@humain/terminal 0.0.2 → 0.0.3-beta.0

package/dist/humain/web-access/vendor/serpdive.js

@@ -0,0 +1,267 @@
+import { existsSync, readFileSync } from "node:fs";
+import { activityMonitor } from "./activity.js";
+import { hasCredentialSource, redactCredential, resolveCredential } from "./credential-source.js";
+import { getWebSearchConfigPath } from "./utils.js";
+const SERPDIVE_API_URL = "https://api.serpdive.com/v1/search";
+const CONFIG_PATH = getWebSearchConfigPath();
+const SEARCH_TIMEOUT_MS = 60_000;
+// Retrieval depth. The default is deliberately the free tier: this provider
+// ships inside someone else's framework, and installing it must never start
+// spending on a user's behalf without them choosing it.
+//   krill — free, fair use, no synthesized answer (see buildAnswer below)
+//   mako  — the fact-carrying sentences of each page, 1 credit
+//   moby  — the full readable content of every page, 1.5 credits
+// Pricing: https://serpdive.com/pricing
+const MODELS = ["krill", "mako", "moby"];
+const DEFAULT_MODEL = "krill";
+let cachedConfig = null;
+function loadConfig() {
+    if (cachedConfig)
+        return cachedConfig;
+    if (!existsSync(CONFIG_PATH)) {
+        cachedConfig = {};
+        return cachedConfig;
+    }
+    const raw = readFileSync(CONFIG_PATH, "utf-8");
+    try {
+        cachedConfig = JSON.parse(raw);
+        return cachedConfig;
+    }
+    catch (err) {
+        const message = err instanceof Error ? err.message : String(err);
+        throw new Error(`Failed to parse ${CONFIG_PATH}: ${message}`);
+    }
+}
+async function getApiKey(signal) {
+    return resolveCredential({
+        provider: "SERPdive",
+        configuredValue: loadConfig().serpdiveApiKey,
+        environmentValue: process.env.SERPDIVE_API_KEY,
+        signal,
+    });
+}
+async function requireApiKey(signal) {
+    const apiKey = await getApiKey(signal);
+    if (!apiKey) {
+        throw new Error("SERPdive API key not found. Either:\n" +
+            `  1. Create ${CONFIG_PATH} with { "serpdiveApiKey": "your-key" }\n` +
+            "  2. Set SERPDIVE_API_KEY environment variable\n" +
+            "Get a key at https://serpdive.com/dashboard/keys");
+    }
+    return apiKey;
+}
+// An unknown value falls back to the free default rather than failing: a typo
+// in a config file must not cost the user money, and must not break search.
+function resolveModel() {
+    const raw = process.env.SERPDIVE_MODEL ?? loadConfig().serpdiveModel;
+    if (typeof raw !== "string")
+        return DEFAULT_MODEL;
+    const value = raw.trim().toLowerCase();
+    return MODELS.includes(value) ? value : DEFAULT_MODEL;
+}
+function normalizeCount(value) {
+    if (typeof value !== "number" || !Number.isFinite(value))
+        return 5;
+    return Math.max(1, Math.min(Math.floor(value), 20));
+}
+function normalizeDomain(value) {
+    let input = value.trim().toLowerCase();
+    if (!input)
+        return null;
+    if (input.startsWith("-"))
+        input = input.slice(1).trim();
+    if (!input)
+        return null;
+    try {
+        const parsed = input.includes("://") ? new URL(input) : new URL(`https://${input}`);
+        input = parsed.hostname;
+    }
+    catch {
+        input = input.split("/")[0]?.split(":")[0] ?? "";
+    }
+    input = input.replace(/^\.+|\.+$/g, "");
+    return /^[a-z0-9][a-z0-9.-]*\.[a-z]{2,}$/i.test(input) ? input : null;
+}
+function parseDomainFilter(domainFilter) {
+    const filters = { include: [], exclude: [] };
+    if (!domainFilter?.length)
+        return filters;
+    for (const raw of domainFilter) {
+        const domain = normalizeDomain(raw);
+        if (!domain)
+            continue;
+        const target = raw.trim().startsWith("-") ? filters.exclude : filters.include;
+        if (!target.includes(domain))
+            target.push(domain);
+    }
+    return filters;
+}
+function domainMatches(hostname, domain) {
+    return hostname === domain || hostname.endsWith(`.${domain}`);
+}
+// SERPdive exposes no include/exclude domain parameter, so the filter is applied
+// here, on what came back. It can therefore only ever narrow a page of results —
+// it cannot ask the engine for more pages from a given domain.
+function passesDomainFilters(url, filters) {
+    if (filters.include.length === 0 && filters.exclude.length === 0)
+        return true;
+    let hostname;
+    try {
+        hostname = new URL(url).hostname.toLowerCase();
+    }
+    catch {
+        return false;
+    }
+    if (filters.exclude.some((domain) => domainMatches(hostname, domain)))
+        return false;
+    if (filters.include.length === 0)
+        return true;
+    return filters.include.some((domain) => domainMatches(hostname, domain));
+}
+// SERPdive has NO time-range parameter. Recency is expressed inside the question
+// and read by the engine, which biases ranking toward recent pages — it is a
+// hint, never a guaranteed freshness filter, and results outside the window can
+// still come back.
+function applyRecencyHint(query, recencyFilter) {
+    if (!recencyFilter)
+        return query;
+    const hints = {
+        day: "past 24 hours",
+        week: "past week",
+        month: "past month",
+        year: "past year",
+    };
+    const hint = hints[recencyFilter];
+    return hint ? `${query} ${hint}` : query;
+}
+function requestSignal(signal) {
+    const timeout = AbortSignal.timeout(SEARCH_TIMEOUT_MS);
+    return signal ? AbortSignal.any([signal, timeout]) : timeout;
+}
+function errorMessage(err) {
+    return err instanceof Error ? err.message : String(err);
+}
+function mapResults(results, numResults, filters) {
+    if (!Array.isArray(results))
+        return [];
+    const mapped = [];
+    for (const item of results) {
+        if (!item?.url || !passesDomainFilters(item.url, filters))
+            continue;
+        mapped.push({
+            title: item.title || `Source ${mapped.length + 1}`,
+            url: item.url,
+            snippet: typeof item.content === "string" ? item.content.replace(/\s+/g, " ").trim() : "",
+        });
+        if (mapped.length >= numResults)
+            break;
+    }
+    return mapped;
+}
+function mapInlineContent(results, filters) {
+    if (!Array.isArray(results))
+        return [];
+    return results.flatMap((item) => {
+        if (!item?.url || !passesDomainFilters(item.url, filters))
+            return [];
+        if (typeof item.content !== "string" || item.content.trim().length === 0)
+            return [];
+        return [{
+                url: item.url,
+                title: item.title || "",
+                content: item.content,
+                error: null,
+            }];
+    });
+}
+// The free krill tier returns extracted content but no synthesized answer, so
+// one is assembled from the sources — the same shape brave.ts and searxng.ts
+// produce for providers that do not synthesize. mako and moby ask the API for a
+// real answer and use it when it comes back.
+function buildAnswer(apiAnswer, results) {
+    if (typeof apiAnswer === "string" && apiAnswer.trim().length > 0)
+        return apiAnswer;
+    return results
+        .map((result) => {
+        if (result.snippet)
+            return `${result.snippet}\nSource: ${result.title} (${result.url})`;
+        return `Source: ${result.title} (${result.url})`;
+    })
+        .join("\n\n");
+}
+export function isSerpdiveAvailable() {
+    return hasCredentialSource({
+        provider: "SERPdive",
+        configuredValue: loadConfig().serpdiveApiKey,
+        environmentValue: process.env.SERPDIVE_API_KEY,
+    });
+}
+export async function searchWithSerpdive(query, options = {}) {
+    const apiKey = await requireApiKey(options.signal);
+    const numResults = normalizeCount(options.numResults);
+    const filters = parseDomainFilter(options.domainFilter);
+    const model = resolveModel();
+    const body = {
+        query: applyRecencyHint(query, options.recencyFilter),
+        model,
+        // max_results is a CAP, never a minimum: the engine returns what it
+        // judges relevant, up to this many. Asking for more does not produce more.
+        max_results: Math.min(numResults, 10),
+        // krill has no answer synthesis — asking for one there is silently ignored
+        // by the API, so it is not asked for at all.
+        ...(model === "krill" ? {} : { answer: true }),
+    };
+    const activityId = activityMonitor.logStart({ type: "api", query });
+    let response;
+    try {
+        response = await fetch(SERPDIVE_API_URL, {
+            method: "POST",
+            headers: {
+                "Authorization": `Bearer ${apiKey}`,
+                "Content-Type": "application/json",
+            },
+            body: JSON.stringify(body),
+            signal: requestSignal(options.signal),
+        });
+    }
+    catch (err) {
+        const message = errorMessage(err);
+        const redactedMessage = redactCredential(message, apiKey);
+        if (redactedMessage.toLowerCase().includes("abort"))
+            activityMonitor.logComplete(activityId, 0);
+        else
+            activityMonitor.logError(activityId, redactedMessage);
+        if (redactedMessage === message)
+            throw err;
+        const redactedError = new Error(redactedMessage);
+        if (err instanceof Error)
+            redactedError.name = err.name;
+        throw redactedError;
+    }
+    if (!response.ok) {
+        activityMonitor.logComplete(activityId, response.status);
+        const errorText = redactCredential(await response.text(), apiKey);
+        throw new Error(`SERPdive API error ${response.status}: ${errorText.slice(0, 300)}`);
+    }
+    let data;
+    try {
+        data = await response.json();
+    }
+    catch (err) {
+        activityMonitor.logComplete(activityId, response.status);
+        throw new Error(`SERPdive API returned invalid JSON: ${errorMessage(err)}`);
+    }
+    activityMonitor.logComplete(activityId, response.status);
+    const results = mapResults(data.results, numResults, filters);
+    const result = {
+        answer: buildAnswer(data.answer, results),
+        results,
+    };
+    if (options.includeContent) {
+        const inlineContent = mapInlineContent(data.results, filters);
+        if (inlineContent.length > 0)
+            result.inlineContent = inlineContent;
+    }
+    return result;
+}
+//# sourceMappingURL=serpdive.js.map

package/dist/humain/web-access/vendor/serpdive.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"serpdive.js","sourceRoot":"","sources":["../../../../src/humain/web-access/vendor/serpdive.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,UAAU,EAAE,YAAY,EAAE,MAAM,SAAS,CAAC;AACnD,OAAO,EAAE,eAAe,EAAE,MAAM,eAAe,CAAC;AAGhD,OAAO,EAAE,mBAAmB,EAAE,gBAAgB,EAAE,iBAAiB,EAAE,MAAM,wBAAwB,CAAC;AAClG,OAAO,EAAE,sBAAsB,EAAE,MAAM,YAAY,CAAC;AAEpD,MAAM,gBAAgB,GAAG,oCAAoC,CAAC;AAC9D,MAAM,WAAW,GAAG,sBAAsB,EAAE,CAAC;AAC7C,MAAM,iBAAiB,GAAG,MAAM,CAAC;AAEjC,4EAA4E;AAC5E,4EAA4E;AAC5E,wDAAwD;AACxD,4EAA0E;AAC1E,iEAA+D;AAC/D,mEAAiE;AACjE,wCAAwC;AACxC,MAAM,MAAM,GAAG,CAAC,OAAO,EAAE,MAAM,EAAE,MAAM,CAAU,CAAC;AAElD,MAAM,aAAa,GAAkB,OAAO,CAAC;AAyB7C,IAAI,YAAY,GAA2B,IAAI,CAAC;AAEhD,SAAS,UAAU,GAAoB;IACtC,IAAI,YAAY;QAAE,OAAO,YAAY,CAAC;IACtC,IAAI,CAAC,UAAU,CAAC,WAAW,CAAC,EAAE,CAAC;QAC9B,YAAY,GAAG,EAAE,CAAC;QAClB,OAAO,YAAY,CAAC;IACrB,CAAC;IAED,MAAM,GAAG,GAAG,YAAY,CAAC,WAAW,EAAE,OAAO,CAAC,CAAC;IAC/C,IAAI,CAAC;QACJ,YAAY,GAAG,IAAI,CAAC,KAAK,CAAC,GAAG,CAAoB,CAAC;QAClD,OAAO,YAAY,CAAC;IACrB,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACd,MAAM,OAAO,GAAG,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;QACjE,MAAM,IAAI,KAAK,CAAC,mBAAmB,WAAW,KAAK,OAAO,EAAE,CAAC,CAAC;IAC/D,CAAC;AAAA,CACD;AAED,KAAK,UAAU,SAAS,CAAC,MAAoB,EAA0B;IACtE,OAAO,iBAAiB,CAAC;QACxB,QAAQ,EAAE,UAAU;QACpB,eAAe,EAAE,UAAU,EAAE,CAAC,cAAc;QAC5C,gBAAgB,EAAE,OAAO,CAAC,GAAG,CAAC,gBAAgB;QAC9C,MAAM;KACN,CAAC,CAAC;AAAA,CACH;AAED,KAAK,UAAU,aAAa,CAAC,MAAoB,EAAmB;IACnE,MAAM,MAAM,GAAG,MAAM,SAAS,CAAC,MAAM,CAAC,CAAC;IACvC,IAAI,CAAC,MAAM,EAAE,CAAC;QACb,MAAM,IAAI,KAAK,CACd,uCAAuC;YACvC,eAAe,WAAW,0CAA0C;YACpE,kDAAkD;YAClD,kDAAkD,CAClD,CAAC;IACH,CAAC;IACD,OAAO,MAAM,CAAC;AAAA,CACd;AAED,8EAA8E;AAC9E,4EAA4E;AAC5E,SAAS,YAAY,GAAkB;IACtC,MAAM,GAAG,GAAG,OAAO,CAAC,GAAG,CAAC,cAAc,IAAI,UAAU,EAAE,CAAC,aAAa,CAAC;IACrE,IAAI,OAAO,GAAG,KAAK,QAAQ;QAAE,OAAO,aAAa,CAAC;IAClD,MAAM,KAAK,GAAG,GAAG,CAAC,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC;IACvC,OAAQ,MAA4B,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,KAAsB,CAAC,CAAC,CAAC,aAAa,CAAC;AAAA,CAC9F;AAED,SAAS,cAAc,CAAC,KAAyB,EAAU;IAC1D,IAAI,OAAO,KAAK,KAAK,QAAQ,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,KAAK,CAAC;QAAE,OAAO,CAAC,CAAC;IACnE,OAAO,IAAI,CAAC,GAAG,CAAC,CAAC,EAAE,IAAI,CAAC,GAAG,CAAC,IAAI,CAAC,KAAK,CAAC,KAAK,CAAC,EAAE,EAAE,CAAC,CAAC,CAAC;AAAA,CACpD;AAED,SAAS,eAAe,CAAC,KAAa,EAAiB;IACtD,IAAI,KAAK,GAAG,KAAK,CAAC,IAAI,EAAE,CAAC,WAAW,EAAE,CAAC;IACvC,IAAI,CAAC,KAAK;QAAE,OAAO,IAAI,CAAC;IACxB,IAAI,KAAK,CAAC,UAAU,CAAC,GAAG,CAAC;QAAE,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;IACzD,IAAI,CAAC,KAAK;QAAE,OAAO,IAAI,CAAC;IACxB,IAAI,CAAC;QACJ,MAAM,MAAM,GAAG,KAAK,CAAC,QAAQ,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,IAAI,GAAG,CAAC,WAAW,KAAK,EAAE,CAAC,CAAC;QACpF,KAAK,GAAG,MAAM,CAAC,QAAQ,CAAC;IACzB,CAAC;IAAC,MAAM,CAAC;QACR,KAAK,GAAG,KAAK,CAAC,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,EAAE,KAAK,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,IAAI,EAAE,CAAC;IAClD,CAAC;IACD,KAAK,GAAG,KAAK,CAAC,OAAO,CAAC,YAAY,EAAE,EAAE,CAAC,CAAC;IACxC,OAAO,mCAAmC,CAAC,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,IAAI,CAAC;AAAA,CACtE;AAOD,SAAS,iBAAiB,CAAC,YAAkC,EAAiB;IAC7E,MAAM,OAAO,GAAkB,EAAE,OAAO,EAAE,EAAE,EAAE,OAAO,EAAE,EAAE,EAAE,CAAC;IAC5D,IAAI,CAAC,YAAY,EAAE,MAAM;QAAE,OAAO,OAAO,CAAC;IAC1C,KAAK,MAAM,GAAG,IAAI,YAAY,EAAE,CAAC;QAChC,MAAM,MAAM,GAAG,eAAe,CAAC,GAAG,CAAC,CAAC;QACpC,IAAI,CAAC,MAAM;YAAE,SAAS;QACtB,MAAM,MAAM,GAAG,GAAG,CAAC,IAAI,EAAE,CAAC,UAAU,CAAC,GAAG,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC,CAAC,CAAC,OAAO,CAAC,OAAO,CAAC;QAC9E,IAAI,CAAC,MAAM,CAAC,QAAQ,CAAC,MAAM,CAAC;YAAE,MAAM,CAAC,IAAI,CAAC,MAAM,CAAC,CAAC;IACnD,CAAC;IACD,OAAO,OAAO,CAAC;AAAA,CACf;AAED,SAAS,aAAa,CAAC,QAAgB,EAAE,MAAc,EAAW;IACjE,OAAO,QAAQ,KAAK,MAAM,IAAI,QAAQ,CAAC,QAAQ,CAAC,IAAI,MAAM,EAAE,CAAC,CAAC;AAAA,CAC9D;AAED,iFAAiF;AACjF,mFAAiF;AACjF,+DAA+D;AAC/D,SAAS,mBAAmB,CAAC,GAAW,EAAE,OAAsB,EAAW;IAC1E,IAAI,OAAO,CAAC,OAAO,CAAC,MAAM,KAAK,CAAC,IAAI,OAAO,CAAC,OAAO,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,IAAI,CAAC;IAC9E,IAAI,QAAgB,CAAC;IACrB,IAAI,CAAC;QACJ,QAAQ,GAAG,IAAI,GAAG,CAAC,GAAG,CAAC,CAAC,QAAQ,CAAC,WAAW,EAAE,CAAC;IAChD,CAAC;IAAC,MAAM,CAAC;QACR,OAAO,KAAK,CAAC;IACd,CAAC;IACD,IAAI,OAAO,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,aAAa,CAAC,QAAQ,EAAE,MAAM,CAAC,CAAC;QAAE,OAAO,KAAK,CAAC;IACpF,IAAI,OAAO,CAAC,OAAO,CAAC,MAAM,KAAK,CAAC;QAAE,OAAO,IAAI,CAAC;IAC9C,OAAO,OAAO,CAAC,OAAO,CAAC,IAAI,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC,aAAa,CAAC,QAAQ,EAAE,MAAM,CAAC,CAAC,CAAC;AAAA,CACzE;AAED,iFAAiF;AACjF,+EAA6E;AAC7E,gFAAgF;AAChF,mBAAmB;AACnB,SAAS,gBAAgB,CAAC,KAAa,EAAE,aAA6C,EAAU;IAC/F,IAAI,CAAC,aAAa;QAAE,OAAO,KAAK,CAAC;IACjC,MAAM,KAAK,GAA2B;QACrC,GAAG,EAAE,eAAe;QACpB,IAAI,EAAE,WAAW;QACjB,KAAK,EAAE,YAAY;QACnB,IAAI,EAAE,WAAW;KACjB,CAAC;IACF,MAAM,IAAI,GAAG,KAAK,CAAC,aAAa,CAAC,CAAC;IAClC,OAAO,IAAI,CAAC,CAAC,CAAC,GAAG,KAAK,IAAI,IAAI,EAAE,CAAC,CAAC,CAAC,KAAK,CAAC;AAAA,CACzC;AAED,SAAS,aAAa,CAAC,MAAoB,EAAe;IACzD,MAAM,OAAO,GAAG,WAAW,CAAC,OAAO,CAAC,iBAAiB,CAAC,CAAC;IACvD,OAAO,MAAM,CAAC,CAAC,CAAC,WAAW,CAAC,GAAG,CAAC,CAAC,MAAM,EAAE,OAAO,CAAC,CAAC,CAAC,CAAC,CAAC,OAAO,CAAC;AAAA,CAC7D;AAED,SAAS,YAAY,CAAC,GAAY,EAAU;IAC3C,OAAO,GAAG,YAAY,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,OAAO,CAAC,CAAC,CAAC,MAAM,CAAC,GAAG,CAAC,CAAC;AAAA,CACxD;AAED,SAAS,UAAU,CAClB,OAAqC,EACrC,UAAkB,EAClB,OAAsB,EACM;IAC5B,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,OAAO,CAAC;QAAE,OAAO,EAAE,CAAC;IACvC,MAAM,MAAM,GAA8B,EAAE,CAAC;IAC7C,KAAK,MAAM,IAAI,IAAI,OAAO,EAAE,CAAC;QAC5B,IAAI,CAAC,IAAI,EAAE,GAAG,IAAI,CAAC,mBAAmB,CAAC,IAAI,CAAC,GAAG,EAAE,OAAO,CAAC;YAAE,SAAS;QACpE,MAAM,CAAC,IAAI,CAAC;YACX,KAAK,EAAE,IAAI,CAAC,KAAK,IAAI,UAAU,MAAM,CAAC,MAAM,GAAG,CAAC,EAAE;YAClD,GAAG,EAAE,IAAI,CAAC,GAAG;YACb,OAAO,EAAE,OAAO,IAAI,CAAC,OAAO,KAAK,QAAQ,CAAC,CAAC,CAAC,IAAI,CAAC,OAAO,CAAC,OAAO,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC,IAAI,EAAE,CAAC,CAAC,CAAC,EAAE;SACzF,CAAC,CAAC;QACH,IAAI,MAAM,CAAC,MAAM,IAAI,UAAU;YAAE,MAAM;IACxC,CAAC;IACD,OAAO,MAAM,CAAC;AAAA,CACd;AAED,SAAS,gBAAgB,CAAC,OAAqC,EAAE,OAAsB,EAAsB;IAC5G,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,OAAO,CAAC;QAAE,OAAO,EAAE,CAAC;IACvC,OAAO,OAAO,CAAC,OAAO,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC;QAChC,IAAI,CAAC,IAAI,EAAE,GAAG,IAAI,CAAC,mBAAmB,CAAC,IAAI,CAAC,GAAG,EAAE,OAAO,CAAC;YAAE,OAAO,EAAE,CAAC;QACrE,IAAI,OAAO,IAAI,CAAC,OAAO,KAAK,QAAQ,IAAI,IAAI,CAAC,OAAO,CAAC,IAAI,EAAE,CAAC,MAAM,KAAK,CAAC;YAAE,OAAO,EAAE,CAAC;QACpF,OAAO,CAAC;gBACP,GAAG,EAAE,IAAI,CAAC,GAAG;gBACb,KAAK,EAAE,IAAI,CAAC,KAAK,IAAI,EAAE;gBACvB,OAAO,EAAE,IAAI,CAAC,OAAO;gBACrB,KAAK,EAAE,IAAI;aACX,CAAC,CAAC;IAAA,CACH,CAAC,CAAC;AAAA,CACH;AAED,8EAA8E;AAC9E,+EAA6E;AAC7E,gFAAgF;AAChF,6CAA6C;AAC7C,SAAS,WAAW,CAAC,SAAoC,EAAE,OAAkC,EAAU;IACtG,IAAI,OAAO,SAAS,KAAK,QAAQ,IAAI,SAAS,CAAC,IAAI,EAAE,CAAC,MAAM,GAAG,CAAC;QAAE,OAAO,SAAS,CAAC;IACnF,OAAO,OAAO;SACZ,GAAG,CAAC,CAAC,MAAM,EAAE,EAAE,CAAC;QAChB,IAAI,MAAM,CAAC,OAAO;YAAE,OAAO,GAAG,MAAM,CAAC,OAAO,aAAa,MAAM,CAAC,KAAK,KAAK,MAAM,CAAC,GAAG,GAAG,CAAC;QACxF,OAAO,WAAW,MAAM,CAAC,KAAK,KAAK,MAAM,CAAC,GAAG,GAAG,CAAC;IAAA,CACjD,CAAC;SACD,IAAI,CAAC,MAAM,CAAC,CAAC;AAAA,CACf;AAED,MAAM,UAAU,mBAAmB,GAAY;IAC9C,OAAO,mBAAmB,CAAC;QAC1B,QAAQ,EAAE,UAAU;QACpB,eAAe,EAAE,UAAU,EAAE,CAAC,cAAc;QAC5C,gBAAgB,EAAE,OAAO,CAAC,GAAG,CAAC,gBAAgB;KAC9C,CAAC,CAAC;AAAA,CACH;AAED,MAAM,CAAC,KAAK,UAAU,kBAAkB,CAAC,KAAa,EAAE,OAAO,GAA0B,EAAE,EAA2B;IACrH,MAAM,MAAM,GAAG,MAAM,aAAa,CAAC,OAAO,CAAC,MAAM,CAAC,CAAC;IACnD,MAAM,UAAU,GAAG,cAAc,CAAC,OAAO,CAAC,UAAU,CAAC,CAAC;IACtD,MAAM,OAAO,GAAG,iBAAiB,CAAC,OAAO,CAAC,YAAY,CAAC,CAAC;IACxD,MAAM,KAAK,GAAG,YAAY,EAAE,CAAC;IAC7B,MAAM,IAAI,GAA4B;QACrC,KAAK,EAAE,gBAAgB,CAAC,KAAK,EAAE,OAAO,CAAC,aAAa,CAAC;QACrD,KAAK;QACL,oEAAoE;QACpE,2EAA2E;QAC3E,WAAW,EAAE,IAAI,CAAC,GAAG,CAAC,UAAU,EAAE,EAAE,CAAC;QACrC,6EAA2E;QAC3E,6CAA6C;QAC7C,GAAG,CAAC,KAAK,KAAK,OAAO,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,EAAE,MAAM,EAAE,IAAI,EAAE,CAAC;KAC9C,CAAC;IAEF,MAAM,UAAU,GAAG,eAAe,CAAC,QAAQ,CAAC,EAAE,IAAI,EAAE,KAAK,EAAE,KAAK,EAAE,CAAC,CAAC;IACpE,IAAI,QAAkB,CAAC;IACvB,IAAI,CAAC;QACJ,QAAQ,GAAG,MAAM,KAAK,CAAC,gBAAgB,EAAE;YACxC,MAAM,EAAE,MAAM;YACd,OAAO,EAAE;gBACR,eAAe,EAAE,UAAU,MAAM,EAAE;gBACnC,cAAc,EAAE,kBAAkB;aAClC;YACD,IAAI,EAAE,IAAI,CAAC,SAAS,CAAC,IAAI,CAAC;YAC1B,MAAM,EAAE,aAAa,CAAC,OAAO,CAAC,MAAM,CAAC;SACrC,CAAC,CAAC;IACJ,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACd,MAAM,OAAO,GAAG,YAAY,CAAC,GAAG,CAAC,CAAC;QAClC,MAAM,eAAe,GAAG,gBAAgB,CAAC,OAAO,EAAE,MAAM,CAAC,CAAC;QAC1D,IAAI,eAAe,CAAC,WAAW,EAAE,CAAC,QAAQ,CAAC,OAAO,CAAC;YAAE,eAAe,CAAC,WAAW,CAAC,UAAU,EAAE,CAAC,CAAC,CAAC;;YAC3F,eAAe,CAAC,QAAQ,CAAC,UAAU,EAAE,eAAe,CAAC,CAAC;QAC3D,IAAI,eAAe,KAAK,OAAO;YAAE,MAAM,GAAG,CAAC;QAC3C,MAAM,aAAa,GAAG,IAAI,KAAK,CAAC,eAAe,CAAC,CAAC;QACjD,IAAI,GAAG,YAAY,KAAK;YAAE,aAAa,CAAC,IAAI,GAAG,GAAG,CAAC,IAAI,CAAC;QACxD,MAAM,aAAa,CAAC;IACrB,CAAC;IAED,IAAI,CAAC,QAAQ,CAAC,EAAE,EAAE,CAAC;QAClB,eAAe,CAAC,WAAW,CAAC,UAAU,EAAE,QAAQ,CAAC,MAAM,CAAC,CAAC;QACzD,MAAM,SAAS,GAAG,gBAAgB,CAAC,MAAM,QAAQ,CAAC,IAAI,EAAE,EAAE,MAAM,CAAC,CAAC;QAClE,MAAM,IAAI,KAAK,CAAC,sBAAsB,QAAQ,CAAC,MAAM,KAAK,SAAS,CAAC,KAAK,CAAC,CAAC,EAAE,GAAG,CAAC,EAAE,CAAC,CAAC;IACtF,CAAC;IAED,IAAI,IAAsB,CAAC;IAC3B,IAAI,CAAC;QACJ,IAAI,GAAG,MAAM,QAAQ,CAAC,IAAI,EAAsB,CAAC;IAClD,CAAC;IAAC,OAAO,GAAG,EAAE,CAAC;QACd,eAAe,CAAC,WAAW,CAAC,UAAU,EAAE,QAAQ,CAAC,MAAM,CAAC,CAAC;QACzD,MAAM,IAAI,KAAK,CAAC,uCAAuC,YAAY,CAAC,GAAG,CAAC,EAAE,CAAC,CAAC;IAC7E,CAAC;IAED,eAAe,CAAC,WAAW,CAAC,UAAU,EAAE,QAAQ,CAAC,MAAM,CAAC,CAAC;IACzD,MAAM,OAAO,GAAG,UAAU,CAAC,IAAI,CAAC,OAAO,EAAE,UAAU,EAAE,OAAO,CAAC,CAAC;IAC9D,MAAM,MAAM,GAAmB;QAC9B,MAAM,EAAE,WAAW,CAAC,IAAI,CAAC,MAAM,EAAE,OAAO,CAAC;QACzC,OAAO;KACP,CAAC;IACF,IAAI,OAAO,CAAC,cAAc,EAAE,CAAC;QAC5B,MAAM,aAAa,GAAG,gBAAgB,CAAC,IAAI,CAAC,OAAO,EAAE,OAAO,CAAC,CAAC;QAC9D,IAAI,aAAa,CAAC,MAAM,GAAG,CAAC;YAAE,MAAM,CAAC,aAAa,GAAG,aAAa,CAAC;IACpE,CAAC;IACD,OAAO,MAAM,CAAC;AAAA,CACd","sourcesContent":["import { existsSync, readFileSync } from \"node:fs\";\nimport { activityMonitor } from \"./activity.ts\";\nimport type { ExtractedContent } from \"./extract.ts\";\nimport type { SearchOptions, SearchResponse } from \"./perplexity.ts\";\nimport { hasCredentialSource, redactCredential, resolveCredential } from \"./credential-source.ts\";\nimport { getWebSearchConfigPath } from \"./utils.ts\";\n\nconst SERPDIVE_API_URL = \"https://api.serpdive.com/v1/search\";\nconst CONFIG_PATH = getWebSearchConfigPath();\nconst SEARCH_TIMEOUT_MS = 60_000;\n\n// Retrieval depth. The default is deliberately the free tier: this provider\n// ships inside someone else's framework, and installing it must never start\n// spending on a user's behalf without them choosing it.\n//   krill — free, fair use, no synthesized answer (see buildAnswer below)\n//   mako  — the fact-carrying sentences of each page, 1 credit\n//   moby  — the full readable content of every page, 1.5 credits\n// Pricing: https://serpdive.com/pricing\nconst MODELS = [\"krill\", \"mako\", \"moby\"] as const;\ntype SerpdiveModel = (typeof MODELS)[number];\nconst DEFAULT_MODEL: SerpdiveModel = \"krill\";\n\ninterface WebSearchConfig {\n\tserpdiveApiKey?: unknown;\n\tserpdiveModel?: unknown;\n}\n\ninterface SerpdiveResult {\n\turl?: string;\n\ttitle?: string | null;\n\tdate?: string;\n\tcontent?: string;\n}\n\ninterface SerpdiveResponse {\n\tquery?: string;\n\tmodel?: string;\n\tanswer?: string | null;\n\tresults?: SerpdiveResult[];\n}\n\ninterface SerpdiveSearchOptions extends SearchOptions {\n\tincludeContent?: boolean;\n}\n\nlet cachedConfig: WebSearchConfig | null = null;\n\nfunction loadConfig(): WebSearchConfig {\n\tif (cachedConfig) return cachedConfig;\n\tif (!existsSync(CONFIG_PATH)) {\n\t\tcachedConfig = {};\n\t\treturn cachedConfig;\n\t}\n\n\tconst raw = readFileSync(CONFIG_PATH, \"utf-8\");\n\ttry {\n\t\tcachedConfig = JSON.parse(raw) as WebSearchConfig;\n\t\treturn cachedConfig;\n\t} catch (err) {\n\t\tconst message = err instanceof Error ? err.message : String(err);\n\t\tthrow new Error(`Failed to parse ${CONFIG_PATH}: ${message}`);\n\t}\n}\n\nasync function getApiKey(signal?: AbortSignal): Promise<string | null> {\n\treturn resolveCredential({\n\t\tprovider: \"SERPdive\",\n\t\tconfiguredValue: loadConfig().serpdiveApiKey,\n\t\tenvironmentValue: process.env.SERPDIVE_API_KEY,\n\t\tsignal,\n\t});\n}\n\nasync function requireApiKey(signal?: AbortSignal): Promise<string> {\n\tconst apiKey = await getApiKey(signal);\n\tif (!apiKey) {\n\t\tthrow new Error(\n\t\t\t\"SERPdive API key not found. Either:\\n\" +\n\t\t\t`  1. Create ${CONFIG_PATH} with { \"serpdiveApiKey\": \"your-key\" }\\n` +\n\t\t\t\"  2. Set SERPDIVE_API_KEY environment variable\\n\" +\n\t\t\t\"Get a key at https://serpdive.com/dashboard/keys\",\n\t\t);\n\t}\n\treturn apiKey;\n}\n\n// An unknown value falls back to the free default rather than failing: a typo\n// in a config file must not cost the user money, and must not break search.\nfunction resolveModel(): SerpdiveModel {\n\tconst raw = process.env.SERPDIVE_MODEL ?? loadConfig().serpdiveModel;\n\tif (typeof raw !== \"string\") return DEFAULT_MODEL;\n\tconst value = raw.trim().toLowerCase();\n\treturn (MODELS as readonly string[]).includes(value) ? value as SerpdiveModel : DEFAULT_MODEL;\n}\n\nfunction normalizeCount(value: number | undefined): number {\n\tif (typeof value !== \"number\" || !Number.isFinite(value)) return 5;\n\treturn Math.max(1, Math.min(Math.floor(value), 20));\n}\n\nfunction normalizeDomain(value: string): string | null {\n\tlet input = value.trim().toLowerCase();\n\tif (!input) return null;\n\tif (input.startsWith(\"-\")) input = input.slice(1).trim();\n\tif (!input) return null;\n\ttry {\n\t\tconst parsed = input.includes(\"://\") ? new URL(input) : new URL(`https://${input}`);\n\t\tinput = parsed.hostname;\n\t} catch {\n\t\tinput = input.split(\"/\")[0]?.split(\":\")[0] ?? \"\";\n\t}\n\tinput = input.replace(/^\\.+|\\.+$/g, \"\");\n\treturn /^[a-z0-9][a-z0-9.-]*\\.[a-z]{2,}$/i.test(input) ? input : null;\n}\n\ninterface DomainFilters {\n\tinclude: string[];\n\texclude: string[];\n}\n\nfunction parseDomainFilter(domainFilter: string[] | undefined): DomainFilters {\n\tconst filters: DomainFilters = { include: [], exclude: [] };\n\tif (!domainFilter?.length) return filters;\n\tfor (const raw of domainFilter) {\n\t\tconst domain = normalizeDomain(raw);\n\t\tif (!domain) continue;\n\t\tconst target = raw.trim().startsWith(\"-\") ? filters.exclude : filters.include;\n\t\tif (!target.includes(domain)) target.push(domain);\n\t}\n\treturn filters;\n}\n\nfunction domainMatches(hostname: string, domain: string): boolean {\n\treturn hostname === domain || hostname.endsWith(`.${domain}`);\n}\n\n// SERPdive exposes no include/exclude domain parameter, so the filter is applied\n// here, on what came back. It can therefore only ever narrow a page of results —\n// it cannot ask the engine for more pages from a given domain.\nfunction passesDomainFilters(url: string, filters: DomainFilters): boolean {\n\tif (filters.include.length === 0 && filters.exclude.length === 0) return true;\n\tlet hostname: string;\n\ttry {\n\t\thostname = new URL(url).hostname.toLowerCase();\n\t} catch {\n\t\treturn false;\n\t}\n\tif (filters.exclude.some((domain) => domainMatches(hostname, domain))) return false;\n\tif (filters.include.length === 0) return true;\n\treturn filters.include.some((domain) => domainMatches(hostname, domain));\n}\n\n// SERPdive has NO time-range parameter. Recency is expressed inside the question\n// and read by the engine, which biases ranking toward recent pages — it is a\n// hint, never a guaranteed freshness filter, and results outside the window can\n// still come back.\nfunction applyRecencyHint(query: string, recencyFilter: SearchOptions[\"recencyFilter\"]): string {\n\tif (!recencyFilter) return query;\n\tconst hints: Record<string, string> = {\n\t\tday: \"past 24 hours\",\n\t\tweek: \"past week\",\n\t\tmonth: \"past month\",\n\t\tyear: \"past year\",\n\t};\n\tconst hint = hints[recencyFilter];\n\treturn hint ? `${query} ${hint}` : query;\n}\n\nfunction requestSignal(signal?: AbortSignal): AbortSignal {\n\tconst timeout = AbortSignal.timeout(SEARCH_TIMEOUT_MS);\n\treturn signal ? AbortSignal.any([signal, timeout]) : timeout;\n}\n\nfunction errorMessage(err: unknown): string {\n\treturn err instanceof Error ? err.message : String(err);\n}\n\nfunction mapResults(\n\tresults: SerpdiveResult[] | undefined,\n\tnumResults: number,\n\tfilters: DomainFilters,\n): SearchResponse[\"results\"] {\n\tif (!Array.isArray(results)) return [];\n\tconst mapped: SearchResponse[\"results\"] = [];\n\tfor (const item of results) {\n\t\tif (!item?.url || !passesDomainFilters(item.url, filters)) continue;\n\t\tmapped.push({\n\t\t\ttitle: item.title || `Source ${mapped.length + 1}`,\n\t\t\turl: item.url,\n\t\t\tsnippet: typeof item.content === \"string\" ? item.content.replace(/\\s+/g, \" \").trim() : \"\",\n\t\t});\n\t\tif (mapped.length >= numResults) break;\n\t}\n\treturn mapped;\n}\n\nfunction mapInlineContent(results: SerpdiveResult[] | undefined, filters: DomainFilters): ExtractedContent[] {\n\tif (!Array.isArray(results)) return [];\n\treturn results.flatMap((item) => {\n\t\tif (!item?.url || !passesDomainFilters(item.url, filters)) return [];\n\t\tif (typeof item.content !== \"string\" || item.content.trim().length === 0) return [];\n\t\treturn [{\n\t\t\turl: item.url,\n\t\t\ttitle: item.title || \"\",\n\t\t\tcontent: item.content,\n\t\t\terror: null,\n\t\t}];\n\t});\n}\n\n// The free krill tier returns extracted content but no synthesized answer, so\n// one is assembled from the sources — the same shape brave.ts and searxng.ts\n// produce for providers that do not synthesize. mako and moby ask the API for a\n// real answer and use it when it comes back.\nfunction buildAnswer(apiAnswer: string | null | undefined, results: SearchResponse[\"results\"]): string {\n\tif (typeof apiAnswer === \"string\" && apiAnswer.trim().length > 0) return apiAnswer;\n\treturn results\n\t\t.map((result) => {\n\t\t\tif (result.snippet) return `${result.snippet}\\nSource: ${result.title} (${result.url})`;\n\t\t\treturn `Source: ${result.title} (${result.url})`;\n\t\t})\n\t\t.join(\"\\n\\n\");\n}\n\nexport function isSerpdiveAvailable(): boolean {\n\treturn hasCredentialSource({\n\t\tprovider: \"SERPdive\",\n\t\tconfiguredValue: loadConfig().serpdiveApiKey,\n\t\tenvironmentValue: process.env.SERPDIVE_API_KEY,\n\t});\n}\n\nexport async function searchWithSerpdive(query: string, options: SerpdiveSearchOptions = {}): Promise<SearchResponse> {\n\tconst apiKey = await requireApiKey(options.signal);\n\tconst numResults = normalizeCount(options.numResults);\n\tconst filters = parseDomainFilter(options.domainFilter);\n\tconst model = resolveModel();\n\tconst body: Record<string, unknown> = {\n\t\tquery: applyRecencyHint(query, options.recencyFilter),\n\t\tmodel,\n\t\t// max_results is a CAP, never a minimum: the engine returns what it\n\t\t// judges relevant, up to this many. Asking for more does not produce more.\n\t\tmax_results: Math.min(numResults, 10),\n\t\t// krill has no answer synthesis — asking for one there is silently ignored\n\t\t// by the API, so it is not asked for at all.\n\t\t...(model === \"krill\" ? {} : { answer: true }),\n\t};\n\n\tconst activityId = activityMonitor.logStart({ type: \"api\", query });\n\tlet response: Response;\n\ttry {\n\t\tresponse = await fetch(SERPDIVE_API_URL, {\n\t\t\tmethod: \"POST\",\n\t\t\theaders: {\n\t\t\t\t\"Authorization\": `Bearer ${apiKey}`,\n\t\t\t\t\"Content-Type\": \"application/json\",\n\t\t\t},\n\t\t\tbody: JSON.stringify(body),\n\t\t\tsignal: requestSignal(options.signal),\n\t\t});\n\t} catch (err) {\n\t\tconst message = errorMessage(err);\n\t\tconst redactedMessage = redactCredential(message, apiKey);\n\t\tif (redactedMessage.toLowerCase().includes(\"abort\")) activityMonitor.logComplete(activityId, 0);\n\t\telse activityMonitor.logError(activityId, redactedMessage);\n\t\tif (redactedMessage === message) throw err;\n\t\tconst redactedError = new Error(redactedMessage);\n\t\tif (err instanceof Error) redactedError.name = err.name;\n\t\tthrow redactedError;\n\t}\n\n\tif (!response.ok) {\n\t\tactivityMonitor.logComplete(activityId, response.status);\n\t\tconst errorText = redactCredential(await response.text(), apiKey);\n\t\tthrow new Error(`SERPdive API error ${response.status}: ${errorText.slice(0, 300)}`);\n\t}\n\n\tlet data: SerpdiveResponse;\n\ttry {\n\t\tdata = await response.json() as SerpdiveResponse;\n\t} catch (err) {\n\t\tactivityMonitor.logComplete(activityId, response.status);\n\t\tthrow new Error(`SERPdive API returned invalid JSON: ${errorMessage(err)}`);\n\t}\n\n\tactivityMonitor.logComplete(activityId, response.status);\n\tconst results = mapResults(data.results, numResults, filters);\n\tconst result: SearchResponse = {\n\t\tanswer: buildAnswer(data.answer, results),\n\t\tresults,\n\t};\n\tif (options.includeContent) {\n\t\tconst inlineContent = mapInlineContent(data.results, filters);\n\t\tif (inlineContent.length > 0) result.inlineContent = inlineContent;\n\t}\n\treturn result;\n}\n"]}

package/dist/humain/web-access/vendor/source-check.d.ts

@@ -0,0 +1,99 @@
+import type { SearchResult } from "./perplexity.ts";
+import type { ExtractedContent } from "./extract.ts";
+export type SourceQuality = "official_docs" | "vendor_docs" | "repo_issue" | "blog" | "forum" | "news" | "unknown";
+export type ClaimStatus = "supported" | "contradicted" | "unclear" | "missing-evidence";
+export type RecencyFilter = "day" | "week" | "month" | "year";
+export interface ResearchSource {
+    rank: number;
+    url: string;
+    title: string;
+    snippet?: string;
+    fetch_timestamp?: number;
+    content_hash?: string;
+    quality: SourceQuality;
+    fetched?: boolean;
+    fetch_error?: string;
+}
+export interface ResearchPassage {
+    passage_id: string;
+    source_url: string;
+    source_rank: number;
+    text: string;
+    extraction_span?: {
+        start: number;
+        end: number;
+    };
+    content_hash?: string;
+}
+export interface ClaimAssessment {
+    claim: string;
+    status: ClaimStatus;
+    supporting_passages: string[];
+    contradicting_passages: string[];
+    rationale: string;
+    confidence: number;
+}
+export interface ResearchArtifact {
+    id: string;
+    type: "research";
+    timestamp: number;
+    query: string;
+    sources: ResearchSource[];
+    passages: ResearchPassage[];
+    claims?: ClaimAssessment[];
+    provider?: string;
+    summary?: string;
+    content_hash?: string;
+    filters?: {
+        recency?: RecencyFilter;
+        domain_include?: string[];
+        domain_exclude?: string[];
+    };
+    errors?: Array<{
+        query: string;
+        error: string;
+    }>;
+}
+export interface ResearchSearchRequest {
+    query: string;
+    numResults: number;
+    recencyFilter?: RecencyFilter;
+    domainFilter?: string[];
+}
+export interface ResearchSearchResult {
+    url: string;
+    title: string;
+    snippet: string;
+    rank: number;
+}
+export interface ResearchSearchResponse {
+    provider: string;
+    results: ResearchSearchResult[];
+    summary?: string;
+}
+export interface ResearchProvider {
+    name: string;
+    search(req: ResearchSearchRequest): Promise<ResearchSearchResponse>;
+}
+export declare function classifySource(url: string): SourceQuality;
+export declare function hashContent(text: string): string;
+export declare function buildPassages(sources: ResearchSource[], fetched?: ExtractedContent[], hint?: string): ResearchPassage[];
+export declare function assessClaim(claim: string, passages: ResearchPassage[]): ClaimAssessment;
+interface RankedSearchResult extends SearchResult {
+    rank?: number;
+}
+export interface BuildArtifactInput {
+    query: string;
+    provider?: string;
+    summary?: string;
+    results: RankedSearchResult[];
+    fetched?: ExtractedContent[];
+    recency?: RecencyFilter;
+    domainFilter?: string[];
+}
+export declare function buildResearchArtifact(input: BuildArtifactInput): ResearchArtifact;
+export declare function withClaimAssessment(artifact: ResearchArtifact, claims: string[]): ResearchArtifact;
+export declare function storeResearchArtifact(artifact: ResearchArtifact): void;
+export declare function getResearchArtifact(id: string): ResearchArtifact | null;
+export {};
+//# sourceMappingURL=source-check.d.ts.map

package/dist/humain/web-access/vendor/source-check.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"source-check.d.ts","sourceRoot":"","sources":["../../../../src/humain/web-access/vendor/source-check.ts"],"names":[],"mappings":"AAGA,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,iBAAiB,CAAC;AACpD,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,cAAc,CAAC;AAErD,MAAM,MAAM,aAAa,GAAG,eAAe,GAAG,aAAa,GAAG,YAAY,GAAG,MAAM,GAAG,OAAO,GAAG,MAAM,GAAG,SAAS,CAAC;AACnH,MAAM,MAAM,WAAW,GAAG,WAAW,GAAG,cAAc,GAAG,SAAS,GAAG,kBAAkB,CAAC;AACxF,MAAM,MAAM,aAAa,GAAG,KAAK,GAAG,MAAM,GAAG,OAAO,GAAG,MAAM,CAAC;AAE9D,MAAM,WAAW,cAAc;IAC9B,IAAI,EAAE,MAAM,CAAC;IACb,GAAG,EAAE,MAAM,CAAC;IACZ,KAAK,EAAE,MAAM,CAAC;IACd,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,eAAe,CAAC,EAAE,MAAM,CAAC;IACzB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,OAAO,EAAE,aAAa,CAAC;IACvB,OAAO,CAAC,EAAE,OAAO,CAAC;IAClB,WAAW,CAAC,EAAE,MAAM,CAAC;CACrB;AAED,MAAM,WAAW,eAAe;IAC/B,UAAU,EAAE,MAAM,CAAC;IACnB,UAAU,EAAE,MAAM,CAAC;IACnB,WAAW,EAAE,MAAM,CAAC;IACpB,IAAI,EAAE,MAAM,CAAC;IACb,eAAe,CAAC,EAAE;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,GAAG,EAAE,MAAM,CAAA;KAAE,CAAC;IACjD,YAAY,CAAC,EAAE,MAAM,CAAC;CACtB;AAED,MAAM,WAAW,eAAe;IAC/B,KAAK,EAAE,MAAM,CAAC;IACd,MAAM,EAAE,WAAW,CAAC;IACpB,mBAAmB,EAAE,MAAM,EAAE,CAAC;IAC9B,sBAAsB,EAAE,MAAM,EAAE,CAAC;IACjC,SAAS,EAAE,MAAM,CAAC;IAClB,UAAU,EAAE,MAAM,CAAC;CACnB;AAED,MAAM,WAAW,gBAAgB;IAChC,EAAE,EAAE,MAAM,CAAC;IACX,IAAI,EAAE,UAAU,CAAC;IACjB,SAAS,EAAE,MAAM,CAAC;IAClB,KAAK,EAAE,MAAM,CAAC;IACd,OAAO,EAAE,cAAc,EAAE,CAAC;IAC1B,QAAQ,EAAE,eAAe,EAAE,CAAC;IAC5B,MAAM,CAAC,EAAE,eAAe,EAAE,CAAC;IAC3B,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,YAAY,CAAC,EAAE,MAAM,CAAC;IACtB,OAAO,CAAC,EAAE;QAAE,OAAO,CAAC,EAAE,aAAa,CAAC;QAAC,cAAc,CAAC,EAAE,MAAM,EAAE,CAAC;QAAC,cAAc,CAAC,EAAE,MAAM,EAAE,CAAA;KAAE,CAAC;IAC5F,MAAM,CAAC,EAAE,KAAK,CAAC;QAAE,KAAK,EAAE,MAAM,CAAC;QAAC,KAAK,EAAE,MAAM,CAAA;KAAE,CAAC,CAAC;CACjD;AAED,MAAM,WAAW,qBAAqB;IACrC,KAAK,EAAE,MAAM,CAAC;IACd,UAAU,EAAE,MAAM,CAAC;IACnB,aAAa,CAAC,EAAE,aAAa,CAAC;IAC9B,YAAY,CAAC,EAAE,MAAM,EAAE,CAAC;CACxB;AACD,MAAM,WAAW,oBAAoB;IAAG,GAAG,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,MAAM,CAAC;IAAC,IAAI,EAAE,MAAM,CAAA;CAAE;AACnG,MAAM,WAAW,sBAAsB;IAAG,QAAQ,EAAE,MAAM,CAAC;IAAC,OAAO,EAAE,oBAAoB,EAAE,CAAC;IAAC,OAAO,CAAC,EAAE,MAAM,CAAA;CAAE;AAC/G,MAAM,WAAW,gBAAgB;IAAG,IAAI,EAAE,MAAM,CAAC;IAAC,MAAM,CAAC,GAAG,EAAE,qBAAqB,GAAG,OAAO,CAAC,sBAAsB,CAAC,CAAA;CAAE;AAavH,wBAAgB,cAAc,CAAC,GAAG,EAAE,MAAM,GAAG,aAAa,CAiBzD;AAED,wBAAgB,WAAW,CAAC,IAAI,EAAE,MAAM,GAAG,MAAM,CAEhD;AAiCD,wBAAgB,aAAa,CAAC,OAAO,EAAE,cAAc,EAAE,EAAE,OAAO,GAAE,gBAAgB,EAAO,EAAE,IAAI,SAAK,GAAG,eAAe,EAAE,CA6BvH;AAwBD,wBAAgB,WAAW,CAAC,KAAK,EAAE,MAAM,EAAE,QAAQ,EAAE,eAAe,EAAE,GAAG,eAAe,CA0BvF;AAED,UAAU,kBAAmB,SAAQ,YAAY;IAChD,IAAI,CAAC,EAAE,MAAM,CAAC;CACd;AAED,MAAM,WAAW,kBAAkB;IAClC,KAAK,EAAE,MAAM,CAAC;IACd,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,OAAO,CAAC,EAAE,MAAM,CAAC;IACjB,OAAO,EAAE,kBAAkB,EAAE,CAAC;IAC9B,OAAO,CAAC,EAAE,gBAAgB,EAAE,CAAC;IAC7B,OAAO,CAAC,EAAE,aAAa,CAAC;IACxB,YAAY,CAAC,EAAE,MAAM,EAAE,CAAC;CACxB;AAED,wBAAgB,qBAAqB,CAAC,KAAK,EAAE,kBAAkB,GAAG,gBAAgB,CAyCjF;AAED,wBAAgB,mBAAmB,CAAC,QAAQ,EAAE,gBAAgB,EAAE,MAAM,EAAE,MAAM,EAAE,GAAG,gBAAgB,CAElG;AAED,wBAAgB,qBAAqB,CAAC,QAAQ,EAAE,gBAAgB,GAAG,IAAI,CAGtE;AAED,wBAAgB,mBAAmB,CAAC,EAAE,EAAE,MAAM,GAAG,gBAAgB,GAAG,IAAI,CAIvE","sourcesContent":["// Structured source checking and machine-readable research artifacts.\nimport { createHash } from \"node:crypto\";\nimport { generateId, getResult, storeResult } from \"./storage.ts\";\nimport type { SearchResult } from \"./perplexity.ts\";\nimport type { ExtractedContent } from \"./extract.ts\";\n\nexport type SourceQuality = \"official_docs\" | \"vendor_docs\" | \"repo_issue\" | \"blog\" | \"forum\" | \"news\" | \"unknown\";\nexport type ClaimStatus = \"supported\" | \"contradicted\" | \"unclear\" | \"missing-evidence\";\nexport type RecencyFilter = \"day\" | \"week\" | \"month\" | \"year\";\n\nexport interface ResearchSource {\n\trank: number;\n\turl: string;\n\ttitle: string;\n\tsnippet?: string;\n\tfetch_timestamp?: number;\n\tcontent_hash?: string;\n\tquality: SourceQuality;\n\tfetched?: boolean;\n\tfetch_error?: string;\n}\n\nexport interface ResearchPassage {\n\tpassage_id: string;\n\tsource_url: string;\n\tsource_rank: number;\n\ttext: string;\n\textraction_span?: { start: number; end: number };\n\tcontent_hash?: string;\n}\n\nexport interface ClaimAssessment {\n\tclaim: string;\n\tstatus: ClaimStatus;\n\tsupporting_passages: string[];\n\tcontradicting_passages: string[];\n\trationale: string;\n\tconfidence: number;\n}\n\nexport interface ResearchArtifact {\n\tid: string;\n\ttype: \"research\";\n\ttimestamp: number;\n\tquery: string;\n\tsources: ResearchSource[];\n\tpassages: ResearchPassage[];\n\tclaims?: ClaimAssessment[];\n\tprovider?: string;\n\tsummary?: string;\n\tcontent_hash?: string;\n\tfilters?: { recency?: RecencyFilter; domain_include?: string[]; domain_exclude?: string[] };\n\terrors?: Array<{ query: string; error: string }>;\n}\n\nexport interface ResearchSearchRequest {\n\tquery: string;\n\tnumResults: number;\n\trecencyFilter?: RecencyFilter;\n\tdomainFilter?: string[];\n}\nexport interface ResearchSearchResult { url: string; title: string; snippet: string; rank: number }\nexport interface ResearchSearchResponse { provider: string; results: ResearchSearchResult[]; summary?: string }\nexport interface ResearchProvider { name: string; search(req: ResearchSearchRequest): Promise<ResearchSearchResponse> }\n\nconst OFFICIAL_DOCS_HOSTS = /^(developers\\.|docs\\.|learn\\.|reference\\.)|\\.github\\.io$/i;\nconst OFFICIAL_DOCS_PATHS = /\\/(docs?|reference)(\\/|\\b)/i;\nconst VENDOR_DOCS_PATHS = /\\/(documentation|docs?)\\//i;\nconst REPO_ISSUE_PATHS = /\\/(issues|pull|pulls)\\//i;\nconst BLOG_HOSTS = /(medium\\.com|substack\\.com|dev\\.to|hashnode\\.)/i;\nconst BLOG_PATHS = /\\/blogs?\\//i;\nconst FORUM_HOSTS = /(stackoverflow\\.com|serverfault\\.com|superuser\\.com|discourse\\.|community\\.)/i;\nconst FORUM_PATHS = /\\/(forum|forums|threads)\\//i;\nconst NEWS_HOSTS = /(reuters\\.com|bloomberg\\.com|techcrunch\\.com|theverge\\.com|arstechnica\\.com|wired\\.com|cnet\\.com|zdnet\\.com)/i;\nconst NEWS_PATHS = /\\/news(\\/|$)/i;\n\nexport function classifySource(url: string): SourceQuality {\n\tlet host = \"\";\n\tlet path = \"\";\n\ttry {\n\t\tconst parsed = new URL(url);\n\t\thost = parsed.hostname;\n\t\tpath = parsed.pathname;\n\t} catch {\n\t\treturn \"unknown\";\n\t}\n\tif (REPO_ISSUE_PATHS.test(path)) return \"repo_issue\";\n\tif (OFFICIAL_DOCS_HOSTS.test(host) || OFFICIAL_DOCS_PATHS.test(path)) return \"official_docs\";\n\tif (VENDOR_DOCS_PATHS.test(path)) return \"vendor_docs\";\n\tif (NEWS_HOSTS.test(host) || NEWS_PATHS.test(path)) return \"news\";\n\tif (FORUM_HOSTS.test(host) || FORUM_PATHS.test(path)) return \"forum\";\n\tif (BLOG_HOSTS.test(host) || BLOG_PATHS.test(path)) return \"blog\";\n\treturn \"unknown\";\n}\n\nexport function hashContent(text: string): string {\n\treturn `sha256:${createHash(\"sha256\").update(text, \"utf8\").digest(\"hex\")}`;\n}\n\ninterface Span { text: string; start: number; end: number }\n\nfunction tokenize(value: string): string[] {\n\treturn [...new Set(value.toLowerCase().split(/[^a-z0-9]+/).filter((term) => term.length > 3))];\n}\n\nfunction extractRelevantSpans(content: string, hint: string): Span[] {\n\tconst sentences: Span[] = [];\n\tconst sentencePattern = /[^.!?]+(?:[.!?]+(?=\\s|$)|$)/g;\n\tfor (const match of content.matchAll(sentencePattern)) {\n\t\tconst raw = match[0];\n\t\tconst text = raw.trim();\n\t\tif (text.length > 0 && text.length <= 400) {\n\t\t\tconst start = (match.index ?? 0) + raw.indexOf(text);\n\t\t\tsentences.push({ text, start, end: start + text.length });\n\t\t}\n\t}\n\tconst terms = tokenize(hint);\n\tif (terms.length === 0) return [];\n\treturn sentences\n\t\t.map((sentence, index) => ({ sentence, index, score: terms.filter((term) => sentence.text.toLowerCase().includes(term)).length }))\n\t\t.filter((item) => item.score > 0)\n\t\t.sort((a, b) => b.score - a.score || a.index - b.index)\n\t\t.slice(0, 3)\n\t\t.map(({ sentence }) => sentence);\n}\n\nfunction passageId(sourceRank: number, index: number): string {\n\treturn `p-${sourceRank}-${index}`;\n}\n\nexport function buildPassages(sources: ResearchSource[], fetched: ExtractedContent[] = [], hint = \"\"): ResearchPassage[] {\n\tconst passages: ResearchPassage[] = [];\n\tconst fetchedByUrl = new Map(fetched.map((item) => [item.url, item]));\n\tfor (const source of sources) {\n\t\tif (source.snippet) {\n\t\t\tpassages.push({\n\t\t\t\tpassage_id: passageId(source.rank, 0),\n\t\t\t\tsource_url: source.url,\n\t\t\t\tsource_rank: source.rank,\n\t\t\t\ttext: source.snippet,\n\t\t\t\tcontent_hash: hashContent(source.snippet),\n\t\t\t});\n\t\t}\n\t\tconst page = fetchedByUrl.get(source.url);\n\t\tif (page && !page.error && page.content) {\n\t\t\tconst passageHint = source.snippet?.trim() || hint;\n\t\t\tfor (const [index, span] of extractRelevantSpans(page.content, passageHint).entries()) {\n\t\t\t\tpassages.push({\n\t\t\t\t\tpassage_id: passageId(source.rank, index + 1),\n\t\t\t\t\tsource_url: source.url,\n\t\t\t\t\tsource_rank: source.rank,\n\t\t\t\t\ttext: span.text,\n\t\t\t\t\textraction_span: { start: span.start, end: span.end },\n\t\t\t\t\tcontent_hash: hashContent(span.text),\n\t\t\t\t});\n\t\t\t}\n\t\t}\n\t}\n\treturn passages;\n}\n\nconst CONTRADICTION_MARKERS = [\"not true\", \"false\", \"incorrect\", \"debunked\", \"retracted\", \"no longer\", \"never\", \"denied\", \"contrary\", \"misleading\"];\nconst SUPPORT_MARKERS = [\"yes\", \"true\", \"correct\", \"confirmed\", \"according to\", \"shows that\", \"demonstrates\", \"reported\", \"verified\", \"established\"];\n\nfunction containsPhrase(value: string, phrase: string): boolean {\n\tconst escaped = phrase.trim().toLowerCase().split(/\\s+/).map((term) => term.replace(/[.*+?^${}()|[\\]\\\\]/g, \"\\\\$&\")).join(\"\\\\s+\");\n\treturn new RegExp(`(?:^|[^a-z0-9])${escaped}(?=$|[^a-z0-9])`, \"i\").test(value);\n}\n\nfunction markerIsNegated(value: string, marker: string): boolean {\n\tconst escaped = marker.trim().toLowerCase().split(/\\s+/).map((term) => term.replace(/[.*+?^${}()|[\\]\\\\]/g, \"\\\\$&\")).join(\"\\\\s+\");\n\tconst markerPattern = new RegExp(`(?:^|[^a-z0-9])${escaped}(?=$|[^a-z0-9])`, \"i\");\n\tconst match = markerPattern.exec(value);\n\tif (!match || match.index === undefined) return false;\n\tconst matchedMarker = match[0].replace(/^[^a-z0-9]+/i, \"\");\n\tconst beforeMarker = value.slice(0, match.index + match[0].length - matchedMarker.length);\n\treturn /(?:^|[^a-z0-9])(?:not|no|never|without)\\s+$/i.test(beforeMarker);\n}\n\nfunction hasPolarityMarker(value: string, markers: string[], allowNegated = false): boolean {\n\treturn markers.some((marker) => containsPhrase(value, marker) && (allowNegated || !markerIsNegated(value, marker)));\n}\n\nexport function assessClaim(claim: string, passages: ResearchPassage[]): ClaimAssessment {\n\tconst terms = tokenize(claim);\n\tif (terms.length === 0 || passages.length === 0) {\n\t\treturn { claim, status: \"missing-evidence\", supporting_passages: [], contradicting_passages: [], rationale: \"No passages available that discuss the claim's terms.\", confidence: 0.2 };\n\t}\n\tconst supporting: string[] = [];\n\tconst contradicting: string[] = [];\n\tfor (const passage of passages) {\n\t\tconst lower = passage.text.toLowerCase();\n\t\tconst overlap = terms.filter((term) => containsPhrase(lower, term)).length;\n\t\tif (overlap < Math.max(2, Math.ceil(terms.length / 4))) continue;\n\t\tconst contra = hasPolarityMarker(lower, CONTRADICTION_MARKERS);\n\t\tconst support = hasPolarityMarker(lower, SUPPORT_MARKERS);\n\t\tif (contra && !support) contradicting.push(passage.passage_id);\n\t\telse if (support && !contra) supporting.push(passage.passage_id);\n\t}\n\tif (contradicting.length > 0 && supporting.length === 0) {\n\t\treturn { claim, status: \"contradicted\", supporting_passages: [], contradicting_passages: contradicting, rationale: `${contradicting.length} passage(s) contradict the claim; none support it.`, confidence: Math.min(0.85, 0.5 + contradicting.length * 0.1) };\n\t}\n\tif (supporting.length > 0 && contradicting.length === 0) {\n\t\treturn { claim, status: \"supported\", supporting_passages: supporting, contradicting_passages: [], rationale: `${supporting.length} passage(s) support the claim; none contradict it.`, confidence: Math.min(0.85, 0.5 + supporting.length * 0.1) };\n\t}\n\tif (supporting.length > 0 || contradicting.length > 0) {\n\t\treturn { claim, status: \"unclear\", supporting_passages: supporting, contradicting_passages: contradicting, rationale: `${supporting.length} supporting and ${contradicting.length} contradicting passage(s); evidence is mixed.`, confidence: 0.4 };\n\t}\n\treturn { claim, status: \"unclear\", supporting_passages: [], contradicting_passages: [], rationale: \"Passages mention the claim's terms but contain no clear support or contradiction markers.\", confidence: 0.3 };\n}\n\ninterface RankedSearchResult extends SearchResult {\n\trank?: number;\n}\n\nexport interface BuildArtifactInput {\n\tquery: string;\n\tprovider?: string;\n\tsummary?: string;\n\tresults: RankedSearchResult[];\n\tfetched?: ExtractedContent[];\n\trecency?: RecencyFilter;\n\tdomainFilter?: string[];\n}\n\nexport function buildResearchArtifact(input: BuildArtifactInput): ResearchArtifact {\n\tconst filters = input.domainFilter ?? [];\n\tconst fetchedByUrl = new Map((input.fetched ?? []).map((page) => [page.url, page]));\n\tconst sources: ResearchSource[] = [];\n\tconst seen = new Set<string>();\n\tfor (const [index, result] of input.results.entries()) {\n\t\tif (seen.has(result.url)) continue;\n\t\tseen.add(result.url);\n\t\tconst page = fetchedByUrl.get(result.url);\n\t\tconst fetched = Boolean(page && !page.error);\n\t\tsources.push({\n\t\t\trank: result.rank ?? index + 1,\n\t\t\turl: result.url,\n\t\t\ttitle: result.title,\n\t\t\tsnippet: result.snippet,\n\t\t\tquality: classifySource(result.url),\n\t\t\tfetched,\n\t\t\t...(page ? { fetch_timestamp: Date.now() } : {}),\n\t\t\t...(page && !page.error ? { content_hash: hashContent(page.content) } : {}),\n\t\t\t...(page?.error ? { fetch_error: page.error } : {}),\n\t\t});\n\t}\n\tconst passages = buildPassages(sources, input.fetched, input.query);\n\tconst domainInclude = filters.filter((domain) => !domain.startsWith(\"-\"));\n\tconst domainExclude = filters.filter((domain) => domain.startsWith(\"-\")).map((domain) => domain.slice(1));\n\treturn {\n\t\tid: generateId(),\n\t\ttype: \"research\",\n\t\ttimestamp: Date.now(),\n\t\tquery: input.query,\n\t\tsources,\n\t\tpassages,\n\t\t...(input.provider !== undefined ? { provider: input.provider } : {}),\n\t\t...(input.summary !== undefined ? { summary: input.summary } : {}),\n\t\t...(passages.length > 0 ? { content_hash: hashContent(passages.map((passage) => passage.text).join(\"\\n\")) } : {}),\n\t\tfilters: {\n\t\t\t...(input.recency !== undefined ? { recency: input.recency } : {}),\n\t\t\tdomain_include: domainInclude,\n\t\t\tdomain_exclude: domainExclude,\n\t\t},\n\t};\n}\n\nexport function withClaimAssessment(artifact: ResearchArtifact, claims: string[]): ResearchArtifact {\n\treturn { ...artifact, claims: claims.map((claim) => assessClaim(claim, artifact.passages)) };\n}\n\nexport function storeResearchArtifact(artifact: ResearchArtifact): void {\n\tif (!artifact.id) throw new Error(\"Research artifact id must not be empty\");\n\tstoreResult(artifact.id, { id: artifact.id, type: \"research\", timestamp: artifact.timestamp, artifact });\n}\n\nexport function getResearchArtifact(id: string): ResearchArtifact | null {\n\tconst data = getResult(id);\n\tif (!data || data.type !== \"research\" || !data.artifact || typeof data.artifact !== \"object\") return null;\n\treturn data.artifact as ResearchArtifact;\n}\n"]}

package/dist/humain/web-access/vendor/source-check.js

@@ -0,0 +1,205 @@
+// Structured source checking and machine-readable research artifacts.
+import { createHash } from "node:crypto";
+import { generateId, getResult, storeResult } from "./storage.js";
+const OFFICIAL_DOCS_HOSTS = /^(developers\.|docs\.|learn\.|reference\.)|\.github\.io$/i;
+const OFFICIAL_DOCS_PATHS = /\/(docs?|reference)(\/|\b)/i;
+const VENDOR_DOCS_PATHS = /\/(documentation|docs?)\//i;
+const REPO_ISSUE_PATHS = /\/(issues|pull|pulls)\//i;
+const BLOG_HOSTS = /(medium\.com|substack\.com|dev\.to|hashnode\.)/i;
+const BLOG_PATHS = /\/blogs?\//i;
+const FORUM_HOSTS = /(stackoverflow\.com|serverfault\.com|superuser\.com|discourse\.|community\.)/i;
+const FORUM_PATHS = /\/(forum|forums|threads)\//i;
+const NEWS_HOSTS = /(reuters\.com|bloomberg\.com|techcrunch\.com|theverge\.com|arstechnica\.com|wired\.com|cnet\.com|zdnet\.com)/i;
+const NEWS_PATHS = /\/news(\/|$)/i;
+export function classifySource(url) {
+    let host = "";
+    let path = "";
+    try {
+        const parsed = new URL(url);
+        host = parsed.hostname;
+        path = parsed.pathname;
+    }
+    catch {
+        return "unknown";
+    }
+    if (REPO_ISSUE_PATHS.test(path))
+        return "repo_issue";
+    if (OFFICIAL_DOCS_HOSTS.test(host) || OFFICIAL_DOCS_PATHS.test(path))
+        return "official_docs";
+    if (VENDOR_DOCS_PATHS.test(path))
+        return "vendor_docs";
+    if (NEWS_HOSTS.test(host) || NEWS_PATHS.test(path))
+        return "news";
+    if (FORUM_HOSTS.test(host) || FORUM_PATHS.test(path))
+        return "forum";
+    if (BLOG_HOSTS.test(host) || BLOG_PATHS.test(path))
+        return "blog";
+    return "unknown";
+}
+export function hashContent(text) {
+    return `sha256:${createHash("sha256").update(text, "utf8").digest("hex")}`;
+}
+function tokenize(value) {
+    return [...new Set(value.toLowerCase().split(/[^a-z0-9]+/).filter((term) => term.length > 3))];
+}
+function extractRelevantSpans(content, hint) {
+    const sentences = [];
+    const sentencePattern = /[^.!?]+(?:[.!?]+(?=\s|$)|$)/g;
+    for (const match of content.matchAll(sentencePattern)) {
+        const raw = match[0];
+        const text = raw.trim();
+        if (text.length > 0 && text.length <= 400) {
+            const start = (match.index ?? 0) + raw.indexOf(text);
+            sentences.push({ text, start, end: start + text.length });
+        }
+    }
+    const terms = tokenize(hint);
+    if (terms.length === 0)
+        return [];
+    return sentences
+        .map((sentence, index) => ({ sentence, index, score: terms.filter((term) => sentence.text.toLowerCase().includes(term)).length }))
+        .filter((item) => item.score > 0)
+        .sort((a, b) => b.score - a.score || a.index - b.index)
+        .slice(0, 3)
+        .map(({ sentence }) => sentence);
+}
+function passageId(sourceRank, index) {
+    return `p-${sourceRank}-${index}`;
+}
+export function buildPassages(sources, fetched = [], hint = "") {
+    const passages = [];
+    const fetchedByUrl = new Map(fetched.map((item) => [item.url, item]));
+    for (const source of sources) {
+        if (source.snippet) {
+            passages.push({
+                passage_id: passageId(source.rank, 0),
+                source_url: source.url,
+                source_rank: source.rank,
+                text: source.snippet,
+                content_hash: hashContent(source.snippet),
+            });
+        }
+        const page = fetchedByUrl.get(source.url);
+        if (page && !page.error && page.content) {
+            const passageHint = source.snippet?.trim() || hint;
+            for (const [index, span] of extractRelevantSpans(page.content, passageHint).entries()) {
+                passages.push({
+                    passage_id: passageId(source.rank, index + 1),
+                    source_url: source.url,
+                    source_rank: source.rank,
+                    text: span.text,
+                    extraction_span: { start: span.start, end: span.end },
+                    content_hash: hashContent(span.text),
+                });
+            }
+        }
+    }
+    return passages;
+}
+const CONTRADICTION_MARKERS = ["not true", "false", "incorrect", "debunked", "retracted", "no longer", "never", "denied", "contrary", "misleading"];
+const SUPPORT_MARKERS = ["yes", "true", "correct", "confirmed", "according to", "shows that", "demonstrates", "reported", "verified", "established"];
+function containsPhrase(value, phrase) {
+    const escaped = phrase.trim().toLowerCase().split(/\s+/).map((term) => term.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")).join("\\s+");
+    return new RegExp(`(?:^|[^a-z0-9])${escaped}(?=$|[^a-z0-9])`, "i").test(value);
+}
+function markerIsNegated(value, marker) {
+    const escaped = marker.trim().toLowerCase().split(/\s+/).map((term) => term.replace(/[.*+?^${}()|[\]\\]/g, "\\$&")).join("\\s+");
+    const markerPattern = new RegExp(`(?:^|[^a-z0-9])${escaped}(?=$|[^a-z0-9])`, "i");
+    const match = markerPattern.exec(value);
+    if (!match || match.index === undefined)
+        return false;
+    const matchedMarker = match[0].replace(/^[^a-z0-9]+/i, "");
+    const beforeMarker = value.slice(0, match.index + match[0].length - matchedMarker.length);
+    return /(?:^|[^a-z0-9])(?:not|no|never|without)\s+$/i.test(beforeMarker);
+}
+function hasPolarityMarker(value, markers, allowNegated = false) {
+    return markers.some((marker) => containsPhrase(value, marker) && (allowNegated || !markerIsNegated(value, marker)));
+}
+export function assessClaim(claim, passages) {
+    const terms = tokenize(claim);
+    if (terms.length === 0 || passages.length === 0) {
+        return { claim, status: "missing-evidence", supporting_passages: [], contradicting_passages: [], rationale: "No passages available that discuss the claim's terms.", confidence: 0.2 };
+    }
+    const supporting = [];
+    const contradicting = [];
+    for (const passage of passages) {
+        const lower = passage.text.toLowerCase();
+        const overlap = terms.filter((term) => containsPhrase(lower, term)).length;
+        if (overlap < Math.max(2, Math.ceil(terms.length / 4)))
+            continue;
+        const contra = hasPolarityMarker(lower, CONTRADICTION_MARKERS);
+        const support = hasPolarityMarker(lower, SUPPORT_MARKERS);
+        if (contra && !support)
+            contradicting.push(passage.passage_id);
+        else if (support && !contra)
+            supporting.push(passage.passage_id);
+    }
+    if (contradicting.length > 0 && supporting.length === 0) {
+        return { claim, status: "contradicted", supporting_passages: [], contradicting_passages: contradicting, rationale: `${contradicting.length} passage(s) contradict the claim; none support it.`, confidence: Math.min(0.85, 0.5 + contradicting.length * 0.1) };
+    }
+    if (supporting.length > 0 && contradicting.length === 0) {
+        return { claim, status: "supported", supporting_passages: supporting, contradicting_passages: [], rationale: `${supporting.length} passage(s) support the claim; none contradict it.`, confidence: Math.min(0.85, 0.5 + supporting.length * 0.1) };
+    }
+    if (supporting.length > 0 || contradicting.length > 0) {
+        return { claim, status: "unclear", supporting_passages: supporting, contradicting_passages: contradicting, rationale: `${supporting.length} supporting and ${contradicting.length} contradicting passage(s); evidence is mixed.`, confidence: 0.4 };
+    }
+    return { claim, status: "unclear", supporting_passages: [], contradicting_passages: [], rationale: "Passages mention the claim's terms but contain no clear support or contradiction markers.", confidence: 0.3 };
+}
+export function buildResearchArtifact(input) {
+    const filters = input.domainFilter ?? [];
+    const fetchedByUrl = new Map((input.fetched ?? []).map((page) => [page.url, page]));
+    const sources = [];
+    const seen = new Set();
+    for (const [index, result] of input.results.entries()) {
+        if (seen.has(result.url))
+            continue;
+        seen.add(result.url);
+        const page = fetchedByUrl.get(result.url);
+        const fetched = Boolean(page && !page.error);
+        sources.push({
+            rank: result.rank ?? index + 1,
+            url: result.url,
+            title: result.title,
+            snippet: result.snippet,
+            quality: classifySource(result.url),
+            fetched,
+            ...(page ? { fetch_timestamp: Date.now() } : {}),
+            ...(page && !page.error ? { content_hash: hashContent(page.content) } : {}),
+            ...(page?.error ? { fetch_error: page.error } : {}),
+        });
+    }
+    const passages = buildPassages(sources, input.fetched, input.query);
+    const domainInclude = filters.filter((domain) => !domain.startsWith("-"));
+    const domainExclude = filters.filter((domain) => domain.startsWith("-")).map((domain) => domain.slice(1));
+    return {
+        id: generateId(),
+        type: "research",
+        timestamp: Date.now(),
+        query: input.query,
+        sources,
+        passages,
+        ...(input.provider !== undefined ? { provider: input.provider } : {}),
+        ...(input.summary !== undefined ? { summary: input.summary } : {}),
+        ...(passages.length > 0 ? { content_hash: hashContent(passages.map((passage) => passage.text).join("\n")) } : {}),
+        filters: {
+            ...(input.recency !== undefined ? { recency: input.recency } : {}),
+            domain_include: domainInclude,
+            domain_exclude: domainExclude,
+        },
+    };
+}
+export function withClaimAssessment(artifact, claims) {
+    return { ...artifact, claims: claims.map((claim) => assessClaim(claim, artifact.passages)) };
+}
+export function storeResearchArtifact(artifact) {
+    if (!artifact.id)
+        throw new Error("Research artifact id must not be empty");
+    storeResult(artifact.id, { id: artifact.id, type: "research", timestamp: artifact.timestamp, artifact });
+}
+export function getResearchArtifact(id) {
+    const data = getResult(id);
+    if (!data || data.type !== "research" || !data.artifact || typeof data.artifact !== "object")
+        return null;
+    return data.artifact;
+}
+//# sourceMappingURL=source-check.js.map