Rhetra Get in touch
Demos · recorded from the running platform

See it running.

Five workspaces on five unrelated domains: a markets desk, a filings desk, a study app, semantic search over aviation safety reports, and a weather-risk desk. Each one is real data in the real product. Each one was assembled from the same five primitives.

01 Markets · investment offices · quant desks

Coverage your headcount can't do

A two-person desk cannot read five hundred charts a day. Every session in the index is debriefed automatically, with notes pinned to the minutes that mattered. Exceptions are emailed and signed off by a human, and the logic stays as plain TypeScript in the workspace.

datasets → ontology → functions → workflows → app → agent
11,289 tickers minute resolution 21M points per series 136-occurrence backtest
See the demo
02 Public filings · equity research

A filings desk from one conversation

Three prompts: get the filings, approve the model, build the app. Every annual and quarterly report AMC has filed since 2013 lands as entities carrying their EDGAR URLs, the schema pauses for a human, and the session ends with a working application on top of it.

prompt → files → schema approval → functions → app
55 primary documents 10-K + 10-Q, 2013 → 2026 1,337-filing index 3 prompts
See the demo
Exam questions clustered around course topics in the graph explorer
03 Education · course material

Five exam PDFs into a study app

Five past papers, as PDFs, and nothing else. One pipeline extracts every question as a typed entity, keeps the image of the page it came from, and links it to the topic it covers. On top of the result sits an app that drills you by topic and tracks mastery. The run cost five cents.

media set → LLM extraction → ontology → links → app → workflow
106 questions 12 topics 31 links $0.0553 · 149.3s
See the demo
Ranked safety incidents returned for a plain-language question about crew fatigue
04 Aviation safety · unstructured text

Semantic search over 585 safety reports

Keyword search only returns reports that use your words. This returns reports that mean the same thing. Embeddings are a node in the pipeline, the vector is a property on the entity, the search is one function, and one checkbox hands it to the agent as a tool.

pipeline → ontology → function → agent
585 reports 1,536 dims minSimilarity 0.25 1,678ms
See the demo
05 Weather · commodities · risk desks

The storm that moves the stock

Nine rare-earth mines watched against live NWS warning zones. A query function scores every site × hazard pair against the operator's own tape and finds the one class that matters: winter. The analyst files the assessment and the decision back into the ontology, on camera.

map → warning zones → event study → action → decision
3,718 NWS products 9 sites, 6 operators winter: −1.8% 5d, t=−2.1 58s, one take
See the demo
The same five primitives

A market desk and a study app are the same build.

Pipelines to read the source, an ontology to type it, functions to hold the logic, workflows to run it, apps and agents on top. What changes between these four is the data and the judgement, which is the part that should be yours.

Get in touch
Pipelinesbatch · streaming · files
Ontologytypes · links · actions
FunctionsTypeScript, typed to the model
Workflowsschedules · approvals
Apps & agentsSDK · MCP
See the platform tour →