@apmantza/greedysearch-pi 1.9.2 → 2.0.0

package/src/search/synthesis.mjs

@@ -1,246 +1,286 @@
-// src/search/synthesis.mjs — Synthesis prompt building, structured JSON parsing,
-// confidence metrics, and payload normalization
-//
-// Extracted from search.mjs to reduce file complexity.
-
-import { ALL_ENGINES } from "./constants.mjs";
-import { trimText } from "./sources.mjs";
-
-export function parseStructuredJson(text) {
-	if (!text) return null;
-	let trimmed = String(text).trim();
-
-	// Look for BEGIN_JSON/END_JSON markers first
-	const beginIdx = trimmed.indexOf("BEGIN_JSON");
-	const endIdx = trimmed.indexOf("END_JSON");
-	if (beginIdx !== -1 && endIdx !== -1 && beginIdx < endIdx) {
-		trimmed = trimmed.slice(beginIdx + "BEGIN_JSON".length, endIdx).trim();
-	} else {
-		// Strip out common LLM preamble text before the actual JSON
-		const jsonStart = trimmed.indexOf("{");
-		if (jsonStart > 0) {
-			trimmed = trimmed.slice(jsonStart);
-		}
-	}
-
-	const candidates = [
-		trimmed,
-		trimmed
-			.replace(/^```json\s*/i, "")
-			.replace(/^```\s*/i, "")
-			.replace(/```$/i, "")
-			.trim(),
-	];
-
-	// Find the outermost JSON object via brace matching (avoids ReDoS-prone .* patterns)
-	const firstBrace = trimmed.indexOf("{");
-	const lastBrace = trimmed.lastIndexOf("}");
-	if (firstBrace !== -1 && lastBrace !== -1 && firstBrace < lastBrace) {
-		candidates.push(trimmed.slice(firstBrace, lastBrace + 1));
-	}
-
-	for (const candidate of candidates) {
-		try {
-			return JSON.parse(candidate);
-		} catch {
-			// try next candidate
-		}
-	}
-	return null;
-}
-
-export function normalizeSynthesisPayload(
-	payload,
-	sources,
-	fallbackAnswer = "",
-) {
-	const sourceIds = new Set(sources.map((source) => source.id));
-	const agreementLevel = [
-		"high",
-		"medium",
-		"low",
-		"mixed",
-		"conflicting",
-	].includes(payload?.agreement?.level)
-		? payload.agreement.level
-		: "mixed";
-	const claims = Array.isArray(payload?.claims)
-		? payload.claims
-				.map((claim) => ({
-					claim: trimText(claim?.claim || "", 260),
-					support: ["strong", "moderate", "weak", "conflicting"].includes(
-						claim?.support,
-					)
-						? claim.support
-						: "moderate",
-					sourceIds: Array.isArray(claim?.sourceIds)
-						? claim.sourceIds.filter((id) => sourceIds.has(id))
-						: [],
-				}))
-				.filter((claim) => claim.claim)
-		: [];
-	const recommendedSources = Array.isArray(payload?.recommendedSources)
-		? payload.recommendedSources.filter((id) => sourceIds.has(id)).slice(0, 6)
-		: [];
-
-	// Clean up fallback answer if it contains preamble text
-	// Use indexOf/lastIndexOf instead of [\s\S]* patterns to avoid ReDoS
-	let cleanFallback = "";
-	if (fallbackAnswer) {
-		const firstBrace = fallbackAnswer.indexOf("{");
-		const lastBrace = fallbackAnswer.lastIndexOf("}");
-		if (firstBrace !== -1 && lastBrace !== -1 && firstBrace < lastBrace) {
-			cleanFallback = fallbackAnswer.slice(firstBrace, lastBrace + 1);
-		} else {
-			cleanFallback = fallbackAnswer;
-		}
-	}
-
-	return {
-		answer: trimText(payload?.answer || cleanFallback || fallbackAnswer, 4000),
-		agreement: {
-			level: agreementLevel,
-			summary: trimText(payload?.agreement?.summary || "", 280),
-		},
-		differences: Array.isArray(payload?.differences)
-			? payload.differences
-					.map((item) => trimText(item, 220))
-					.filter(Boolean)
-					.slice(0, 5)
-			: [],
-		caveats: Array.isArray(payload?.caveats)
-			? payload.caveats
-					.map((item) => trimText(item, 220))
-					.filter(Boolean)
-					.slice(0, 5)
-			: [],
-		claims,
-		recommendedSources,
-	};
-}
-
-export function buildSynthesisPrompt(
-	query,
-	results,
-	sources,
-	{ grounded = false } = {},
-) {
-	const engineSummaries = {};
-	for (const engine of ["perplexity", "bing", "google"]) {
-		const result = results[engine];
-		if (!result) continue;
-		if (result.error) {
-			engineSummaries[engine] = {
-				status: "error",
-				error: String(result.error),
-			};
-			continue;
-		}
-
-		engineSummaries[engine] = {
-			status: "ok",
-			answer: trimText(result.answer || "", grounded ? 4500 : 2200),
-			sourceIds: sources
-				.filter((source) => source.engines.includes(engine))
-				.sort(
-					(a, b) =>
-						(a.perEngine[engine]?.rank || 99) -
-						(b.perEngine[engine]?.rank || 99),
-				)
-				.map((source) => source.id)
-				.slice(0, 6),
-		};
-	}
-
-	// Snippet budget: always include content for fetched sources so Gemini can
-	// make citation decisions based on what the sources actually say, not just
-	// their metadata. Grounded mode gets a larger budget per source.
-	const snippetChars = grounded ? 700 : 300;
-	const sourceRegistry = sources.slice(0, grounded ? 10 : 8).map((source) => ({
-		id: source.id,
-		title: source.title,
-		domain: source.domain,
-		canonicalUrl: source.canonicalUrl,
-		sourceType: source.sourceType,
-		isOfficial: source.isOfficial,
-		engines: source.engines,
-		engineCount: source.engineCount,
-		fetch: source.fetch?.attempted
-			? {
-					ok: source.fetch.ok,
-					publishedTime: source.fetch.publishedTime || "",
-					byline: source.fetch.byline || "",
-					snippet: trimText(source.fetch.snippet || "", snippetChars),
-				}
-			: undefined,
-	}));
-
-	return [
-		"You are a research synthesizer. Combine these search engine results into a single authoritative answer.",
-		"",
-		`Query: ${query}`,
-		"",
-		`Engine summaries:\n${JSON.stringify(engineSummaries, null, 2)}`,
-		"",
-		`Source registry:\n${JSON.stringify(sourceRegistry, null, 2)}`,
-		"",
-		"Instructions:",
-		"- Write a clear, direct answer in markdown (use headers/bullets where they help readability)",
-		"- Cite sources inline as [S1], [S2] etc. when making specific claims",
-		"- Prefer sources with content (fetch.ok=true and non-empty snippet) for citations",
-		"- Note where the engines agree or meaningfully disagree",
-		"- List any important caveats or limitations",
-		"- recommendedSources: the 2-4 source IDs most worth reading for this query",
-		"",
-		"Respond ONLY with a JSON object wrapped in BEGIN_JSON / END_JSON markers:",
-		"",
-		"BEGIN_JSON",
-		JSON.stringify({
-			answer: "<your markdown answer here>",
-			agreement: { level: "high|medium|mixed|conflicting", summary: "<one sentence>" },
-			differences: ["<notable difference between engines, if any>"],
-			caveats: ["<important caveat or limitation>"],
-			recommendedSources: ["S1", "S2"],
-		}, null, 2),
-		"END_JSON",
-	].join("\n");
-}
-
-export function buildConfidence(out) {
-	const sources = Array.isArray(out._sources) ? out._sources : [];
-	const topConsensus = sources.length > 0 ? sources[0]?.engineCount || 0 : 0;
-	const officialSourceCount = sources.filter(
-		(source) => source.isOfficial,
-	).length;
-	const firstPartySourceCount = sources.filter(
-		(source) => source.isOfficial || source.sourceType === "maintainer-blog",
-	).length;
-	const fetchedAttempted = sources.filter(
-		(source) => source.fetch?.attempted,
-	).length;
-	const fetchedSucceeded = sources.filter((source) => source.fetch?.ok).length;
-	const sourceTypeBreakdown = sources.reduce((acc, source) => {
-		acc[source.sourceType] = (acc[source.sourceType] || 0) + 1;
-		return acc;
-	}, {});
-	const synthesisLevel = out._synthesis?.agreement?.level;
-
-	return {
-		sourcesCount: sources.length,
-		topSourceConsensus: topConsensus,
-		agreementLevel:
-			synthesisLevel ||
-			(topConsensus >= 3 ? "high" : topConsensus >= 2 ? "medium" : "low"),
-		enginesResponded: ALL_ENGINES.filter(
-			(engine) => out[engine]?.answer && !out[engine]?.error,
-		),
-		enginesFailed: ALL_ENGINES.filter((engine) => out[engine]?.error),
-		officialSourceCount,
-		firstPartySourceCount,
-		fetchedSourceSuccessRate:
-			fetchedAttempted > 0
-				? Number((fetchedSucceeded / fetchedAttempted).toFixed(2))
-				: 0,
-		sourceTypeBreakdown,
-	};
-}
+// src/search/synthesis.mjs — Synthesis prompt building, structured JSON parsing,
+// confidence metrics, and payload normalization
+//
+// Extracted from search.mjs to reduce file complexity.
+
+import { ALL_ENGINES } from "./constants.mjs";
+import { trimText } from "./sources.mjs";
+
+function escapeControlCharsInsideJsonStrings(text) {
+	let out = "";
+	let inString = false;
+	let escaped = false;
+	for (const char of String(text)) {
+		if (escaped) {
+			out += char;
+			escaped = false;
+			continue;
+		}
+		if (char === "\\") {
+			out += char;
+			escaped = true;
+			continue;
+		}
+		if (char === '"') {
+			inString = !inString;
+			out += char;
+			continue;
+		}
+		if (inString && char === "\n") out += "\\n";
+		else if (inString && char === "\r") out += "\\r";
+		else if (inString && char === "\t") out += "\\t";
+		else out += char;
+	}
+	return out;
+}
+
+export function parseStructuredJson(text) {
+	if (!text) return null;
+	let trimmed = String(text).trim();
+
+	// Look for BEGIN_JSON/END_JSON markers first
+	const beginIdx = trimmed.indexOf("BEGIN_JSON");
+	const endIdx = trimmed.indexOf("END_JSON");
+	if (beginIdx !== -1 && endIdx !== -1 && beginIdx < endIdx) {
+		trimmed = trimmed.slice(beginIdx + "BEGIN_JSON".length, endIdx).trim();
+	} else {
+		// Strip out common LLM preamble text before the actual JSON
+		const jsonStart = trimmed.indexOf("{");
+		if (jsonStart > 0) {
+			trimmed = trimmed.slice(jsonStart);
+		}
+	}
+
+	const candidates = [
+		trimmed,
+		trimmed
+			.replace(/^```json\s*/i, "")
+			.replace(/^```\s*/i, "")
+			.replace(/```$/i, "")
+			.trim(),
+	];
+
+	// Find the outermost JSON object via brace matching (avoids ReDoS-prone .* patterns)
+	const firstBrace = trimmed.indexOf("{");
+	const lastBrace = trimmed.lastIndexOf("}");
+	if (firstBrace !== -1 && lastBrace !== -1 && firstBrace < lastBrace) {
+		candidates.push(trimmed.slice(firstBrace, lastBrace + 1));
+	}
+
+	for (const candidate of [...candidates]) {
+		const repaired = escapeControlCharsInsideJsonStrings(candidate);
+		if (repaired !== candidate) candidates.push(repaired);
+	}
+
+	for (const candidate of candidates) {
+		try {
+			return JSON.parse(candidate);
+		} catch {
+			// try next candidate
+		}
+	}
+	return null;
+}
+
+export function normalizeSynthesisPayload(
+	payload,
+	sources,
+	fallbackAnswer = "",
+) {
+	const sourceIds = new Set(sources.map((source) => source.id));
+	const agreementLevel = [
+		"high",
+		"medium",
+		"low",
+		"mixed",
+		"conflicting",
+	].includes(payload?.agreement?.level)
+		? payload.agreement.level
+		: "mixed";
+	const claims = Array.isArray(payload?.claims)
+		? payload.claims
+				.map((claim) => ({
+					claim: trimText(claim?.claim || "", 260),
+					support: ["strong", "moderate", "weak", "conflicting"].includes(
+						claim?.support,
+					)
+						? claim.support
+						: "moderate",
+					sourceIds: Array.isArray(claim?.sourceIds)
+						? claim.sourceIds.filter((id) => sourceIds.has(id))
+						: [],
+				}))
+				.filter((claim) => claim.claim)
+		: [];
+	const recommendedSources = Array.isArray(payload?.recommendedSources)
+		? payload.recommendedSources.filter((id) => sourceIds.has(id)).slice(0, 6)
+		: [];
+
+	// Clean up fallback answer if it contains preamble text
+	// Use indexOf/lastIndexOf instead of [\s\S]* patterns to avoid ReDoS
+	let cleanFallback = "";
+	if (fallbackAnswer) {
+		const firstBrace = fallbackAnswer.indexOf("{");
+		const lastBrace = fallbackAnswer.lastIndexOf("}");
+		if (firstBrace !== -1 && lastBrace !== -1 && firstBrace < lastBrace) {
+			cleanFallback = fallbackAnswer.slice(firstBrace, lastBrace + 1);
+		} else {
+			cleanFallback = fallbackAnswer;
+		}
+	}
+
+	return {
+		answer: trimText(payload?.answer || cleanFallback || fallbackAnswer, 4000),
+		agreement: {
+			level: agreementLevel,
+			summary: trimText(payload?.agreement?.summary || "", 280),
+		},
+		differences: Array.isArray(payload?.differences)
+			? payload.differences
+					.map((item) => trimText(item, 220))
+					.filter(Boolean)
+					.slice(0, 5)
+			: [],
+		caveats: Array.isArray(payload?.caveats)
+			? payload.caveats
+					.map((item) => trimText(item, 220))
+					.filter(Boolean)
+					.slice(0, 5)
+			: [],
+		claims,
+		recommendedSources,
+	};
+}
+
+export function buildSynthesisPrompt(
+	query,
+	results,
+	sources,
+	{ grounded = false } = {},
+) {
+	const engineSummaries = {};
+	for (const engine of ["perplexity", "bing", "google"]) {
+		const result = results[engine];
+		if (!result) continue;
+		if (result.error) {
+			engineSummaries[engine] = {
+				status: "error",
+				error: String(result.error),
+			};
+			continue;
+		}
+
+		engineSummaries[engine] = {
+			status: "ok",
+			answer: trimText(result.answer || "", grounded ? 4500 : 2200),
+			sourceIds: sources
+				.filter((source) => source.engines.includes(engine))
+				.sort(
+					(a, b) =>
+						(a.perEngine[engine]?.rank || 99) -
+						(b.perEngine[engine]?.rank || 99),
+				)
+				.map((source) => source.id)
+				.slice(0, 6),
+		};
+	}
+
+	// Snippet budget: always include content for fetched sources so Gemini can
+	// make citation decisions based on what the sources actually say, not just
+	// their metadata. Grounded mode gets a larger budget per source.
+	const snippetChars = grounded ? 700 : 300;
+	const sourceRegistry = sources.slice(0, grounded ? 10 : 8).map((source) => ({
+		id: source.id,
+		title: source.title,
+		domain: source.domain,
+		canonicalUrl: source.canonicalUrl,
+		sourceType: source.sourceType,
+		isOfficial: source.isOfficial,
+		engines: source.engines,
+		engineCount: source.engineCount,
+		fetch: source.fetch?.attempted
+			? {
+					ok: source.fetch.ok,
+					publishedTime: source.fetch.publishedTime || "",
+					byline: source.fetch.byline || "",
+					snippet: trimText(source.fetch.snippet || "", snippetChars),
+				}
+			: undefined,
+	}));
+
+	return [
+		"You are a research synthesizer. Combine these search engine results into a single authoritative answer.",
+		"",
+		`Query: ${query}`,
+		"",
+		`Engine summaries:\n${JSON.stringify(engineSummaries, null, 2)}`,
+		"",
+		`Source registry:\n${JSON.stringify(sourceRegistry, null, 2)}`,
+		"",
+		"Instructions:",
+		"- Write a clear, direct answer in markdown (use headers/bullets where they help readability)",
+		"- Cite sources inline as [S1], [S2] etc. when making specific claims",
+		"- Prefer sources with content (fetch.ok=true and non-empty snippet) for citations",
+		"- Note where the engines agree or meaningfully disagree",
+		"- List any important caveats or limitations",
+		"- recommendedSources: the 2-4 source IDs most worth reading for this query",
+		"",
+		"Respond ONLY with a JSON object wrapped in BEGIN_JSON / END_JSON markers:",
+		"",
+		"BEGIN_JSON",
+		JSON.stringify(
+			{
+				answer: "<your markdown answer here>",
+				agreement: {
+					level: "high|medium|mixed|conflicting",
+					summary: "<one sentence>",
+				},
+				differences: ["<notable difference between engines, if any>"],
+				caveats: ["<important caveat or limitation>"],
+				recommendedSources: ["S1", "S2"],
+			},
+			null,
+			2,
+		),
+		"END_JSON",
+	].join("\n");
+}
+
+export function buildConfidence(out) {
+	const sources = Array.isArray(out._sources) ? out._sources : [];
+	const topConsensus = sources.length > 0 ? sources[0]?.engineCount || 0 : 0;
+	const officialSourceCount = sources.filter(
+		(source) => source.isOfficial,
+	).length;
+	const firstPartySourceCount = sources.filter(
+		(source) => source.isOfficial || source.sourceType === "maintainer-blog",
+	).length;
+	const fetchedAttempted = sources.filter(
+		(source) => source.fetch?.attempted,
+	).length;
+	const fetchedSucceeded = sources.filter((source) => source.fetch?.ok).length;
+	const sourceTypeBreakdown = sources.reduce((acc, source) => {
+		acc[source.sourceType] = (acc[source.sourceType] || 0) + 1;
+		return acc;
+	}, {});
+	const synthesisLevel = out._synthesis?.agreement?.level;
+
+	return {
+		sourcesCount: sources.length,
+		topSourceConsensus: topConsensus,
+		agreementLevel:
+			synthesisLevel ||
+			(topConsensus >= 3 ? "high" : topConsensus >= 2 ? "medium" : "low"),
+		enginesResponded: ALL_ENGINES.filter(
+			(engine) => out[engine]?.answer && !out[engine]?.error,
+		),
+		enginesFailed: ALL_ENGINES.filter((engine) => out[engine]?.error),
+		officialSourceCount,
+		firstPartySourceCount,
+		fetchedSourceSuccessRate:
+			fetchedAttempted > 0
+				? Number((fetchedSucceeded / fetchedAttempted).toFixed(2))
+				: 0,
+		sourceTypeBreakdown,
+	};
+}