Give your apps the memory LLMs understand
Every App ID gets its own pgvector-backed knowledge base. Ingest anything, recall it semantically, and get answers grounded in your app's own facts — with citations.
What is the easiest way to add memory to an LLM app?
Intelliverse Memory gives every App ID a hosted pgvector knowledge base behind two API calls: ingest documents or URLs, then search or chat with grounded, cited answers. No vector database to run, no embedding model to pick, per-end-user memory included — it ships with every tier, including Free.
Ingest
POST text, documents, or URLs. We fetch, clean, and chunk them — no pipelines to build, no ETL to babysit.
Embed
Every chunk is embedded and stored in a pgvector database scoped to your App ID. Your apps never share memory.
Recall
Semantic search finds what matters by meaning, not keywords — the exact context your request needs, in milliseconds.
Grounded answers
Chat completions are grounded in your app's own memory and return citations, so every claim traces back to a source.
Two calls. That's the whole integration.
No modelfield, no embedding config, no vector index tuning. Ingest once, then search or chat against your app's memory forever.
# 1. Teach your app something — text, documents, or URLs
curl https://kb.router.intelli-verse-x.ai/v1/kb/ingest \
-H "Authorization: Bearer $INTELLIVERSE_API_KEY" \
-d '{ "app_id": "APP_ID", "documents": [{ "url": "https://docs.myapp.com" }] }'
# 2a. Recall it semantically
curl https://kb.router.intelli-verse-x.ai/v1/kb/search \
-H "Authorization: Bearer $INTELLIVERSE_API_KEY" \
-d '{ "app_id": "APP_ID", "query": "what is our refund window?" }'
# 2b. ...or chat with grounded, cited answers
curl https://kb.router.intelli-verse-x.ai/v1/kb/chat \
-H "Authorization: Bearer $INTELLIVERSE_API_KEY" \
-d '{ "app_id": "APP_ID", "messages": [{"role":"user","content":"How do refunds work?"}] }'Scoped to your App ID
Memory is isolated per app, not per account. Ship ten apps and each one remembers only its own world — billed through the same key.
Zero infrastructure
No vector database to provision, no embedding jobs to schedule, no retrieval pipeline to tune. It's one API call in, one API call out.
Model-less, like everything here
You never pick an embedding model or a chat model. Send the request; the router picks the right brain in your tier and keeps receipts.
Citations built in
Grounded answers cite the exact chunks they used. Fewer hallucinations, and your users can verify every answer.
Per-user memory
Beyond app-level knowledge, each of your end users can get their own remembered facts — scoped, inspectable, and deletable per user for clean data hygiene.
Look inside your app's memory
The dashboard visualizes your knowledge base as a living chunk graph: what your app has learned, how it's connected, and which memories answered which questions.
Your apps deserve to remember
Memory ships with every tier — Free included. Create an app, get an App ID, and give it a memory in the next five minutes.