Relational SQL
Create and query structured business data with tables, indexes, joins, filtering, grouping, ordering, and established InternetSQL database operations.
Database foundation
Core relational and server capabilities
Transactions
Use transaction boundaries, commit, and rollback so relational and supported AI-related data changes follow controlled application workflows.
Distributed server queries
Run qualified distributed queries across InternetSQL server deployments while preserving explicit connection, authorization, and failure boundaries.
Parallel-server operation
Use qualified parallel-server execution and workload paths for deployments that require concurrent database processing and server-side scale.
Embedded and TCP modes
Embed the engine through existing library APIs or connect through authenticated TCP clients and servers using the deployment model appropriate to the application.
Backup, restore, and recovery
Apply documented whole-home backup, checksum verification, separate restore testing, restart, cleanup, and recovery procedures.
Database-native AI
Search, retrieval, and grounded answers
Vector storage and search
Create collections, store embeddings, retain stable vector identifiers, and perform similarity search with COSINE, L2, and DOT metrics.
Hybrid retrieval
Combine semantic ranking, lexical relevance, relational filters, metadata, and source-row joins in inspectable retrieval workflows.
Source-backed RAG
Retrieve authorized records, construct bounded context, generate answers through a configured provider, and return citations tied to source material.
Embedding workflows
Use deterministic offline embeddings for evaluation or protected provider aliases for configured text-embedding services, with batch and error diagnostics.
Ranking and evaluation
Inspect metric ordering, scores, filters, three-weight hybrid ranking, deterministic evaluation sets, and source relationships.
Synchronization and durable jobs
Coordinate committed source state, vector updates, durable jobs, interruption handling, cancellation, restart, and orphan checks.
Integration
AI clients, providers, and application APIs
MCP read tools
Expose governed database reads, vector search, and RAG retrieval to MCP clients over stdio or authenticated HTTP.
Controlled MCP writes
Require named policy, narrow approval, privilege checks, expiration and replay defenses, audit, and a kill switch for enabled write operations.
API-compatible AI providers
Use credential aliases, an isolated provider host, timeouts, retry controls, usage reporting, redaction, and deterministic offline operation.
C and C++ integration
Call the existing DLL/import library or static library through the qualified public C API without introducing a separate AI product library.
Python integration
Use the pure-Python ctypes wrapper for embedded and TCP workflows, relational SQL, transactions, vectors, RAG, diagnostics, and cleanup.
ICPP integration path
Internet C++ is the C++-like scripting/compiler suite and KCPP its backend compiler. Its updated database invocation path remains separately release-gated.
Operations and assurance
Inspection, security, and deployment
All-shell AI tracing
Inspect redacted tokens, embedding batches, three-weight scores, transformer boundaries, provider calls, retrieval, jobs, timing, cancellation, and errors.
Security boundaries
Apply authorization before provider calls, protected credentials, bounded context, injection defenses, fail-closed policy, and redacted audit records.
Concurrency and monitoring
Exercise transaction isolation, interruption, cancellation, restart, job monitoring, and operational-limit checks through qualified workflows.
Turnkey solutions
Start from packaged support, maintenance, policy/records, or embedded OEM applications with setup, sample data, diagnostics, and documentation.
Training and evidence
Use short lessons and 17 evidence-bound long-form demonstrations covering installation, APIs, MCP, RAG, tracing, security, and recovery.
Deployment qualification
Use offline tests, checksums, manifests, documented limitations, and customer-environment validation before accepting a production workload.
Evaluate claims against your workload
Capabilities depend on the selected build, deployment mode, configuration, security policy, provider, and workload. Validate functionality, performance, recovery, and suitability in the intended environment before production use.
Go to DownloadsDiscuss your AI project