@exulu/backend 1.46.0 → 1.47.0

package/changelog-backend-10.11.2025_03.12.2025.md

@@ -1,316 +0,0 @@
-
-----
-
-### feat: add project-scoped context retrieval and enhanced personalization
-
-Date: 2025-12-03
-Author: dclaessen-exulu
-
-This commit introduces a comprehensive project-based information retrieval
-system and several enhancements to the agent and context infrastructure:
-
-**Project Context Retrieval:**
-- Add createProjectRetrievalTool for project-specific information search
-  across multiple contexts
-- Implement project caching mechanism to optimize repeated project queries
-- Support hybrid search across project items with filtering and ranking
-- Enable automatic project tool injection when session has associated
-  project
-
-**Enhanced Personalization:**
-- Add firstname/lastname fields to user authentication for API users
-- Include user information in system prompts for better personalization
-- Make personalization configurable via privacy settings
-
-**Field Processor Improvements:**
-- Add generateEmbeddings configuration option to field processors
-- Support onInsert trigger in addition to existing triggers
-- Enable processors to run without user context for system operations
-- Improve processor execution flow with better logging
-
-**Infrastructure Enhancements:**
-- Add database connection pool logging for debugging
-- Improve file parts processing with better error handling
-- Make convertToolsArrayToObject async to support dynamic tool generation
-- Enhance storage utility to support uploads without user context
-- Better handling of tool variable configs vs execution configs
-- Comment out unused project_id from RBAC schema
-
-**Context and Search Improvements:**
-- Add search method directly to ExuluContext class
-- Expose applyFilters and contextToTableDefinition utilities
-- Support context source execution with exuluConfig parameter
-- Improve error messages and reduce excessive logging
-
-🤖 Generated with [Claude Code](https://claude.com/claude-code)
-
-Co-Authored-By: Claude <noreply@anthropic.com>
-
-----
-
-### feat: add document parsing for AI agent file handling with JWT secret fix
-
-Date: 2025-11-30
-Author: dclaessen-exulu
-
-Enhance agent file processing capabilities and fix authentication:
-
-- Add officeparser integration to extract text from document files
-- Implement processFilePartsInMessages to convert file parts to
-  OpenAI Responses API compatible format
-- Convert document files to text parts with extracted content
-- Keep image files as image parts (natively supported by API)
-- Add message deduplication to prevent duplicate message IDs
-- Change saveChat to process messages sequentially for correct
-  timestamp ordering
-- Fix JWT verification in getToken by converting NEXTAUTH_SECRET
-  to base64url format as required by jose library
-
-This enables agents to process uploaded documents by extracting
-their text content and presenting it in a format compatible with
-the OpenAI Responses API.
-
-🤖 Generated with [Claude Code](https://claude.com/claude-code)
-
-Co-Authored-By: Claude <noreply@anthropic.com>
-
-----
-
-### feat: add PDF preview tool and enhance agent API with cookie authentication and message persistence
-
-Date: 2025-11-26
-Author: dclaessen-exulu
-
-Adds a new preview-pdf tool for viewing PDF documents and significantly enhances the agent
-chat API to support public agents with optional authentication. Implements cookie-based session
-management, message ID tracking to prevent duplicates during chat persistence, and allows
-passing full message arrays for stateless interactions.
-
-Also downgrades AI SDK from v5.0.95 to v5.0.65 for stability, adds cookie-parser dependency,
-and improves context embedding generation with better metadata handling and increased timeout
-limits. Enhances S3 integration with multi-bucket support by making the bucket always be the part of the key before the first slash
-improves vector search result limits, and adds better error handling for encrypted variables
-and external ID lookups in context operations.
-
-🤖 Generated with [Claude Code](https://claude.com/claude-code)
-
-Co-Authored-By: Claude <noreply@anthropic.com>
-
-----
-
-### fix: issue with upsert
-
-Date: 2025-11-19
-Author: dclaessen-exulu
-
-
-----
-
-### feat: upgrade AI SDK to v5.0.95 and add Gemini 3 Pro support with parameterized context sources
-
-Date: 2025-11-19
-Author: dclaessen-exulu
-
-This commit introduces several enhancements to the AI capabilities and context system:
-
-- Upgrade AI SDK from v5.0.56 to v5.0.95 with improved provider utilities
-- Add @vercel/oidc dependency for enhanced authentication support
-- Implement Google Vertex Gemini 3 Pro agent with 1M+ token context window
-- Enhance context source configuration with parameterizable inputs
-- Improve Vertex authentication documentation with detailed setup instructions
-- Add support for dynamic parameters in context source definitions via GraphQL schema
-
-The new Gemini 3 Pro agent provides very high intelligence with moderate speed,
-supporting text, images, files, audio, and video inputs. Context sources can now
-define parameters with names, descriptions, and default values for more flexible
-data retrieval configurations.
-
-Changes include:
-- src/index.ts: Export new vertexGemini3ProAgent
-- src/registry/index.ts: Register Gemini 3 Pro agent
-- src/registry/classes.ts: Add params field to ExuluContextSource config
-- src/registry/utils/graphql.ts: Add ContextSourceParam type and params support
-- src/templates/agents/google/vertex/index.ts: Add Gemini 3 Pro agent, refactor auth docs
-- types/models/context.ts: Add params field to Context interface
-
-🤖 Generated with [Claude Code](https://claude.com/claude-code)
-
-Co-Authored-By: Claude <noreply@anthropic.com>
-
-----
-
-### feat: add Google Vertex AI agent support and enhance embedder configuration system
-
-Date: 2025-11-18
-Author: dclaessen-exulu
-
-Add Google Vertex Gemini 2.5 Flash agent integration with support for
-optional authentication. Implement new embedder_settings table for
-context-specific embedder configuration with variable management.
-
-Major changes:
-- Add vertexGemini25FlashAgent to default agents and registry
-- Implement embedder_settings schema for per-context embedder config
-- Add authenticationInformation field to ExuluAgent class
-- Make providerapikey optional across agent and MCP initialization
-- Add config field to ExuluEmbedder with hydrateEmbedderConfig method
-- Pass context ID and settings to embedder chunker and generation ops
-- Update GraphQL schema to expose authenticationInformation and config
-- Enhance error messages with agent name and ID for better debugging
-
-Breaking changes:
-- ChunkerOperation signature now requires config parameter
-- VectorGenerateOperation signature now requires settings parameter
-- convertToolsArrayToObject contexts parameter now optional
-
-This enables agents that don't require API keys (e.g., Vertex AI with
-workload identity) and allows embedders to retrieve configuration from
-variables per context.
-
-🤖 Generated with [Claude Code](https://claude.com/claude-code)
-
-Co-Authored-By: Claude <noreply@anthropic.com>
-
-----
-
-### fix: project tracking anthropic passthrough and timeout for worker
-
-Date: 2025-11-17
-Author: dclaessen-exulu
-
-
-----
-
-### feat: add MCP prompt library integration and JSON filtering enhancements
-
-Date: 2025-11-17
-Author: dclaessen-exulu
-
-Add comprehensive prompt template management via MCP tools and enhanced JSON field filtering:
-
-Prompt Library Features:
-- Add prompt_library and prompt_favorites database schemas
-- Register getListOfPromptTemplates and getPromptTemplateDetails MCP tools
-- Enable agents to discover and retrieve prompt templates with usage/favorite tracking
-- Support agent-specific prompt assignment via assigned_agents JSON field
-
-GraphQL & Filtering Improvements:
-- Add JSON field containment support using PostgreSQL @> operator
-- Enhance filter operators to handle JSON equality and IN operations with jsonb casting
-- Pass table schema context to applyFilters for type-aware query building
-
-Bug Fixes & Refinements:
-- Fix upsert validation to require id or external_id
-- Add source update statistics tracking for API and job triggers
-- Improve eval function result metadata structure with function_results array
-- Add default scoring method fallback to average for eval runs
-- Fix typo: rename eval to evaluation in bullmq decorator
-
-Configuration:
-- Add .mcp.json for exulu-mcp-server-default-coding-agent integration
-
-🤖 Generated with [Claude Code](https://claude.com/claude-code)
-
-Co-Authored-By: Claude <noreply@anthropic.com>
-
-----
-
-### feat: add todo system, enhanced roles, session metadata, and scheduling support
-
-Date: 2025-11-14
-Author: dclaessen-exulu
-
-This commit introduces several significant features and improvements to the Exulu backend:
-
-- Implement TodoWrite and TodoRead tools for session-based task tracking
-- Store todos in agent_sessions.metadata JSON field
-- Add access control checks for todo operations
-- Tools require authenticated session context
-
-- Create default "admin" role with full write permissions
-- Create "default" role with agent write + read-only for other resources
-- Add support for "evals" and "api" permissions in role structure
-- Automatically provision both roles during database initialization
-
-- Add metadata JSON column to agent_sessions table
-- Pass sessionID throughout tool execution pipeline
-- Implement getSession helper for consistent session retrieval
-- Enable session-aware tool execution with user context
-
-- Add cron-like scheduling support for context data sources
-- Implement queue configuration options
-- Add retry logic with exponential/linear backoff strategies
-- Enable automated data ingestion workflows
-
-- Fix Redis URL construction when using authentication
-- Improve handling of username/password credentials
-- Better fallback values for missing environment variables
-
-- Support config-based tool description overrides
-- Pass sessionID to all tool executions
-- Enable tools to access session metadata
-- Improve tool configuration hydration
-
-- Add better-auth (v1.3.34) with WebAuthn support
-- Include cryptography libraries (@noble/ciphers, @noble/hashes)
-- Add @simplewebauthn packages for authentication flows
-- Include kysely for type-safe database queries
-
-🤖 Generated with [Claude Code](https://claude.com/claude-code)
-
-Co-Authored-By: Claude <noreply@anthropic.com>
-
-----
-
-### fix: add ExuluItem as export
-
-Date: 2025-11-11
-Author: dclaessen-exulu
-
-
-----
-
-### feat: implement scheduled data sources for contexts
-
-Date: 2025-11-11
-Author: dclaessen-exulu
-
-Add support for scheduled data sources within ExuluContext that automatically
-fetch and ingest items at regular intervals.
-
-Key changes:
-- Introduce ExuluContextSource type with configurable cron schedules, retry
-  logic, and backoff strategies
-- Add source execution handler in BullMQ workers to process scheduled jobs
-- Create automatic job schedulers for each context source with configurable
-  retry attempts (default: 3) and exponential backoff (default: 2000ms)
-- Rename Context.process() to processField() for clarity
-- Add executeSource() method to handle source execution and item creation
-- Track source metadata in BullMQ job data for job identification
-
-Sources enable automated data ingestion workflows where external data can be
-pulled into contexts on a schedule without manual intervention. Each source
-execution creates items in the context and optionally schedules follow-up
-processing jobs for embeddings and chunking.
-
-🤖 Generated with [Claude Code](https://claude.com/claude-code)
-
-Co-Authored-By: Claude <noreply@anthropic.com>
-
-----
-
-### fix: minor exulu upgrade for redis connection
-
-Date: 2025-11-10
-Author: dclaessen-exulu
-
-
-----
-
-### fix: redis auth url
-
-Date: 2025-11-10
-Author: dclaessen-exulu
-
-
-----

package/types/models/agent-backend.ts

@@ -1,15 +0,0 @@
-// Represents an agent backend
-// which is created and run in
-// the agents service. Agents
-// registered in Exulu must have
-// a connected backend agent that
-// does the actual work.
-export interface AgentBackend {
-    id: string
-    name: string
-    description: string
-    enable_batch: boolean
-    inputSchema: any;
-    slug: string
-    type: string
-}