@umituz/web-ai 1.0.8 → 1.0.9

package/dist/core/index.d.cts

@@ -233,6 +233,12 @@ interface ChunkOptions {
 }
 declare function chunkText(text: string, options?: ChunkOptions): string[];
 
+declare const TEXT_EXTRACT_PER_FILE_LIMIT = 24000;
+declare const TEXT_EXTRACT_TOTAL_LIMIT = 48000;
+declare function isTextLike(mimeType: string, name: string): boolean;
+declare function extractText(bytes: Uint8Array): string;
+declare function capText(text: string, limit: number): string;
+
 interface AccessTimestamped {
     readonly lastAccess: number;
 }
@@ -476,7 +482,7 @@ interface MemoryLike {
     readonly pinned: boolean;
     readonly updatedAt: Date | string | number;
 }
-declare function selectMemories(rows: readonly MemoryLike[], maxItems?: number, totalLimit?: number): string[];
+declare function selectMemories(rows: readonly MemoryLike[], maxItems?: number, totalLimit?: number, capper?: (text: string, limit: number) => string): string[];
 
 declare const RESEARCH_QUERY_MIN = 4;
 declare const RESEARCH_QUERY_MAX = 400;
@@ -626,4 +632,4 @@ declare const DEFAULT_RETRY_MAX_DELAY_MS = 10000;
 declare const DEFAULT_RETRY_BACKOFF_MULTIPLIER = 2;
 declare const DEFAULT_RETRY_JITTER_FACTOR = 0.1;
 
-export { AIError, type AIErrorBody, AIErrorMetadata, AIErrorType, type AccessCounted, type AccessTimestamped, type AgentContext, AiError, type AiErrorCode, type ApiKeyValidationRule, type AssistantToolCall, CHUNK_MAX_CHARS, CHUNK_OVERLAP_CHARS, CHUNK_TEXT_LIMIT, CODE_MESSAGES, ChatMessage, type ChunkOptions, CompositeExecutor, type ContextBudget, type ContextMessage, type ContextRole, DEFAULT_AI_IDENTITY, DEFAULT_API_KEY_RULE, DEFAULT_EMBED_BATCH_SIZE, DEFAULT_ERROR_CLASSIFICATION_RULE, DEFAULT_GENERATION, DEFAULT_HTTP_STATUS_RULE, DEFAULT_RETRY_BACKOFF_MULTIPLIER, DEFAULT_RETRY_INITIAL_DELAY_MS, DEFAULT_RETRY_JITTER_FACTOR, DEFAULT_RETRY_MAX_ATTEMPTS, DEFAULT_RETRY_MAX_DELAY_MS, DEFAULT_TASK_POLICIES, DebounceManager, type DebounceOptions, DedupeExecutor, type DedupeExecutorOptions, type DiffChangeCount, type DiffLine, type EmbeddingsBatchAdapter, type ErrorClassificationRule, type ExponentialBackoffParams, type FailoverErrorLike, type FetchLike, GENERATION_BOUNDS, type GenerationSettings, GroqError, type GroqErrorMetadata, AIErrorType as GroqErrorType, type HttpStatusClassificationRule, type IRequestExecutor, JSON_CODE_BLOCK_PATTERN, JSON_OBJECT_PATTERN, JsonResponseParser, type LeadingEdgeParams, type LoggedError, MEMORY_MAX_ITEMS, MEMORY_TOTAL_LIMIT, MICROS_PER_UNIT, type MarkdownSegment, type MemoryLike, type ModelPricing, type ModelWithProvider, type NormalizedSseChunk, type ParsedSseEvent, type ProjectContext, type ProviderErrorShape, QueueExecutor, RESEARCH_MAX_OUTPUT_TOKENS, RESEARCH_QUERY_MAX, RESEARCH_QUERY_MIN, RESEARCH_REPORT_MIN_CHARS, RESEARCH_SNIPPET_MAX_CHARS, RESEARCH_TEMPERATURE, RequestDeduplicator, RequestQueue, type ResearchSource, type ResearchStreamEvent, RetryExecutor, RetryManager, RetryOptions, SEARCH_MAX_RESULTS, SSE_BRIDGE_EVENTS, SSE_DONE_MARKER, SSE_PREFIX, STATUS_TO_CODE, SYSTEM_PROMPT_VERSION, type SearchTopic, type SseDeltaExtractor, type SseResponseOptions, type StreamCallbacks, type StreamEvent, type StreamingParserCallbacks, StreamingResponseParser, StructuredPromptBuilder, type SystemPromptInput, TRANSIENT_ERROR_CODES, type TaskPlan, type TaskPolicy, type TaskState, ThrottleManager, type ThrottleOptions, TokenUsage, ToolCallAccumulator, type ToolSchema, type ToolValidationResult, type ToolValidatorFunction, type TrailingDelayParams, type UsageRecordData, type UsageRecordInput, type VectorCandidate, type VectorMatch, type WebSearchInput, type WebSearchResponse, type WebSearchResult, type WireToolCallDelta, type ZodLikeIssue, type ZodLikeSchema, accumulateReport, aiErrorMessage, asObjectArray, asRecord, asStringArray, asTrimmedOptionalString, batchEmbed, buildCacheKey, buildContext, buildHttpStatusFallback, buildResearchMessages, buildSystemPrompt, buildUsageData, calculateCacheHitRate, calculateExponentialBackoff, calculateTrailingDelay, canFailover, canStartTask, chunkText, clampScore, classifyHttpStatus, computeCostMicros, cosineSimilarity, countChanges, createSseResponse, debounce, debounceManager, defaultSseDeltaExtractor, deriveSnippetTitle, describeAIErrorType, diffLines, dotProduct, errorFromStatus, errorFromUnknown, estimateMessageTokens, estimateTokens, extractErrorMessageFromJsonBody, extractErrorMessageFromResponse, extractJsonArray, extractJsonCandidate, extractJsonFromCodeFence, extractJsonObject, extractSseDataPayload, failoverCandidate, findTopKSimilar, formatCost, formatErrorForLogging, formatSseEvent, generateRandomBase36Suffix, isAuthErrorType, isRetryableErrorType, isSseDataLine, isSseDoneMarker, isTransientError, isValidApiKey, isWithinWindow, magnitude, nextAttemptDelayMs, normalizeGenerationSettings, normalizeSources, normalizeSseChunk, oneOfCatalog, parseSseEvent, planTaskFailure, readSseResponse, repairCommonJsonDefects, replayFirst, resolveUserFriendlyErrorMessage, resolveUserFriendlyMessage, searchWeb, selectLeastRecentlyUsedKey, selectMemories, shouldCallOnLeadingEdge, splitMarkdown, splitSseLines, sseStreamResponse, throttle, throttleManager, validateApiKey, validateToolArgs };
+export { AIError, type AIErrorBody, AIErrorMetadata, AIErrorType, type AccessCounted, type AccessTimestamped, type AgentContext, AiError, type AiErrorCode, type ApiKeyValidationRule, type AssistantToolCall, CHUNK_MAX_CHARS, CHUNK_OVERLAP_CHARS, CHUNK_TEXT_LIMIT, CODE_MESSAGES, ChatMessage, type ChunkOptions, CompositeExecutor, type ContextBudget, type ContextMessage, type ContextRole, DEFAULT_AI_IDENTITY, DEFAULT_API_KEY_RULE, DEFAULT_EMBED_BATCH_SIZE, DEFAULT_ERROR_CLASSIFICATION_RULE, DEFAULT_GENERATION, DEFAULT_HTTP_STATUS_RULE, DEFAULT_RETRY_BACKOFF_MULTIPLIER, DEFAULT_RETRY_INITIAL_DELAY_MS, DEFAULT_RETRY_JITTER_FACTOR, DEFAULT_RETRY_MAX_ATTEMPTS, DEFAULT_RETRY_MAX_DELAY_MS, DEFAULT_TASK_POLICIES, DebounceManager, type DebounceOptions, DedupeExecutor, type DedupeExecutorOptions, type DiffChangeCount, type DiffLine, type EmbeddingsBatchAdapter, type ErrorClassificationRule, type ExponentialBackoffParams, type FailoverErrorLike, type FetchLike, GENERATION_BOUNDS, type GenerationSettings, GroqError, type GroqErrorMetadata, AIErrorType as GroqErrorType, type HttpStatusClassificationRule, type IRequestExecutor, JSON_CODE_BLOCK_PATTERN, JSON_OBJECT_PATTERN, JsonResponseParser, type LeadingEdgeParams, type LoggedError, MEMORY_MAX_ITEMS, MEMORY_TOTAL_LIMIT, MICROS_PER_UNIT, type MarkdownSegment, type MemoryLike, type ModelPricing, type ModelWithProvider, type NormalizedSseChunk, type ParsedSseEvent, type ProjectContext, type ProviderErrorShape, QueueExecutor, RESEARCH_MAX_OUTPUT_TOKENS, RESEARCH_QUERY_MAX, RESEARCH_QUERY_MIN, RESEARCH_REPORT_MIN_CHARS, RESEARCH_SNIPPET_MAX_CHARS, RESEARCH_TEMPERATURE, RequestDeduplicator, RequestQueue, type ResearchSource, type ResearchStreamEvent, RetryExecutor, RetryManager, RetryOptions, SEARCH_MAX_RESULTS, SSE_BRIDGE_EVENTS, SSE_DONE_MARKER, SSE_PREFIX, STATUS_TO_CODE, SYSTEM_PROMPT_VERSION, type SearchTopic, type SseDeltaExtractor, type SseResponseOptions, type StreamCallbacks, type StreamEvent, type StreamingParserCallbacks, StreamingResponseParser, StructuredPromptBuilder, type SystemPromptInput, TEXT_EXTRACT_PER_FILE_LIMIT, TEXT_EXTRACT_TOTAL_LIMIT, TRANSIENT_ERROR_CODES, type TaskPlan, type TaskPolicy, type TaskState, ThrottleManager, type ThrottleOptions, TokenUsage, ToolCallAccumulator, type ToolSchema, type ToolValidationResult, type ToolValidatorFunction, type TrailingDelayParams, type UsageRecordData, type UsageRecordInput, type VectorCandidate, type VectorMatch, type WebSearchInput, type WebSearchResponse, type WebSearchResult, type WireToolCallDelta, type ZodLikeIssue, type ZodLikeSchema, accumulateReport, aiErrorMessage, asObjectArray, asRecord, asStringArray, asTrimmedOptionalString, batchEmbed, buildCacheKey, buildContext, buildHttpStatusFallback, buildResearchMessages, buildSystemPrompt, buildUsageData, calculateCacheHitRate, calculateExponentialBackoff, calculateTrailingDelay, canFailover, canStartTask, capText, chunkText, clampScore, classifyHttpStatus, computeCostMicros, cosineSimilarity, countChanges, createSseResponse, debounce, debounceManager, defaultSseDeltaExtractor, deriveSnippetTitle, describeAIErrorType, diffLines, dotProduct, errorFromStatus, errorFromUnknown, estimateMessageTokens, estimateTokens, extractErrorMessageFromJsonBody, extractErrorMessageFromResponse, extractJsonArray, extractJsonCandidate, extractJsonFromCodeFence, extractJsonObject, extractSseDataPayload, extractText, failoverCandidate, findTopKSimilar, formatCost, formatErrorForLogging, formatSseEvent, generateRandomBase36Suffix, isAuthErrorType, isRetryableErrorType, isSseDataLine, isSseDoneMarker, isTextLike, isTransientError, isValidApiKey, isWithinWindow, magnitude, nextAttemptDelayMs, normalizeGenerationSettings, normalizeSources, normalizeSseChunk, oneOfCatalog, parseSseEvent, planTaskFailure, readSseResponse, repairCommonJsonDefects, replayFirst, resolveUserFriendlyErrorMessage, resolveUserFriendlyMessage, searchWeb, selectLeastRecentlyUsedKey, selectMemories, shouldCallOnLeadingEdge, splitMarkdown, splitSseLines, sseStreamResponse, throttle, throttleManager, validateApiKey, validateToolArgs };

package/dist/core/index.d.ts

@@ -233,6 +233,12 @@ interface ChunkOptions {
 }
 declare function chunkText(text: string, options?: ChunkOptions): string[];
 
+declare const TEXT_EXTRACT_PER_FILE_LIMIT = 24000;
+declare const TEXT_EXTRACT_TOTAL_LIMIT = 48000;
+declare function isTextLike(mimeType: string, name: string): boolean;
+declare function extractText(bytes: Uint8Array): string;
+declare function capText(text: string, limit: number): string;
+
 interface AccessTimestamped {
     readonly lastAccess: number;
 }
@@ -476,7 +482,7 @@ interface MemoryLike {
     readonly pinned: boolean;
     readonly updatedAt: Date | string | number;
 }
-declare function selectMemories(rows: readonly MemoryLike[], maxItems?: number, totalLimit?: number): string[];
+declare function selectMemories(rows: readonly MemoryLike[], maxItems?: number, totalLimit?: number, capper?: (text: string, limit: number) => string): string[];
 
 declare const RESEARCH_QUERY_MIN = 4;
 declare const RESEARCH_QUERY_MAX = 400;
@@ -626,4 +632,4 @@ declare const DEFAULT_RETRY_MAX_DELAY_MS = 10000;
 declare const DEFAULT_RETRY_BACKOFF_MULTIPLIER = 2;
 declare const DEFAULT_RETRY_JITTER_FACTOR = 0.1;
 
-export { AIError, type AIErrorBody, AIErrorMetadata, AIErrorType, type AccessCounted, type AccessTimestamped, type AgentContext, AiError, type AiErrorCode, type ApiKeyValidationRule, type AssistantToolCall, CHUNK_MAX_CHARS, CHUNK_OVERLAP_CHARS, CHUNK_TEXT_LIMIT, CODE_MESSAGES, ChatMessage, type ChunkOptions, CompositeExecutor, type ContextBudget, type ContextMessage, type ContextRole, DEFAULT_AI_IDENTITY, DEFAULT_API_KEY_RULE, DEFAULT_EMBED_BATCH_SIZE, DEFAULT_ERROR_CLASSIFICATION_RULE, DEFAULT_GENERATION, DEFAULT_HTTP_STATUS_RULE, DEFAULT_RETRY_BACKOFF_MULTIPLIER, DEFAULT_RETRY_INITIAL_DELAY_MS, DEFAULT_RETRY_JITTER_FACTOR, DEFAULT_RETRY_MAX_ATTEMPTS, DEFAULT_RETRY_MAX_DELAY_MS, DEFAULT_TASK_POLICIES, DebounceManager, type DebounceOptions, DedupeExecutor, type DedupeExecutorOptions, type DiffChangeCount, type DiffLine, type EmbeddingsBatchAdapter, type ErrorClassificationRule, type ExponentialBackoffParams, type FailoverErrorLike, type FetchLike, GENERATION_BOUNDS, type GenerationSettings, GroqError, type GroqErrorMetadata, AIErrorType as GroqErrorType, type HttpStatusClassificationRule, type IRequestExecutor, JSON_CODE_BLOCK_PATTERN, JSON_OBJECT_PATTERN, JsonResponseParser, type LeadingEdgeParams, type LoggedError, MEMORY_MAX_ITEMS, MEMORY_TOTAL_LIMIT, MICROS_PER_UNIT, type MarkdownSegment, type MemoryLike, type ModelPricing, type ModelWithProvider, type NormalizedSseChunk, type ParsedSseEvent, type ProjectContext, type ProviderErrorShape, QueueExecutor, RESEARCH_MAX_OUTPUT_TOKENS, RESEARCH_QUERY_MAX, RESEARCH_QUERY_MIN, RESEARCH_REPORT_MIN_CHARS, RESEARCH_SNIPPET_MAX_CHARS, RESEARCH_TEMPERATURE, RequestDeduplicator, RequestQueue, type ResearchSource, type ResearchStreamEvent, RetryExecutor, RetryManager, RetryOptions, SEARCH_MAX_RESULTS, SSE_BRIDGE_EVENTS, SSE_DONE_MARKER, SSE_PREFIX, STATUS_TO_CODE, SYSTEM_PROMPT_VERSION, type SearchTopic, type SseDeltaExtractor, type SseResponseOptions, type StreamCallbacks, type StreamEvent, type StreamingParserCallbacks, StreamingResponseParser, StructuredPromptBuilder, type SystemPromptInput, TRANSIENT_ERROR_CODES, type TaskPlan, type TaskPolicy, type TaskState, ThrottleManager, type ThrottleOptions, TokenUsage, ToolCallAccumulator, type ToolSchema, type ToolValidationResult, type ToolValidatorFunction, type TrailingDelayParams, type UsageRecordData, type UsageRecordInput, type VectorCandidate, type VectorMatch, type WebSearchInput, type WebSearchResponse, type WebSearchResult, type WireToolCallDelta, type ZodLikeIssue, type ZodLikeSchema, accumulateReport, aiErrorMessage, asObjectArray, asRecord, asStringArray, asTrimmedOptionalString, batchEmbed, buildCacheKey, buildContext, buildHttpStatusFallback, buildResearchMessages, buildSystemPrompt, buildUsageData, calculateCacheHitRate, calculateExponentialBackoff, calculateTrailingDelay, canFailover, canStartTask, chunkText, clampScore, classifyHttpStatus, computeCostMicros, cosineSimilarity, countChanges, createSseResponse, debounce, debounceManager, defaultSseDeltaExtractor, deriveSnippetTitle, describeAIErrorType, diffLines, dotProduct, errorFromStatus, errorFromUnknown, estimateMessageTokens, estimateTokens, extractErrorMessageFromJsonBody, extractErrorMessageFromResponse, extractJsonArray, extractJsonCandidate, extractJsonFromCodeFence, extractJsonObject, extractSseDataPayload, failoverCandidate, findTopKSimilar, formatCost, formatErrorForLogging, formatSseEvent, generateRandomBase36Suffix, isAuthErrorType, isRetryableErrorType, isSseDataLine, isSseDoneMarker, isTransientError, isValidApiKey, isWithinWindow, magnitude, nextAttemptDelayMs, normalizeGenerationSettings, normalizeSources, normalizeSseChunk, oneOfCatalog, parseSseEvent, planTaskFailure, readSseResponse, repairCommonJsonDefects, replayFirst, resolveUserFriendlyErrorMessage, resolveUserFriendlyMessage, searchWeb, selectLeastRecentlyUsedKey, selectMemories, shouldCallOnLeadingEdge, splitMarkdown, splitSseLines, sseStreamResponse, throttle, throttleManager, validateApiKey, validateToolArgs };
+export { AIError, type AIErrorBody, AIErrorMetadata, AIErrorType, type AccessCounted, type AccessTimestamped, type AgentContext, AiError, type AiErrorCode, type ApiKeyValidationRule, type AssistantToolCall, CHUNK_MAX_CHARS, CHUNK_OVERLAP_CHARS, CHUNK_TEXT_LIMIT, CODE_MESSAGES, ChatMessage, type ChunkOptions, CompositeExecutor, type ContextBudget, type ContextMessage, type ContextRole, DEFAULT_AI_IDENTITY, DEFAULT_API_KEY_RULE, DEFAULT_EMBED_BATCH_SIZE, DEFAULT_ERROR_CLASSIFICATION_RULE, DEFAULT_GENERATION, DEFAULT_HTTP_STATUS_RULE, DEFAULT_RETRY_BACKOFF_MULTIPLIER, DEFAULT_RETRY_INITIAL_DELAY_MS, DEFAULT_RETRY_JITTER_FACTOR, DEFAULT_RETRY_MAX_ATTEMPTS, DEFAULT_RETRY_MAX_DELAY_MS, DEFAULT_TASK_POLICIES, DebounceManager, type DebounceOptions, DedupeExecutor, type DedupeExecutorOptions, type DiffChangeCount, type DiffLine, type EmbeddingsBatchAdapter, type ErrorClassificationRule, type ExponentialBackoffParams, type FailoverErrorLike, type FetchLike, GENERATION_BOUNDS, type GenerationSettings, GroqError, type GroqErrorMetadata, AIErrorType as GroqErrorType, type HttpStatusClassificationRule, type IRequestExecutor, JSON_CODE_BLOCK_PATTERN, JSON_OBJECT_PATTERN, JsonResponseParser, type LeadingEdgeParams, type LoggedError, MEMORY_MAX_ITEMS, MEMORY_TOTAL_LIMIT, MICROS_PER_UNIT, type MarkdownSegment, type MemoryLike, type ModelPricing, type ModelWithProvider, type NormalizedSseChunk, type ParsedSseEvent, type ProjectContext, type ProviderErrorShape, QueueExecutor, RESEARCH_MAX_OUTPUT_TOKENS, RESEARCH_QUERY_MAX, RESEARCH_QUERY_MIN, RESEARCH_REPORT_MIN_CHARS, RESEARCH_SNIPPET_MAX_CHARS, RESEARCH_TEMPERATURE, RequestDeduplicator, RequestQueue, type ResearchSource, type ResearchStreamEvent, RetryExecutor, RetryManager, RetryOptions, SEARCH_MAX_RESULTS, SSE_BRIDGE_EVENTS, SSE_DONE_MARKER, SSE_PREFIX, STATUS_TO_CODE, SYSTEM_PROMPT_VERSION, type SearchTopic, type SseDeltaExtractor, type SseResponseOptions, type StreamCallbacks, type StreamEvent, type StreamingParserCallbacks, StreamingResponseParser, StructuredPromptBuilder, type SystemPromptInput, TEXT_EXTRACT_PER_FILE_LIMIT, TEXT_EXTRACT_TOTAL_LIMIT, TRANSIENT_ERROR_CODES, type TaskPlan, type TaskPolicy, type TaskState, ThrottleManager, type ThrottleOptions, TokenUsage, ToolCallAccumulator, type ToolSchema, type ToolValidationResult, type ToolValidatorFunction, type TrailingDelayParams, type UsageRecordData, type UsageRecordInput, type VectorCandidate, type VectorMatch, type WebSearchInput, type WebSearchResponse, type WebSearchResult, type WireToolCallDelta, type ZodLikeIssue, type ZodLikeSchema, accumulateReport, aiErrorMessage, asObjectArray, asRecord, asStringArray, asTrimmedOptionalString, batchEmbed, buildCacheKey, buildContext, buildHttpStatusFallback, buildResearchMessages, buildSystemPrompt, buildUsageData, calculateCacheHitRate, calculateExponentialBackoff, calculateTrailingDelay, canFailover, canStartTask, capText, chunkText, clampScore, classifyHttpStatus, computeCostMicros, cosineSimilarity, countChanges, createSseResponse, debounce, debounceManager, defaultSseDeltaExtractor, deriveSnippetTitle, describeAIErrorType, diffLines, dotProduct, errorFromStatus, errorFromUnknown, estimateMessageTokens, estimateTokens, extractErrorMessageFromJsonBody, extractErrorMessageFromResponse, extractJsonArray, extractJsonCandidate, extractJsonFromCodeFence, extractJsonObject, extractSseDataPayload, extractText, failoverCandidate, findTopKSimilar, formatCost, formatErrorForLogging, formatSseEvent, generateRandomBase36Suffix, isAuthErrorType, isRetryableErrorType, isSseDataLine, isSseDoneMarker, isTextLike, isTransientError, isValidApiKey, isWithinWindow, magnitude, nextAttemptDelayMs, normalizeGenerationSettings, normalizeSources, normalizeSseChunk, oneOfCatalog, parseSseEvent, planTaskFailure, readSseResponse, repairCommonJsonDefects, replayFirst, resolveUserFriendlyErrorMessage, resolveUserFriendlyMessage, searchWeb, selectLeastRecentlyUsedKey, selectMemories, shouldCallOnLeadingEdge, splitMarkdown, splitSseLines, sseStreamResponse, throttle, throttleManager, validateApiKey, validateToolArgs };

package/dist/core/index.js

@@ -1132,6 +1132,83 @@ ${paragraph}`;
   return chunks.map((chunk) => chunk.length > maxChars ? chunk.slice(0, maxChars) : chunk).filter((chunk) => chunk.length > 0);
 }
 
+// src/core/parsers/text-extract.ts
+var TEXT_MIME_PREFIX = "text/";
+var TEXT_MIME_TYPES = /* @__PURE__ */ new Set([
+  "application/json",
+  "application/xml",
+  "application/yaml",
+  "application/x-yaml",
+  "application/javascript",
+  "application/typescript",
+  "application/x-sh",
+  "application/sql",
+  "text/csv"
+]);
+var TEXT_EXTENSIONS = /* @__PURE__ */ new Set([
+  "txt",
+  "md",
+  "markdown",
+  "csv",
+  "tsv",
+  "json",
+  "xml",
+  "yaml",
+  "yml",
+  "toml",
+  "ini",
+  "js",
+  "mjs",
+  "cjs",
+  "ts",
+  "tsx",
+  "jsx",
+  "py",
+  "rb",
+  "go",
+  "rs",
+  "java",
+  "kt",
+  "swift",
+  "c",
+  "h",
+  "cpp",
+  "hpp",
+  "cs",
+  "php",
+  "sh",
+  "bash",
+  "zsh",
+  "sql",
+  "html",
+  "css",
+  "scss",
+  "env"
+]);
+var TEXT_EXTRACT_PER_FILE_LIMIT = 24e3;
+var TEXT_EXTRACT_TOTAL_LIMIT = 48e3;
+function isTextLike(mimeType, name) {
+  if (mimeType.startsWith(TEXT_MIME_PREFIX) || TEXT_MIME_TYPES.has(mimeType)) return true;
+  const extension = name.split(".").pop()?.toLowerCase() ?? "";
+  return TEXT_EXTENSIONS.has(extension);
+}
+function extractText(bytes) {
+  let out = "";
+  const decoder = new TextDecoder("utf-8", { fatal: false });
+  const chunkSize = 65536;
+  for (let offset = 0; offset < bytes.length; offset += chunkSize) {
+    out += decoder.decode(bytes.subarray(offset, Math.min(offset + chunkSize, bytes.length)), {
+      stream: offset + chunkSize < bytes.length
+    });
+  }
+  return out.split("\0").join("");
+}
+function capText(text, limit) {
+  if (text.length <= limit) return text;
+  return `${text.slice(0, limit)}
+[truncated]`;
+}
+
 // src/core/calculations/random-base36-suffix.ts
 var BASE_36_RADIX = 36;
 var DECIMAL_OFFSET = 2;
@@ -1579,11 +1656,7 @@ function resolveTimestamp(value) {
   const parsed = Date.parse(value);
   return Number.isNaN(parsed) ? 0 : parsed;
 }
-function truncateString(str, maxLength) {
-  if (str.length <= maxLength) return str;
-  return str.slice(0, Math.max(maxLength, 0));
-}
-function selectMemories(rows, maxItems = MEMORY_MAX_ITEMS, totalLimit = MEMORY_TOTAL_LIMIT) {
+function selectMemories(rows, maxItems = MEMORY_MAX_ITEMS, totalLimit = MEMORY_TOTAL_LIMIT, capper = capText) {
   const ordered = [...rows].sort((a, b) => {
     if (a.pinned !== b.pinned) return a.pinned ? -1 : 1;
     return resolveTimestamp(b.updatedAt) - resolveTimestamp(a.updatedAt);
@@ -1594,7 +1667,7 @@ function selectMemories(rows, maxItems = MEMORY_MAX_ITEMS, totalLimit = MEMORY_T
     if (out.length >= maxItems || used >= totalLimit) break;
     const content = row.content.trim();
     if (content.length === 0) continue;
-    const capped = truncateString(content, totalLimit - used);
+    const capped = capper(content, totalLimit - used);
     out.push(capped);
     used += capped.length;
   }
@@ -1826,6 +1899,6 @@ function validateToolArgs(toolKey, args, schemas) {
   return { ok: false, error: `Invalid schema definition for tool: ${toolKey}` };
 }
 
-export { AIError, AIErrorType, AiError, CHUNK_MAX_CHARS, CHUNK_OVERLAP_CHARS, CHUNK_TEXT_LIMIT, CODE_MESSAGES, CacheManager, CompositeExecutor, DEFAULT_AI_IDENTITY, DEFAULT_API_KEY_RULE, DEFAULT_EMBED_BATCH_SIZE, DEFAULT_ERROR_CLASSIFICATION_RULE, DEFAULT_GENERATION, DEFAULT_HTTP_STATUS_RULE, DEFAULT_RETRY_BACKOFF_MULTIPLIER, DEFAULT_RETRY_INITIAL_DELAY_MS, DEFAULT_RETRY_JITTER_FACTOR, DEFAULT_RETRY_MAX_ATTEMPTS, DEFAULT_RETRY_MAX_DELAY_MS, DEFAULT_TASK_POLICIES, DebounceManager, DedupeExecutor, GENERATION_BOUNDS, GroqError, AIErrorType as GroqErrorType, JSON_CODE_BLOCK_PATTERN, JSON_OBJECT_PATTERN, JsonResponseParser, MEMORY_MAX_ITEMS, MEMORY_TOTAL_LIMIT, MICROS_PER_UNIT, QueueExecutor, RESEARCH_MAX_OUTPUT_TOKENS, RESEARCH_QUERY_MAX, RESEARCH_QUERY_MIN, RESEARCH_REPORT_MIN_CHARS, RESEARCH_SNIPPET_MAX_CHARS, RESEARCH_TEMPERATURE, RequestDeduplicator, RequestQueue, RetryExecutor, RetryManager, SEARCH_MAX_RESULTS, SSE_BRIDGE_EVENTS, SSE_DONE_MARKER, SSE_PREFIX, STATUS_TO_CODE, SYSTEM_PROMPT_VERSION, StreamingResponseParser, StructuredPromptBuilder, TEXT_TIERS, TRANSIENT_ERROR_CODES, ThrottleManager, ToolCallAccumulator, USER_FRIENDLY_ERROR_MESSAGES, accumulateReport, aiErrorMessage, asObjectArray, asRecord, asStringArray, asTrimmedOptionalString, batchEmbed, buildCacheKey, buildContext, buildHttpStatusFallback, buildResearchMessages, buildSystemPrompt, buildUsageData, cacheManager, calculateCacheHitRate, calculateExponentialBackoff, calculateTrailingDelay, canFailover, canStartTask, chunkText, clampScore, classifyHttpStatus, computeCostMicros, cosineSimilarity, countChanges, createSseResponse, debounce, debounceManager, defaultSseDeltaExtractor, deriveSnippetTitle, describeAIErrorType, diffLines, dotProduct, errorFromStatus, errorFromUnknown, estimateMessageTokens, estimateTokens, extractErrorMessageFromJsonBody, extractErrorMessageFromResponse, extractJsonArray, extractJsonCandidate, extractJsonFromCodeFence, extractJsonObject, extractSseDataPayload, failoverCandidate, findTopKSimilar, formatCost, formatErrorForLogging, formatSseEvent, generateRandomBase36Suffix, isAuthErrorType, isRetryableErrorType, isSseDataLine, isSseDoneMarker, isTextTier, isTransientError, isValidApiKey, isWithinWindow, magnitude, mergeTierOverrides, nextAttemptDelayMs, normalizeGenerationSettings, normalizeSources, normalizeSseChunk, oneOfCatalog, parseSseEvent, planTaskFailure, readSseResponse, repairCommonJsonDefects, replayFirst, requestDeduplicator, requestQueue, resolveModel, resolveUserFriendlyErrorMessage, resolveUserFriendlyMessage, retryManager, searchWeb, selectLeastRecentlyUsedKey, selectMemories, shouldCallOnLeadingEdge, splitMarkdown, splitSseLines, sseStreamResponse, throttle, throttleManager, validateApiKey, validateToolArgs, withRetry };
+export { AIError, AIErrorType, AiError, CHUNK_MAX_CHARS, CHUNK_OVERLAP_CHARS, CHUNK_TEXT_LIMIT, CODE_MESSAGES, CacheManager, CompositeExecutor, DEFAULT_AI_IDENTITY, DEFAULT_API_KEY_RULE, DEFAULT_EMBED_BATCH_SIZE, DEFAULT_ERROR_CLASSIFICATION_RULE, DEFAULT_GENERATION, DEFAULT_HTTP_STATUS_RULE, DEFAULT_RETRY_BACKOFF_MULTIPLIER, DEFAULT_RETRY_INITIAL_DELAY_MS, DEFAULT_RETRY_JITTER_FACTOR, DEFAULT_RETRY_MAX_ATTEMPTS, DEFAULT_RETRY_MAX_DELAY_MS, DEFAULT_TASK_POLICIES, DebounceManager, DedupeExecutor, GENERATION_BOUNDS, GroqError, AIErrorType as GroqErrorType, JSON_CODE_BLOCK_PATTERN, JSON_OBJECT_PATTERN, JsonResponseParser, MEMORY_MAX_ITEMS, MEMORY_TOTAL_LIMIT, MICROS_PER_UNIT, QueueExecutor, RESEARCH_MAX_OUTPUT_TOKENS, RESEARCH_QUERY_MAX, RESEARCH_QUERY_MIN, RESEARCH_REPORT_MIN_CHARS, RESEARCH_SNIPPET_MAX_CHARS, RESEARCH_TEMPERATURE, RequestDeduplicator, RequestQueue, RetryExecutor, RetryManager, SEARCH_MAX_RESULTS, SSE_BRIDGE_EVENTS, SSE_DONE_MARKER, SSE_PREFIX, STATUS_TO_CODE, SYSTEM_PROMPT_VERSION, StreamingResponseParser, StructuredPromptBuilder, TEXT_EXTRACT_PER_FILE_LIMIT, TEXT_EXTRACT_TOTAL_LIMIT, TEXT_TIERS, TRANSIENT_ERROR_CODES, ThrottleManager, ToolCallAccumulator, USER_FRIENDLY_ERROR_MESSAGES, accumulateReport, aiErrorMessage, asObjectArray, asRecord, asStringArray, asTrimmedOptionalString, batchEmbed, buildCacheKey, buildContext, buildHttpStatusFallback, buildResearchMessages, buildSystemPrompt, buildUsageData, cacheManager, calculateCacheHitRate, calculateExponentialBackoff, calculateTrailingDelay, canFailover, canStartTask, capText, chunkText, clampScore, classifyHttpStatus, computeCostMicros, cosineSimilarity, countChanges, createSseResponse, debounce, debounceManager, defaultSseDeltaExtractor, deriveSnippetTitle, describeAIErrorType, diffLines, dotProduct, errorFromStatus, errorFromUnknown, estimateMessageTokens, estimateTokens, extractErrorMessageFromJsonBody, extractErrorMessageFromResponse, extractJsonArray, extractJsonCandidate, extractJsonFromCodeFence, extractJsonObject, extractSseDataPayload, extractText, failoverCandidate, findTopKSimilar, formatCost, formatErrorForLogging, formatSseEvent, generateRandomBase36Suffix, isAuthErrorType, isRetryableErrorType, isSseDataLine, isSseDoneMarker, isTextLike, isTextTier, isTransientError, isValidApiKey, isWithinWindow, magnitude, mergeTierOverrides, nextAttemptDelayMs, normalizeGenerationSettings, normalizeSources, normalizeSseChunk, oneOfCatalog, parseSseEvent, planTaskFailure, readSseResponse, repairCommonJsonDefects, replayFirst, requestDeduplicator, requestQueue, resolveModel, resolveUserFriendlyErrorMessage, resolveUserFriendlyMessage, retryManager, searchWeb, selectLeastRecentlyUsedKey, selectMemories, shouldCallOnLeadingEdge, splitMarkdown, splitSseLines, sseStreamResponse, throttle, throttleManager, validateApiKey, validateToolArgs, withRetry };
 //# sourceMappingURL=index.js.map
 //# sourceMappingURL=index.js.map