Add synthetic taker flow, timestamps, open interest, and metadata to indicator harness; improve error messages and theme tweaks

- Generate buy/sell volume split with random fractions
- Add nanosecond timestamps for OHLC extremes within 1-minute bars
- Include open interest as separate random walk
- Add num_trades and quote_volume derived fields
- Derive extra_columns dynamically from indicator input requirements instead of hardcoded volume check
- Improve input_series error message clarity
- Adjust ChatPanel background color for user messages

sandbox/dexorder/tools/evaluate_indicator.py

@@ -179,7 +179,10 @@ async def evaluate_indicator(
 
         # Determine required columns
         input_cols = _INPUTS.get(name_lower, ("close",))
-    needs_volume = "volume" in input_cols
+
+    # Derive extra_columns from whatever the indicator actually needs
+    _STANDARD_OHLC = {"open", "high", "low", "close"}
+    extra_columns = [col for col in input_cols if col not in _STANDARD_OHLC]
 
     # Fetch OHLC
     try:
@@ -190,7 +193,7 @@ async def evaluate_indicator(
             period_seconds=period_seconds,
             start_time=from_time,
             end_time=to_time,
-            extra_columns=["volume"] if needs_volume else [],
+            extra_columns=extra_columns,
         )
     except Exception as exc:
         log.exception("evaluate_indicator: OHLC fetch failed")

sandbox/dexorder/tools/indicator_harness.py

@@ -36,9 +36,9 @@ def make_synthetic_ohlcv(n: int = 200):
 
     rng = np.random.default_rng(42)
 
-    # Realistic BTC-style price random walk
-    returns = rng.normal(0, 0.015, n)
-    closes = 40_000.0 * np.cumprod(1.0 + returns)
+    # Realistic BTC-style price random walk (log-normal GBM — always positive)
+    log_returns = rng.normal(0, 0.015, n)
+    closes = 40_000.0 * np.exp(np.cumsum(log_returns))
 
     opens = np.empty(n)
     opens[0] = closes[0]
@@ -49,12 +49,43 @@ def make_synthetic_ohlcv(n: int = 200):
     lows  = np.minimum(opens, closes) * (1.0 - noise)
     volumes = rng.uniform(1e6, 1e8, n)
 
+    # Taker flow: randomly split volume into buy/sell side
+    buy_frac = rng.uniform(0.2, 0.8, n)
+    buy_vols = volumes * buy_frac
+    sell_vols = volumes - buy_vols
+
+    # Synthetic bar timestamps (1-minute bars starting 2024-01-01 UTC, in nanoseconds)
+    bar_ns = 60 * 1_000_000_000
+    epoch_start = 1_704_067_200_000_000_000  # 2024-01-01 00:00:00 UTC in ns
+    bar_starts = epoch_start + np.arange(n) * bar_ns
+    # Each extreme occurs at a random point within its bar
+    open_times  = bar_starts + rng.integers(0, bar_ns, n)
+    high_times  = bar_starts + rng.integers(0, bar_ns, n)
+    low_times   = bar_starts + rng.integers(0, bar_ns, n)
+    close_times = bar_starts + rng.integers(0, bar_ns, n)
+
+    # Futures open interest: separate slow random walk (log-normal — always positive)
+    oi_log_returns = rng.normal(0, 0.003, n)
+    open_interest = 5e8 * np.exp(np.cumsum(oi_log_returns))
+
+    num_trades = rng.integers(500, 5000, n).astype(float)
+    quote_volume = closes * volumes
+
     return pd.DataFrame({
-        "open":   opens,
-        "high":   highs,
-        "low":    lows,
-        "close":  closes,
-        "volume": volumes,
+        "open":          opens,
+        "high":          highs,
+        "low":           lows,
+        "close":         closes,
+        "volume":        volumes,
+        "buy_vol":       buy_vols,
+        "sell_vol":      sell_vols,
+        "open_time":     open_times.astype(float),
+        "high_time":     high_times.astype(float),
+        "low_time":      low_times.astype(float),
+        "close_time":    close_times.astype(float),
+        "open_interest": open_interest,
+        "num_trades":    num_trades,
+        "quote_volume":  quote_volume,
     })
 
 
@@ -175,7 +206,13 @@ def run(impl_path: Path, metadata_path: Path) -> dict:
     args = []
     for col in input_series:
         if col not in df.columns:
-            return {"success": False, "error": f"input_series '{col}' not in synthetic df columns {list(df.columns)}"}
+            return {
+                "success": False,
+                "error": (
+                    f"input_series '{col}' is not a valid OHLC column. "
+                    f"Available columns: {list(df.columns)}"
+                ),
+            }
         args.append(df[col])
 
     # --- Execute ---

web/src/components/ChatPanel.vue

@@ -601,7 +601,7 @@ const chatStyles = {
   },
   message: {
     background: '#141414',
-    backgroundMe: '#089981',
+    backgroundMe: '#1c1c1c',
     color: '#dbdbdb',
     colorStarted: '#8a8a8a',
     backgroundDeleted: '#0f0f0f',