aether-quant 0.5.0__tar.gz → 0.6.0__tar.gz

{aether_quant-0.5.0 → aether_quant-0.6.0}/.gitignore

@@ -19,9 +19,7 @@ env/
 !.env.live.example
 ib_config.py
 *_config.local.py
-config.json
 config.local.json
-lean.json
 lean.local.json
 
 *.key

{aether_quant-0.5.0 → aether_quant-0.6.0}/PKG-INFO

@@ -1,5 +1,5 @@
 Metadata-Version: 2.4
 Name: aether-quant
-Version: 0.5.0
+Version: 0.6.0
 Summary: Aether Quant convenience CLI
 Requires-Python: >=3.10

{aether_quant-0.5.0 → aether_quant-0.6.0}/README.md

@@ -10,11 +10,23 @@
 
 <p align="center">
   <img src="https://img.shields.io/badge/python-3.10%2B-FF8C00?style=flat-square&labelColor=1A1A1A&logo=python&logoColor=white" alt="Python 3.10+">
-  <!-- AQ:TEST_BADGE_START --><img src="https://img.shields.io/badge/tests-583%2F583%20passing-brightgreen?style=flat-square&labelColor=1A1A1A" alt="583 of 583 tests passing"><!-- AQ:TEST_BADGE_END -->
+  <!-- AQ:TEST_BADGE_START --><img src="https://img.shields.io/badge/tests-611%2F611%20passing-brightgreen?style=flat-square&labelColor=1A1A1A" alt="611 of 611 tests passing"><!-- AQ:TEST_BADGE_END -->
   <img src="https://img.shields.io/pypi/v/aether-quant?style=flat-square&labelColor=1A1A1A&color=FF8C00" alt="PyPI version">
   <img src="https://img.shields.io/badge/docker-ghcr.io%2Faether--quant-2496ED?style=flat-square&labelColor=1A1A1A&logo=docker&logoColor=white" alt="Docker image on GHCR">
 </p>
 
+<p align="center">
+  <img src="https://img.shields.io/badge/PyTorch-4B5563?style=flat-square&labelColor=1A1A1A&logo=pytorch&logoColor=white" alt="PyTorch">
+  <img src="https://img.shields.io/badge/scikit--learn-4B5563?style=flat-square&labelColor=1A1A1A&logo=scikitlearn&logoColor=white" alt="scikit-learn">
+  <img src="https://img.shields.io/badge/QuantConnect%20Lean-4B5563?style=flat-square&labelColor=1A1A1A" alt="QuantConnect Lean">
+  <img src="https://img.shields.io/badge/FastAPI-4B5563?style=flat-square&labelColor=1A1A1A&logo=fastapi&logoColor=white" alt="FastAPI">
+  <img src="https://img.shields.io/badge/React-4B5563?style=flat-square&labelColor=1A1A1A&logo=react&logoColor=white" alt="React">
+  <img src="https://img.shields.io/badge/TypeScript-4B5563?style=flat-square&labelColor=1A1A1A&logo=typescript&logoColor=white" alt="TypeScript">
+  <img src="https://img.shields.io/badge/Redis-4B5563?style=flat-square&labelColor=1A1A1A&logo=redis&logoColor=white" alt="Redis">
+  <img src="https://img.shields.io/badge/PostgreSQL-4B5563?style=flat-square&labelColor=1A1A1A&logo=postgresql&logoColor=white" alt="PostgreSQL">
+  <img src="https://img.shields.io/badge/GitHub%20Actions-4B5563?style=flat-square&labelColor=1A1A1A&logo=githubactions&logoColor=white" alt="GitHub Actions">
+</p>
+
 Aether Quant is not a single static strategy — it's a **dynamic system**: a
 Mixture-of-Experts ensemble (bullish/bearish/sideways/volatility specialists)
 routed by a learned gating network, a market-regime detector, a 3D market
@@ -514,6 +526,8 @@ aq api
 aq webui
 aq docker up [--lean|--all]
 aq docker build
+aq config [get <dotted.key>|set <dotted.key> <value>|keys [<dotted.prefix>]]
+aq lean [get <dotted.key>|set <dotted.key> <value>|keys [<dotted.prefix>]]
 aq retrain <plan|train|validate|backtest|commit|promote|rollback|status> [...]
 aq trade-lock --on|--off|--auto|--status
 aq fetch <crypto|stock> --ticker <TICKER> --start <YYYY-MM-DD> --end <YYYY-MM-DD> [--apply]
@@ -531,6 +545,8 @@ wrapper around a command already documented elsewhere in this README:
 - **`aq webui`** — starts the webui dev server (`npm run dev`).
 - **`aq docker up [--lean|--all]`** — starts local infrastructure (default: Redis + PostgreSQL only).
 - **`aq docker build`** — rebuilds the `aether-quant` app image.
+- **`aq config`** — reads or edits `config.json` directly, no manual file editing needed. Bare `aq config` pretty-prints the whole file; `aq config keys [<dotted.prefix>]` lists every leaf key path (handy for finding the right key in a deeply nested file); `aq config get <dotted.key>` prints one value (scalar, or a whole nested section as JSON); `aq config set <dotted.key> <value>` writes it — the value is parsed as JSON first (so `true`/`123`/`0.5`/`["a","b"]` become their real types automatically), falling back to a plain string otherwise. Every `set` backs up the previous file to `config.json.bak` first and prints old → new so a mistake is immediately visible; changing a value's type (e.g. bool → string) prints a warning but still writes it, since this command intentionally gives full access to every key, not just a safe subset.
+- **`aq lean`** — the exact same `get`/`set`/`keys` tool as `aq config`, just pointed at `lean.json` (the QuantConnect Lean CLI's own config file — broker credentials, environments, data providers) instead. `aq lean set ib-trading-mode live`, `aq lean keys environments.live-paper`, etc.
 - **`aq retrain <stage>`** — dispatches to `python -m retraining.orchestrator <stage> ...` for a single manual pipeline stage.
 - **`aq trade-lock --on|--off|--auto|--status`** — manually overrides `main.py`'s sticky total-drawdown trade lock (see `development/v2_architecture.md`'s Manual Trade-Lock Override Contract). `--off` deliberately clears an otherwise-permanent lock; `--auto` returns to fully automatic behavior.
 - **`aq fetch <crypto|stock> --ticker <TICKER> --start <YYYY-MM-DD> --end <YYYY-MM-DD> [--apply]`** — fetches historical OHLCV from Yahoo Finance for a ticker that isn't in `config.json` yet, formats it into Lean's zip/CSV convention, and writes it to the right spot under `data/` (`data/crypto/coinbase/daily/<ticker>_trade.zip` or `data/equity/usa/daily/<ticker>.zip`). On `--apply`, it also appends a new asset block to `config.json`'s `phase1.universe.assets[]` — no manual editing needed. Dry run by default (no `--apply`): reports what would happen, writes nothing. Never runs `train.py` itself — once applied, run `python train.py --dataset-only` (then `python train.py` when ready) yourself to actually train on the new ticker. `crypto`/`stock` today; a `derivative` asset class is planned for V3.
@@ -706,5 +722,5 @@ architecture docs already identify:
 ---
 
 <div align="center">
-  <sub>Built with Python · PyTorch · QuantConnect Lean · FastAPI · React · Redis · PostgreSQL · Docker</sub>
+  <sub>Aether Quant</sub>
 </div>

{aether_quant-0.5.0 → aether_quant-0.6.0}/aether_quant.egg-info/PKG-INFO

@@ -1,5 +1,5 @@
 Metadata-Version: 2.4
 Name: aether-quant
-Version: 0.5.0
+Version: 0.6.0
 Summary: Aether Quant convenience CLI
 Requires-Python: >=3.10

{aether_quant-0.5.0 → aether_quant-0.6.0}/aether_quant.egg-info/SOURCES.txt

@@ -8,8 +8,10 @@ Dockerfile.retraining_worker
 Dockerfile.workers
 README.md
 aq_cli.py
+config.json
 docker-compose.yml
 generate_backtest_report.py
+lean.json
 main.py
 pyproject.toml
 risk_controls.py

{aether_quant-0.5.0 → aether_quant-0.6.0}/aether_quant.egg-info/scm_file_list.json

@@ -9,8 +9,10 @@
     ".env.compose.example",
     "Dockerfile.retraining_worker",
     "train_topology.py",
+    "config.json",
     "train.py",
     ".env.live.example",
+    "lean.json",
     ".dockerignore",
     "generate_backtest_report.py",
     ".gitattributes",

{aether_quant-0.5.0 → aether_quant-0.6.0}/aether_quant.egg-info/scm_version.json

@@ -1,7 +1,7 @@
 {
-  "tag": "0.5.0",
+  "tag": "0.6.0",
   "distance": 0,
-  "node": "gb230bc437a96905925c7aa521e7006da67cfb98c",
+  "node": "g5cffff5750320cc223918c89314db15f96f5365b",
   "dirty": false,
   "branch": "HEAD",
   "node_date": "2026-07-07"

{aether_quant-0.5.0 → aether_quant-0.6.0}/analyzer/README.md

@@ -53,3 +53,29 @@ risk engine, and before action categorization / Lean order placement.
   actions are written into the per-asset `signal_payload` for
   dashboard/Grafana visibility regardless of whether a real order is
   placed.
+- **Real composite scoring (additive, config-gated).** Until now, every
+  priority tier here checked exactly one raw field in isolation against a
+  fixed threshold (confidence alone, `topology_risk` alone,
+  `liquidity_action` alone) — pure if/elif routing, no aggregation. New
+  `compute_signal_quality_score(confidence, regime_confidence, topology,
+  liquidity)` computes a real bounded `[0,1]` composite — a small,
+  hand-tuned weighted blend of raw model confidence (0.45), regime
+  confidence (0.20), topology peer-support (0.20, penalized when
+  `topology_risk` is `isolated`/`elevated`), and liquidity friction (0.15,
+  penalized by `participation_rate`) — mirroring `moe/gating.py`'s
+  `_quality_multiplier`/`_performance_score` style: real math over
+  already-available fields, not a trained model. `MarketAnalysisDecision`
+  gains `signal_quality_score`/`signal_quality_breakdown` fields that are
+  **always** computed and populated on every decision, regardless of any
+  flag — visible in `visualization/state.json` immediately for dashboard
+  use. It only ever changes routing when
+  `phase_v2.market_analyzer.use_composite_signal_score` is explicitly
+  `true` (default `false`) — in that case it replaces raw `confidence`
+  with the composite score in the `trade` gate (priority 7) and the
+  `simulate`-vs-`observe` split (priority 8) only. The hard safety-override
+  tiers (trade-lock, `risk_off` regime, elevated topology, liquidity
+  blocked — priorities 1-6) are **never** affected by this flag, regardless
+  of its value — they stay simple categorical gates on purpose, same
+  reasoning as the topology-elevated/isolated rules above. Default `false`
+  means output is byte-identical to pre-this-change behavior everywhere
+  the flag isn't explicitly turned on.

{aether_quant-0.5.0 → aether_quant-0.6.0}/analyzer/__init__.py

@@ -4,10 +4,12 @@ from .market_analyzer import (
     ACTIONS,
     MarketAnalysisDecision,
     build_market_analysis_decision,
+    compute_signal_quality_score,
 )
 
 __all__ = [
     "ACTIONS",
     "MarketAnalysisDecision",
     "build_market_analysis_decision",
+    "compute_signal_quality_score",
 ]

{aether_quant-0.5.0 → aether_quant-0.6.0}/analyzer/market_analyzer.py

@@ -4,7 +4,7 @@ action."""
 
 from __future__ import annotations
 
-from dataclasses import asdict, dataclass
+from dataclasses import asdict, dataclass, field
 
 
 ACTIONS = ("observe", "simulate", "trade", "reduce_risk", "retrain_candidate")
@@ -21,11 +21,57 @@ class MarketAnalysisDecision:
     topology_considered: bool
     liquidity_considered: bool
     reasons: list[str]
+    signal_quality_score: float = 0.0
+    signal_quality_breakdown: dict = field(default_factory=dict)
 
     def to_dict(self) -> dict:
         return asdict(self)
 
 
+def _clamp01(value: float) -> float:
+    return max(0.0, min(1.0, value))
+
+
+def compute_signal_quality_score(
+    confidence: float,
+    regime_confidence: float,
+    topology: dict,
+    liquidity: dict,
+) -> tuple[float, dict]:
+    """Bounded [0,1] composite of raw model confidence, regime confidence,
+    topology peer-support and liquidity friction. Always computed and
+    exposed (signal_quality_score/signal_quality_breakdown) for dashboard
+    visibility; only influences routing in build_market_analysis_decision()
+    when use_composite_signal_score=True - see analyzer/README.md. Small
+    hand-tuned weights, mirroring moe/gating.py's
+    _quality_multiplier/_performance_score style - not a trained model."""
+    confidence_component = _clamp01(float(confidence))
+    regime_component = _clamp01(float(regime_confidence))
+
+    correlation_strength = _clamp01(float(topology.get("correlation_strength", 0.0) or 0.0))
+    topology_penalty = {"isolated": 0.3, "elevated": 0.6}.get(str(topology.get("topology_risk", "unknown")), 1.0)
+    topology_component = correlation_strength * topology_penalty if topology else regime_component
+
+    participation_rate = _clamp01(float(liquidity.get("participation_rate", 0.0) or 0.0))
+    liquidity_component = _clamp01(1.0 - participation_rate) if liquidity else 1.0
+
+    weights = {"confidence": 0.45, "regime": 0.20, "topology": 0.20, "liquidity": 0.15}
+    score = (
+        weights["confidence"] * confidence_component
+        + weights["regime"] * regime_component
+        + weights["topology"] * topology_component
+        + weights["liquidity"] * liquidity_component
+    )
+    breakdown = {
+        "confidence_component": confidence_component,
+        "regime_component": regime_component,
+        "topology_component": topology_component,
+        "liquidity_component": liquidity_component,
+        "weights": dict(weights),
+    }
+    return _clamp01(score), breakdown
+
+
 def build_market_analysis_decision(
     signal_name: str,
     confidence: float,
@@ -41,6 +87,7 @@ def build_market_analysis_decision(
     min_confidence_to_trade: float = 0.12,
     retrain_min_regime_confidence: float = 0.20,
     low_regime_confidence_threshold: float = 0.35,
+    use_composite_signal_score: bool = False,
 ) -> MarketAnalysisDecision:
     reasons: list[str] = []
     decision_source = str(gating.get("decision_source", "unknown"))
@@ -57,6 +104,17 @@ def build_market_analysis_decision(
     else:
         reasons.append("topology_absent_v2_11_pending")
 
+    # Always computed and exposed for dashboard visibility, regardless of
+    # use_composite_signal_score - see compute_signal_quality_score()'s
+    # docstring and analyzer/README.md. trade_metric is what priorities 7/8
+    # actually gate on below: the composite score when the flag is on,
+    # otherwise raw confidence - byte-identical to pre-flag behavior when
+    # use_composite_signal_score=False (the default).
+    signal_quality_score, signal_quality_breakdown = compute_signal_quality_score(
+        confidence, regime_confidence, topology, liquidity
+    )
+    trade_metric = signal_quality_score if use_composite_signal_score else confidence
+
     # Priority 1: reduce_risk - portfolio-level risk lock always wins.
     if trade_lock_active:
         reasons.append(trade_lock_reason or "risk_lock_active")
@@ -70,6 +128,8 @@ def build_market_analysis_decision(
             topology_considered=topology_considered,
             liquidity_considered=liquidity_considered,
             reasons=reasons,
+            signal_quality_score=signal_quality_score,
+            signal_quality_breakdown=signal_quality_breakdown,
         )
 
     # Priority 2: reduce_risk - asset-level risk regime override even
@@ -86,6 +146,8 @@ def build_market_analysis_decision(
             topology_considered=topology_considered,
             liquidity_considered=liquidity_considered,
             reasons=reasons,
+            signal_quality_score=signal_quality_score,
+            signal_quality_breakdown=signal_quality_breakdown,
         )
 
     # Priority 3: reduce_risk - elevated cross-sectional volatility pressure
@@ -106,6 +168,8 @@ def build_market_analysis_decision(
             topology_considered=topology_considered,
             liquidity_considered=liquidity_considered,
             reasons=reasons,
+            signal_quality_score=signal_quality_score,
+            signal_quality_breakdown=signal_quality_breakdown,
         )
 
     # Priority 4: retrain_candidate - zero experts contributing AND the
@@ -125,6 +189,8 @@ def build_market_analysis_decision(
             topology_considered=topology_considered,
             liquidity_considered=liquidity_considered,
             reasons=reasons,
+            signal_quality_score=signal_quality_score,
+            signal_quality_breakdown=signal_quality_breakdown,
         )
 
     # Priority 5: simulate - liquidity blocked (zero volume or DDV below
@@ -142,6 +208,8 @@ def build_market_analysis_decision(
             topology_considered=topology_considered,
             liquidity_considered=liquidity_considered,
             reasons=reasons,
+            signal_quality_score=signal_quality_score,
+            signal_quality_breakdown=signal_quality_breakdown,
         )
 
     # Priority 6: simulate - thin market; participation rate would be
@@ -158,15 +226,20 @@ def build_market_analysis_decision(
             topology_considered=topology_considered,
             liquidity_considered=liquidity_considered,
             reasons=reasons,
+            signal_quality_score=signal_quality_score,
+            signal_quality_breakdown=signal_quality_breakdown,
         )
 
     # Priority 7: trade - only for trading-eligible assets with an actionable
     # directional signal, sufficient confidence, no topology isolation, and
     # no liquidity block (redundant guard — tiers 5/6 already caught those).
+    # Gates on trade_metric, not raw confidence directly - see trade_metric's
+    # definition above (identical to confidence unless
+    # use_composite_signal_score=True).
     if (
         trading_eligible
         and signal_name in {"buy", "sell"}
-        and confidence >= min_confidence_to_trade
+        and trade_metric >= min_confidence_to_trade
         and topology_risk != "isolated"
         and liquidity_action not in {"block", "simulate_instead"}
     ):
@@ -181,6 +254,8 @@ def build_market_analysis_decision(
             topology_considered=topology_considered,
             liquidity_considered=liquidity_considered,
             reasons=reasons,
+            signal_quality_score=signal_quality_score,
+            signal_quality_breakdown=signal_quality_breakdown,
         )
 
     # Priority 8: simulate vs observe.
@@ -201,6 +276,8 @@ def build_market_analysis_decision(
             topology_considered=topology_considered,
             liquidity_considered=liquidity_considered,
             reasons=reasons,
+            signal_quality_score=signal_quality_score,
+            signal_quality_breakdown=signal_quality_breakdown,
         )
 
     reasons.append("no_actionable_edge")
@@ -214,4 +291,6 @@ def build_market_analysis_decision(
         topology_considered=topology_considered,
         liquidity_considered=liquidity_considered,
         reasons=reasons,
+        signal_quality_score=signal_quality_score,
+        signal_quality_breakdown=signal_quality_breakdown,
     )

{aether_quant-0.5.0 → aether_quant-0.6.0}/aq_cli.py

@@ -33,6 +33,7 @@ import sys
 import time
 import urllib.request
 import uuid
+from collections.abc import Iterator
 from datetime import date
 from importlib.metadata import version as installed_version
 from pathlib import Path
@@ -42,6 +43,7 @@ from risk.manual_override import read_manual_trade_lock_override, write_manual_t
 
 ROOT_DIR = Path(__file__).resolve().parent
 CONFIG_PATH = ROOT_DIR / "config.json"
+LEAN_JSON_PATH = ROOT_DIR / "lean.json"
 WEBUI_DIR = ROOT_DIR / "webui"
 README_PATH = ROOT_DIR / "README.md"
 
@@ -393,6 +395,115 @@ def cmd_trade_lock(args: argparse.Namespace) -> int:
     return 0
 
 
+class ConfigPathError(Exception):
+    """Raised by _get_config_value/_set_config_value for a bad dotted path."""
+
+
+def _get_config_value(config: dict, dotted_path: str) -> object:
+    node = config
+    walked: list[str] = []
+    for segment in dotted_path.split("."):
+        walked.append(segment)
+        if not isinstance(node, dict) or segment not in node:
+            raise ConfigPathError(f"no such config key: {'.'.join(walked)!r}")
+        node = node[segment]
+    return node
+
+
+def _coerce_config_value(raw: str) -> object:
+    """JSON-first parsing so true/false/123/0.5/[...]/{...} all become their
+    real type automatically; anything that isn't valid JSON on its own
+    (e.g. a bare word) is kept as a plain string."""
+    try:
+        return json.loads(raw)
+    except json.JSONDecodeError:
+        return raw
+
+
+def _set_config_value(config: dict, dotted_path: str, raw_value: str) -> tuple[object, object, bool]:
+    """Mutates `config` in place. Returns (old_value, new_value, type_changed) -
+    deliberately does not refuse to overwrite a list/dict: the caller wants
+    full read/write access to every key, not just scalars. Safety instead
+    comes from always reporting old -> new (and a type-change warning) to
+    the caller, plus the automatic config.json.bak snapshot cmd_config()
+    writes before every set."""
+    *parents, leaf = dotted_path.split(".")
+    node = config
+    for segment in parents:
+        if not isinstance(node, dict) or segment not in node:
+            raise ConfigPathError(f"no such config key: {dotted_path!r}")
+        node = node[segment]
+    if not isinstance(node, dict) or leaf not in node:
+        raise ConfigPathError(f"no such config key: {dotted_path!r}")
+    old_value = node[leaf]
+    new_value = _coerce_config_value(raw_value)
+    node[leaf] = new_value
+    return old_value, new_value, type(old_value) is not type(new_value)
+
+
+def _iter_leaf_paths(node: object, prefix: str = "") -> Iterator[str]:
+    """Recursively yields every dot-joined leaf path under `node`. A "leaf"
+    is any non-dict value (or an empty dict) - list-valued keys show up as
+    one leaf, never expanded per-element."""
+    if isinstance(node, dict) and node:
+        for key, value in node.items():
+            yield from _iter_leaf_paths(value, f"{prefix}.{key}" if prefix else key)
+    else:
+        yield prefix
+
+
+def _dispatch_json_config_command(args: argparse.Namespace, json_path: Path, command_attr: str) -> int:
+    """Shared dispatch for `aq config`/`aq lean` - both are the same
+    dump/get/set/keys tool over a single flat JSON file, just pointed at a
+    different path. See cmd_config()/cmd_lean()."""
+    data = json.loads(json_path.read_text(encoding="utf-8"))
+    command = getattr(args, command_attr, None)
+
+    try:
+        if command is None:
+            print(json.dumps(data, indent=2))
+            return 0
+
+        if command == "get":
+            value = _get_config_value(data, args.dotted_path)
+            print(value if isinstance(value, str) else json.dumps(value, indent=2))
+            return 0
+
+        if command == "keys":
+            root = _get_config_value(data, args.dotted_prefix) if args.dotted_prefix else data
+            for path in _iter_leaf_paths(root, args.dotted_prefix or ""):
+                print(path)
+            return 0
+
+        if command == "set":
+            shutil.copy2(json_path, json_path.with_suffix(".json.bak"))
+            old_value, new_value, type_changed = _set_config_value(data, args.dotted_path, args.value)
+            json_path.write_text(json.dumps(data, indent=4) + "\n", encoding="utf-8")
+            print(f"{args.dotted_path}: {old_value!r} -> {new_value!r}")
+            if type_changed:
+                print(
+                    f"WARNING: type changed from {type(old_value).__name__} to {type(new_value).__name__} "
+                    f"for {args.dotted_path}",
+                    file=sys.stderr,
+                )
+            return 0
+    except ConfigPathError as error:
+        print(f"error: {error}", file=sys.stderr)
+        return 1
+
+    return 1
+
+
+def cmd_config(args: argparse.Namespace) -> int:
+    return _dispatch_json_config_command(args, CONFIG_PATH, "config_command")
+
+
+def cmd_lean(args: argparse.Namespace) -> int:
+    return _dispatch_json_config_command(args, LEAN_JSON_PATH, "lean_command")
+
+    return 1
+
+
 def cmd_fetch(args: argparse.Namespace) -> int:
     report = fetch_adhoc_asset(args.asset_class, args.ticker, args.start, args.end, apply=args.apply)
     label = "APPLY" if args.apply else "DRY RUN"
@@ -464,12 +575,51 @@ def build_parser() -> argparse.ArgumentParser:
     docker_build_parser = docker_subparsers.add_parser("build", help="Rebuild the aether-quant app image")
     docker_build_parser.set_defaults(func=cmd_docker_build)
 
+    config_parser = subparsers.add_parser("config", help="Show or edit config.json")
+    config_parser.set_defaults(func=cmd_config)
+    config_subparsers = config_parser.add_subparsers(dest="config_command")
+
+    config_get_parser = config_subparsers.add_parser("get", help="Print a config.json value")
+    config_get_parser.add_argument("dotted_path")
+
+    config_keys_parser = config_subparsers.add_parser("keys", help="List leaf key paths (optionally scoped to a prefix)")
+    config_keys_parser.add_argument("dotted_prefix", nargs="?", default=None)
+
+    config_set_parser = config_subparsers.add_parser("set", help="Set a config.json value (JSON-parsed, string fallback)")
+    config_set_parser.add_argument("dotted_path")
+    config_set_parser.add_argument("value")
+
+    lean_parser = subparsers.add_parser("lean", help="Show or edit lean.json (same shape as `aq config`)")
+    lean_parser.set_defaults(func=cmd_lean)
+    lean_subparsers = lean_parser.add_subparsers(dest="lean_command")
+
+    lean_get_parser = lean_subparsers.add_parser("get", help="Print a lean.json value")
+    lean_get_parser.add_argument("dotted_path")
+
+    lean_keys_parser = lean_subparsers.add_parser("keys", help="List leaf key paths (optionally scoped to a prefix)")
+    lean_keys_parser.add_argument("dotted_prefix", nargs="?", default=None)
+
+    lean_set_parser = lean_subparsers.add_parser("set", help="Set a lean.json value (JSON-parsed, string fallback)")
+    lean_set_parser.add_argument("dotted_path")
+    lean_set_parser.add_argument("value")
+
     retrain_parser = subparsers.add_parser(
         "retrain", help="Thin dispatcher to python -m retraining.orchestrator <stage> ..."
     )
     retrain_parser.add_argument(
         "stage",
-        choices=["plan", "train", "train_topology", "validate", "backtest", "commit", "promote", "rollback", "status"],
+        choices=[
+            "plan",
+            "train",
+            "train_topology",
+            "train_gating",
+            "validate",
+            "backtest",
+            "commit",
+            "promote",
+            "rollback",
+            "status",
+        ],
     )
     retrain_parser.add_argument("retrain_args", nargs=argparse.REMAINDER, help="Passed through verbatim, e.g. --version-id <uuid>")
     retrain_parser.set_defaults(func=cmd_retrain)