Skip to content

ENTERPRISE.md

Latest commit

 

History

History
141 lines (108 loc) · 5.33 KB

ENTERPRISE.md

File metadata and controls

141 lines (108 loc) · 5.33 KB
Raw

Reattend Enterprise

Organizational memory that never forgets. Self-healing knowledge for companies and government.

Forked from Reattend Personal (~/Desktop/Final Reattend/reattend/). Shares the core ingestion pipeline, ask engine, and DeepThink. Adds enterprise layers on top.

What this solves

Organizations lose $31.5B/year from poor knowledge sharing. 42% of institutional knowledge lives only in individual employees' heads. When someone leaves, transfers, or retires — their context goes with them.

Reattend Enterprise ensures nothing is ever lost. Every decision, every context, every piece of institutional knowledge is captured, linked, and accessible — regardless of who joins or leaves.

How it's different from Glean

Glean Reattend Enterprise
Core Search across tools Memory graph with temporal reasoning
When someone leaves Their docs are searchable Their decisions + context are preserved as institutional memory
Analytics Search analytics Decision analytics — what was decided, by whom, was it reversed
On-premise No Yes — Rabbit on their servers
Self-healing No Auto-detects stale policies, contradictions, knowledge gaps
Price $25/user/month $8/user/month

Architecture

enterprise.reattend.com (SaaS control plane)
  ├── Admin console (org setup, users, roles, analytics)
  ├── Connector management (which tools to sync)
  └── Billing & licensing

Customer's infrastructure (on-prem or private cloud)
  ├── Rabbit v2.1 (ingestion) — their GPU or managed
  ├── Memory graph (Postgres + pgvector)
  ├── Claude API or Rabbit (answering)
  └── SSO/SAML integration

What's inherited from personal Reattend

  • Ingestion pipeline (Groq primary, Rabbit batch)
  • Ask engine (Claude Sonnet for answering)
  • DeepThink (strategic thinking partner)
  • FTS5 full-text search
  • FastEmbed local embeddings
  • Workspace system (already supports teams)
  • Invite flow (email invite → accept → join)
  • Role system (owner/admin/member)

What needs to be built

Phase 1: Foundation (Week 1-2)

  • Org hierarchy: Organization → Department → Division → Team
  • Enhanced RBAC: who sees what, at what level
  • Audit trail: every query logged, every access tracked
  • Admin dashboard: usage, health, member management
  • SSO/SAML integration (Azure AD, Okta, Google Workspace)

Phase 2: Enterprise Features (Week 3-4)

  • Decision tracking: who decided, when, context, outcome, was it reversed
  • Transfer protocol: when employee leaves, knowledge stays with the role
  • Self-healing: auto-detect stale policies, contradicting documents, orphaned knowledge
  • Knowledge gap detection: "Department X has no documented decisions for 6 months"
  • Enterprise connectors: SharePoint, Teams, Confluence, SAP

Phase 3: Analytics (Week 5-6)

  • Decision analytics: "This decision was made 3 times and reversed each time"
  • Knowledge health score per department
  • Turnover impact analysis: "When Officer A left, 47% of project context had no successor"
  • Compliance tracking: which policies are referenced vs outdated
  • Export/reporting for leadership

Pricing

Tier Users Price Features
Starter 10-50 $3/user/month Core memory, 5 connectors, basic analytics
Business 50-500 $8/user/month Full memory graph, all connectors, advanced analytics
Enterprise 500+ Custom On-prem, SSO/SAML, audit trail, dedicated support
Government Department Custom Fully on-prem, air-gapped, Hindi support

Infrastructure

Component SaaS On-prem
Ingestion Groq (Llama 3.3 70B) Rabbit v2.1 (customer's GPU)
Answering Claude Sonnet Rabbit v2.1
Embeddings FastEmbed (local) FastEmbed (local)
Database SQLite → Postgres Postgres
Vector search sqlite-vec → pgvector pgvector
Auth NextAuth + SSO SSO/SAML only
Hosting DigitalOcean droplet Customer's servers

Team

  • Partha (Product): Builds the product with AI
  • Cofounder 2 (Marketing): Reattend personal marketing + Reddit + user acquisition
  • Cofounder 3 (Enterprise Sales): Enterprise client acquisition, government relationships

Key files from personal Reattend

These are the core files that power the enterprise version:

src/lib/ai/llm.ts          — LLM provider factory (Groq, Claude, Rabbit)
src/lib/ai/agents.ts        — Ingestion pipeline (triage, extract, embed, link)
src/lib/ai/prompts.ts       — Triage prompts
src/lib/db/index.ts          — Database + FTS5 + vec search
src/lib/db/schema.ts         — Drizzle schema (workspaces, members, records, etc.)
src/lib/jobs/worker.ts       — Job queue with retry/backoff
src/app/api/ask/route.ts     — Ask endpoint (Claude answering)
src/app/api/deepthink/route.ts — DeepThink strategic advisor
src/app/api/records/route.ts  — Memory CRUD
src/app/api/upload/route.ts   — File upload + bulk import

Environment variables needed

# Ingestion (Groq — primary)
GROQ_API_KEY=

# Answering (Claude)
ANTHROPIC_API_KEY=

# Rabbit (batch re-enrichment, on-prem)
RABBIT_API_URL=
RABBIT_API_KEY=

# Auth
NEXTAUTH_SECRET=
NEXTAUTH_URL=

# Database
DATABASE_URL=

# Email
RESEND_API_KEY=