Parinita Almanac
Sovereign reference layer — you consult the Almanac, you receive signed entries.
Almanac is the sovereign vector and retrieval-augmented generation layer. Its flagship workload is AI RAG — supplying signed entries to downstream models routed through Conduit. It isn't competing on raw nearest-neighbor latency against Pinecone or Weaviate; it's competing on verifiable sovereignty.
What it does
-
Signed entries, not chunks
Users don't query a database — they consult the Almanac. Results are signed entries with Chrysalis receipts anchored to the customer's 101-LLC jurisdictional entity.
-
Two surfaces, one engine
Almanac Search is the lower-level retrieval API for teams building their own RAG. Almanac RAG is the higher-level managed pipeline that combines retrieval, reranking, and prompt construction with synthesis through Conduit.
-
Anchored on Plane 4
In-memory index serving and search on Plane 4 (Intel Sierra Forest); persistent indices and source corpora on Plane 5 (NVMe). The Tracker agent is the consultation entry point.
-
Four-layer tenant isolation
Legal (101-LLC binding), network (Crucible per-packet identity), storage (per-tenant Plane 5 partition, jurisdiction-bound), cryptographic (HSM-signed receipts on every consultation).
-
Built for compliance retrieval
ITAR / HIPAA / financial workloads where provenance is a compliance primitive — every consultation produces a cryptographic citation that's independently verifiable on Chrysalis.
-
Post-quantum receipts
ML-DSA-65 (CRYSTALS-Dilithium, FIPS 204) for receipt signing on the roadmap. Multi-modal entries (image/audio/video) and federated cross-jurisdiction consultation coming Q4 2026.
How it works
The Almanac introduces a vocabulary that mirrors how regulated buyers think about authoritative reference. Users don’t query a database; they consult the Almanac. Results aren’t retrieved chunks; they’re signed entries. Every consultation produces a Chrysalis receipt anchored to the 101-LLC jurisdictional structure that underpins the Parinita platform.
Almanac exposes two surfaces over a single underlying engine. Almanac Search is the lower-level search and retrieval API for organizations building their own RAG pipelines. Almanac RAG is the higher-level managed retrieval-augmented-generation pipeline that combines retrieval, reranking, and prompt construction with synthesis through Conduit. Both surfaces produce signed entries; both run on the same plane infrastructure.
Tenant separation is enforced at four independent layers — legal, network, storage, and cryptographic — the same model Sovereign uses for data, applied to vector indices and source corpora.
When to use it
- Regulated AI deployments where the retrieval step needs verifiable provenance, not just the inference step.
- Compliance-bound RAG — ITAR research, healthcare clinical decision support, financial filings analysis, legal discovery — where every cited source must be admissible evidence.
- Multi-jurisdiction operations where vector indices themselves carry residency requirements.
What it isn’t
A nearest-neighbor benchmark winner. Pinecone, Weaviate, Qdrant, Milvus, and Chroma will likely beat Almanac on raw latency. Almanac wins on verifiable provenance per consultation — if that’s not your compliance posture, use one of the others.
Related products
Part of the Parinita AI Edge
Bring Parinita Almanac into your stack.
Every Parinita product runs on the same 9-plane fabric across 101 edge POPs. Talk to us about a pilot, or see how the pieces fit together.