Platform Overview

Conduit is a complete industrial intelligence platform that combines three capabilities into a unified system: data querying, context intelligence, and AI-mediated collaboration. Together, these transform how industrial organizations access data, build knowledge, and make decisions.

The Three Pillars

1. Data Querying (NQE)

The Natural Query Engine (NQE) lets anyone query operational technology data using natural language -- no SQL, no SPL, no specialized query syntax required.

• Natural language interface: Ask questions in plain English
• Golden Templates: Learned query patterns that improve accuracy over time
• Multi-source federation: Query across Splunk, MQTT, MCP Gateway, and DuckDB without data movement
• Multi-LLM support: Claude, OpenAI, Azure OpenAI, or self-hosted Ollama
• Confidence scoring: Every compiled query includes a confidence score

"What was the average temperature of Tank 1 over the last 24 hours?"
-> Compiled to SPL with 0.95 confidence (Golden Template match)

2. Context Intelligence (Context Engine)

The Context Engine builds persistent organizational knowledge from every interaction that flows through the platform.

• 16 expertise domains: Manufacturing, engineering, operations, quality, safety, and 11 more
• Three context types: Individual (portable), Relational (company-owned), Organizational (company-owned)
• Composite scoring: Frequency, recency, consistency, and depth weighted into expertise levels
• Temporal decay: Configurable half-life ensures context stays current
• Expert discovery: Real-time API to find who knows what

3. AI Collaboration

AI-Mediated Collaboration routes questions to the right domain experts through the organizational hierarchy with AI assistance at every level.

• Hierarchical query routing: Questions flow down to experts, answers flow back up
• Intent lineage: Full transparency into how questions evolved
• Thread-level isolation: Each participant only sees their own thread
• Conversation expiration: 30-day default TTL; context persists, conversations do not
• Continue Intent: Human-controlled routing at every step

Architecture at a Glance

Conduit is built as a modern, multi-service platform designed for industrial environments:

| Component | Details | | -------------------- | ----------------------------------------------------------------------------- | | Backend services | 11 distinct services (auth, NQE, context, collaboration, adapters, etc.) | | Databases | 3 (Neo4j for graphs, PostgreSQL for relational, pgvector for semantic search) | | LLM providers | Multi-provider support (Claude, OpenAI, Azure, Ollama, Mock) | | Codebase | 43,000+ lines of code | | Test coverage | 1,237+ tests across unit, integration, and E2E | | API | RESTful with JWT authentication (HS256, 15-min expiry) | | Deployment | Self-hosted, private cloud, or air-gapped |

Key Statistics

| Metric | Value | | ---------------------------- | ------------------------------------------------------------------ | | Expertise domains | 16 (manufacturing, engineering, operations, quality, safety, etc.) | | NQE query templates | 10 Golden Template patterns | | Production adapters | 4 (Splunk, MQTT, MCP Gateway, DuckDB) | | Neo4j node types | 7 | | Neo4j relationship types | 14 | | PostgreSQL tables | 19 (with Row-Level Security) | | Vector dimensions | 1536 (pgvector embeddings) | | Multi-tenant | Yes, with full data isolation |

How It All Fits Together

The three pillars are not isolated features -- they form a reinforcing loop:

                    +-------------------+
                    |   Data Querying   |
                    |      (NQE)        |
                    +--------+----------+
                             |
                    Query patterns feed
                    expertise signals
                             |
                             v
+-------------------+    +-------------------+
| AI Collaboration  |<-->| Context Engine    |
| (Routing +        |    | (Expertise +      |
|  Answers)         |    |  Knowledge Graph) |
+-------------------+    +-------------------+

Data Querying feeds the Context Engine: Every NQE query generates expertise signals. When an engineer queries manufacturing data, the Context Engine records that as evidence of manufacturing domain knowledge.

Context Engine powers AI Collaboration: When a question needs routing, the Context Engine identifies the best expert based on real demonstrated expertise, not job titles.

AI Collaboration drives better queries: As experts answer questions through the collaboration system, NQE learns new Golden Templates and the Context Engine refines its expertise model.

Privacy by Design

Conduit's Privacy & Security Model is built into the architecture, not bolted on:

• "Record context, not conversations": Raw interactions expire; abstract knowledge persists
• Thread isolation: No upstream visibility in collaboration threads
• Individual ownership: Your expertise context is yours, portable and exportable
• Self-hosted: Your data stays in your infrastructure
• Compliance-ready: Full audit trail for all significant actions

Getting Started

Ready to explore? Start with these guides:

| Guide | Description | | ------------------------------------------ | ------------------------------------------------ | | Getting Started | Install and configure Conduit in minutes | | Context Engine | Deep dive into organizational knowledge building | | AI Collaboration | Learn how intelligent question routing works | | Privacy Model | Understand the security and privacy architecture | | NQE | Master natural language data querying | | Architecture | Explore the system design and infrastructure |