Built on FabrCore

Pro Solutions

Vulcan365 builds production-grade AI products on top of our open-source FabrCore platform. Each solution starts as an experimental project, is hardened through real-world use, and progresses to general availability.

Product Lifecycle

Every solution follows the same path from lab to production.

Experimental

Active development. Core architecture proven, API surface may change. Available for early adopters and feedback.

Preview

Feature-complete. Hardened through real-world testing. API stabilizing. Available for production pilots.

General Availability

Stable API. Production support. Documentation complete. Ready for enterprise deployment.

Solutions

Experimental

Swarm

Distributed Multi-Agent Task Orchestration

Submit a natural-language goal and Swarm builds a dependency-aware task plan, discovers your available agents, fans work out across them in parallel waves, and adapts in real time when things go wrong — with graduated recovery from automatic retries through SME consultation to human escalation.

  • Six-agent coordination layer: orchestrator, planner, supervisor, workers, blackboard, factory
  • DAG-based parallel execution with automatic context propagation
  • Zero-registration client agent discovery from live metadata and health state
  • Mid-execution replanning when the original plan hits a wall
  • Nine configurable termination guards ensure every plan completes
  • Full state persistence — resume from host restart without data loss
Experimental

Knowledge Graph

Hybrid Vector + Graph Semantic Search

A production-grade RAG system that combines vector-based semantic search with SQL Server native graph traversal. Ingest documents, automatically extract typed entities and relationships using LLM-driven pipelines, and retrieve with multi-hop graph traversal — all behind mandatory scope-based access control.

  • Four search modes: entity search, chunk search, relationship traversal, and hybrid (vector + graph combined)
  • LLM-driven entity extraction with 12 relationship types and 10+ entity categories
  • Scope-pinned access control — mandatory on every query, validated at both ends of every edge
  • Domain > Category taxonomy with provenance on every result
  • SQL Server 2025 native graph tables with VECTOR(1536) embeddings
  • Domain intent classification for query-time routing
Experimental

Long-Term Memory

Three-Temperature Knowledge Persistence

A durable knowledge management system that gives AI agents persistent, structured memory across conversations and sessions. Three temperature tiers — hot (always-loaded index), warm (on-demand LLM-selected recall), and cold (archival vector search) — with intelligent consolidation that merges duplicates, prunes stale observations, and resolves contradictions automatically.

  • Hot/warm/cold tiers with bounded token budgets and automatic eviction
  • Knowledge graph storage: typed entities (Fact, Rule, Instruction, Observation) with weighted relationships
  • Entity matching on save — prevents duplicates by merging similar knowledge with LLM content synthesis
  • 4-pass consolidation: deduplication, staleness pruning, contradiction resolution, index truncation
  • Synthetic imagining — proactively discovers relevant memories from conversation context
  • 3-stage retrieval pipeline: header scan → LLM relevance selection → full content load with graph traversal

Interested in our solutions?

Whether you want early access to an experimental product or need a custom AI solution built on FabrCore, we’d love to hear from you.