AI Product Index

Apache-2.0 native Python agent runtime and LangGraph alternative with its own Graph engine, semantic cache, Route Ledger, memory, and ports for PrismGuard security and PrismRAG retrieval. It does not require LangGraph in production.

Self-hosted prompt-injection firewall and ChorusGraph security plug-in. Returns auditable allow/block with resolution_gate and decision_source. Opt-in profiles: web_chat, light/heavy ONNX, and 0.1.10 domain_pilot after train-first.

Zero-trust tool-argument gate. AI/OCR may propose dollar parameters; only an Ed25519-signed ParameterManifest that clears the hard security check may enter tax, underwriting, claims, or payment engines. Formerly ParamGate. Works with any orchestrator — ChorusGraph optional.

Runtime zero-trust gateway between agent extractions and enterprise DAGs. Canonicalize → policy/schema → seal ParameterManifest → KMS-attest → ACCEPT / REVIEW / REFUSE. Apache-2.0 open core; depends prismmanifest>=0.3.4.

Pre-deploy adversarial CI harness. Runs G4 corpora (digit drops, prompt injections, layout/OCR drift) in pytest or CI and fails the build on false accepts — before poisoned tool calls reach money engines.

Self-hosted anti-hallucination verdict engine: cause-side forensics (Tier-0 halt before tokens) plus effect-side grounding (Tiers 1–4) in one auditable ShineVerdict. PASS means grounded in the supplied preload — not world-true.

Taxonomy-controlled Graph RAG library (PyPI: prismrag-patch). Client-defined word-to-category mapping, dual vectors, Louvain communities, BFS graph search, and auditable retrieval for pgvector, ChromaDB, Pinecone, and Weaviate.

Dynamic wavepacket memory for RAG: embeddings as z = A·e^(iφ) — amplitude frozen, phase dynamic. Two-level group-then-chunk retrieval. Source DB stays read-only.

Vector database migration tool (PyPI: insight-vector-bridge). Move embeddings with semantic validation, not hope.

Self-hosted AI Operations Platform — the optional paid ops plane for the Prism stack. Side 1 (insightits.com) issues offline Ed25519 JWTs (iss=insightits.com). Verification never phones home. Connected installs may optionally POST /api/choruscontrol/validate about every 14 days.

Supporting navigation hub for six PyPI pages. Not an Open-Source AI Infrastructure category and not a pip meta-package. PrismManifest, PrismRAG, PrismResonance, and VectorBridge stay first-class in the five-category tree; PrismLang and PrismLib stay supporting.

Middleware layer for LangGraph — not a replacement. Add @prism_node on existing graphs for leaner inter-agent state, audit trail, and tenant isolation. Zero LLM dependency for routing.

Open-source library with PrismCache (in-process semantic LLM cache) and PrismDriver (WAL-streamed DB driver). Full stack ships as prismlib-plus. ChorusMesh is the paid cluster companion on the same landing — different from CHORUS Fabric and ChorusControl.

Deterministic, bitemporal memory and execution engine for multi-turn AI agents — digest → sleep → recall with byte-identical replay. Current PyPI pin 0.4.1 (same runtime as 0.4.0).

Intent-gated 1024d retrieval using positional subspace multiplexing in one tensor — causal / taxonomy channels. Pin vectorprism==0.1.0.

Patent-pending tensor-native agent protocol that streams float32 embeddings directly over gRPC — no text, no tokens, no JSON.

Attention lab for PrismManifest: 10-second punch on a fixed Form 1040 pack — $450,000 vs AI $45,000, blocked, wrong by $60,750.

Money Path Demo for PrismManifest: Document → untrusted tool-call proposal → PrismManifest gate → deterministic engine. Not an agent runtime.

VectorPrism Demo page: two separate retrievers on the same docs. Agent A is VectorPrism (encode_query + PSMRetrievalEngine.search). Agent B is plain all-mpnet-base-v2 cosine — not VectorPrism. Plus a live PrismManifest money/digit gate. README does not require HashingEncoder (plumbing/CI only). Full Agent A experience is frozen 768d encoder plus all six trained MultiTaskProjectionAdapter heads. This lab runs that path on all-mpnet-base-v2 (lab contrast pack, not the finance adversarial .pt).

a canned-fixture evidence lab that runs the in-process Ledgerlock gate on three labeled documents so buyers can inspect the text, watch ACCEPT / REVIEW / REFUSE, and optionally approve or reject a REVIEW path. Not a live borrower file and not file upload.

Prism Pack Family board: FinancePackBench job-by-job wins for the wired stack PrismGuard → ChorusGraph → PrismShine. Not a pip package.

ML score + why for slop vs builder/product signal — posts (PP-DETECT) or README/PDF (4-layer product index). Open API, JWT or API key.

Presentation layer for AI agents — embeddable photoreal talking-face engine. Host keeps LLM and TTS; ChorusFace lip-syncs and streams MJPEG. Beta 0.2.0b1, one fixed TickFeed avatar.

a cryptographic authorization gate for AI-assisted residential mortgage income and asset verification that sits between extraction and the LOS write and issues a signed AuditBundle. Design-partner / evidence demo — not SOC 2 and not a live loan-file production claim.

AI concierge for hotels: room bookings, amenities, housekeeping requests, and 24/7 guest support—like a front desk that never sleeps.

AI coach for gyms and fitness apps: class schedules, membership questions, workout tips, and nutrition guidance—24/7 motivation for your members.

AI assistant for agencies: property search, MLS-style answers, viewing schedules, and lead qualification for buyers and renters.

AI host for restaurants: menu questions, reservations, hours, dietary needs, and specials—so your team can focus on the floor.

AI legal information assistant for California law topics—helping visitors understand processes, terms, and when to consult a licensed attorney.

AI guide for hospitals: departments, visiting hours, appointment information, and services—helping patients find answers without overloading staff.

AI shopping assistant: product discovery, sizes, order status, returns, and recommendations—like a sales associate on every page.

AI for manufacturers and industrial firms: product specs, services, RFQs, and technical FAQs—so sales and support scale without extra headcount.

AI reception for dental practices: appointments, services, insurance FAQs, and post-visit instructions—friendly support that fits your clinic.