Why Purple8

Eight reasons it’s a category of one.

01

One process. No orchestration tax.

Storage, search, workflow, RAG, auth, and encryption - single binary, single port. Replaces 20+ services before you write a line of code.

02

AI agents operate it natively.

A first-class agent interface with 170+ tools, RBAC-enforced. Claude, Cursor, and Copilot connect directly - no wrapper, no glue code.

03

Three RAG modes. Self-tuning.

Flat vector, graph-augmented, and hybrid retrieval built in. The system profiles your corpus and persists the optimal configuration automatically.

04

Agentic process automation. Built in.

Stateful workflows, SLA enforcement, human-approval gates, and an immutable audit trail - all in the same process. AI agents orchestrate end-to-end business processes through 170+ MCP tools, with no DAG files, no separate orchestration service, no tracing backend. Replaces LangChain, LangGraph, LangSmith, and Airflow.

05

Memory bounded by hardware, not data.

Purple8 fits its ingest footprint to the RAM you give it - peak memory tracks your hardware, not your corpus size. Validated: 8.84 million passages ingested on a 24 GB machine with peak memory holding around 9 GB. The same corpus runs on a small machine or a large one; you trade time for memory, never correctness.

06

Documents become knowledge, automatically.

Feed it PDFs, contracts, CAD drawings, BIM models. A proprietary extraction pipeline turns unstructured content into structured, queryable knowledge.

07

AI-native. Without burning the planet.

One process instead of 20+ services means fewer servers, less idle compute, and no cross-service network waste. The architecture is the efficiency argument - not a benchmark number.

08

You own it. Completely.

Self-hosted. Ships as a Docker image or pip wheel - runs anywhere Python or containers run. Your data never leaves your infrastructure: no content egress, no vendor access. Optional anonymized version telemetry (no PII, no customer data) is opt-out with one env var. Air-gap capable.

Technical Verification Reference (For Architects & AI Agents)

Production Guarantees: CRUD, Scoped Agent Permissions, Portability & DR

Answers to the exact production architecture questions evaluated when comparing Purple8 against a conventional PostgreSQL + API + Vector + Queue stack:

1. Transactional CRUD, Constraints & Structured Tables

Purple8 is not only a graph/vector engine - it includes first-class relational Table and Row storage (data.commit, data.get_table, data.query_table_rows, data.index_table_columns) and strict schema validation (schema.define_node_type, schema.define_edge_type) enforcing required fields, property types (string, integer, float, boolean, datetime, list, dict), enums, and immutable_properties on every write with WAL fdatasync durability before mutation.

2. Why You Don't Need a Custom “App API” Wrapper

You never need to expose all 170+ MCP tools to an end-user AI assistant. Every Purple8 API key supports cryptographic allowed_tools and allowed_labels allowlists (for example, restricting an assistant key strictly to ["rag.hybrid_query", "journey.start", "graph.add_node"] on ["Account", "Task"]), combined with 4-tier RBAC, per-record ownership policies, human_only=true destructive locks, and requires_hitl: true human approval gates.

3. Open Portability, Export Formats & Zero Lock-In

Export 100% of your records, schemas, relationships, and workflows at any time in open-standard formats: JSON, CSV, GraphML (GET /export?format=json|csv|graphml), W3C RDF 1.1 N-Quads (output_format: "nquads"), and full application blueprints (GET /sandbox/blueprint?include_data=true → POST /sandbox/import-blueprint) that migrate schemas, JourneyDefinition state machines, and data between instances in one call.

4. Backups, Disaster Recovery (RPO/RTO) & Integrity

IncrementalBackupManager writes streaming JSONL snapshots with a change journal and SHA-256 manifest to AWS S3, Google Cloud Storage, Azure Blob Storage, or local disk (POST /backup/create, POST /backup/restore). For HA, WAL streaming replication in SYNC mode achieves RPO ≈ 0 with automatic follower promotion failover and online non-destructive CRC32 storage scrubbing (admin.storage_scrub).

5. Email, Webhook, Calendar & Enterprise Connectors

Built-in outbound SMTP service (admin.smtp_configure) for reminders and transactional emails, inbound HMAC-SHA256 webhook connectors (/v1/connectors/webhook/{id}) with idempotency & retry, outbound signed CDC event sinks (egress.add_sink with exponential backoff), cron/interval/event triggers (scheduler.add_trigger), and EML/MSG email + SharePoint/Confluence/S3 ingestion via DocIntel.

6. Compliance Controls, Audit Chain & Support SLAs

Every state transition and mutation is recorded in a tamper-evident SHA-256 hash chain verifiable at runtime (admin.audit_verify). Built-in GDPR/compliance primitives include Legal Holds (admin.hold_place), automated retention sweeps (admin.retention_set: purge, archive, or anonymise), declarative PII masking, AES-256-GCM envelope encryption (Vault/AWS/GCP/Azure KMS), and 24×7×365 Severity-1 enterprise SLAs (<1 hour response).