Documentation · AlphaEngine CLI + quantOS
Local-First Execution. Shared System of Record.
Run parameter sweeps and proprietary simulators locally with AlphaEngine, then sync sealed studies (schema 1.2), ideas, runs, and trial ledgers to quantOS with `ae push` and `ae pull`.
Terminal Quickstart · ae push / ae pull
# 1. Install AlphaEngine 1.1.2 (free for anyone) pip install alphaengine==1.1.2 # 2. Initialize local research repo & seal a study ae init ae seal study.json ae verify study.json # 3. Firm orgs: mint an ae_live_ key in /portal/data (Connect) export QUANTOS_API_KEY="ae_live_..." ae remote add quantos https://api.qosai.app # 4. Sync sealed studies, ideas, runs & trial ledger ae push ae pull
01 · Architecture
Local-first vs. synced workspace.
The alphaengine library and its math run offline; login, workflows, reporting and sealing use the network; full return series are sent only with --share-series. The package requires numpy>=1.24 and scipy>=1.10 (our server pins numpy==2.2.6 and scipy==1.15.3). Firms get ae_live_ keys through Connect; retail has no CLI; anyone can pip install alphaengine free.
| Workflow Layer | Local-First (`alphaengine` offline) | Synced (`ae push` / `ae pull` + quantOS) |
|---|---|---|
| Parameter sweeps & custom simulators | Runs 100% locally in Python (`alphaengine.sweep`, `backtest_fn`). No account or network connection required. | Synced sealed studies (`schema_version: 1.2`) update the shared Experiment Ledger trial counter and Deflated Sharpe. |
| Canonical study sealing (`content_sha256`) | `ae seal study.json` and `ae verify study.json` compute and check the 64-hex SHA-256 over canonical JSON offline. | `POST /api/workspace/sync` recomputes the canonical JSON hash server-side, rejects mismatches, and marks verified studies. |
| Hypothesis Board & kill rules | Drafted locally as idea objects (`thesis`, `kill_rule`, `status`, `family`) in your project repo. | `ae push` maps ideas onto the permanent Hypothesis Board (`open`, `testing`, `killed`, `promoted`) and preserves killed theses. |
| Team handoffs & allocator verification | Local JSON artifacts on disk (`study.json`). | `ae pull` fetches teammate runs, ideas, and studies; verified studies display a 'Sealed with AlphaEngine' badge on `/shared/reports/[token]`. |
02 · Data Custody Boundary
Exactly what `ae push` sends — and never sends.
The sync endpoint (`POST /api/workspace/sync`) enforces a strict structural allowlist and series length guard on the server.
- Sealed studies (Schema 1.2)
- Canonical JSON metadata, 64-hex `content_sha256`, `n_trials` + `n_trials_source` (`derived_from_grid` | `asserted`), 64-hex `data_hash` / `dataset_hashes`, `deflated_sharpe`, `psr_vs_zero`, `sr0_expected_max`, `min_track_record_length`, `selection` (`pbo` via CSCV), `surface`, `performance`, and optional `trial_ledger` (`params_hash` + Sharpe per trial).
- Idea objects
- `id`, `title`, `thesis` (claim), `kill_rule` (pre-committed falsifier), `status` (`open`, `testing`, `killed`, `promoted`), `killed_reason`, and `family` (factor mechanism).
- Run records & trial ledger entries
- `id`, `workflow`, `label`, `verdict`, summary statistics (`sharpe`, `total_return_pct`, `max_drawdown_pct`), `n_trials`, `data_hash`, and `params_hash` entries.
- Derived 1D figure curves
- Capped at 512 points (`MAX_LIST_LEN = 512`) by default. Full-length 1D series are stored only when the caller explicitly opts in (`--share-series` / `user_opted_in: true`).
- Raw datasets & price/return matrices
- Never accepted or stored. Payloads containing `raw_dataset`, `raw_data`, `price_matrix`, `return_matrix`, `ohlcv`, `dataframe`, or 2D return matrices are rejected by both client and server guards.
- Proprietary strategy code (`.py`)
- Your `backtest_fn` and signal generation code execute strictly on your machine and never cross the wire.
- Unhashed parameter grids
- `store_params` is off by default; the trial ledger records `params_hash` so search breadth is auditable without exposing proprietary parameter values.
- Local LLM / model provider keys
- `ANTHROPIC_API_KEY`, `OPENAI_API_KEY`, `GEMINI_API_KEY`, and local gateway credentials remain in your local environment.
03 · Server-Enforced Entitlements
Plan limits for `/api/workspace/sync` and `/pull`.
Entitlements are resolved server-side from your account profile. Users cannot modify their own tier. Research records only — never investment advice.
| Plan Tier | Sync Scope | Allowance | Behavior |
|---|---|---|---|
| AlphaEngine OSS (Free) | Local Python library (`pip install alphaengine==1.1.2`) | 403 Forbidden on cloud sync (links to https://qosai.app/access) | Free for anyone to install. The alphaengine library and its math run offline; login, workflows, reporting and sealing use the network; full return series are sent only with --share-series. |
| Retail Starter ($79), Pro ($149) & Power ($299) | Web-only (`/retail`) · No CLI | 403 Forbidden on CLI sync & key minting | Retail has no CLI workflows and no `ae_live_` keys (anyone can still `pip install alphaengine` free for local math). |
| Firm Core ($10K), Pro ($15K) & Enterprise ($25K) | `team` · Org-scoped `ae_live_` keys via Connect | Unlimited team workspace sync & pull | Firms get org-scoped `ae_live_` keys through Connect in `/portal/data`, with full `ae push` and `ae pull` across all pod members and zero seat counting. |
Need to upgrade your plan or request Firm Pod access? Visit qosai.app/access or return to the complete documentation index.