@triplef/agent 0.1.17 → 0.1.19

package/dist/schemas/index.d.ts

@@ -1,5 +1,6 @@
 import { z } from 'zod';
-export { B as BROWSER_TOOL_NAMES, D as DEFAULT_VARIANT_ID, F as FormatZodShapeOptions, a as IMAGE_TASK_TEMPLATES, b as ImageTaskTemplate, I as IntentResult, c as IntentSchema, M as MEMORY_TOOL_NAMES, P as ProviderConfig, d as TOOL_DESCRIPTIONS, e as TOOL_NAMES, T as TemplateName, f as ToolName, g as VARIANT_NAMES, V as VariantName, h as formatZodShape, i as isImageTaskTemplate } from '../intent.schema-Bp3p7h7P.js';
+export { E as EncyclopediaDocumentInput, a as EncyclopediaDocumentResult, b as EncyclopediaSearchHit, c as EncyclopediaSearchInput } from '../encyclopedia-search.model-oFW8_rUT.js';
+export { B as BROWSER_TOOL_NAMES, D as DEFAULT_VARIANT_ID, E as ENCYCLOPEDIA_TOOL_NAMES, F as FormatZodShapeOptions, a as IMAGE_TASK_TEMPLATES, b as ImageTaskTemplate, I as IntentResult, c as IntentSchema, M as MEMORY_TOOL_NAMES, P as ProviderConfig, d as TOOL_DESCRIPTIONS, e as TOOL_NAMES, T as TemplateName, f as ToolName, g as VARIANT_NAMES, V as VariantName, h as formatZodShape, i as isImageTaskTemplate } from '../format-zod-shape.helper-BviIexpu.js';
 
 declare const BLOCKED_IMAGE_HOSTS: Set<string>;
 
@@ -22,6 +23,9 @@ interface EncyclopediaSourceDocument {
     url?: string;
     title?: string;
     content: string;
+    mimeType?: string;
+    sizeBytes?: number;
+    originalHash?: string;
 }
 interface EncyclopediaSelectedChunk {
     url?: string;
@@ -186,11 +190,6 @@ declare const memoryCognitionProfileSchema: z.ZodObject<{
     corrections: z.ZodOptional<z.ZodNullable<z.ZodRecord<z.ZodString, z.ZodString>>>;
 }, z.core.$strip>;
 type MemoryCognitionProfile = z.infer<typeof memoryCognitionProfileSchema>;
-interface MergedCognitionProfile {
-    profile?: MemoryCognitionProfile;
-    removals: string[];
-}
-declare function mergeCognitionProfiles(current: MemoryCognitionProfile | undefined, patch: MemoryCognitionProfile): MergedCognitionProfile;
 declare function isAllFieldsNullWipe(current: MemoryCognitionProfile | undefined, patch: MemoryCognitionProfile): boolean;
 declare function parseStoredProfile(text: string | undefined): MemoryCognitionProfile;
 declare const memoryProfileInsightSchema: z.ZodObject<{
@@ -267,6 +266,29 @@ declare function clampEpisodeRecencyMidpoint(value: number): number;
 declare function clampEpisodeProbeLimit(value: number): number;
 declare function clampEpisodeScoreThreshold(value: number): number;
 
+interface MergedCognitionProfile {
+    profile?: MemoryCognitionProfile;
+    removals: string[];
+}
+declare function mergeCognitionProfiles(current: MemoryCognitionProfile | undefined, patch: MemoryCognitionProfile): MergedCognitionProfile;
+
+declare const ResearchTriageDecisionSchema: z.ZodObject<{
+    url: z.ZodString;
+    close: z.ZodBoolean;
+    reason: z.ZodString;
+    followUpTopics: z.ZodOptional<z.ZodArray<z.ZodString>>;
+}, z.core.$strip>;
+declare const ResearchTriageSchema: z.ZodObject<{
+    decisions: z.ZodArray<z.ZodObject<{
+        url: z.ZodString;
+        close: z.ZodBoolean;
+        reason: z.ZodString;
+        followUpTopics: z.ZodOptional<z.ZodArray<z.ZodString>>;
+    }, z.core.$strip>>;
+}, z.core.$strip>;
+type ResearchTriageDecision = z.infer<typeof ResearchTriageDecisionSchema>;
+type ResearchTriage = z.infer<typeof ResearchTriageSchema>;
+
 declare const cardSchema: z.ZodObject<{
     url: z.ZodString;
     title: z.ZodOptional<z.ZodString>;
@@ -873,4 +895,4 @@ declare const videolistSchema: z.ZodObject<{
     }, z.core.$strip>>>;
 }, z.core.$strip>;
 
-export { BLOCKED_IMAGE_HOSTS, BLOCKED_URL_HOSTS, COGNITION_LIMIT_DEFAULT, COGNITION_LIMIT_MAX, COGNITION_LIMIT_MIN, COGNITION_PURGE_BATCH, CONVICTION_TAGS, CONVICTION_TEXT_LIMIT, type ConsolidationVerdict, ConsolidationVerdictSchema, DEFAULT_MEDIA_COUNT, EMBEDDABLE_VIDEO_PROVIDER_CLAUSE, EMBEDDABLE_VIDEO_PROVIDER_LABELS, EPISODE_PROBE_LIMIT_DEFAULT, EPISODE_PROBE_LIMIT_MIN, EPISODE_RECENCY_MIDPOINT_DEFAULT, EPISODE_RECENCY_MIDPOINT_MAX, EPISODE_RECENCY_MIDPOINT_MIN, EPISODE_RECENCY_SCALE_SECONDS_DEFAULT, EPISODE_RECENCY_SCALE_SECONDS_MAX, EPISODE_RECENCY_SCALE_SECONDS_MIN, EPISODE_RECENCY_WEIGHT_DEFAULT, EPISODE_RECENCY_WEIGHT_MAX, EPISODE_RECENCY_WEIGHT_MIN, EPISODE_SCORE_THRESHOLD_DEFAULT, EPISODE_SCORE_THRESHOLD_MAX, EPISODE_SCORE_THRESHOLD_MIN, EPISODE_TAGS, EPISODE_TEXT_LIMIT, type EncyclopediaClassification, EncyclopediaClassifySchema, type EncyclopediaSearchResult, type EncyclopediaSelectInput, type EncyclopediaSelectResult, type EncyclopediaSelectedChunk, type EncyclopediaSourceDocument, type ExtractedFact, ExtractedFactSchema, ExtractionSchema, FACT_KINDS, FACT_STABILITIES, HERO_VIDEO_TITLE_ISSUE, INSIGHTS_MAX_PER_TURN, INSIGHT_TAGS, INSIGHT_TEXT_LIMIT, MORE_MEDIA_COUNT, type MemoryClusterSummary, MemoryClusterSummarySchema, type MemoryCognitionProfile, type MemoryEnrichment, MemoryEnrichmentSchema, type MemoryExtraction, type MemoryProfileInsight, type MemoryProfileResponse, NON_PAGE_EXTENSIONS, cardSchema, categorizeTools, clampCognitionLimit, clampEpisodeProbeLimit, clampEpisodeRecencyMidpoint, clampEpisodeRecencyScaleSeconds, clampEpisodeRecencyWeight, clampEpisodeScoreThreshold, compareSchema, createTextItemSchema, deriveSchemaKeys, describeSchema, discardedReferenceSchema, formatZodIssues, galleryItemSchema, hasBlockedImageHost, heroVideoHasTitle, imagelistSchema, internationalCoverageSchema, isAllFieldsNullWipe, isPrivateOrLocalhost, isTrustedImageUrl, isTrustedUrl, markerSchema, memoryProfileResponseSchema, mergeCognitionProfiles, normalizeInsightPath, ocrSchema, parseStoredProfile, productSchema, referenceGalleryItemSchema, referenceLineSchema, relatedStorySchema, safeMediaUrl, safeMediaUrlOrEmpty, safeUrl, safeVideoUrl, safeVideoUrlOrEmpty, shoplistSchema, sourceSchema, stockmarketItemSchema, stockmarketListSchema, summarySchema, videoGalleryItemSchema, videolistSchema };
+export { BLOCKED_IMAGE_HOSTS, BLOCKED_URL_HOSTS, COGNITION_LIMIT_DEFAULT, COGNITION_LIMIT_MAX, COGNITION_LIMIT_MIN, COGNITION_PURGE_BATCH, CONVICTION_TAGS, CONVICTION_TEXT_LIMIT, type ConsolidationVerdict, ConsolidationVerdictSchema, DEFAULT_MEDIA_COUNT, EMBEDDABLE_VIDEO_PROVIDER_CLAUSE, EMBEDDABLE_VIDEO_PROVIDER_LABELS, EPISODE_PROBE_LIMIT_DEFAULT, EPISODE_PROBE_LIMIT_MIN, EPISODE_RECENCY_MIDPOINT_DEFAULT, EPISODE_RECENCY_MIDPOINT_MAX, EPISODE_RECENCY_MIDPOINT_MIN, EPISODE_RECENCY_SCALE_SECONDS_DEFAULT, EPISODE_RECENCY_SCALE_SECONDS_MAX, EPISODE_RECENCY_SCALE_SECONDS_MIN, EPISODE_RECENCY_WEIGHT_DEFAULT, EPISODE_RECENCY_WEIGHT_MAX, EPISODE_RECENCY_WEIGHT_MIN, EPISODE_SCORE_THRESHOLD_DEFAULT, EPISODE_SCORE_THRESHOLD_MAX, EPISODE_SCORE_THRESHOLD_MIN, EPISODE_TAGS, EPISODE_TEXT_LIMIT, type EncyclopediaClassification, EncyclopediaClassifySchema, type EncyclopediaSearchResult, type EncyclopediaSelectInput, type EncyclopediaSelectResult, type EncyclopediaSelectedChunk, type EncyclopediaSourceDocument, type ExtractedFact, ExtractedFactSchema, ExtractionSchema, FACT_KINDS, FACT_STABILITIES, HERO_VIDEO_TITLE_ISSUE, INSIGHTS_MAX_PER_TURN, INSIGHT_TAGS, INSIGHT_TEXT_LIMIT, MORE_MEDIA_COUNT, type MemoryClusterSummary, MemoryClusterSummarySchema, type MemoryCognitionProfile, type MemoryEnrichment, MemoryEnrichmentSchema, type MemoryExtraction, type MemoryProfileInsight, type MemoryProfileResponse, type MergedCognitionProfile, NON_PAGE_EXTENSIONS, type ResearchTriage, type ResearchTriageDecision, ResearchTriageDecisionSchema, ResearchTriageSchema, cardSchema, categorizeTools, clampCognitionLimit, clampEpisodeProbeLimit, clampEpisodeRecencyMidpoint, clampEpisodeRecencyScaleSeconds, clampEpisodeRecencyWeight, clampEpisodeScoreThreshold, compareSchema, createTextItemSchema, deriveSchemaKeys, describeSchema, discardedReferenceSchema, formatZodIssues, galleryItemSchema, hasBlockedImageHost, heroVideoHasTitle, imagelistSchema, internationalCoverageSchema, isAllFieldsNullWipe, isPrivateOrLocalhost, isTrustedImageUrl, isTrustedUrl, markerSchema, memoryProfileResponseSchema, mergeCognitionProfiles, normalizeInsightPath, ocrSchema, parseStoredProfile, productSchema, referenceGalleryItemSchema, referenceLineSchema, relatedStorySchema, safeMediaUrl, safeMediaUrlOrEmpty, safeUrl, safeVideoUrl, safeVideoUrlOrEmpty, shoplistSchema, sourceSchema, stockmarketItemSchema, stockmarketListSchema, summarySchema, videoGalleryItemSchema, videolistSchema };

package/dist/schemas/index.mjs

@@ -182,6 +182,7 @@ var MEMORY_TOOL_NAMES = [
   "memory-cognition-remember",
   "memory-cognition-forget"
 ];
+var ENCYCLOPEDIA_TOOL_NAMES = ["encyclopedia-search", "encyclopedia-read"];
 var TOOL_NAMES = [
   "webFetch",
   "brightDataWebSearch",
@@ -629,44 +630,6 @@ var memoryCognitionProfileSchema = z.object({
    */
   corrections: z.record(z.string(), z.string()).nullish()
 });
-function normalizeCognitionProfile(profile) {
-  if (!profile) return void 0;
-  const cleaned = {};
-  for (const [key, value] of Object.entries(profile)) {
-    const kept = cleanProfileValue(value);
-    if (kept !== void 0) cleaned[key] = kept;
-  }
-  return Object.keys(cleaned).length > 0 ? cleaned : void 0;
-}
-function cleanProfileValue(value) {
-  if (value === null || value === void 0) return void 0;
-  if (typeof value === "string") return value.trim() || void 0;
-  if (Array.isArray(value)) {
-    const items = value.map((entry) => typeof entry === "string" ? entry.trim() : entry).filter(Boolean);
-    return items.length > 0 ? items : void 0;
-  }
-  if (typeof value !== "object") return void 0;
-  const inner = Object.fromEntries(
-    Object.entries(value).map(([k, v]) => [k, cleanProfileValue(v)]).filter(([, v]) => v !== void 0)
-  );
-  return Object.keys(inner).length > 0 ? inner : void 0;
-}
-function mergeCognitionProfiles(current, patch) {
-  const patchRecord = patch;
-  const removals = Object.keys(patchRecord).filter(
-    (key) => patchRecord[key] === null || patchRecord[key] === void 0
-  );
-  const base = current ?? {};
-  const merged = merge(
-    omit(base, removals),
-    omit(patch, removals),
-    true
-  );
-  const profile = normalizeCognitionProfile(merged) ?? {};
-  const before = normalizeCognitionProfile(base) ?? {};
-  const changed = JSON.stringify(profile) !== JSON.stringify(before);
-  return { profile: changed ? profile : void 0, removals };
-}
 function isAllFieldsNullWipe(current, patch) {
   if (!current) return false;
   const currentKeys = Object.keys(current);
@@ -773,6 +736,61 @@ function clampEpisodeScoreThreshold(value) {
   if (!Number.isFinite(value)) return EPISODE_SCORE_THRESHOLD_DEFAULT;
   return Math.min(EPISODE_SCORE_THRESHOLD_MAX, Math.max(EPISODE_SCORE_THRESHOLD_MIN, value));
 }
+function normalizeCognitionProfile(profile) {
+  if (!profile) return void 0;
+  const cleaned = {};
+  for (const [key, value] of Object.entries(profile)) {
+    const kept = cleanProfileValue(value);
+    if (kept !== void 0) cleaned[key] = kept;
+  }
+  return Object.keys(cleaned).length > 0 ? cleaned : void 0;
+}
+function cleanProfileValue(value) {
+  if (value === null || value === void 0) return void 0;
+  if (typeof value === "string") return value.trim() || void 0;
+  if (Array.isArray(value)) {
+    const items = value.map((entry) => typeof entry === "string" ? entry.trim() : entry).filter(Boolean);
+    return items.length > 0 ? items : void 0;
+  }
+  if (typeof value !== "object") return void 0;
+  const inner = Object.fromEntries(
+    Object.entries(value).map(([k, v]) => [k, cleanProfileValue(v)]).filter(([, v]) => v !== void 0)
+  );
+  return Object.keys(inner).length > 0 ? inner : void 0;
+}
+function mergeCognitionProfiles(current, patch) {
+  const patchRecord = patch;
+  const removals = Object.keys(patchRecord).filter(
+    (key) => patchRecord[key] === null || patchRecord[key] === void 0
+  );
+  const base = current ?? {};
+  const merged = merge(
+    omit(base, removals),
+    omit(patch, removals),
+    true
+  );
+  const profile = normalizeCognitionProfile(merged) ?? {};
+  const before = normalizeCognitionProfile(base) ?? {};
+  const changed = JSON.stringify(profile) !== JSON.stringify(before);
+  return { profile: changed ? profile : void 0, removals };
+}
+var ResearchTriageDecisionSchema = z.object({
+  /** The gap's url — echoed back so the worker maps the verdict to its candidate. */
+  url: z.string(),
+  /** True = fetch and persist this page; false = skip (transient/low-value). */
+  close: z.boolean(),
+  /** One-sentence reason — logged for observability, never shown to users. */
+  reason: z.string(),
+  /**
+   * New topics this page (or snippet) references that the knowledge base does
+   * not yet cover — the next deep-dive's search queries (the Z candidates).
+   * Empty when the page is self-contained.
+   */
+  followUpTopics: z.array(z.string().min(1).max(200)).max(3).optional()
+});
+var ResearchTriageSchema = z.object({
+  decisions: z.array(ResearchTriageDecisionSchema).max(50)
+});
 var cardSchema = z.object(
   {
     url: safeUrl({ message: "cards entries must have a valid url" }),
@@ -1153,4 +1171,4 @@ var videolistSchema = z.object({
   internationalCoverage: internationalCoverageSchema.optional()
 });
 
-export { BLOCKED_IMAGE_HOSTS, BLOCKED_URL_HOSTS, BROWSER_TOOL_NAMES, COGNITION_LIMIT_DEFAULT, COGNITION_LIMIT_MAX, COGNITION_LIMIT_MIN, COGNITION_PURGE_BATCH, CONVICTION_TAGS, CONVICTION_TEXT_LIMIT, ConsolidationVerdictSchema, DEFAULT_MEDIA_COUNT, DEFAULT_VARIANT_ID, EMBEDDABLE_VIDEO_PROVIDER_CLAUSE, EMBEDDABLE_VIDEO_PROVIDER_LABELS, EPISODE_PROBE_LIMIT_DEFAULT, EPISODE_PROBE_LIMIT_MIN, EPISODE_RECENCY_MIDPOINT_DEFAULT, EPISODE_RECENCY_MIDPOINT_MAX, EPISODE_RECENCY_MIDPOINT_MIN, EPISODE_RECENCY_SCALE_SECONDS_DEFAULT, EPISODE_RECENCY_SCALE_SECONDS_MAX, EPISODE_RECENCY_SCALE_SECONDS_MIN, EPISODE_RECENCY_WEIGHT_DEFAULT, EPISODE_RECENCY_WEIGHT_MAX, EPISODE_RECENCY_WEIGHT_MIN, EPISODE_SCORE_THRESHOLD_DEFAULT, EPISODE_SCORE_THRESHOLD_MAX, EPISODE_SCORE_THRESHOLD_MIN, EPISODE_TAGS, EPISODE_TEXT_LIMIT, EncyclopediaClassifySchema, ExtractedFactSchema, ExtractionSchema, FACT_KINDS, FACT_STABILITIES, HERO_VIDEO_TITLE_ISSUE, IMAGE_TASK_TEMPLATES, INSIGHTS_MAX_PER_TURN, INSIGHT_TAGS, INSIGHT_TEXT_LIMIT, IntentSchema, MEMORY_TOOL_NAMES, MORE_MEDIA_COUNT, MemoryClusterSummarySchema, MemoryEnrichmentSchema, NON_PAGE_EXTENSIONS, TOOL_DESCRIPTIONS, TOOL_NAMES, VARIANT_NAMES, cardSchema, categorizeTools, clampCognitionLimit, clampEpisodeProbeLimit, clampEpisodeRecencyMidpoint, clampEpisodeRecencyScaleSeconds, clampEpisodeRecencyWeight, clampEpisodeScoreThreshold, compareSchema, createTextItemSchema, deriveSchemaKeys, describeSchema, discardedReferenceSchema, formatZodIssues, formatZodShape, galleryItemSchema, hasBlockedImageHost, heroVideoHasTitle, imagelistSchema, internationalCoverageSchema, isAllFieldsNullWipe, isImageTaskTemplate, isPrivateOrLocalhost, isTrustedImageUrl, isTrustedUrl, markerSchema, memoryProfileResponseSchema, mergeCognitionProfiles, normalizeInsightPath, ocrSchema, parseStoredProfile, productSchema, referenceGalleryItemSchema, referenceLineSchema, relatedStorySchema, safeMediaUrl, safeMediaUrlOrEmpty, safeUrl, safeVideoUrl, safeVideoUrlOrEmpty, shoplistSchema, sourceSchema, stockmarketItemSchema, stockmarketListSchema, summarySchema, videoGalleryItemSchema, videolistSchema };
+export { BLOCKED_IMAGE_HOSTS, BLOCKED_URL_HOSTS, BROWSER_TOOL_NAMES, COGNITION_LIMIT_DEFAULT, COGNITION_LIMIT_MAX, COGNITION_LIMIT_MIN, COGNITION_PURGE_BATCH, CONVICTION_TAGS, CONVICTION_TEXT_LIMIT, ConsolidationVerdictSchema, DEFAULT_MEDIA_COUNT, DEFAULT_VARIANT_ID, EMBEDDABLE_VIDEO_PROVIDER_CLAUSE, EMBEDDABLE_VIDEO_PROVIDER_LABELS, ENCYCLOPEDIA_TOOL_NAMES, EPISODE_PROBE_LIMIT_DEFAULT, EPISODE_PROBE_LIMIT_MIN, EPISODE_RECENCY_MIDPOINT_DEFAULT, EPISODE_RECENCY_MIDPOINT_MAX, EPISODE_RECENCY_MIDPOINT_MIN, EPISODE_RECENCY_SCALE_SECONDS_DEFAULT, EPISODE_RECENCY_SCALE_SECONDS_MAX, EPISODE_RECENCY_SCALE_SECONDS_MIN, EPISODE_RECENCY_WEIGHT_DEFAULT, EPISODE_RECENCY_WEIGHT_MAX, EPISODE_RECENCY_WEIGHT_MIN, EPISODE_SCORE_THRESHOLD_DEFAULT, EPISODE_SCORE_THRESHOLD_MAX, EPISODE_SCORE_THRESHOLD_MIN, EPISODE_TAGS, EPISODE_TEXT_LIMIT, EncyclopediaClassifySchema, ExtractedFactSchema, ExtractionSchema, FACT_KINDS, FACT_STABILITIES, HERO_VIDEO_TITLE_ISSUE, IMAGE_TASK_TEMPLATES, INSIGHTS_MAX_PER_TURN, INSIGHT_TAGS, INSIGHT_TEXT_LIMIT, IntentSchema, MEMORY_TOOL_NAMES, MORE_MEDIA_COUNT, MemoryClusterSummarySchema, MemoryEnrichmentSchema, NON_PAGE_EXTENSIONS, ResearchTriageDecisionSchema, ResearchTriageSchema, TOOL_DESCRIPTIONS, TOOL_NAMES, VARIANT_NAMES, cardSchema, categorizeTools, clampCognitionLimit, clampEpisodeProbeLimit, clampEpisodeRecencyMidpoint, clampEpisodeRecencyScaleSeconds, clampEpisodeRecencyWeight, clampEpisodeScoreThreshold, compareSchema, createTextItemSchema, deriveSchemaKeys, describeSchema, discardedReferenceSchema, formatZodIssues, formatZodShape, galleryItemSchema, hasBlockedImageHost, heroVideoHasTitle, imagelistSchema, internationalCoverageSchema, isAllFieldsNullWipe, isImageTaskTemplate, isPrivateOrLocalhost, isTrustedImageUrl, isTrustedUrl, markerSchema, memoryProfileResponseSchema, mergeCognitionProfiles, normalizeInsightPath, ocrSchema, parseStoredProfile, productSchema, referenceGalleryItemSchema, referenceLineSchema, relatedStorySchema, safeMediaUrl, safeMediaUrlOrEmpty, safeUrl, safeVideoUrl, safeVideoUrlOrEmpty, shoplistSchema, sourceSchema, stockmarketItemSchema, stockmarketListSchema, summarySchema, videoGalleryItemSchema, videolistSchema };

package/dist/tools/index.d.ts

@@ -1,5 +1,6 @@
 import { z } from 'zod';
 import { Tool } from 'ai';
+import { a as EncyclopediaDocumentResult, b as EncyclopediaSearchHit } from '../encyclopedia-search.model-oFW8_rUT.js';
 import { M as MemoryPoint } from '../memory-point.model-C4wvvjAY.js';
 export { a as MemoryRole } from '../memory-point.model-C4wvvjAY.js';
 
@@ -283,6 +284,40 @@ declare const SEARCH_RETRIES = 1;
 declare const STANDALONE_QUERY_DESCRIPTION = "A standalone, self-contained search query that explicitly names the subject (title, entity, brand, person, place, or topic). Never copy the user message verbatim: rewrite short follow-ups (e.g. \"what do the reviews say?\") into a full query that names the established subject from the conversation (e.g. \"Neverness to Everness NTE reviews\").";
 declare const STANDALONE_QUERY_TOOL_CLAUSE = "Always pass a standalone query that names the subject explicitly \u2014 never the user message verbatim.";
 
+declare const encyclopediaReadSchema: z.ZodObject<{
+    url: z.ZodString;
+    startChunk: z.ZodOptional<z.ZodNumber>;
+    maxChars: z.ZodOptional<z.ZodNumber>;
+}, z.core.$strip>;
+type EncyclopediaReadToolInput = z.infer<typeof encyclopediaReadSchema>;
+
+interface EncyclopediaReadDeps {
+    readDocument: (input: {
+        url: string;
+        offset?: number;
+        maxChars?: number;
+    }) => Promise<EncyclopediaDocumentResult | null>;
+}
+declare function createEncyclopediaReadTool(deps: EncyclopediaReadDeps): Tool;
+
+declare const encyclopediaSearchSchema: z.ZodObject<{
+    query: z.ZodString;
+    topK: z.ZodOptional<z.ZodNumber>;
+    url: z.ZodOptional<z.ZodString>;
+    domain: z.ZodOptional<z.ZodString>;
+}, z.core.$strip>;
+type EncyclopediaSearchToolInput = z.infer<typeof encyclopediaSearchSchema>;
+
+interface EncyclopediaSearchDeps {
+    search: (input: {
+        text: string;
+        url?: string;
+        domain?: string;
+        limit?: number;
+    }) => Promise<EncyclopediaSearchHit[]>;
+}
+declare function createEncyclopediaSearchTool(deps: EncyclopediaSearchDeps): Tool;
+
 declare function applyLocaleParams(body: Record<string, unknown>, lang?: string): void;
 
 type SearchRecency = keyof typeof QDR_BY_RECENCY;
@@ -643,6 +678,8 @@ declare const summarizeFound: ToolSummaryFn;
 declare const defaultSummarize: ToolSummaryFn;
 declare function withSummary(tool: Tool, summarize?: ToolSummaryFn): ToolWithSummary;
 
+declare function extractArticleText(html: string): string;
+
 declare const webFetchSchema: z.ZodObject<{
     url: z.ZodString;
 }, z.core.$strip>;
@@ -705,4 +742,4 @@ interface VideoStats {
     lang?: string;
 }
 
-export { BRIGHT_DATA_TIMEOUT_MS, type BrightDataConfig, type BrightDataImageSearchInput, type BrightDataImageSearchResponse, type BrightDataNewsSearchInput, type BrightDataNewsSearchResponse, type BrightDataPlacesSearchInput, type BrightDataPlacesSearchResponse, type BrightDataShoppingSearchInput, type BrightDataShoppingSearchResponse, type BrightDataVideoSearchInput, type BrightDataVideoSearchResponse, type BrightDataWebSearchInput, type BrightDataWebSearchResponse, type BrightDataWebpageScrapeInput, type BrightDataWebpageScrapeResponse, type EodhdConfig, type EodhdNewsConfig, type FilterVariant, HEADERS, MIN_IMAGE_HEIGHT, MIN_IMAGE_WIDTH, type MarketDailyBar, type MemoryCognitionForgetInput, type MemoryCognitionRememberInput, type MemoryPartitionDeleteInput, type MemoryPartitionRecallInput, type MemoryPartitionRememberInput, MemoryPoint, type MemoryToolScope, type ProviderEndpointConfig, RECENCY_DESCRIPTION, type ResolvableOffer, SEARCH_RETRIES, SEARCH_TIMEOUT_MS, SOURCE, STANDALONE_QUERY_DESCRIPTION, STANDALONE_QUERY_TOOL_CLAUSE, type SearchRecency, type SerperBusinessReviewsSearchInput, type SerperBusinessReviewsSearchResponse, type SerperConfig, type SerperImageSearchInput, type SerperImageSearchResponse, type SerperNewsSearchInput, type SerperNewsSearchResponse, type SerperPlacesSearchInput, type SerperPlacesSearchResponse, type SerperShoppingSearchInput, type SerperShoppingSearchResponse, type SerperVideoSearchInput, type SerperVideoSearchResponse, type SerperWebSearchInput, type SerperWebSearchResponse, type SerperWebpageScrapeInput, type SerperWebpageScrapeResponse, type SourcesConfig, type ToolConfigSnapshot, type ToolDependencies, type ToolLogger, type ToolSummaryFn, type ToolWithSummary, type Variant, type VideoStats, type WebFetchInput, type YoutubeConfig, type YoutubeSearchResponse, type YoutubeVideoSearchInput, type YoutubeVideosResponse, applyLocaleParams, applyRecencyParam, brightDataImageSearchSchema, brightDataNewsSearchSchema, brightDataPlacesSearchSchema, brightDataShoppingSearchSchema, brightDataVideoSearchSchema, brightDataWebSearchSchema, brightDataWebpageScrapeSchema, buildGoogleUrl, buildYoutubeThumbnailUrl, createBrightDataImageSearch, createBrightDataNewsSearch, createBrightDataPlacesSearch, createBrightDataShoppingSearch, createBrightDataVideoSearch, createBrightDataWebSearch, createBrightDataWebpageScrape, createMemoryCognitionForgetTool, createMemoryCognitionRememberTool, createMemoryPartitionDeleteTool, createMemoryPartitionRecallTool, createMemoryPartitionRememberTool, createSerperBusinessReviewsSearch, createSerperImageSearch, createSerperNewsSearch, createSerperPlacesSearch, createSerperShoppingSearch, createSerperVideoSearch, createSerperWebSearch, createSerperWebpageScrape, createVariantRequestTool, createWebFetchTool, createYoutubeVideoSearch, defaultSummarize, engineEnabled, fetchWithTimeout, isGoogleHostUrl, localizedQuerySuffix, meetsMinimumImageDimensions, memoryCognitionForgetSchema, memoryCognitionRememberSchema, memoryPartitionDeleteSchema, memoryPartitionRecallSchema, memoryPartitionRememberSchema, pickMerchantResult, requestBrightData, resolveSerperShopOfferLinks, serperBusinessReviewsSearchSchema, serperImageSearchSchema, serperNewsSearchSchema, serperPlacesSearchSchema, serperShoppingSearchSchema, serperVideoSearchSchema, serperWebSearchSchema, serperWebpageScrapeSchema, storeHostToken, summarizeContent, summarizeFound, summarizeResults, tbsSizeLabelForPixels, variantRequestSchema, webFetchSchema, withSummary, youtubeVideoSearchSchema };
+export { BRIGHT_DATA_TIMEOUT_MS, type BrightDataConfig, type BrightDataImageSearchInput, type BrightDataImageSearchResponse, type BrightDataNewsSearchInput, type BrightDataNewsSearchResponse, type BrightDataPlacesSearchInput, type BrightDataPlacesSearchResponse, type BrightDataShoppingSearchInput, type BrightDataShoppingSearchResponse, type BrightDataVideoSearchInput, type BrightDataVideoSearchResponse, type BrightDataWebSearchInput, type BrightDataWebSearchResponse, type BrightDataWebpageScrapeInput, type BrightDataWebpageScrapeResponse, type EncyclopediaReadToolInput, type EncyclopediaSearchToolInput, type EodhdConfig, type EodhdNewsConfig, type FilterVariant, HEADERS, MIN_IMAGE_HEIGHT, MIN_IMAGE_WIDTH, type MarketDailyBar, type MemoryCognitionForgetInput, type MemoryCognitionRememberInput, type MemoryPartitionDeleteInput, type MemoryPartitionRecallInput, type MemoryPartitionRememberInput, MemoryPoint, type MemoryToolScope, type ProviderEndpointConfig, RECENCY_DESCRIPTION, type ResolvableOffer, SEARCH_RETRIES, SEARCH_TIMEOUT_MS, SOURCE, STANDALONE_QUERY_DESCRIPTION, STANDALONE_QUERY_TOOL_CLAUSE, type SearchRecency, type SerperBusinessReviewsSearchInput, type SerperBusinessReviewsSearchResponse, type SerperConfig, type SerperImageSearchInput, type SerperImageSearchResponse, type SerperNewsSearchInput, type SerperNewsSearchResponse, type SerperPlacesSearchInput, type SerperPlacesSearchResponse, type SerperShoppingSearchInput, type SerperShoppingSearchResponse, type SerperVideoSearchInput, type SerperVideoSearchResponse, type SerperWebSearchInput, type SerperWebSearchResponse, type SerperWebpageScrapeInput, type SerperWebpageScrapeResponse, type SourcesConfig, type ToolConfigSnapshot, type ToolDependencies, type ToolLogger, type ToolSummaryFn, type ToolWithSummary, type Variant, type VideoStats, type WebFetchInput, type YoutubeConfig, type YoutubeSearchResponse, type YoutubeVideoSearchInput, type YoutubeVideosResponse, applyLocaleParams, applyRecencyParam, brightDataImageSearchSchema, brightDataNewsSearchSchema, brightDataPlacesSearchSchema, brightDataShoppingSearchSchema, brightDataVideoSearchSchema, brightDataWebSearchSchema, brightDataWebpageScrapeSchema, buildGoogleUrl, buildYoutubeThumbnailUrl, createBrightDataImageSearch, createBrightDataNewsSearch, createBrightDataPlacesSearch, createBrightDataShoppingSearch, createBrightDataVideoSearch, createBrightDataWebSearch, createBrightDataWebpageScrape, createEncyclopediaReadTool, createEncyclopediaSearchTool, createMemoryCognitionForgetTool, createMemoryCognitionRememberTool, createMemoryPartitionDeleteTool, createMemoryPartitionRecallTool, createMemoryPartitionRememberTool, createSerperBusinessReviewsSearch, createSerperImageSearch, createSerperNewsSearch, createSerperPlacesSearch, createSerperShoppingSearch, createSerperVideoSearch, createSerperWebSearch, createSerperWebpageScrape, createVariantRequestTool, createWebFetchTool, createYoutubeVideoSearch, defaultSummarize, encyclopediaReadSchema, encyclopediaSearchSchema, engineEnabled, extractArticleText, fetchWithTimeout, isGoogleHostUrl, localizedQuerySuffix, meetsMinimumImageDimensions, memoryCognitionForgetSchema, memoryCognitionRememberSchema, memoryPartitionDeleteSchema, memoryPartitionRecallSchema, memoryPartitionRememberSchema, pickMerchantResult, requestBrightData, resolveSerperShopOfferLinks, serperBusinessReviewsSearchSchema, serperImageSearchSchema, serperNewsSearchSchema, serperPlacesSearchSchema, serperShoppingSearchSchema, serperVideoSearchSchema, serperWebSearchSchema, serperWebpageScrapeSchema, storeHostToken, summarizeContent, summarizeFound, summarizeResults, tbsSizeLabelForPixels, variantRequestSchema, webFetchSchema, withSummary, youtubeVideoSearchSchema };

package/dist/tools/index.mjs

@@ -719,6 +719,65 @@ function createBrightDataWebpageScrape(deps) {
     }
   });
 }
+var encyclopediaReadSchema = z.object({
+  url: z.string().min(1).max(2e3).describe(
+    "The exact url of the document to read \u2014 taken from an encyclopedia-search result (uploaded documents included)."
+  ),
+  startChunk: z.number().int().min(0).optional().describe(
+    "Chunk index to start reading from (default 0). To continue a previous read, pass the next chunk index named in its footer."
+  ),
+  maxChars: z.number().int().min(1e3).max(48e3).optional().describe("Maximum chars of content to return for this read (default 12000).")
+});
+function createEncyclopediaReadTool(deps) {
+  return tool({
+    description: "Read one knowledge-base document in original order \u2014 the deep-dive follow-up to encyclopedia-search. Pass a result\u2019s exact url to pull the surrounding and remaining content of that document (a fetched page, or the extracted text of an uploaded PDF, receipt, or instruction document) without a live fetch. Use it when a knowledge-base passage looks incomplete, when the user asks for more detail or a deep dive into one stored document, or to scan the whole document section by section. Returns a verbatim window; when the footer says more content follows, call again with startChunk set to the named next index to continue reading.",
+    inputSchema: encyclopediaReadSchema,
+    execute: async ({ url, startChunk, maxChars }) => {
+      const result = await deps.readDocument({ url, offset: startChunk, maxChars });
+      if (!result) {
+        return "No document found in the knowledge base for this url \u2014 pass the exact url of an encyclopedia-search result.";
+      }
+      const footer = result.hasMore ? `More of this document follows \u2014 call encyclopedia-read again with startChunk=${result.toChunk + 1} to continue.` : `End of document (${result.totalChunks} chunks total).`;
+      const title = result.title ? ` "${result.title}"` : "";
+      return `YOUR KNOWLEDGE BASE \u2014 DOCUMENT${title} \u2014 ${result.url}
+Verbatim content, chunks ${result.fromChunk + 1}\u2013${result.toChunk + 1} of ${result.totalChunks} (cite this url when you use it; upload urls open/download the original file):
+${result.content}
+
+${footer}`;
+    }
+  });
+}
+var encyclopediaSearchSchema = z.object({
+  query: z.string().min(1).max(1e3).describe(
+    'What to look up in the knowledge base, as a natural-language question or keywords, e.g. "receipt hardware store march" or "graphrag retriever patterns".'
+  ),
+  topK: z.number().int().min(1).max(10).optional().describe("Maximum number of passages to return (default 5)."),
+  url: z.string().min(1).max(2e3).optional().describe(
+    "Restrict the search to ONE document \u2014 pass the exact url of a previous knowledge-base result to search only within that document\u2019s embeddings."
+  ),
+  domain: z.string().min(1).max(200).optional().describe('Restrict the search to one source domain, e.g. "reddit.com".')
+});
+function createEncyclopediaSearchTool(deps) {
+  return tool({
+    description: "Search your knowledge base (encyclopedia): VERBATIM passages from webpages you fetched and documents the user uploaded in past sessions \u2014 each with its source url (upload urls open/download the original file) and the fetch date as the freshness signal. Consult it whenever a request might be answerable from knowledge you already have, and cite the url/title of what you use. It complements live web search, never replaces it: weigh the dates, and still search the web when recency, updates, or coverage gaps matter \u2014 fresh fetches are how this knowledge base closes its gaps and stays current. Scope options: pass a known result url to search within that one document only, or a domain to search one source. Passages may be fragments \u2014 pass a result url to the encyclopedia-read tool to pull the surrounding or full document (PDF, receipt, instructions).",
+    inputSchema: encyclopediaSearchSchema,
+    execute: async ({ query, topK, url, domain }) => {
+      const hits = await deps.search({ text: query, url, domain, limit: topK ?? 5 });
+      if (hits.length === 0) {
+        return url ? "No matching knowledge found inside that document \u2014 check the url, or fall back to a live web search/fetch." : "No matching knowledge found in your knowledge base (no previously fetched page or uploaded document matched). Answer from a live web search instead \u2014 its sources also grow the knowledge base.";
+      }
+      const lines = hits.map((hit) => {
+        const title = hit.title ? `"${hit.title}"` : "Untitled";
+        const fetched = hit.fetchedAt ? `, fetched ${hit.fetchedAt.slice(0, 10)}` : "";
+        const position = hit.sourceType === "result" ? "search snippet \u2014 NOT full text (fetch the page for more)" : `chunk ${hit.chunkIndex + 1} of ${hit.chunkCount}`;
+        return `- ${title} \u2014 ${hit.url} [${position}; relevance ${hit.score.toFixed(2)}${fetched}]
+  ${hit.content}`;
+      });
+      return `YOUR KNOWLEDGE BASE (encyclopedia) \u2014 verbatim passages from pages you fetched and documents the user uploaded in earlier sessions. Cite the url/title of what you use so the user can open the source (upload urls open/download the original file). Full-text chunks are fragments of their document \u2014 pass a result's exact url to the encyclopedia-read tool when you need the surrounding or full document:
+${lines.join("\n")}`;
+    }
+  });
+}
 
 // src/tools/helpers/is-google-host-url.helper.ts
 function isGoogleHostUrl(url) {
@@ -1606,14 +1665,11 @@ var defaultSummarize = (data) => {
   if (typeof data.content === "string") return summarizeContent(data);
   return {};
 };
-function withSummary(tool24, summarize = defaultSummarize) {
-  const t = tool24;
+function withSummary(tool26, summarize = defaultSummarize) {
+  const t = tool26;
   t.summarize = summarize;
   return t;
 }
-var webFetchSchema = z.object({
-  url: z.string().describe("The URL to fetch content from")
-});
 function extractArticleText(html) {
   if (!html.trim()) return "";
   const { document } = parseHTML(html);
@@ -1625,8 +1681,9 @@ function extractArticleText(html) {
   }
   return turndown.turndown(document.body?.innerHTML ?? "");
 }
-
-// src/tools/web-fetch/web-fetch.tool.ts
+var webFetchSchema = z.object({
+  url: z.string().describe("The URL to fetch content from")
+});
 function createWebFetchTool() {
   return tool({
     description: "Fetch the full content of a specific URL (e.g. a search result page). Returns the main article text as Markdown (boilerplate removed). Prefer fetching the most relevant search results to ground the answer in full source text.",
@@ -1792,4 +1849,4 @@ function createYoutubeVideoSearch(deps) {
   });
 }
 
-export { BRIGHT_DATA_TIMEOUT_MS, HEADERS, MIN_IMAGE_HEIGHT, MIN_IMAGE_WIDTH, RECENCY_DESCRIPTION, SEARCH_RETRIES, SEARCH_TIMEOUT_MS, SOURCE, STANDALONE_QUERY_DESCRIPTION, STANDALONE_QUERY_TOOL_CLAUSE, applyLocaleParams, applyRecencyParam, brightDataImageSearchSchema, brightDataNewsSearchSchema, brightDataPlacesSearchSchema, brightDataShoppingSearchSchema, brightDataVideoSearchSchema, brightDataWebSearchSchema, brightDataWebpageScrapeSchema, buildGoogleUrl, buildYoutubeThumbnailUrl, createBrightDataImageSearch, createBrightDataNewsSearch, createBrightDataPlacesSearch, createBrightDataShoppingSearch, createBrightDataVideoSearch, createBrightDataWebSearch, createBrightDataWebpageScrape, createMemoryCognitionForgetTool, createMemoryCognitionRememberTool, createMemoryPartitionDeleteTool, createMemoryPartitionRecallTool, createMemoryPartitionRememberTool, createSerperBusinessReviewsSearch, createSerperImageSearch, createSerperNewsSearch, createSerperPlacesSearch, createSerperShoppingSearch, createSerperVideoSearch, createSerperWebSearch, createSerperWebpageScrape, createVariantRequestTool, createWebFetchTool, createYoutubeVideoSearch, defaultSummarize, engineEnabled, fetchWithTimeout, isGoogleHostUrl, localizedQuerySuffix, meetsMinimumImageDimensions, memoryCognitionForgetSchema, memoryCognitionRememberSchema, memoryPartitionDeleteSchema, memoryPartitionRecallSchema, memoryPartitionRememberSchema, pickMerchantResult, requestBrightData, resolveSerperShopOfferLinks, serperBusinessReviewsSearchSchema, serperImageSearchSchema, serperNewsSearchSchema, serperPlacesSearchSchema, serperShoppingSearchSchema, serperVideoSearchSchema, serperWebSearchSchema, serperWebpageScrapeSchema, storeHostToken, summarizeContent, summarizeFound, summarizeResults, tbsSizeLabelForPixels, variantRequestSchema, webFetchSchema, withSummary, youtubeVideoSearchSchema };
+export { BRIGHT_DATA_TIMEOUT_MS, HEADERS, MIN_IMAGE_HEIGHT, MIN_IMAGE_WIDTH, RECENCY_DESCRIPTION, SEARCH_RETRIES, SEARCH_TIMEOUT_MS, SOURCE, STANDALONE_QUERY_DESCRIPTION, STANDALONE_QUERY_TOOL_CLAUSE, applyLocaleParams, applyRecencyParam, brightDataImageSearchSchema, brightDataNewsSearchSchema, brightDataPlacesSearchSchema, brightDataShoppingSearchSchema, brightDataVideoSearchSchema, brightDataWebSearchSchema, brightDataWebpageScrapeSchema, buildGoogleUrl, buildYoutubeThumbnailUrl, createBrightDataImageSearch, createBrightDataNewsSearch, createBrightDataPlacesSearch, createBrightDataShoppingSearch, createBrightDataVideoSearch, createBrightDataWebSearch, createBrightDataWebpageScrape, createEncyclopediaReadTool, createEncyclopediaSearchTool, createMemoryCognitionForgetTool, createMemoryCognitionRememberTool, createMemoryPartitionDeleteTool, createMemoryPartitionRecallTool, createMemoryPartitionRememberTool, createSerperBusinessReviewsSearch, createSerperImageSearch, createSerperNewsSearch, createSerperPlacesSearch, createSerperShoppingSearch, createSerperVideoSearch, createSerperWebSearch, createSerperWebpageScrape, createVariantRequestTool, createWebFetchTool, createYoutubeVideoSearch, defaultSummarize, encyclopediaReadSchema, encyclopediaSearchSchema, engineEnabled, extractArticleText, fetchWithTimeout, isGoogleHostUrl, localizedQuerySuffix, meetsMinimumImageDimensions, memoryCognitionForgetSchema, memoryCognitionRememberSchema, memoryPartitionDeleteSchema, memoryPartitionRecallSchema, memoryPartitionRememberSchema, pickMerchantResult, requestBrightData, resolveSerperShopOfferLinks, serperBusinessReviewsSearchSchema, serperImageSearchSchema, serperNewsSearchSchema, serperPlacesSearchSchema, serperShoppingSearchSchema, serperVideoSearchSchema, serperWebSearchSchema, serperWebpageScrapeSchema, storeHostToken, summarizeContent, summarizeFound, summarizeResults, tbsSizeLabelForPixels, variantRequestSchema, webFetchSchema, withSummary, youtubeVideoSearchSchema };

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "@triplef/agent",
   "description": "tripleF (3F) agent domain — structured-output schemas, prompt builders, and model tools shared across the apps.",
-  "version": "0.1.17",
+  "version": "0.1.19",
   "packageManager": "pnpm@11.25.0",
   "license": "MIT",
   "type": "module",
@@ -60,7 +60,7 @@
     "turndown": "^7.2.4"
   },
   "peerDependencies": {
-    "ai": "^7.0.90",
+    "ai": "^7.0.91",
     "zod": "^4.5.4"
   },
   "peerDependenciesMeta": {