Turn documents into
intelligence in minutes.
Fully managed vector storage with built-in embedding models. Upload, chunk, embed, and query—all on sovereign Australian infrastructure.
Build RAG in minutes, not months
Create a vector store, upload documents, and query with context-aware generation—all through the API you already know.
Connect your data sources
Pull documents from Google Drive, S3, Notion, Confluence, Slack, or upload directly via API.
Data sources
Google Drive
AWS S3
Notion
Confluence
Slack
Dropbox
Automatic document chunking
Documents are split into semantically meaningful chunks optimised for retrieval quality.
# Upload a file — chunking is automatic client.vector_stores.files.create( vector_store_id=store.id, file=open("knowledge-base.pdf", "rb"), ) # Check chunking progress store = client.vector_stores.retrieve(store.id) print(store.file_counts) # {completed: 3, in_progress: 1, total: 4}
High-dimensional embeddings
Each chunk is converted to a 4,096-dimensional vector using E5-Mistral-7B-Instruct.
Vector embeddings
Indexed vector storage
Vectors are indexed in your isolated store with configurable distance metrics.
import openai client = openai.OpenAI( base_url="https://api.scx.ai/v1", api_key="YOUR_API_KEY", ) store = client.vector_stores.create( name="support-docs", distance="Cosine", ) print(store.id) # "vs_abc123..."
Semantic similarity search
Send a natural language query. We embed it and find the most relevant chunks.
Search results
Context-aware generation
Retrieved context is injected into the LLM prompt for grounded, accurate responses.
# Retrieved chunks become system context response = client.chat.completions.create( model="DeepSeek-V3.1", messages=[ {"role": "system", "content": results.to_context()}, {"role": "user", "content": "How do I reset my password?"}, ], ) print(response.choices[0].message.content) # "To reset your password, navigate to # Settings → Security → Reset password..."
Why teams choose SCX.ai for RAG
Fully managed vector infrastructure so you can focus on your application, not your database.
Zero infrastructure to manage
No Pinecone, no Weaviate, no pgvector setup. Create a vector store in one API call and start querying immediately.
Built-in embedding pipeline
Upload raw documents—we handle chunking, embedding with E5-Mistral-7B-Instruct (4,096 dimensions), and indexing automatically.
Semantic similarity search
Query with natural language and retrieve the most relevant chunks. Choose from Cosine, Euclidean, or Dot Product distance metrics.
Sovereign by default
Documents, embeddings, and queries stay on-shore in Australian infrastructure. Full audit logging and per-tenant isolation.
OpenAI-compatible API
Use the same client libraries you already know. Two lines to switch from OpenAI or any other provider.
Real-time ingestion tracking
Monitor file processing, vector counts, and storage usage in real time. Know exactly where every document stands.
Monitor your vector stores
Real-time visibility into file processing, vector counts, and storage usage across every store.
customer-support-docs
Ready12
Files
48,392
Vectors
186 MB
Storage
| File | Status | Chunks | Vectors | Uploaded |
|---|---|---|---|---|
| support-guide-v3.pdf | Completed | 142 | 142 | 2h ago |
| faq-database.json | Completed | 89 | 89 | 2h ago |
| product-manual.pdf | Processing | 67 / 92 | 67 | Just now |
| release-notes-q4.md | Pending | — | — | Just now |
Built for every RAG workload
From customer-facing chatbots to internal compliance tools, managed vector storage handles the infrastructure so you can focus on your application.
Customer Support
Ground your support bots in real documentation. Reduce hallucinations and improve first-contact resolution.
Enterprise Search
Make internal knowledge bases searchable with natural language. Surface the right document in milliseconds.
Compliance & Legal
Query regulatory documents, policies, and contracts with precision. Maintain full audit trails.
Start building RAG applications today
Create your first vector store in under a minute. No infrastructure to manage, no embeddings to host, no similarity search to optimise.
Get your API key