digiquant
olympus

A quant research deskin a glass boxyou own

The research stack an institutional desk would build — Atlas researches and Hermes sizes the risk, through backtest to a tearsheet. Open-source and self-hosted, so work that once needed a team runs for one.

Scroll to explore

marketsconnecting to the live feed…

// by the numbers

The desk, in four numbers.

No projections — every figure is a property of the shipped stack: the subsystems, the pipeline, and the published tearsheets. Live stays zero because there is no execution path: every broker adapter is a stub.

subsystems
pipeline stages
backtested trades
live orders
// the pipeline

Research in, a tested strategy out.

digiquant is not a hub of services routing messages around; it’s a linear research workflow. You start in a chat, and each stage hands its output to the next until a strategy is ready to run.

  1. 01

    Research

    Ask in plain language. An LLM research loop pulls free macro and market data and proposes directions to test.

    chat · LLM
  2. 02

    Indicators

    Compose validated indicators — moving averages, RSI, ADF, DPSD — from the shared, unit-tested library.

    indicators lib
  3. 03

    Strategy

    Wire indicators into a rules-based strategy with explicit entries, exits, sizing, and risk.

    strategy spec
  4. 04

    Signals

    Generate entry and exit signals across historical bars — deterministic and reproducible.

    signal gen
  5. 05

    Optimize

    Search the parameter space with Optuna against your own objective. Results are in-sample — the tearsheets say so too.

    Optuna
  6. 06

    Backtest

    Replay on a NautilusTrader core with Pine-faithful indicators — RSI on Wilder's smoothing, Bollinger on population sigma — a full trade ledger, and a tearsheet.

    NautilusTrader
  7. 07

    Export

    Export the strategy to TradingView as Pine v5. Live execution is not built: there is no paper or live path, and every broker adapter is a stub.

    Kairos · planned
Olympus · research → portfolio
The research desk in a box.
01
Atlas
02
Hermes
03
Kairos
01 — Atlas · research

Reads the market into ranked, sourced theses.

Ten phases turn alt-data, institutional flow and macro into evidence-linked theses — every claim traceable to its source.

02 — Hermes · deliberation

Debates the thesis, sizes the conviction.

Thesis review to committed run — multi-agent deliberation, PM direction and risk sizing, with the dissent on record.

03 — Kairos · execution (planned)

The stage after the backtest. Not built.

Kairos names the execution stage, not a package: there is no digiquant.olympus.kairos, and every broker adapter under digiquant/src/digiquant/brokers/ is a stub that raises NotImplementedError.

Atlas
00
Preflight
config + data-layer check
01
Triage
what changed since last run
02
Alt-data
sentiment, flows, on-chain
03
Institutional
positioning & 13F flow
04
Macro
rates, liquidity, regime
05
Asset class
cross-asset context
06
Equities
sector & single-name
07
Consolidate
merge the evidence
08
Synthesis
ranked theses
09
Publish
to the thesis store
Hermes
h1
Thesis review
inherit & re-score
h2
Market thesis
exploration
h3
Vehicle map
thesis → instruments
h4
Screener
opportunity filter
h5
Asset analyst
per-name workup
h6
Deliberation
multi-agent debate
h7
PM direction
allocate & gate
h7e
Risk sizing
½-Kelly, ceilings
h9
Commit run
persist & evolve
Kairos
In development

Atlas and Hermes run today. Kairos — live execution — is next.

Open the Olympus dashboard
// olympus · live book

The research book, valued live.

The paper portfolio Atlas and Hermes maintain, marked against the same price feed. Positions and weights are the published book; prices tick live during market hours, and P&L is stated as of the last close.

Loading the book…
// pre-built strategy library

Research-grade systems, ready to explore.

Browse calibrated backtests from the digiquant library — equity, drawdown, trade logs, and full tearsheets for every release. More assets join the catalog as they clear the pipeline.

Full strategy library
CAGR
Max drawdown
Profit factor
Win rate
Avg trade return
Trades / yr

CAGR
Max drawdown
Profit factor
Win rate
Avg trade return
Trades / yr

CAGR
Max drawdown
Profit factor
Win rate
Avg trade return
Trades / yr

  1. BTC
  2. ETH
  3. SOL

// pricing

Own it, or have it run for you.

digiquant is open core, and it is built on the same digithings modules you can deploy yourself. Self-host the whole stack at no cost, join the waitlist for managed Olympus, or talk to us about enterprise — the same engine either way.

Self-hosted

Free · MIT

Run the full stack on your own infrastructure with your own model keys.

  • All core services
  • Bring your own key
  • Community support

Managed

Coming soon

Hosted digiquant with managed upgrades and observability. In development.

  • Everything in Self-hosted
  • Managed upgrades
  • Hosted tracing (digismith)
Join the waitlist

Enterprise

Contact

Custom deployment, SLAs, and support for regulated environments.

  • Everything in Managed
  • SLA + priority support
  • Deployment assistance
Contact us

Questions

What do I need to self-host?

A container runtime (or a Python environment) and access to an LLM — any LiteLLM-supported provider or a local model. The rest ships in the open-core stack.

How is NautilusTrader licensed?

digiquant builds on NautilusTrader (open source) for all backtest, optimize, and live paths — see the NautilusTrader repository for its current license terms.

Do I bring my own model keys?

Yes. Self-hosting uses your own provider keys (any LiteLLM-supported provider) or a local model; keys stay on your infrastructure.

Are there usage limits?

No artificial request caps on the self-hosted stack. Throughput is bounded only by your own infrastructure and provider limits.