Meet TORQ Console v0.80
Governed intelligence for business, finance, and strategic decisions
TORQ Console is an advisory workbench, not a chat app. Every request is risk-classified, routed through the L27 model cascade, remembered across sessions, and rendered on a purpose-built canvas.
No credit card required · Cancel anytime · Built on your context, not a template
TORQ AI— Prince Flowers
Position sits inside your stored capital cap and DCA rules. Live quotes are pulled only because this query needs market context. Downside is leveraged to BTC; the offramp remains the exit target already in memory — not a generic model guess. Risk: Elevated but consistent with the written strategy.
TORQ AI— Prince Flowers
Position sits inside your stored capital cap and DCA rules. Live quotes are pulled only because this query needs market context. Downside is leveraged to BTC; the offramp remains the exit target already in memory — not a generic model guess. Risk: Elevated but consistent with the written strategy.
The story behind TORQ Console
Founder-built. Operator-tested.
Barry Flowers, Founder · TORQ Business Solutions · Austin, TX · Building since 2024
TORQ Console was built to manage a real portfolio and real operating decisions before it ever touched anyone else's. Not a proof of concept. A production workbench running against live positions, tax context, documents, and execution — built by one operator who needed institutional-grade advisory without a $500/hr bench on speed dial.
v0.80 is the current line: Prince Flowers as the lead advisor, an 8,381-example fine-tune for local identity, L17–L28 governance, L27 cost routing, and memory that compounds across sessions. Built from scratch for operators who make decisions that matter.
TORQ Console is the system that didn't exist when the work started. Now it's available to you.
Prince Flowers / Qwen3-8B identity trained on real advisory, market, document, draft, and run work — not generic finance text.
Enforcement, context, decision memory, risk, DefendSwarm, model routing, and an experience engine that scores live outcomes.
Structured canvases own the center. Chat is the fallback. Every high-stakes path is risk-first.
The Problem
The old model is broken.
You deserve better.
You're paying $300/hr for advice you need daily
Top financial advisors and consultants bill $250–$500/hr — and they're not available at 11pm when you're making the call. Your business doesn't wait for office hours.
Your data lives in 12 different places
Brokerage accounts. Bank portals. Spreadsheets. PDFs. Google Docs. No single view means no single truth — and you're the one stitching it all together manually.
Generic AI gives generic answers — your business isn't generic
ChatGPT doesn't know your portfolio, your risk tolerance, your tax situation, or your five-year plan. It gives you the same answer it gives everyone else. That's not intelligence — that's autocomplete.
Use Cases
One governed system.
Six canvases. Cyber-range when you need it.
Every request is classified into a render type and opened on the matching canvas — Advisory, Market, Documents, Draft, Runs, or Cyber Range. Chat is the fallback, not the product.
Investment Planning
Portfolio strategy calibrated to your holdings, risk cap, time horizon, and tax situation. TORQ AI loads stored context before it answers — then routes market questions to live quotes instead of inventing prices.
Market Intelligence
Live price feeds, chart capture, and web search — only when the query needs them. Non-market work skips the network and keeps token cost down.
Business Advisory
Strategic briefs, growth planning, and executive summaries on a dedicated advisory canvas. Not chat. A board-ready deliverable.
Document Review
Contracts, financials, trusts, leases. Routed into a review workspace that flags risk, surfaces terms, and produces a structured summary.
Accounting & Tax
Entity structure, deduction mapping, and tax strategy grounded in persistent memory — current rules you store, not last year's training cutoff.
Estate Planning
Asset protection, succession, wealth transfer. Family structure and asset mix persist across sessions so the next conversation does not start from zero.
Drafting & Reports
Memos, briefs, and reports in Draft Workshop. Export PDF, DOCX, Markdown, or Slack in one click.
Multi-Agent Execution
Mission-graph runs across steps, sources, and decisions inside one governed system. Work that used to take a junior analyst a day completes as a run timeline.
Cyber Range
DefendSwarm specialists scan, score, and brief security findings on a dedicated canvas — gated by L21 scope and approval, not dumped into chat.
How It Works
From question
to governed decision — in seconds.
Three steps on the surface. Risk, enforcement, routing, and memory underneath so you do not have to operate the stack.
Connect Your Context
Holdings, strategy, risk cap, tax situation, entity structure, watchlists. Stored in user context, Dream Memory, and T63 unified semantic memory — retrieved on every turn, wrapped so recalled facts cannot be treated as instructions.
Supabase + pgvector · progressive recall · prompt-injection envelope
Ask in Plain English
The 8-stage classifier reads intent in under a millisecond. L25 scores risk first. L21 enforces policy. L27 picks the cheapest model that can do the job — local, GLM, or frontier Claude — and only turns on web search or live quotes when the query needs them.
Risk-first · tool-gated serving · no prompt tricks
Act on Structured Intelligence
Output lands on the matching canvas, not a chat dump. Export PDF, DOCX, Markdown, or Slack. High-stakes changes leave an append-only audit trail. Outcomes feed the L28 experience engine so routing and depth improve with use.
Structured canvases · audit trail · torq_score feedback
Live path: risk → enforcement → classification → L27 routing → structured canvas. L26 DefendSwarm and L18 promotion sit on the same ladder.
The TORQ System
Not a wrapped LLM.
A governed operating system.
Five integrated layers. Together they turn advisory thinking into a production system that can operate, remember, and improve — without treating chat as the product.
Memory
Persistent system state
Three stores, one operator experience: user context (holdings, strategy), Dream Memory (consolidated facts on a schedule), and T63 unified semantic memory with vector recall. Recalled facts are wrapped so they cannot be treated as commands.
Routing
L27 4-tier cascade
Classifier first. Then the cheapest capable model: local Ollama when enabled, GLM-5-Turbo for mid work, frontier Claude for high-stakes, search, and vision. Tools and live quotes are gated — no silent price invention.
Workspaces
Purpose-built canvases
Advisory, Market, Documents, Draft, Execution Runs, Cyber Range. Chat is the fallback renderer. Each canvas has its own rail, export path, and output shape.
Governance
L17–L28 ladder
Risk before the model call. L21 enforcement with single-use approvals. L18 promotion policy. L26 DefendSwarm on the cyber canvas. Append-only audit. Mutations do not ship themselves.
Output & Evolution
Production-ready, scored
PDF, DOCX, Markdown, Slack. L28 scores live outcomes (torq_score). Routing and depth adapt under thresholds and rollback — not unbounded self-rewrite.
Features
Engineered from first principles.
Not assembled from parts.
Every layer exists because a real decision required it. v0.80 is the current production line.
TORQ AI / Prince Flowers
Lead AdvisorPurpose-built advisory agent, not a generic wrapper. Local identity is a fine-tuned Qwen3-8B trained on 8,381 curated examples. Production answers run through L27 so simple work stays cheap and high-stakes work reaches frontier Claude — with your portfolio and strategy loaded first.
Compounding Memory
Dream + T63User context for holdings and rules. Dream Memory consolidates durable facts on a schedule. T63 unified memory stores embeddings and retrieves by meaning, then wraps the result so recalled text cannot become instructions. Sessions start aware. Off-topic recall stays quiet.
Purpose-Built Workspaces
Render TypesAdvisory brief, market insight, document review, draft output, run timeline, cyber-range brief. Chat is the fallback. An 8-stage classifier picks the canvas before the model is called.
Conditional Live Intelligence
ProvenanceLive quotes, TradingView snapshots, vision on chart uploads, and web search — only when the serving gate says tools are required. Live quotes are stamped. Tool-less paths are not allowed to invent a price.
L27 Multi-Model Routing
4-TierLocal Ollama (config-gated), GLM-5-Turbo, optional NIM, frontier Claude Opus. Stakes and search needs escalate. Prompt caching and prefix-stable memory keep repeat work cheap.
Learning Loop
L28Classifications, interactions, and outcomes persist. The experience engine scores responses. Adaptive depth: short answers for simple questions, structured deep analysis when you ask for it.
Export Everything
Production-ReadyPDF for print, DOCX for Word, Markdown for repos, Slack for the team. Draft workspace ships the Slack button next to the file exports.
L17–L28 Governance
Risk-FirstL25 classifies risk before the model call. L21 runs Observe → Enforce with single-use approval tokens. L22.5 compacts long context. L22.6 records decision projections. L18 governs promotions. Audit is append-only.
DefendSwarm / Cyber Range
L26Scoped security scanning with specialist lanes, evidence, and a dedicated brief canvas. Findings are gated — the scanner does not silently swap in a fixture when you asked for a real target.
Who It's For
Built for operators who make decisions that matter
TORQ AI was built by one investor-operator, for other investor-operators. If you've ever wished you had a senior analyst, a tax strategist, and a financial advisor in the same room at 11pm — this is for you.
The individual investor managing real money
You're not a hobbyist. You're managing a real portfolio — maybe $100K, maybe $1M+ — with a strategy, a timeline, and tax implications that matter. You've outgrown generic robo-advisors and you can't justify $500/hr for every question you need answered.
TORQ AI knows your positions, your DCA schedule, your cost basis, your risk tolerance, and your exit targets before you finish typing. It doesn't give you the answer it gives everyone else. It gives you the answer for your situation.
- Portfolio risk analysis with your actual holdings loaded
- Live market data and technical signal interpretation
- Position sizing, DCA logic, rebalancing — grounded in your strategy
The business owner wearing 12 hats
You're running operations, managing cash flow, reviewing contracts, and trying to build a long-term strategy — simultaneously. You can't afford a full-time CFO, legal team, and strategy consultant. But you're making decisions that require all three.
TORQ routes your question to the right intelligence layer automatically. Legal question? Document Review canvas. Cash flow analysis? Advisory Desk. Competitive brief for your board? Execution Run, delivered in minutes.
- Strategic briefs without the consulting bill
- Contract and document review that flags what matters
- Tax guidance grounded in current IRS law, not last year's training data
The operator or CFO who needs real-time clarity
You need a single view across market conditions, P&L, competitor moves, and regulatory landscape — simultaneously. You need it fast and you need it in a format you can act on. Not a chat thread you have to reformat before sending to anyone.
TORQ's multi-agent Execution Runs handle complex, multi-source research workflows. Every output is structured, export-ready, and built for a professional audience. PDF for your attorney. DOCX for your board. Slack for your team.
- Multi-agent execution for complex, multi-step analysis
- Outputs formatted for real distribution — not just for you to read
- Full audit trail on every decision and governance event
The wealth manager ready to scale
You're already good at what you do. TORQ makes you faster. Pull real-time data mid-client-call. Draft estate planning summaries in seconds. Prep a portfolio brief before the meeting instead of during it.
Every session remembers your client context — no re-entering positions, no re-explaining the family structure.
- Client briefs and portfolio summaries at speed
- Estate and succession planning frameworks on demand
- Document review for client financial packages
One thing all four have in common: they're making decisions with real consequences and they don't have time for tools that make them do the thinking twice. TORQ AI does the intelligence work. You make the call.
Pricing
Simple to start.
Scales with your usage.
$50/month gets you the workbench. Heavy users pay a fair overage — light users are not subsidizing power users.
Includes up to 300 queries/month
Start with 48 hours FREE — no credit card required
- All workspaces: Advisory, Market, Docs, Draft, Runs, Cyber Range, Chat fallback
- TORQ AI / Prince Flowers — routed through L27
- Dream Memory + T63 unified semantic memory
- Live quotes, chart vision, and gated web search
- PDF, DOCX, Markdown, and Slack export
- L17–L28 governance: risk, enforcement, audit
- Up to 300 queries/month included
- Priority support
- Early-access rate lock — base price not raised on this plan
Usage & Overages
The $50 base plan covers up to 300 queries/month — enough for most operators.
Overages billed at $0.15 per query above 300. Early access users stay on the locked base rate.
Pro Plan
Coming SoonUnlimited queries, team seats, API access, custom memory configurations, and dedicated support.
Join Pro Waitlist- No credit card for trial
- Cancel anytime
- Full workbench from day one
- Costs stay visible — $50 + $0.15 overage