Key facts
| RAG endpoints | POST /v1/embeddings, POST /v1/rag/collections and POST /v1/rag/query |
| Embedding models | plugsky-embed-v1 (1536 dimensions, OpenAI-ada compatible) and plugsky-embed-large (3072 dimensions) |
| Retrieval modes | Keyword, vector and hybrid search with optional cross-encoder reranking |
| Citations | Every query returns ranked chunks with source attribution |
| Ingestion | Documents are chunked, embedded and indexed automatically per collection |
| Batch limits | Up to 2,048 inputs per embeddings request, max 8,191 tokens each |
| Data handling | Per-collection encryption at rest; API data is not used to train models |
| Deployment | Hosted, VPC, on-prem and air-gapped options; 30+ models behind one API |
TL;DR
- Ground answers in policy, KYC and credit documents with citations reviewers can check.
- Three OpenAI-compatible endpoints replace a multi-vendor RAG stack.
- Hybrid retrieval with optional reranking handles dense policy language better than keywords alone.
- VPC, on-prem or air-gapped deployment with BYOK and SIEM audit export covers model risk review.
- Pilot on one low-risk corpus and measure retrieval quality, not answer fluency.
How it works, step by step
- Choose one corpus and write down the questions the assistant must answer.
- Build an evaluation set with the expected source documents for each question.
- Create a collection, ingest the corpus and confirm chunking suits the document type.
- Start with hybrid retrieval plus reranking for dense policy and legal text.
- Return citations with every answer and require human review for customer-facing output.
- Move the workload to a region-locked plane or your VPC and switch on audit export.
- Review access, retention and model pinning with compliance before scaling out.
Original data
Try it yourself
Open the RAG architecture builder →
Why banks need RAG grounded in their own content
Banks already own the corpus; the problem is access. Policy manuals, KYC files, credit memos, regulator circulars and product terms sit across document stores and core systems, and a general-purpose model cannot see any of it. RAG retrieves the passages that matter and hands them to a model that answers with citations — exactly what model-risk, compliance and internal audit functions want to review.
A generic chatbot answer is hard to defend. A retrieved, attributable answer is easier to check, challenge and log, which is why grounding is usually the first AI pattern banks put into production.
The Plugsky RAG stack: embeddings, collections, queries
You do not have to assemble a vector database, an embedder, a retriever and a reranker separately. Documents are chunked, embedded and indexed automatically per collection, retrieval supports keyword, vector and hybrid modes, and every query returns ranked chunks with citations. The endpoints stay OpenAI-compatible — POST /v1/embeddings, POST /v1/rag/collections, POST /v1/rag/query — and the embeddings call also works standalone if you keep your own vector store. For a bank review process the key property is attribution: each answer traces to a passage and a document, and per-collection encryption keeps the corpus protected at rest.
The controls banks should configure first
The controls a bank will ask about are deployment, keys and evidence. On Enterprise, RAG runs in your VPC, on-prem or air-gapped with BYOK through AWS KMS, Azure Key Vault, HashiCorp Vault or an on-prem HSM. Per-request logs record model, tokens, latency, user and region and export to your SIEM with retention up to 7 years, which supports access reviews and incident forensics.
Pin model versions where reproducibility matters, keep a human in the loop for customer-facing answers, and align processing location with the data-protection regime that applies to each legal entity and jurisdiction.
A practical pilot path
Pick one corpus with clear value and low risk — a retail policy manual or an internal procedures handbook — and build a small evaluation set of real questions with expected sources. Measure retrieval accuracy and citation quality, not just answer fluency. Prototype on the free plan with plugsky-micro and plugsky-lite, then move the workload to a region-locked plane and Enterprise deployment when it proves out. See the live pricing page for current plans.
Honest comparison
| Capability | Plugsky | Assembling your own stack | Keyword search only |
|---|---|---|---|
| Retrieval | Keyword, vector and hybrid with optional reranking | You integrate and tune each component | Keyword matching, no semantics |
| Citations | Ranked chunks with source attribution | You build attribution yourself | Document links, not passages |
| Embeddings | plugsky-embed-v1 (1536d) and plugsky-embed-large (3072d) | Model choice spread across vendors | Not applicable |
| Deployment | Hosted, VPC, on-prem and air-gapped | Depends on every component | Usually on-prem already |
| Data handling | Per-collection encryption; API data not used to train models | You audit every vendor | Existing controls apply |
| Time to first answer | One collection and a query call | Weeks of integration work | Days, with lower answer quality |
Frequently asked questions
Can RAG run inside our own environment?
Yes on Enterprise: in your VPC, on-prem or fully air-gapped, with BYOK key custody. The API stays OpenAI-compatible, so applications do not change when the endpoint moves.
How do citations work?
Each query returns ranked chunks with source attribution, which you can pass to any chat model or show directly to a reviewer next to the generated answer.
Do we have to change our application code?
No. The RAG endpoints are OpenAI-compatible, so existing SDKs, prompts and evaluation harnesses keep working when you point the base URL at Plugsky.
Is there a free plan?
Yes — plugsky-micro and plugsky-lite with 2 API keys and no credit card. A 14-day full-access trial is also available.
How is pricing structured?
Self-serve plans are flat monthly with unlimited fair-use usage and no per-token charges or overage fees. See the live pricing page for current plans.
Does Plugsky train on our data?
Plugsky states that API data is not used to train models, and collections are encrypted at rest; on Enterprise you can keep everything inside your own VPC.