FAQ

OilGas Nanobot Swarm is a hierarchical multi-agent AI system built for petroleum engineering. It uses a 3-tier Queen → L1 Domain Leads → L2 Sub-Agents architecture to decompose complex engineering problems into specialized tasks and synthesize expert-level answers. It runs on Ollama Cloud and NVIDIA NIM — no GPU or local model setup required.

• Reservoir engineers — IPR analysis, material balance, decline curves
• Drilling engineers — ECD, kick tolerance, casing design verification
• Production engineers — well surveillance, optimization, artificial lift
• HSE professionals — PSM compliance audits, HAZOP support
• Operations teams — daily field briefings, digital morning reports
• Technical managers — economic screening, capital allocation

The web demo at oilgas-nanobot-swarm.vibecaas.app is completely free — no account required. The API is open and requires no key for /swarm/run. Self-hosting is also free (MIT license). The upcoming Pro Edition will offer dedicated capacity, Redis memory, scheduled teams, and MS 365 integration.

The deployed demo uses Ollama Cloud as the primary backend (fast, ~3–5s) with NVIDIA NIM as fallback. The full self-hosted stack supports Anthropic Claude, local vLLM, and any OpenAI-compatible endpoint. Specific model names are intentionally not exposed in the public API to protect infrastructure details.

For sensitive data, deploy the self-hosted version on your own infrastructure using Docker Compose, Render, or Railway. The codebase is fully open source and can run air-gapped with a local vLLM server.

7 built-in engineering calculators:

• reservoir_pressure_calc — hydrostatic gradient, BHP, fracture gradient (Hubbert & Willis)
• drilling_engineering_calc — ECD, kick tolerance, mud weight window
• production_engineering_calc — PI, Vogel IPR, Darcy flow, artificial lift
• pipeline_hydraulics_calc — Darcy-Weisbach pressure drop, flow regime, line sizing
• well_control_calc — MAASP, kill mud weight, driller's method
• formation_evaluation_calc — Archie Sw, Wyllie porosity, GR shale volume
• oilgas_regulatory_reference — API standards, BSEE, OSHA PSM, EPA Quad O

See Engineering Tools docs for full details.

The standard industry formula: Where MW = mud weight, APL = annular pressure loss, TVD = true vertical depth.

Vogel (1968) for solution-gas drive wells: Darcy radial flow for tight/conventional wells. The system generates a full 11-point IPR curve from a single test point.

The system references: API 6A, 7-1, 10A, 16A, 17D, 570, 620/650 · BSEE 30 CFR 250 · OSHA 29 CFR 1910.119 (PSM) · EPA 40 CFR Part 60 Subpart OOOO/OOOOa · NORSOK D-010 · ISO 16530.

1. Go to the Console section on the dashboard
2. Click a team card to pre-load an example goal, or type your own
3. Click ▶ Run — no login or API key needed for the demo
4. Results appear in the console output with step-by-step calculations

Teams are pre-configured agent workflows for common engineering tasks. Use them by specifying "team": "team-name" in your API call or clicking a team card:

• well-engineering-review — comprehensive well design review
• reservoir-analysis — formation evaluation + IPR
• oilgas-field-briefing — daily ops summary
• hse-compliance-audit — PSM 14-element audit

See the Teams section for all 9 available teams.

Go to the Builder section:

• Agent Builder — describe what you want one agent to do → generates name, system prompt, tools list, and copy-paste Python code
• Team Builder — describe a multi-step workflow → generates the full agent team with orchestration, roles, and schedule expression

The generated Python code can be pasted directly into nanobot/teams/oilgas_teams.py in a self-hosted deployment.

The demo runs on Vercel's serverless platform with a shared AI backend. Response times:

• Simple calculation: 3–8 seconds
• Complex analysis (hierarchical mode): 10–25 seconds
• Cold start (first request after inactivity): add 2–4 seconds

For faster, dedicated responses — including background scheduled teams — contact info@neuralquantum.ai about the Pro Edition.

No. The POST /swarm/run and POST /v1/chat/completions endpoints are fully public — no API key, no account. Just send a POST request with Content-Type: application/json and a goal field.

Yes — the API is OpenAI-compatible:

The public demo is subject to Vercel's serverless function limits and the upstream AI provider rate limits. For production use cases with guaranteed availability and no rate limits, use the self-hosted deployment or contact us about the Pro Edition.

Python: JavaScript:

The engineering calculators use standard industry equations (Archie 1942, Vogel 1968, Hubbert & Willis 1957, Darcy-Weisbach, etc.) implemented in Python. Numerical accuracy is high for inputs within normal operating ranges.

Use it for preliminary screening, education, and generating starting-point analyses that a licensed engineer then reviews and validates.

Yes. Large language models can hallucinate, misinterpret units, or apply incorrect formula variants. Always check: (1) the equation shown matches the standard reference, (2) inputs are in the correct units, (3) the result is in a physically reasonable range. The system is designed to show its work — if a step looks wrong, it is wrong.

Three options:

• Vercel (fastest) — one-click deploy, no server needed. Set OLLAMA_API_KEY and you're live in minutes.
• Docker Compose — full stack locally. docker compose up -d starts gateway + Redis.
• Render / Railway — includes Redis for session memory and background scheduler.

See the Deployment Guide for step-by-step instructions.

No. The default configuration uses Ollama Cloud and NVIDIA NIM as cloud AI backends — no GPU, no local model downloads. If you want to run a local model with a GPU, set VLLM_URL to point to your vLLM server.

Minimum required: one of

• OLLAMA_API_KEY — from ollama.com (recommended, fast)
• NVIDIA_API_KEY — from build.nvidia.com
• ANTHROPIC_API_KEY — for Claude (full-stack deployments)

GATEWAY_API_KEY is optional — set it to restrict who can call admin endpoints.

• Full 3-tier hierarchical swarm (Queen → L1 → L2 agents)
• Redis-backed session memory and conversation history
• Obsidian-compatible knowledge vault with vector search
• Background job scheduler (daily briefings, weekly reviews)
• Microsoft 365 integration (email, calendar, contacts sync)
• Custom agent and team creation
• Priority support and dedicated capacity
• On-premise / air-gapped deployment support

Email info@neuralquantum.ai with your use case and team size. We'll set up a demo and discuss deployment options (cloud or on-premise).

Yes — the complete codebase is MIT-licensed and available on GitHub. The full hierarchical swarm, Redis state, scheduler, and knowledge vault all work in self-hosted mode. You only need your own AI API keys.