alphaspec 0.1.0__py3-none-any.whl

alphaspec/_data/docs/MARKETS.md

@@ -0,0 +1,29 @@
+# Market registry
+
+Market ids are lower snake case. A runtime declares which ids it supports and rejects the
+rest. New ids are added here by pull request.
+
+| id | Instruments | Calendar / time zone | Symbol form | Notes |
+|---|---|---|---|---|
+| `cn_stock` | China A-shares | SSE/SZSE, Asia/Shanghai | `600000.SH`, `000001.SZ`, `830799.BJ` | T+1, daily price limits, 100-share lots |
+| `cn_etf` | China ETFs | SSE/SZSE, Asia/Shanghai | `510300.SH` | |
+| `cn_convbond` | China convertible bonds | SSE/SZSE, Asia/Shanghai | `113039.SH` | quantity in units of 10 bonds |
+| `cn_future` | China futures | CFFEX/SHFE/DCE/CZCE/INE/GFEX, Asia/Shanghai | `RB2505.SHF` | night sessions belong to the next trading day |
+| `cn_option` | China options | SSE/SZSE/CFFEX, Asia/Shanghai | `10008001.SH` | contract multiplier |
+| `hk_stock` | Hong Kong equities | HKEX, Asia/Hong_Kong | `00700.HK` | board lots vary per stock; half-day sessions |
+| `us_equity` | US equities and ETFs | NYSE/Nasdaq, America/New_York | `AAPL` | |
+| `crypto_spot` | crypto spot pairs | 24/7, UTC | `BTCUSDT` | |
+
+## Indicator profiles
+
+Every catalog entry has a profile:
+
+| profile | Meaning | Allowed on |
+|---|---|---|
+| `core` | computable from OHLCV bars | any market |
+| `ext.cn` | encodes China exchange rules: price-limit streaks, sealed boards, call auction, main-force money flow | `cn_*` |
+| `ext.l2` | needs order-book or trade-by-trade data | markets where the runtime's data source provides it |
+| `ext.data` | needs a runtime dataset beyond bars (fundamentals, index series) | where the runtime provides it |
+
+Market rules are scoped, not deleted. A new market brings its own rules the same way:
+as a profile, with lint rules stating where it applies.

alphaspec/_data/docs/RULES.md

@@ -0,0 +1,133 @@
+# Lint rules
+
+Schema validation tells you a document is well-formed. These rules tell you it will do
+what you think. Every rule here exists because a runtime accepted the mistake, ran it,
+and produced wrong results without raising an error.
+
+Severity: **error** = a compliant runtime must reject the document.
+**warning** = accepted, but almost certainly not what the author meant.
+
+Each rule has at least one case in `conformance/cases/` named after it.
+
+## Artifact rules
+
+### AS001
+**Indicator reference does not resolve** · error
+
+A rule refers to `{"ind": "alias"}` but `dsl.indicators` has no such alias (`close` is
+built in). The reference runtime compiles the node to nothing; the strategy runs, never
+enters, and reports nothing wrong.
+
+### AS002
+**Parameter reference does not resolve** · error
+
+`{"ref": "name"}` where `dsl.params` has no `name`.
+
+### AS003
+**Higher-timeframe indicator uses LIVE bars (look-ahead)** · error
+
+An indicator on a timeframe above the primary one with `barPolicy: "LIVE"`. Live, that
+reads the still-forming bar, which is legitimate. In a backtest, that bar has already
+finished: a daily bar read at 10:00 contains the 15:00 close. Backtest results look
+excellent and cannot be reproduced. Use `CLOSED` (the default): only completed
+higher-timeframe bars are visible.
+
+### AS004
+**Indicator references form a cycle** · error
+
+`a` uses `b` and `b` uses `a`.
+
+### AS005
+**Unknown indicator type** · error
+
+The type is not in `catalog/indicators.json`.
+
+### AS006
+**Unknown indicator parameter** · warning
+
+A key the indicator type does not take, e.g. `length` on `EMA` (which takes `period`).
+The runtime ignores it and uses the default, so the strategy runs with a different
+parameter than the one written.
+
+### AS007
+**Indicator not available on this market** · error
+
+An indicator from the `ext.cn` profile (price-limit streaks, sealed limit-up boards,
+call auction, main-force money flow) on a market outside `cn_*`. These encode China
+exchange rules and have no meaning elsewhere.
+
+### AS008
+**timeframe/barPolicy ignored for primary-series-only indicator** · warning
+
+Some indicators (listed in the catalog under `primarySeriesOnly`) always compute on the
+primary series with their own daily aggregation. A `timeframe` or `barPolicy` on them
+has no effect.
+
+### AS009
+**Indicator timeframe finer than primary timeframe** · error
+
+A 5m indicator on a 1d strategy. A finer series cannot be derived from coarser bars.
+
+### AS010
+**tick timeframe cannot be bar-backtested** · warning
+
+`tick` means "evaluate on every market-data frame". That timing dimension does not
+exist in bar history, so a bar backtest of a tick strategy would be a different strategy.
+
+### AS011
+**Symbol missing venue suffix** · error
+
+On `cn_*` and `hk_*` markets the code is only unique together with its venue:
+`600000.SH`, `000001.SZ`, `00700.HK`. A bare code is stored and looked up as a different
+series, so data silently splits in two.
+
+### AS012
+**searchTrials required for generated output** · error
+
+`producer.kind` is `AI_AGENT` or `FACTOR_MINER` and `searchTrials` is missing. Anyone who
+tries a thousand variants will find one with a great backtest. The number of trials is
+the input to correcting for that (deflated Sharpe); without it the gate cannot tell
+search luck from signal.
+
+### AS013
+**No exit rule** · warning
+
+Positions close only through stop rules or rebalancing.
+
+### AS014
+**Intraday indicator on a too-coarse primary timeframe** · error
+
+Indicators such as `DAY_HIGH` or `MINUTES_SINCE_OPEN` measure within a session. On a
+daily strategy there is nothing inside the bar to measure.
+
+### AS015
+**Indicator defined but never used** · warning
+
+Usually a leftover from editing, or an entry/exit rule that references the wrong alias.
+
+## Score rules
+
+### AS101
+**Duplicate symbol in submission** · error
+
+### AS102
+**Every score is null** · error
+
+### AS103
+**asOf is in the future** · error
+
+`asOf` is the last trading day whose information the scores used. A future date means
+the producer's clock or calendar is wrong, and every downstream timing is suspect.
+
+### AS104
+**Zeros look like missing values** · warning
+
+At least 20% of scores are exactly `0` and none are `null`. If `0` means "no data",
+send `null`: in a cross-sectional ranking `0` is a real position, and those symbols
+will be bought or sold as if the model had an opinion about them.
+
+### AS105
+**Weights sum above 1** · error
+
+Target weights must sum to at most 1; the remainder is cash. Leverage is a runtime and
+account decision, not a field in the submission.

alphaspec/_data/docs/RUNTIME_API.md

@@ -0,0 +1,80 @@
+# Runtime agent API (v1)
+
+What a runtime exposes to agents and SDKs. The reference runtime (EasyQuant) implements it
+under `/api/agent/v1`. Authenticate every request with an API key created in the runtime's
+account page:
+
+```
+Authorization: Bearer eqk_<keyId>_<secret>
+```
+
+An API key can submit documents, run evaluations and read results. It **cannot** place
+orders, operate accounts or manage keys; the runtime rejects it everywhere outside this
+prefix.
+
+| Method & path | Success | Other responses |
+|---|---|---|
+| `GET /whoami` | 200 identity, contract version, what the key can and cannot do | 401 invalid / revoked / expired key |
+| `POST /artifacts` — body: an `artifact` document | 201 `{artifactId, warnings}` | 422 `{issues}` (same codes as the local validator) · 400 unusable universe |
+| `GET /artifacts` | 200 your submissions, newest first, with the latest verdict | |
+| `POST /artifacts/{id}/evaluate` | 202 `{evaluationId, status}` | 429 daily evaluation quota used up · 503 queue full |
+| `GET /artifacts/{id}/gate-result` | 200 `gate-result` document | 202 still running · 409 `{status: FAILED, error}` |
+| `POST /artifacts/{id}/promote` — body optional: `{name}` | 201 `{strategyId, created, enabled: false}` · 200 if already promoted | 400 gate not passed · runtime product rules (market, pool size, names) apply as in its UI |
+
+## Evaluation
+
+The runtime picks the windows; the submitter cannot. The reference runtime runs **one**
+continuous backtest and reports statistics for two segments of it, so positions carry
+across the boundary and indicators warm up once:
+
+| Strategy timeframe | In-sample | Out-of-sample |
+|---|---|---|
+| `1d`, `1w` | 4 years | the most recent 1 year |
+| intraday | 18 months | the most recent 6 months |
+
+Performance figures are recomputed by the runtime (`computedBy: PLATFORM`) and annualised
+on trading time (A-shares 242 days a year, Hong Kong 247, US 252; crypto 24×7). Checks beyond
+the basic in-sample return / drawdown / Sharpe thresholds:
+
+Each check names what its number is in `metric` (`excessReturnVsIndex`, `sharpe`,
+`deflatedSharpeProbability`, ...). Read it rather than inferring from the code: the same check can
+change measure.
+
+- `GATE_OUT_OF_SAMPLE` — the out-of-sample segment must **beat the market index over the same
+  period** (`metric: excessReturnVsIndex`, a fraction, threshold 0) and stay within the drawdown limit. A single
+  year in a falling market makes absolute Sharpe look bad without saying anything about the
+  strategy; relative performance does. Markets without a benchmark index in the runtime fall
+  back to an absolute Sharpe floor, and the hint says so.
+- `GATE_MULTIPLE_TESTING` — Deflated Sharpe on the in-sample segment using `producer.searchTrials`;
+  the reference runtime passes it at 0.90. For a daily strategy (4-year in-sample) that is roughly an
+  annualised Sharpe of 0.64 with 1 trial, 0.90 with 2, 1.24 with 5 and 1.43 with 10: every extra
+  candidate you tried raises the bar. The gate only promotes to paper trading, which is the real
+  out-of-sample test.
+- `GATE_LOOKAHEAD` — the structural look-ahead checks passed at intake.
+
+"Out-of-sample" here means *chosen by the platform*, not *unseen at submission*: recent
+market data is public. The only truly unseen period is the one after you submit, which is
+what paper trading is for.
+
+## Promotion
+
+`promote` turns a gate-passed artifact into a paper-trading strategy in your runtime
+account. It is created **disabled**; you start it yourself in the runtime's UI.
+
+The reference runtime stores strategies with plain values only: on intake, `dsl.params` is
+inlined (`{"ref": "x"}` becomes the value, or `{"const": value}` where an operand is expected)
+and dropped. The promoted strategy therefore carries no `params`; your submitted document, with
+its parameters, stays attached to the artifact.
+
+## Python
+
+```python
+from alphaspec import Client
+
+c = Client()                       # ALPHASPEC_BASE_URL, ALPHASPEC_API_KEY
+aid = c.submit(doc)["artifactId"]  # validates locally first
+c.evaluate(aid)
+gate = c.wait(aid)                 # polls until done
+if gate["passed"]:
+    c.promote(aid, name="my strategy")
+```

alphaspec/_data/docs/SPEC.md

@@ -0,0 +1,99 @@
+# AlphaSpec 0.1 — semantics
+
+The JSON Schemas in `schema/0.1/` define the shape of each document. This page defines
+what the fields **mean**, where the shape alone cannot. Where this page and a runtime
+disagree, the runtime is non-compliant.
+
+## 1. Documents
+
+Every document has `contractVersion` and `kind`.
+
+| kind | Written by | Purpose |
+|---|---|---|
+| `artifact` | strategy author / agent | a complete declarative strategy (L2) |
+| `score` | model / research pipeline | cross-sectional scores or target weights (L1) |
+| `gate-result` | runtime | the verdict on a submission |
+| `evidence` | runtime | what happened when it ran |
+
+A runtime must refuse a `contractVersion` it does not implement rather than guess.
+
+## 2. Time
+
+### 2.1 Market-local calendar
+All dates (`asOf`, sample windows) are trading days in the **market's own calendar and
+time zone**: `Asia/Shanghai` for `cn_*`, `America/New_York` for `us_equity`, and so on.
+"Today" for a market is never the server's date.
+
+### 2.2 `asOf` for scores
+A score with `asOf = T` may use information available at T's close and nothing later.
+It is executed no earlier than the next session after T. A runtime receiving a score
+after its rebalance cutoff for T+1 must say what it did (skipped, used it late) in
+evidence; it must not silently trade on stale scores.
+
+### 2.3 Bars
+A bar is identified by its **end**: the 1d bar of T covers T's whole session and becomes
+visible only after the session closes. A 15m bar labelled 10:00 covers 09:45–10:00.
+
+### 2.4 `barPolicy`
+On an indicator whose `timeframe` is coarser than the artifact's primary timeframe:
+
+- `CLOSED` (default): the indicator sees only completed bars of its timeframe. At 10:00
+  on day T, a daily indicator sees days up to T-1.
+- `LIVE`: the indicator includes the forming bar, updated with intraday data.
+
+`LIVE` is meaningful in live trading, and look-ahead in a backtest unless the runtime
+rebuilds the forming bar strictly from data before each evaluation point. Version 0.1
+forbids it for higher timeframes (rule AS003).
+
+## 3. Instruments
+
+### 3.1 Identity
+An instrument is `(market, symbol)`. On `cn_*` and `hk_*` markets the venue suffix is
+part of the symbol (`600000.SH`); elsewhere the symbol is the venue's own ticker
+(`AAPL`, `BTCUSDT`).
+
+### 3.2 Units
+Quantities in evidence are in `caliber.volumeUnit`. Derivatives are counted in
+contracts; notional = quantity × price × contract multiplier, and
+`caliber.multiplierApplied` states whether reported notionals already include it.
+Monetary fields are in `caliber.currency`.
+
+## 4. Rule tree (artifact `dsl`)
+
+- `indicators` maps aliases to indicator instances. Rules reference them as
+  `{"ind": "alias"}`. `close` is a built-in alias.
+- Operands are `{"ind"}`, `{"const"}`, `{"ref"}`, inline arithmetic
+  `{"op": "SUB", "left", "right"}`, or rule statistics `{"op": "COUNT" | "BARS_SINCE", "rule", "bars"}`.
+  **A bare number is not an operand**; write `{"const": 0}`.
+- Scalar settings (`period`, `bars`, `tolerancePct`, ...) take a number, `{"const"}` or `{"ref"}`.
+- `{"ref": "name"}` resolves against `dsl.params`. Put anything you intend to tune there.
+- Logical nodes: `AND`/`OR` take `left`/`right` rules; `NOT` and `DAY_LATCH` take `rule`.
+- Comparisons: `GT GTE LT LTE EQ NEQ CROSS_UP CROSS_DOWN` take `left`/`right` operands.
+- Temporal: `HOLD_FOR` (held on each of the last N bars), `WITHIN` (held on at least one
+  of the last N bars), `SEQUENCE` (`then` holds now and `first` held within the previous
+  N bars), `DAY_LATCH` (stays true for the rest of the session once true).
+- Level events: `TOUCH BREAK RECLAIM RETEST HOLD` against any operand as `level`.
+
+Entry fires on the bar where `entry` evaluates true; how and at what price the order is
+placed is the runtime's execution model and is reported in evidence, not specified here.
+
+## 5. Gates
+
+A gate result with `computedBy: "SELF_REPORTED"` carries numbers the submitter supplied.
+Public rankings and any automatic promotion to live trading must use `PLATFORM` results
+only. Additional checks for generated output (`producer.kind` `AI_AGENT` or
+`FACTOR_MINER`):
+
+- `GATE_OUT_OF_SAMPLE`: the most recent segment of the runtime's evaluation, judged relative to
+  the market over the same period (beat the index, stay within the drawdown limit).
+- `GATE_MULTIPLE_TESTING`: Sharpe deflated by `producer.searchTrials`.
+- `GATE_LOOKAHEAD`: the structural checks of section 2 hold.
+
+Deflation alone does not catch leakage; it is necessary, not sufficient.
+
+## 6. What a document must never contain
+
+- Orders, accounts, broker routes. A spec document states intent; execution goes through
+  the runtime's single order path, the same one a human uses.
+- Code, or strings that are executed.
+- Performance figures as trusted input.

alphaspec/_data/examples/artifacts/01-ema-cross-us-equity.json

@@ -0,0 +1,20 @@
+{
+  "contractVersion": "0.1",
+  "kind": "artifact",
+  "title": "EMA 20/60 crossover",
+  "note": "The smallest useful artifact: two indicators, one entry, one exit, a fixed symbol list.",
+  "market": "us_equity",
+  "timeframe": "1d",
+  "dsl": {
+    "version": 1,
+    "indicators": {
+      "fast": { "type": "EMA", "period": 20 },
+      "slow": { "type": "EMA", "period": 60 }
+    },
+    "entry": { "type": "CROSS_UP", "left": { "ind": "fast" }, "right": { "ind": "slow" } },
+    "exit": { "type": "CROSS_DOWN", "left": { "ind": "fast" }, "right": { "ind": "slow" } }
+  },
+  "universe": { "mode": "SYMBOLS", "symbols": ["AAPL", "MSFT", "NVDA"] },
+  "riskProfile": { "maxSymbolExposurePct": 20, "maxTotalExposurePct": 60 },
+  "producer": { "kind": "HUMAN", "name": "example" }
+}

alphaspec/_data/examples/artifacts/02-rsi-reversion-params.json

@@ -0,0 +1,25 @@
+{
+  "contractVersion": "0.1",
+  "kind": "artifact",
+  "title": "RSI mean reversion with named parameters",
+  "note": "Thresholds live in dsl.params and are referenced with {\"ref\": ...}. A parameter scan or an agent iterating on this artifact changes one number in one place, and the diff says exactly that.",
+  "market": "crypto_spot",
+  "timeframe": "1h",
+  "dsl": {
+    "version": 1,
+    "params": { "rsiPeriod": 14, "oversold": 28, "exitLevel": 55 },
+    "indicators": {
+      "rsi": { "type": "RSI", "period": { "ref": "rsiPeriod" } }
+    },
+    "entry": { "type": "CROSS_UP", "left": { "ind": "rsi" }, "right": { "ref": "oversold" } },
+    "exit": { "type": "GTE", "left": { "ind": "rsi" }, "right": { "ref": "exitLevel" } }
+  },
+  "universe": { "mode": "SYMBOLS", "symbols": ["BTCUSDT", "ETHUSDT"] },
+  "riskProfile": { "maxOrderNotional": 5000, "maxDailyOrders": 12 },
+  "producer": {
+    "kind": "AI_AGENT",
+    "name": "example-agent@0.1.0",
+    "modelId": "claude-opus-5-5",
+    "searchTrials": 64
+  }
+}

alphaspec/_data/examples/artifacts/03-mtf-trend-pullback.json

@@ -0,0 +1,25 @@
+{
+  "contractVersion": "0.1",
+  "kind": "artifact",
+  "title": "Daily trend filter, 15-minute pullback entry",
+  "note": "Multi-timeframe without look-ahead: the daily EMA is read with barPolicy CLOSED, i.e. only from completed daily bars. With LIVE, a backtest would see today's close at 10:00 (rule AS003 rejects that).",
+  "market": "hk_stock",
+  "timeframe": "15m",
+  "dsl": {
+    "version": 1,
+    "indicators": {
+      "dailyClose": { "type": "PRICE", "field": "CLOSE", "timeframe": "1d", "barPolicy": "CLOSED" },
+      "dailyEma": { "type": "EMA", "period": 50, "timeframe": "1d", "barPolicy": "CLOSED" },
+      "price": { "type": "PRICE", "field": "CLOSE" },
+      "ema20": { "type": "EMA", "period": 20 }
+    },
+    "entry": {
+      "type": "AND",
+      "left": { "type": "GT", "left": { "ind": "dailyClose" }, "right": { "ind": "dailyEma" } },
+      "right": { "type": "CROSS_UP", "left": { "ind": "price" }, "right": { "ind": "ema20" } }
+    },
+    "exit": { "type": "LT", "left": { "ind": "dailyClose" }, "right": { "ind": "dailyEma" } }
+  },
+  "universe": { "mode": "SYMBOLS", "symbols": ["00700.HK", "09988.HK"] },
+  "producer": { "kind": "HUMAN", "name": "example" }
+}

alphaspec/_data/examples/artifacts/04-breakout-retest-sequence.json

@@ -0,0 +1,58 @@
+{
+  "contractVersion": "0.1",
+  "kind": "artifact",
+  "title": "Channel breakout, then retest within 10 bars",
+  "note": "Temporal composition: SEQUENCE requires the breakout first and the successful retest afterwards. Level events (BREAK, RETEST) take any operand as the level, here a 20-day Donchian upper band.",
+  "market": "cn_stock",
+  "timeframe": "1d",
+  "dsl": {
+    "version": 1,
+    "params": {
+      "retestBars": 10
+    },
+    "indicators": {
+      "upper": {
+        "type": "DONCHIAN_UPPER",
+        "period": 20
+      }
+    },
+    "entry": {
+      "type": "SEQUENCE",
+      "first": {
+        "type": "BREAK",
+        "level": {
+          "ind": "upper"
+        },
+        "direction": "UP",
+        "minVolumeRatio": 1.5
+      },
+      "then": {
+        "type": "RETEST",
+        "level": {
+          "ind": "upper"
+        },
+        "direction": "UP",
+        "tolerancePct": 1.0
+      },
+      "bars": {
+        "ref": "retestBars"
+      }
+    },
+    "exit": {
+      "type": "BREAK",
+      "level": {
+        "ind": "upper"
+      },
+      "direction": "DOWN",
+      "tolerancePct": 3.0
+    }
+  },
+  "universe": {
+    "mode": "UNIVERSE",
+    "universeId": "csi300"
+  },
+  "producer": {
+    "kind": "HUMAN",
+    "name": "example"
+  }
+}

alphaspec/_data/examples/artifacts/05-cn-limit-up-pullback.json

@@ -0,0 +1,33 @@
+{
+  "contractVersion": "0.1",
+  "kind": "artifact",
+  "title": "Pullback after a limit-up streak (China extension)",
+  "note": "Uses LIMIT_UP_DAYS from the ext.cn profile. Market rules are not removed from the spec, they are scoped: the same artifact on us_equity is rejected by AS007 because price limits do not exist there.",
+  "market": "cn_stock",
+  "timeframe": "5m",
+  "dsl": {
+    "version": 1,
+    "params": { "drawdownFactor": 0.65 },
+    "indicators": {
+      "streak": { "type": "LIMIT_UP_DAYS", "lookbackDays": 10, "limitDays": 3, "matchMode": "AT_LEAST" },
+      "dailyHigh": { "type": "HIGHEST", "in": { "ind": "high1d" }, "period": 10, "timeframe": "1d", "barPolicy": "CLOSED" },
+      "high1d": { "type": "PRICE", "field": "HIGH", "timeframe": "1d", "barPolicy": "CLOSED" },
+      "drawdownLine": { "type": "MUL", "left": { "ind": "dailyHigh" }, "right": { "ref": "drawdownFactor" } },
+      "price": { "type": "PRICE", "field": "CLOSE" },
+      "ema20": { "type": "EMA", "period": 20 }
+    },
+    "entry": {
+      "type": "AND",
+      "left": {
+        "type": "AND",
+        "left": { "type": "GT", "left": { "ind": "streak" }, "right": { "const": 0 } },
+        "right": { "type": "LTE", "left": { "ind": "price" }, "right": { "ind": "drawdownLine" } }
+      },
+      "right": { "type": "CROSS_UP", "left": { "ind": "price" }, "right": { "ind": "ema20" } }
+    },
+    "exit": { "type": "CROSS_DOWN", "left": { "ind": "price" }, "right": { "ind": "ema20" } }
+  },
+  "universe": { "mode": "ALL_MARKET" },
+  "riskProfile": { "maxSymbolExposurePct": 8, "maxTotalExposurePct": 60, "maxImpactSlippageBps": 30 },
+  "producer": { "kind": "HUMAN", "name": "example" }
+}

alphaspec/_data/examples/scores/01-model-scores-cn-stock.json

@@ -0,0 +1,17 @@
+{
+  "contractVersion": "0.1",
+  "kind": "score",
+  "asOf": "2026-09-30",
+  "market": "cn_stock",
+  "scoreName": "lgbm-v7",
+  "direction": "DESC",
+  "scores": [
+    { "symbol": "600519.SH", "score": 1.8423 },
+    { "symbol": "000858.SZ", "score": 1.5107 },
+    { "symbol": "601318.SH", "score": null },
+    { "symbol": "300750.SZ", "score": -0.9912 }
+  ],
+  "execution": { "topN": 30, "weighting": "SCORE_PROPORTIONAL", "universePolicy": "SUBMITTED",
+                 "riskProfile": { "maxSymbolExposurePct": 5, "maxTotalExposurePct": 80 } },
+  "producer": { "kind": "INSTITUTION", "name": "example-desk", "searchTrials": 1 }
+}

alphaspec/_data/examples/scores/02-target-weights-us-equity.json

@@ -0,0 +1,13 @@
+{
+  "contractVersion": "0.1",
+  "kind": "score",
+  "asOf": "2026-09-30",
+  "market": "us_equity",
+  "scoreName": "risk-parity-sleeve",
+  "weights": [
+    { "symbol": "SPY", "weight": 0.40 },
+    { "symbol": "TLT", "weight": 0.35 },
+    { "symbol": "GLD", "weight": 0.15 }
+  ],
+  "producer": { "kind": "HUMAN", "name": "example" }
+}