World markets, knowledge, energy, and signals in one command layer.
Compass is being rebuilt into a mass intelligence platform for global markets, commodities, trade, economic releases, quarterly reports, books, public datasets, verified web sources, social operations, and mini data center telemetry.
Market Atlas
Markets
Equities, bonds, FX, crypto, real estate, private markets
Economics
Rates, inflation, jobs, GDP, housing, money supply, credit
Trade
Imports, exports, tariffs, ports, logistics, shipping routes
Commodities
Energy, metals, agriculture, water, carbon, batteries
Industry
Manufacturing, robotics, logistics, software, healthcare, defense
Vision Intelligence
Cameras, inspection, object tracking, defect detection, occupancy, machine state
Knowledge
Books, papers, standards, datasets, technical documents
The operating stack inside Compass
Separate modules, one platform: internet library, live data mesh, LLM lab, social manager, digital warehouse, and source verification.
Internet Library
Crawled and curated corpus with provenance, license, hash, topic, and retrieval metadata.
Live Data Mesh
Connectors for official APIs, filings, datasets, news, trade, weather, energy, and market feeds.
Compass LLM Lab
From-scratch tokenizer, tiny transformer, token mini-GPT, domain corpus, eval gates, checkpoints, and inference routing.
Digital Warehouse
Compute, storage, power, model, workload, telemetry, cost, and energy control plane.
Social Manager
Local-first agent for trends, drafts, approvals, posting, memory, and analytics.
Source Verification
Primary-source preference, hashes, trust levels, timestamps, citations, and license checks.
Compass Network
Private-first node network for data, compute, agents, vision events, provenance, credits, and future federation.
Vision Hardware Intelligence
Roboflow Supervision path for camera streams, annotations, detection datasets, and edge hardware signals.
Machine vision joins the intelligence layer
Compass now carries a Roboflow Supervision path for cameras, inspection lanes, warehouse awareness, and hardware telemetry. This is the bridge from public data intelligence into physical-world sensing.
python vision/supervision-lab/edge_node.py --source 0 --profile mini-datacenter --dry-run --detections-json vision/supervision-lab/sample_detections.json --annotate --no-post --output-dir runsFallback upstream profile. Live GitHub and PyPI metadata is requested by the public vision API.
aisles, shelves, dock doors, people, pallets
parts, defects, labels, fixture state, tooling
rack state, cables, LEDs, airflow, thermal cameras
Frames seen, stream uptime, dropped frame rate, and camera reachability.
Counts and dwell time for defined dock, aisle, fixture, rack, and cable zones.
Pass/fail, defect class, visual alarm, service event, and rework candidates.
Node, workload, estimated watts, camera source, and local inference status.
Build sequence
This path turns Compass into a Bloomberg-level intelligence system without pretending a frontier LLM or full web crawler appears overnight.