }) => status === "passed")
|
|
? "passed"
|
|
: "failed",
|
|
manifestFailures,
|
|
claims,
|
|
};
|
|
};
|