Expand model tag support: add GLM-5.1, simplify Anthropic IDs, scan tags anywhere in message

- Flink update_bars debouncing
- update_bars subscription idempotency bugfix
- Price decimal correction bugfix of previous commit
- Add GLM-5.1 model tag alongside renamed GLM-5
- Use short Anthropic model IDs (sonnet/haiku/opus) instead of full version strings
- Allow @tags anywhere in message content, not just at start
- Return hasOtherContent flag instead of trimmed rest string
- Only trigger greeting stream when tag has no other content
- Update workspace knowledge base references to platform/workspace and platform/shapes
- Hierarchical knowledge base catalog
- 151 Trading Strategies knowledge base articles
- Shapes knowledge base article
- MutateShapes tool instead of workspace patch

gateway/knowledge/trading/strategies/futures/contrarian-trading.md

@@ -0,0 +1,79 @@
+---
+description: "Futures mean-reversion strategy that buys recent underperformers and sells recent outperformers relative to an equally-weighted futures market index, with an extension using volume and open interest filters."
+tags: [futures, mean-reversion, contrarian, market-index, dollar-neutral]
+---
+
+# Contrarian Trading (Mean-Reversion)
+
+**Section**: 10.3 | **Asset Class**: Futures | **Type**: Mean-Reversion / Contrarian
+
+## Overview
+Analogous to the equity mean-reversion strategy (Section 3.9), this futures strategy bets that recent losers will rebound and recent winners will give back gains. Returns of individual futures are measured relative to an equally-weighted market index, and capital is allocated inversely to the deviation from that index. The result is a dollar-neutral, automatically constructed contrarian portfolio rebalanced weekly.
+
+## Construction / Mechanics
+Within a universe of N futures labeled i = 1,...,N, define the "market index" return as the equally-weighted average:
+
+```
+R_m = (1/N) Σ R_i                                            (469)
+```
+
+where R_i are individual futures returns, typically measured over the last one week.
+
+The capital allocation weights are:
+
+```
+w_i = -γ [R_i - R_m]                                         (470)
+```
+
+where γ > 0 is fixed via the dollar-neutral normalization condition:
+
+```
+Σ |w_i| = 1                                                   (471)
+```
+
+- Futures below the market index (R_i < R_m): positive weight (long)
+- Futures above the market index (R_i > R_m): negative weight (short)
+- The portfolio is automatically dollar-neutral (Σ w_i = 0)
+- The strategy buys losers and sells winners relative to the market index
+
+**Volatility adjustment**: To mitigate overinvestment in volatile futures, suppress w_i by 1/σ_i or 1/σ_i², where σ_i are the historical volatilities.
+
+## Return Profile
+Profits when futures returns mean-revert toward the market index over a one-week horizon. Returns are driven by short-term overreaction and subsequent correction. The strategy is market-neutral at the index level.
+
+## Key Parameters / Signals
+| Parameter | Description |
+|-----------|-------------|
+| R_i | Individual futures return over the last week |
+| R_m | Equally-weighted market index return (Eq. 469) |
+| w_i = -γ[R_i - R_m] | Allocation weight; negative for winners, positive for losers |
+| γ | Scaling parameter fixed by Eq. (471) |
+| σ_i | Historical volatility; used to suppress w_i optionally |
+| Rebalancing | Weekly |
+
+## Variations
+
+### 10.3.1 Contrarian Trading — Market Activity
+Volume and open interest filters can improve the basic mean-reversion signal. Define:
+
+```
+v_i = ln(V_i / V_i')                                         (472)
+u_i = ln(U_i / U_i')                                         (473)
+```
+
+where V_i is total volume for futures i over the last week, V_i' is total volume over the prior week, and U_i, U_i' are the analogous open interest quantities.
+
+**Construction:**
+1. Take the upper half of futures by volume factor v_i (higher recent volume relative to prior week).
+2. Within that subset, take the lower half by open interest factor u_i.
+3. Apply the contrarian weights from Eq. (470) to this filtered subset.
+
+**Rationale:**
+- Larger volume changes indicate greater overreaction (a stronger snap-back is expected).
+- A decrease in open interest (low u_i) signals hedger withdrawal and suggests a deeper market for the mean-reversion to work.
+
+## Notes
+- The simple weighting scheme (Eq. 470) can overinvest in highly volatile futures; volatility scaling (1/σ_i or 1/σ_i²) is recommended in practice.
+- Weekly rebalancing incurs transaction costs; the net alpha must exceed round-trip costs across all positions.
+- Contrarian strategies can suffer sustained losses during trending regimes; combining with a trend-following overlay (Section 10.4) may reduce drawdowns.
+- The market-index return R_m links this strategy to the broader futures universe; changing the universe composition changes the benchmark and alters all weights.