Information Management
Cortex Guide: Grounded Answers Over Your Own Content
Upload documents and datasets, then ask plain-language questions and get accurate, grounded answers — combining AI retrieval with exact numeric computation, so explanations come from your content and numbers are computed, not guessed.
The Problem
Why Cortex Guide Exists?
General LLMs guess; your team needs answers grounded in your own data.
Ask a generic model about your documents and you get plausible-sounding fiction. Knowledge Base grounds every answer in the user's uploaded content, and routes numeric questions to a computation engine so counts and filters are exact — critical for credible scientific data.
Watch the 2-min demo — no form needed
Six things Cortex Guide does
Secure project spaces
Every user gets secure sign-in and their own project spaces, so uploaded content and conversations stay private and organised per project.
- Secure authentication
- Per-user project isolation
- Organised content per project
- Access-controlled sharing
Document & dataset ingestion
Bring in general documents for Q&A and scientific datasets — including NetCDF weather and climate data — so both narrative and numeric content live in one queryable place.
- Document upload and indexing
- Scientific NetCDF dataset support
- Per-project content library
- Automatic processing on upload
Grounded RAG chat
A retrieval-augmented pipeline answers qualitative questions using the user's own material, so responses are grounded and traceable rather than invented.
- Retrieval-augmented generation
- Answers grounded in your uploads
- Handles explanation and metadata questions
- Dynamic per-project welcome messages
Exact numeric computation
Numeric and filtering questions — like listing every record at a given longitude — go to a dedicated query engine and are computed exactly, instead of being estimated by a language model.
- Direct computation over datasets
- Exact filtering and counts
- NetCDF query engine
- Credible answers for scientific data
Smart query routing
Each question is classified: metadata and explanation questions go to the RAG pipeline, while numeric or filtering questions go to the computation engine — natural-language flexibility with computed precision.
- Question classifier
- RAG for qualitative questions
- Compute engine for quantitative ones
- Best of both in one interface
Shareable chat links
Share a conversation by link so recipients can see the result without creating an account — useful for handing findings to stakeholders quickly.
- Shareable conversation links
- No account required to view
- Read-only stakeholder access
- Polished share view
Different Teams, Same Platform
Research teams
Query papers and datasets together, with exact numbers on the scientific data.
Climate & weather
Ask plain-language and precise numeric questions over NetCDF data.
Internal knowledge
Give staff grounded answers from your own docs instead of guesswork.
Consulting
Share a grounded answer with a client by link, no account needed.
Compliance & policy
Answer questions from policy documents with traceable sources.
Data teams
Combine narrative Q&A with exact filtering in one chat interface.
NOTABLE TECHNICAL HIGHLIGHT
Natural language flexibility, computed precision
Each question is routed through a classifier: explanation questions hit an LLM-powered RAG pipeline, while numeric or filtering questions hit a dedicated NetCDF query engine — combining the flexibility of language with the precision of direct computation.
Frontend
Next.js chat interface with a shareable view.
API
Node.js / Express backend API.
Rag service
FastAPI (Python) with ChromaDB and xarray.
Database
PostgreSQL 16.
Pipeline
GitLab CI/CD builds, containerises, and pushes every change.
Deploy
Docker Compose — portable to any Docker-capable host.
“We finally trust the numbers. Ask what a dataset means and it explains; ask how many records match a filter and it computes the exact answer instead of guessing.”
1
chat interface for both qualitative and quantitative questions.
Test It Yourself
Talk to it — don't just read about it
Ask the agent a question grounded in real content and hear it answer — then ask it anything about the platform.
We'll call within ~30 seconds. The call takes about 2 minutes.
Explore The Rest Of The Suite
Cortex Space
A unified platform that centralises property portfolios, HOAs, municipal compliance, and financial workflows — and turns raw property data into a trackable, auditable task queue instead of a passive database.
Cortex Workbench
One place for the whole team to run accounts, deals, and conversations. Workbench pulls contacts, pipeline,...
Cortex Clinic
Clinic automation from WhatsApp booking to the dental chair, real-time OPD queue, clinical workspace, and billing, live in production.
Ready to see it in your workflow?
Tell us the operation you want to automate and we'll show you the product that fits — running on your scenario.
Book a live call