VedRaq: The Active Brain for Your Entire Company

The Causal Intelligence Platform & Unified Context Layer for Modern Enterprise Agents.

Inspired by the vision of a single source of truth for organizational memory, VedRaq actively ingests, classifies, maps, and coordinates the scattered knowledge, workflows, and agents across your company.

---

The Problem

• Company knowledge is scattered everywhere: Essential facts, decisions, and documentation live isolated in Slack, Notion, Jira, emails, and databases. There is no single source of truth.
• AI agents operate in silos: Agents can run tactical scripts, but they lack overall context of company policies, historical incidents, active workflows, and cross-team dependencies.
• No central coordinator: As companies deploy multiple AI agents, there is no shared memory layer or policy-grounded execution bus to monitor correctness, manage conflict resolutions, or maintain safety guardrails.

---

The Solution: VedRaq

An active operational brain that runs alongside your enterprise tools. It:

1. Reads & Reacts: Ingests live streams of organizational events (messages, code, commits, tickets, docs).
2. Constructs Causal Graphs: Maps how decisions connect to outcomes, teams to processes, and agents to resources.
3. Orchestrates Agents: Evaluates agent behaviors against policies, shares high-speed operational snapshots, and flags anomalies.
4. Keeps Humans in the Loop: Creates active approval paths for low-confidence claims and operational conflicts.

Unified Core Tech Stack

Frontend & Visuals

• Next.js 15 (App Router) - Server-rendered routing framework.
• React 19 & Tailwind CSS v4 - Component architecture and sleek, modern layout engines.
• ShadCN UI - Beautiful Radix UI primitives.
• React Flow - Direct node-based visualization of corporate SOPs, causal graphs, and active pipelines.

High-Performance Backend

• FastAPI - Async routing, validation, and high-throughput ingestion interfaces.
• LangChain & LangGraph - Complex agent orchestration, structured loops, and state machines.
• Claude API (Anthropic) - Advanced semantic analysis, agent reasoning, and structural synthesis.
• Socket.IO - Live event pushes, agent logs streaming, and user console telemetry.

Memory & Event Infrastructures

• Kafka (Upstash) - Resilient, parallel event stream and adapter bus.
• Redis (Upstash) - Real-time hot memory, caching state snapshots, and agent event pub/sub.
• Neo4j - Causal knowledge graph representing entities, timelines, and dependencies.
• PostgreSQL (Supabase) - Cold relational storage, vector lookup (pgvector), and tenant metadata.

---

Core Infrastructure Engines

1. Ingestion Engine: Kafka (Upstash)

• Zero Event Loss: Slack logs, Jira updates, email webhooks, and commit histories are pushed instantly into high-durability queues.
• Pipeline Isolation: Decoupled architecture prevents slow downstream LLM pipelines from bottlenecking fast ingest paths.
• Parallel Fan-out: Stream events are simultaneously routed to multiple consumers (embeddings creation, raw storage, live active agent triggers).

2. Live State Engine: Redis (Upstash)

• Hot Memory State: Stores fast-changing, high-access runtime info: who is on call, open incident tickets, active task queues, and agent session context.
• Sub-Millisecond Read Latency: Empowers reasoning steps to query active company states instantly without requesting heavy database joins.
• Agent-to-Agent Pub/Sub: Enables instant cross-agent coordination and decentralized feedback loops.

3. Causal Intelligence Engine: Neo4j Knowledge Graph

• Relational Mapping: Instead of isolated document indexes, Neo4j maps relationships between teams, resources, actions, policies, and systems.
• Causal Outage/Incident Diagnosis: Allows deep graph traversals to immediately resolve complex organizational questions:

  "What software deploy caused this support spike, who authorized it, and what policy covers this service?"

• Dynamic Policy Dependency Checks: Alerts teams when a change in one policy implicitly conflicts with steps in another process.

---

Design & Aesthetic Reference

Our landing page and console design take direct inspiration from top-tier, developer-centric active AI platforms:

• Sim.ai — Dark modes, neon glassmorphism nodes, and beautiful active animation trails.
• Retool — Highly interactive dashboards, smooth drag-and-drop mechanics, and custom panel configurations.
• Vellum.ai — Advanced playground panels, side-by-side prompt tuning tools, and clean, high-density visualization components.

---

Setting Up the Development Server

1. Frontend Landing Page Setup

Move to the landing-page directory and spin up the development compiler:

cd landing-page
npm install
npm run dev

2. Next Steps

• Spin up PostgreSQL with the pgvector extension enabled.
• Run a local Neo4j desktop instance or link to a cloud Aura database.
• Configure Upstash Kafka & Redis tokens inside a root .env file.