@exulu/backend 1.68.0 → 1.69.0

package/dist/index.d.cts

@@ -1,5 +1,6 @@
 import * as _opentelemetry_sdk_node from '@opentelemetry/sdk-node';
 import * as knex from 'knex';
+import { Knex } from 'knex';
 import { RedisClientType } from 'redis';
 import * as bullmq from 'bullmq';
 import { Queue } from 'bullmq';
@@ -148,6 +149,7 @@ interface ExuluAgent {
     instructions?: string;
     feedback?: boolean;
     suggestions_enabled?: boolean;
+    sandbox_enabled?: boolean;
     slug?: string;
     tools?: {
         id: string;
@@ -786,6 +788,9 @@ declare class ExuluContext {
             after?: number;
         };
         entityFilter?: EntityFilter;
+        /** Precomputed query embedding; forwarded to vectorSearch to skip re-embedding (caller must use
+         *  this context's embedding model). Spread into vectorSearch via `...options` below. */
+        queryEmbedding?: number[];
     }) => Promise<{
         itemFilters: SearchFilters;
         chunkFilters: SearchFilters;
@@ -828,9 +833,11 @@ declare class ExuluContext {
     getItem: ({ item }: {
         item: Item;
     }) => Promise<Item>;
-    getItems: ({ filters, fields, }: {
+    getItems: ({ filters, fields, user, role, }: {
         filters?: any[];
         fields?: string[];
+        user?: User;
+        role?: string;
     }) => Promise<Item[]>;
     embeddings: {
         generate: {
@@ -1804,6 +1811,50 @@ declare class RecursiveChunker extends BaseChunker {
     toString(): string;
 }
 
+interface AuthorizedReadOpts {
+    itemIds?: string[];
+    externalIds?: string[];
+    chunkIndexRange?: {
+        from?: number;
+        to?: number;
+    };
+}
+/**
+ * Supported, RBAC-safe read API for retrieval clients (e.g. the agentic harness).
+ * Relevance + visibility still go through ExuluContext.search(); this namespace
+ * adds the read shapes search() cannot express, always over authorized rows.
+ */
+declare const ExuluReadApi: {
+    getTableName: (id: string) => string;
+    getChunksTableName: (id: string) => string;
+    getEntitiesTableName: (id: string) => string;
+    getChunkEntitiesTableName: (id: string) => string;
+    entitiesAvailable: (context: {
+        id: string;
+        entities?: unknown;
+    }) => Promise<boolean>;
+    authorizedRead: (context: {
+        id: string;
+        fields?: unknown[];
+        entities?: unknown;
+    }, user: User, role: string, opts?: AuthorizedReadOpts) => Promise<VectorSearchChunkResult[]>;
+    embedQuery: (context: {
+        id: string;
+        name?: string;
+        embedder: {
+            model: string;
+        };
+    }, text: string, opts?: {
+        user?: User;
+        role?: string;
+        inputType?: "document" | "query";
+    }) => Promise<number[]>;
+};
+
+declare function postgresClient(): Promise<{
+    db: Knex;
+}>;
+
 /**
  * Represents the essential data for a sentence within a text.
  *
@@ -2286,6 +2337,13 @@ type DocumentProcessorConfig = {
          * vertex_ai) can be switched without code changes.
          */
         model?: string;
+        /**
+         * Maximum pages per OCR request for the "mistral" processor.
+         * Vertex AI OCR rejects documents over 30 pages; the PDF is split into
+         * chunks of this size and each chunk is OCR'd independently.
+         * Defaults to 25 (safely under the Vertex AI 30-page limit).
+         */
+        maxPagesPerChunk?: number;
     };
     /**
      * Optional cost-attribution context, forwarded to LiteLLM as spend tags
@@ -2522,4 +2580,4 @@ declare const ExuluPython: {
     instructions: typeof getPythonSetupInstructions;
 };
 
-export { type ChunkerOperation, type ChunkerResponse, type JOB_STATUS as EXULU_JOB_STATUS, JOB_STATUS_ENUM as EXULU_JOB_STATUS_ENUM, type STATISTICS_TYPE as EXULU_STATISTICS_TYPE, STATISTICS_TYPE_ENUM as EXULU_STATISTICS_TYPE_ENUM, type ExuluAgent, ExuluApp, ExuluAuthentication, ExuluChunkers, ExuluContext, type ExuluContextEmbedder, ExuluDatabase, ExuluDefaultProviders, ExuluDefaultTools, ExuluDocumentProcessor, ExuluEval, type Item as ExuluItem, ExuluJobs, type ExuluOauthConfig, type ExuluOauthToolContext, ExuluOtel, ExuluProvider, ExuluPython, queues as ExuluQueues, ExuluReranker, ExuluTool, trajectoryRegistry as ExuluTrajectoryRegistry, ExuluVariables, defaultChunker };
+export { type ChunkerOperation, type ChunkerResponse, type JOB_STATUS as EXULU_JOB_STATUS, JOB_STATUS_ENUM as EXULU_JOB_STATUS_ENUM, type STATISTICS_TYPE as EXULU_STATISTICS_TYPE, STATISTICS_TYPE_ENUM as EXULU_STATISTICS_TYPE_ENUM, type ExuluAgent, ExuluApp, ExuluAuthentication, ExuluChunkers, ExuluContext, type ExuluContextEmbedder, ExuluDatabase, ExuluDefaultProviders, ExuluDefaultTools, ExuluDocumentProcessor, ExuluEval, type Item as ExuluItem, ExuluJobs, type ExuluOauthConfig, type ExuluOauthToolContext, ExuluOtel, ExuluProvider, ExuluPython, queues as ExuluQueues, ExuluReadApi, ExuluReranker, ExuluTool, trajectoryRegistry as ExuluTrajectoryRegistry, ExuluVariables, type VectorSearchChunkResult, defaultChunker, postgresClient };

package/dist/index.d.ts

@@ -1,5 +1,6 @@
 import * as _opentelemetry_sdk_node from '@opentelemetry/sdk-node';
 import * as knex from 'knex';
+import { Knex } from 'knex';
 import { RedisClientType } from 'redis';
 import * as bullmq from 'bullmq';
 import { Queue } from 'bullmq';
@@ -148,6 +149,7 @@ interface ExuluAgent {
     instructions?: string;
     feedback?: boolean;
     suggestions_enabled?: boolean;
+    sandbox_enabled?: boolean;
     slug?: string;
     tools?: {
         id: string;
@@ -786,6 +788,9 @@ declare class ExuluContext {
             after?: number;
         };
         entityFilter?: EntityFilter;
+        /** Precomputed query embedding; forwarded to vectorSearch to skip re-embedding (caller must use
+         *  this context's embedding model). Spread into vectorSearch via `...options` below. */
+        queryEmbedding?: number[];
     }) => Promise<{
         itemFilters: SearchFilters;
         chunkFilters: SearchFilters;
@@ -828,9 +833,11 @@ declare class ExuluContext {
     getItem: ({ item }: {
         item: Item;
     }) => Promise<Item>;
-    getItems: ({ filters, fields, }: {
+    getItems: ({ filters, fields, user, role, }: {
         filters?: any[];
         fields?: string[];
+        user?: User;
+        role?: string;
     }) => Promise<Item[]>;
     embeddings: {
         generate: {
@@ -1804,6 +1811,50 @@ declare class RecursiveChunker extends BaseChunker {
     toString(): string;
 }
 
+interface AuthorizedReadOpts {
+    itemIds?: string[];
+    externalIds?: string[];
+    chunkIndexRange?: {
+        from?: number;
+        to?: number;
+    };
+}
+/**
+ * Supported, RBAC-safe read API for retrieval clients (e.g. the agentic harness).
+ * Relevance + visibility still go through ExuluContext.search(); this namespace
+ * adds the read shapes search() cannot express, always over authorized rows.
+ */
+declare const ExuluReadApi: {
+    getTableName: (id: string) => string;
+    getChunksTableName: (id: string) => string;
+    getEntitiesTableName: (id: string) => string;
+    getChunkEntitiesTableName: (id: string) => string;
+    entitiesAvailable: (context: {
+        id: string;
+        entities?: unknown;
+    }) => Promise<boolean>;
+    authorizedRead: (context: {
+        id: string;
+        fields?: unknown[];
+        entities?: unknown;
+    }, user: User, role: string, opts?: AuthorizedReadOpts) => Promise<VectorSearchChunkResult[]>;
+    embedQuery: (context: {
+        id: string;
+        name?: string;
+        embedder: {
+            model: string;
+        };
+    }, text: string, opts?: {
+        user?: User;
+        role?: string;
+        inputType?: "document" | "query";
+    }) => Promise<number[]>;
+};
+
+declare function postgresClient(): Promise<{
+    db: Knex;
+}>;
+
 /**
  * Represents the essential data for a sentence within a text.
  *
@@ -2286,6 +2337,13 @@ type DocumentProcessorConfig = {
          * vertex_ai) can be switched without code changes.
          */
         model?: string;
+        /**
+         * Maximum pages per OCR request for the "mistral" processor.
+         * Vertex AI OCR rejects documents over 30 pages; the PDF is split into
+         * chunks of this size and each chunk is OCR'd independently.
+         * Defaults to 25 (safely under the Vertex AI 30-page limit).
+         */
+        maxPagesPerChunk?: number;
     };
     /**
      * Optional cost-attribution context, forwarded to LiteLLM as spend tags
@@ -2522,4 +2580,4 @@ declare const ExuluPython: {
     instructions: typeof getPythonSetupInstructions;
 };
 
-export { type ChunkerOperation, type ChunkerResponse, type JOB_STATUS as EXULU_JOB_STATUS, JOB_STATUS_ENUM as EXULU_JOB_STATUS_ENUM, type STATISTICS_TYPE as EXULU_STATISTICS_TYPE, STATISTICS_TYPE_ENUM as EXULU_STATISTICS_TYPE_ENUM, type ExuluAgent, ExuluApp, ExuluAuthentication, ExuluChunkers, ExuluContext, type ExuluContextEmbedder, ExuluDatabase, ExuluDefaultProviders, ExuluDefaultTools, ExuluDocumentProcessor, ExuluEval, type Item as ExuluItem, ExuluJobs, type ExuluOauthConfig, type ExuluOauthToolContext, ExuluOtel, ExuluProvider, ExuluPython, queues as ExuluQueues, ExuluReranker, ExuluTool, trajectoryRegistry as ExuluTrajectoryRegistry, ExuluVariables, defaultChunker };
+export { type ChunkerOperation, type ChunkerResponse, type JOB_STATUS as EXULU_JOB_STATUS, JOB_STATUS_ENUM as EXULU_JOB_STATUS_ENUM, type STATISTICS_TYPE as EXULU_STATISTICS_TYPE, STATISTICS_TYPE_ENUM as EXULU_STATISTICS_TYPE_ENUM, type ExuluAgent, ExuluApp, ExuluAuthentication, ExuluChunkers, ExuluContext, type ExuluContextEmbedder, ExuluDatabase, ExuluDefaultProviders, ExuluDefaultTools, ExuluDocumentProcessor, ExuluEval, type Item as ExuluItem, ExuluJobs, type ExuluOauthConfig, type ExuluOauthToolContext, ExuluOtel, ExuluProvider, ExuluPython, queues as ExuluQueues, ExuluReadApi, ExuluReranker, ExuluTool, trajectoryRegistry as ExuluTrajectoryRegistry, ExuluVariables, type VectorSearchChunkResult, defaultChunker, postgresClient };