data timeout fixes; research agent improvements

2026-04-24 20:43:42 -04:00
parent 1800363566
commit 319d81c41f
37 changed files with 672 additions and 280 deletions
--- a/bin/deploy
+++ b/bin/deploy
@@ -109,72 +109,6 @@ if [ "$PROJECT" != "lifecycle-sidecar" ]; then
  rsync -a --checksum --delete protobuf/ $PROJECT/protobuf/
 fi

-# For gateway: copy Python API files for research subagent
-if [ "$PROJECT" == "gateway" ]; then
-  echo "Copying Python API files for research subagent..."
-
-  # Create api-source directory
-  mkdir -p gateway/src/harness/subagents/research/api-source
-
-  # Copy all Python API files (for easy future expansion)
-  cp sandbox/dexorder/api/*.py gateway/src/harness/subagents/research/api-source/
-
-  # Generate api-reference.md with verbatim Python source code
-  API_REF="gateway/src/harness/subagents/research/memory/api-reference.md"
-
-  cat > "$API_REF" << 'HEADER'
-# Dexorder Research API Reference
-
-This file contains the complete Python API source code with full docstrings.
-These files are copied verbatim from `sandbox/dexorder/api/`.
-
-The API provides access to market data and charting capabilities for research scripts.
-
---
-
-## Overview
-
-Research scripts access the API via:
-```python
-from dexorder.api import get_api
-api = get_api()
-```
-
-The API instance provides:
- `api.data` - DataAPI for fetching OHLC market data
- `api.charting` - ChartingAPI for creating financial charts
-
---
-
-## Complete API Source Code
-
-The following sections contain the verbatim Python source files with complete
-type hints, docstrings, and examples.
-
-HEADER
-
-  # Append each Python file
-  for py_file in api.py data_api.py charting_api.py __init__.py; do
-    if [ -f "sandbox/dexorder/api/$py_file" ]; then
-      echo "" >> "$API_REF"
-      echo "### $py_file" >> "$API_REF"
-      echo '```python' >> "$API_REF"
-      cat "sandbox/dexorder/api/$py_file" >> "$API_REF"
-      echo '```' >> "$API_REF"
-      echo "" >> "$API_REF"
-    fi
-  done
-
-  cat >> "$API_REF" << 'FOOTER'
-
---
-
-For practical usage patterns and complete working examples, see `usage-examples.md`.
-FOOTER
-
-  echo "Generated api-reference.md with Python API source code"
-fi
-
 docker build $NO_CACHE -f $PROJECT/Dockerfile --build-arg="CONFIG=$CONFIG" --build-arg="DEPLOYMENT=$DEPLOYMENT" -t dexorder/ai-$PROJECT:latest $PROJECT || exit 1

 # Cleanup is handled by trap
--- a/bin/dev
+++ b/bin/dev
@@ -511,9 +511,9 @@ deep_restart() {
            # Force restart iceberg-catalog since it depends on postgres and minio
            echo -e "${GREEN}→${NC} Force restarting iceberg-catalog (depends on postgres/minio)..."
            kubectl delete pod -l app=iceberg-catalog 2>/dev/null || true
-            # Remove all sandbox deployments and services to free quota
-            echo -e "${GREEN}→${NC} Removing all sandbox deployments and services..."
-            kubectl delete deployments,services --all -n sandbox 2>/dev/null || true
+            # Remove all sandbox deployments, services, and PVCs to fully reset user state
+            echo -e "${GREEN}→${NC} Removing all sandbox deployments, services, and PVCs..."
+            kubectl delete deployments,services,pvc --all -n sandbox 2>/dev/null || true
            ;;
        *)
            echo -e "${RED}Error: Unknown service '$service'${NC}"
--- a/deploy/k8s/base/gateway.yaml
+++ b/deploy/k8s/base/gateway.yaml
@@ -13,6 +13,10 @@ spec:
      protocol: TCP
      port: 3000
      targetPort: http
+    - name: zmq-events
+      protocol: TCP
+      port: 5571
+      targetPort: 5571
  type: ClusterIP
 ---
 apiVersion: apps/v1
@@ -64,6 +68,9 @@ spec:
            - name: http
              containerPort: 3000
              protocol: TCP
+            - name: zmq-events
+              containerPort: 5571
+              protocol: TCP

          volumeMounts:
            - name: config
--- a/deploy/k8s/base/network-policies.yaml
+++ b/deploy/k8s/base/network-policies.yaml
@@ -69,6 +69,8 @@ spec:
      ports:
        - protocol: TCP
          port: 3000
+        - protocol: TCP
+          port: 5571
    # External HTTPS (for exchange APIs, LLM APIs)
    - to:
        - ipBlock:
@@ -102,3 +104,5 @@ spec:
      ports:
        - protocol: TCP
          port: 3000
+        - protocol: TCP
+          port: 5571
--- a/doc/test_prompt.md
+++ b/doc/test_prompt.md
@@ -1 +1,11 @@
+---
+
+For the following series of analysis questions, use 5 years of 15 minute data from `ETH/USDT.BINANCE`
+
 what conclusions can you make by analyzing historical data on ETH price direction changes near market session overlaps and market sessions changes on monday and tuesday?
+
+---
+
+do the same price direction change analysis but specifically compre a two hr range (1 hour before and after 9am EST, NY open; 11am to 1pm range, and 4:30 to 7pm EST range to look for potential price direction changes and compare relative probibilty for monday and tuesday for all other times between wednesday and sunday..
+
+---
--- a/flink/src/main/java/com/dexorder/flink/iceberg/SchemaInitializer.java
+++ b/flink/src/main/java/com/dexorder/flink/iceberg/SchemaInitializer.java
@@ -123,10 +123,10 @@ public class SchemaInitializer {
    /**
     * Initialize the OHLC table if it doesn't exist.
     */
-    // Bump this when the schema changes. Tables with a different (or missing) version
-    // will be dropped and recreated. Increment by 1 for each incompatible change.
+    // Bump this when the schema changes. Increment by 1 for each change.
    // v1: open/high/low/close required; ingestor forward-fills interior gaps with previous close
-    private static final String OHLC_SCHEMA_VERSION = "1";
+    // v2: added num_trades and quote_volume (appended; backward-compatible via Iceberg schema evolution)
+    private static final String OHLC_SCHEMA_VERSION = "2";
    private static final String SCHEMA_VERSION_PROP = "app.schema.version";

    private void initializeOhlcTable() {
@@ -154,11 +154,13 @@ public class SchemaInitializer {
            if (tableExists) {
                Table existing = catalog.loadTable(tableId);
                String existingVersion = existing.properties().get(SCHEMA_VERSION_PROP);
+                LOG.info("Table {} already exists at schema version {}", tableId, existingVersion);
                if (!OHLC_SCHEMA_VERSION.equals(existingVersion)) {
-                    LOG.warn("Table {} has schema version '{}', expected '{}' — skipping (manual migration required if needed)",
-                            tableId, existingVersion, OHLC_SCHEMA_VERSION);
+                    LOG.info("Evolving table {} from version '{}' to '{}'", tableId, existingVersion, OHLC_SCHEMA_VERSION);
+                    evolveOhlcSchema(existing);
+                    existing.updateProperties().set(SCHEMA_VERSION_PROP, OHLC_SCHEMA_VERSION).commit();
+                    LOG.info("Schema evolution complete for {}", tableId);
                }
-                LOG.info("Table {} already exists at schema version {} — skipping creation", tableId, existingVersion);
                return;
            }

@@ -195,7 +197,11 @@ public class SchemaInitializer {

                // Metadata fields
                optional(16, "request_id", Types.StringType.get(), "Request ID that generated this data"),
-                required(17, "ingested_at", Types.LongType.get(), "Timestamp when data was ingested by Flink (nanoseconds since epoch)")
+                required(17, "ingested_at", Types.LongType.get(), "Timestamp when data was ingested by Flink (nanoseconds since epoch)"),
+
+                // Extended exchange fields — appended for backward-compatible schema evolution (v2)
+                optional(18, "num_trades",   Types.LongType.get(), "Number of trades in the candle"),
+                optional(19, "quote_volume", Types.LongType.get(), "Total quote asset volume (scaled by price precision)")
            );

            // Create the table with partitioning and properties
@@ -218,6 +224,30 @@ public class SchemaInitializer {
        }
    }

+    /**
+     * Add any columns missing from a v1 OHLC table to bring it to v2.
+     * Iceberg schema evolution is safe and non-destructive — existing rows get null for new columns.
+     */
+    private void evolveOhlcSchema(Table table) {
+        org.apache.iceberg.UpdateSchema update = table.updateSchema();
+        boolean changed = false;
+        java.util.Set<String> existing = new java.util.HashSet<>();
+        for (org.apache.iceberg.types.Types.NestedField f : table.schema().columns()) {
+            existing.add(f.name());
+        }
+        if (!existing.contains("num_trades")) {
+            update.addColumn("num_trades", Types.LongType.get(), "Number of trades in the candle");
+            changed = true;
+        }
+        if (!existing.contains("quote_volume")) {
+            update.addColumn("quote_volume", Types.LongType.get(), "Total quote asset volume (scaled by price precision)");
+            changed = true;
+        }
+        if (changed) {
+            update.commit();
+        }
+    }
+
    /**
     * Initialize the symbol_metadata table if it doesn't exist.
     */
--- a/flink/src/main/java/com/dexorder/flink/publisher/OHLCBatchDeserializer.java
+++ b/flink/src/main/java/com/dexorder/flink/publisher/OHLCBatchDeserializer.java
@@ -69,7 +69,13 @@ public class OHLCBatchDeserializer implements DeserializationSchema<OHLCBatchWra
                    row.getHigh(),
                    row.getLow(),
                    row.getClose(),
-                    row.hasVolume() ? row.getVolume()  : null
+                    row.hasVolume()      ? row.getVolume()      : null,
+                    row.hasBuyVol()      ? row.getBuyVol()      : null,
+                    row.hasSellVol()     ? row.getSellVol()     : null,
+                    row.hasOpenTime()    ? row.getOpenTime()    : null,
+                    row.hasCloseTime()   ? row.getCloseTime()   : null,
+                    row.hasNumTrades()   ? row.getNumTrades()   : null,
+                    row.hasQuoteVolume() ? row.getQuoteVolume() : null
            ));
        }

--- a/flink/src/main/java/com/dexorder/flink/publisher/OHLCBatchWrapper.java
+++ b/flink/src/main/java/com/dexorder/flink/publisher/OHLCBatchWrapper.java
@@ -116,57 +116,58 @@ public class OHLCBatchWrapper implements Serializable {

    /**
     * Single OHLC row. open/high/low/close/volume are nullable to support gap bars
-     * (periods where no trades occurred).
+     * (periods where no trades occurred). All extended fields are nullable and only
+     * populated when the exchange provides them (e.g. Binance klines).
     */
    public static class OHLCRow implements Serializable {
        private static final long serialVersionUID = 1L;

        private final long timestamp;
        private final String ticker;
-        private final Long open;   // null for gap bars
-        private final Long high;   // null for gap bars
-        private final Long low;    // null for gap bars
-        private final Long close;  // null for gap bars
-        private final Long volume; // null when no volume data
+        private final Long open;       // null for gap bars
+        private final Long high;       // null for gap bars
+        private final Long low;        // null for gap bars
+        private final Long close;      // null for gap bars
+        private final Long volume;
+        private final Long buyVol;
+        private final Long sellVol;
+        private final Long openTime;
+        private final Long closeTime;
+        private final Long numTrades;
+        private final Long quoteVolume;

        public OHLCRow(long timestamp, String ticker, Long open, Long high,
-                      Long low, Long close, Long volume) {
-            this.timestamp = timestamp;
-            this.ticker = ticker;
-            this.open = open;
-            this.high = high;
-            this.low = low;
-            this.close = close;
-            this.volume = volume;
+                      Long low, Long close, Long volume,
+                      Long buyVol, Long sellVol, Long openTime, Long closeTime,
+                      Long numTrades, Long quoteVolume) {
+            this.timestamp   = timestamp;
+            this.ticker      = ticker;
+            this.open        = open;
+            this.high        = high;
+            this.low         = low;
+            this.close       = close;
+            this.volume      = volume;
+            this.buyVol      = buyVol;
+            this.sellVol     = sellVol;
+            this.openTime    = openTime;
+            this.closeTime   = closeTime;
+            this.numTrades   = numTrades;
+            this.quoteVolume = quoteVolume;
        }

-        public long getTimestamp() {
-            return timestamp;
-        }
-
-        public String getTicker() {
-            return ticker;
-        }
-
-        public Long getOpen() {
-            return open;
-        }
-
-        public Long getHigh() {
-            return high;
-        }
-
-        public Long getLow() {
-            return low;
-        }
-
-        public Long getClose() {
-            return close;
-        }
-
-        public Long getVolume() {
-            return volume;
-        }
+        public long getTimestamp()   { return timestamp; }
+        public String getTicker()    { return ticker; }
+        public Long getOpen()        { return open; }
+        public Long getHigh()        { return high; }
+        public Long getLow()         { return low; }
+        public Long getClose()       { return close; }
+        public Long getVolume()      { return volume; }
+        public Long getBuyVol()      { return buyVol; }
+        public Long getSellVol()     { return sellVol; }
+        public Long getOpenTime()    { return openTime; }
+        public Long getCloseTime()   { return closeTime; }
+        public Long getNumTrades()   { return numTrades; }
+        public Long getQuoteVolume() { return quoteVolume; }

        public boolean isGapBar() {
            return open == null && high == null && low == null && close == null;
@@ -180,6 +181,9 @@ public class OHLCBatchWrapper implements Serializable {
                    (isGapBar() ? ", gap=true" :
                        ", open=" + open + ", high=" + high + ", low=" + low + ", close=" + close) +
                    ", volume=" + volume +
+                    ", buyVol=" + buyVol +
+                    ", sellVol=" + sellVol +
+                    ", numTrades=" + numTrades +
                    '}';
        }
    }
--- a/flink/src/main/java/com/dexorder/flink/sink/IcebergOHLCSink.java
+++ b/flink/src/main/java/com/dexorder/flink/sink/IcebergOHLCSink.java
@@ -79,10 +79,9 @@ public class IcebergOHLCSink {

            // Emit one RowData for each OHLC row in the batch
            for (OHLCBatchWrapper.OHLCRow row : batch.getRows()) {
-                GenericRowData rowData = new GenericRowData(RowKind.INSERT, 17);
+                GenericRowData rowData = new GenericRowData(RowKind.INSERT, 19);

                // Natural key fields (ticker, period_seconds, timestamp)
-                // Used by equality delete files for deduplication
                rowData.setField(0, StringData.fromString(ticker));
                rowData.setField(1, periodSeconds);
                rowData.setField(2, row.getTimestamp());
@@ -95,22 +94,26 @@ public class IcebergOHLCSink {

                // Volume data
                rowData.setField(7, row.getVolume());
-                rowData.setField(8, null);  // buy_vol (TODO: extract from protobuf)
-                rowData.setField(9, null);  // sell_vol
+                rowData.setField(8, row.getBuyVol());
+                rowData.setField(9, row.getSellVol());

                // Timing data
-                rowData.setField(10, null);  // open_time
-                rowData.setField(11, null);  // high_time
-                rowData.setField(12, null);  // low_time
-                rowData.setField(13, null);  // close_time
+                rowData.setField(10, row.getOpenTime());
+                rowData.setField(11, null);               // high_time — not provided by exchanges
+                rowData.setField(12, null);               // low_time  — not provided by exchanges
+                rowData.setField(13, row.getCloseTime());

                // Additional fields
-                rowData.setField(14, null);  // open_interest
+                rowData.setField(14, null);  // open_interest (futures only, not yet fetched)

                // Metadata fields
                rowData.setField(15, StringData.fromString(requestId));
                rowData.setField(16, ingestedAt);

+                // Extended exchange fields (appended at end for backward-compatible schema evolution)
+                rowData.setField(17, row.getNumTrades());
+                rowData.setField(18, row.getQuoteVolume());
+
                out.collect(rowData);
            }

--- a/gateway/knowledge/README.md
+++ b/gateway/knowledge/README.md
@@ -17,14 +17,21 @@ Documents should be in Markdown format with:
 - Code examples where relevant
 - Cross-references to other docs

+### Frontmatter Fields
+
+`description` (required) — One or two sentences describing what the article covers. This is injected into every agent's system prompt as a KB catalog so agents know what to look up without making an extra tool call.
+
+`tags` (optional) — List of topic tags for categorization.
+
 ### Example with Frontmatter

 ```markdown
 ---
-tags: [trading, risk-management, position-sizing]
+description: "Patterns for writing custom Python indicator scripts that compute values from OHLCV data and plot live on the chart."
+tags: [indicators, python, development]
 ---

-# Risk Management
+# Custom Indicator Development

 Content here...
 ```
--- a/gateway/knowledge/api-reference.md
+++ b/gateway/knowledge/api-reference.md
@@ -1,3 +1,7 @@
+---
+description: "Complete Python API reference (DataAPI and ChartingAPI) with full source and docstrings for use in research scripts."
+---
+
 # Dexorder Research API Reference

 This file contains the complete Python API source code with full docstrings.
@@ -124,11 +128,14 @@ class DataAPI(ABC):
                       - pandas Timestamp: pd.Timestamp("2021-12-20")
            end_time: End of time range. Same formats as start_time.
            extra_columns: Optional additional columns to include beyond the standard
-                          OHLC columns. Available options:
-                          - "volume" - Total volume (decimal float)
-                          - "buy_vol" - Buy-side volume (decimal float)
-                          - "sell_vol" - Sell-side volume (decimal float)
-                          - "open_time", "high_time", "low_time", "close_time" (timestamps)
+                          OHLC columns. Available options (all populated for Binance data):
+                          - "volume" - Total base-asset volume (decimal float)
+                          - "buy_vol" - Taker buy volume in base asset (decimal float)
+                          - "sell_vol" - Taker sell volume in base asset (decimal float)
+                          - "quote_volume" - Total quote-asset volume, e.g. USDT (decimal float)
+                          - "num_trades" - Number of trades in the candle (integer)
+                          - "open_time", "close_time" (nanosecond timestamps; Binance only)
+                          - "high_time", "low_time" (not provided by any exchange; always null)
                          - "open_interest" (for futures markets)
                          - "ticker", "period_seconds"

@@ -201,8 +208,8 @@ class DataAPI(ABC):
            length: Number of most recent candles to return (default: 1)
            extra_columns: Optional list of additional column names to include.
                          Same column options as historical_ohlc:
-                          - "volume", "buy_vol", "sell_vol"
-                          - "open_time", "high_time", "low_time", "close_time"
+                          - "volume", "buy_vol", "sell_vol", "quote_volume", "num_trades"
+                          - "open_time", "close_time", "high_time", "low_time"
                          - "open_interest", "ticker", "period_seconds"

        Returns:
--- a/gateway/knowledge/indicators/indicator-development.md
+++ b/gateway/knowledge/indicators/indicator-development.md
@@ -1,3 +1,7 @@
+---
+description: "API and patterns for writing custom Python indicator scripts that compute values from OHLCV data and plot live on the chart."
+---
+
 # Custom Indicator Development

 Custom indicators are Python scripts saved in the `indicator` category. They compute values from OHLCV data and are plotted live on the TradingView chart alongside built-in indicators.
--- a/gateway/knowledge/pandas-ta-reference.md
+++ b/gateway/knowledge/pandas-ta-reference.md
@@ -1,3 +1,7 @@
+---
+description: "Full catalog of technical indicators available via pandas-ta, with parameters and usage for research scripts and custom indicators."
+---
+
 # pandas-ta Reference for Research Scripts

 This catalog applies to both research scripts and custom indicators. For usage in research scripts see [`usage-examples.md`](usage-examples.md). For writing custom indicator scripts (with metadata for the TradingView plotter) see [`indicators/indicator-development.md`](indicators/indicator-development.md).
--- a/gateway/knowledge/platform/mcp-integration.md
+++ b/gateway/knowledge/platform/mcp-integration.md
@@ -1,3 +1,7 @@
+---
+description: "User sandbox lifecycle, persistent script storage categories, and session management for indicator, strategy, and research scripts."
+---
+
 # User Sandbox

 Each user has a dedicated sandbox environment that persists their data across sessions.
--- a/gateway/knowledge/platform/workspace.md
+++ b/gateway/knowledge/platform/workspace.md
@@ -1,3 +1,7 @@
+---
+description: "Workspace store schema, available stores, and WorkspaceRead/WorkspacePatch usage for reading and updating the user's UI state."
+---
+
 # Workspace

 The Workspace is the user's current UI context — what they are looking at, what is selected, and what persistent state belongs to their session. It is a collection of named **stores** that are kept in sync between the web client, the gateway, and the user's sandbox container.
--- a/gateway/knowledge/strategies/strategy-development.md
+++ b/gateway/knowledge/strategies/strategy-development.md
@@ -1,3 +1,7 @@
+---
+description: "PandasStrategy class API, order placement, backtesting, and paper trading patterns for automated crypto strategy development."
+---
+
 # Strategy Development Guide

 Strategies on Dexorder are `PandasStrategy` subclasses that receive a live stream of OHLCV bars and call `self.buy()` / `self.sell()` / `self.flatten()` to place orders.
--- a/gateway/knowledge/trading/technical-analysis.md
+++ b/gateway/knowledge/trading/technical-analysis.md
@@ -1,3 +1,7 @@
+---
+description: "Technical analysis fundamentals covering chart patterns, trend analysis, and TA concepts applicable to crypto OHLCV data."
+---
+
 # Technical Analysis Fundamentals

 Technical analysis is the study of historical price and volume data to identify patterns and predict future market movements.
--- a/gateway/knowledge/usage-examples.md
+++ b/gateway/knowledge/usage-examples.md
@@ -1,3 +1,7 @@
+---
+description: "Example research scripts showing common patterns for data retrieval, statistical analysis, and visualization."
+---
+
 # Research Script API Usage

 See [`api-reference.md`](api-reference.md) for the full DataAPI and ChartingAPI source with all method signatures and docstrings. See [`pandas-ta-reference.md`](pandas-ta-reference.md) for the indicator catalog.
@@ -75,11 +79,16 @@ print(df.head())

 ### Available Extra Columns

- `"volume"` - Total volume
- `"buy_vol"` - Buy-side volume
- `"sell_vol"` - Sell-side volume
- `"open_time"`, `"high_time"`, `"low_time"`, `"close_time"` - Timestamps for each price point
- `"open_interest"` - Open interest (for futures)
+All columns below are fully populated for Binance data. Other exchanges provide only `volume`.
+
+- `"volume"` - Total base-asset volume
+- `"buy_vol"` - Taker buy volume (base asset) — order flow long pressure
+- `"sell_vol"` - Taker sell volume (base asset) — order flow short pressure
+- `"quote_volume"` - Total quote-asset volume (e.g. USDT traded)
+- `"num_trades"` - Number of individual trades in the candle
+- `"open_time"`, `"close_time"` - Exact candle open/close timestamps (Binance only)
+- `"high_time"`, `"low_time"` - Not provided by any exchange (always null)
+- `"open_interest"` - Open interest (futures only)
 - `"ticker"` - Market identifier
 - `"period_seconds"` - Period in seconds

--- a/gateway/prompt/agent-main.md
+++ b/gateway/prompt/agent-main.md
@@ -1,6 +1,8 @@
 ---
 dynamic_imports:
  - user-preferences
+  - research-summary
+  - research-scripts
 ---

 # Main Agent Instructions
@@ -60,7 +62,9 @@ Use research for exploratory or one-off analysis. Use indicator whenever the use

 ## Pre-delegation Checks

-Before calling research, call `PythonList(category="research")` to check if a relevant script already exists. If it does, pass its name to the research instruction so the agent updates it rather than creating a duplicate.
+Before calling research, check the **Existing Research Scripts** list above. If a relevant script already exists, pass its exact name to the research instruction so the agent updates it rather than creating a duplicate.
+
+**Iterating on an idea across turns**: When the user refines, tweaks, or asks follow-up questions about an analysis already performed this session (e.g. "now do it with a 30-day window", "can you add a volume subplot", "try with ETH instead"), pass the **same script name** as before in the research instruction. The agent will update the existing script in place. Old versions are preserved in git history and do not need to be kept as separate scripts.

 Before calling strategy, call `PythonList(category="strategy")` similarly.

--- a/gateway/prompt/agent-research.md
+++ b/gateway/prompt/agent-research.md
@@ -9,6 +9,7 @@ static_imports:
 dynamic_imports:
  - conda-environment
  - custom-indicators
+  - research-scripts
 ---
 # Research Script Assistant

@@ -22,6 +23,22 @@ Create Python scripts that:
 - Generate professional charts using matplotlib via the ChartingAPI
 - All matplotlib figures are automatically captured and sent to the user as images

+## Exploratory Mindset
+
+Go beyond the literal request. The user's question is a starting point, not a ceiling. Adjacent analysis — things the user didn't ask for but that naturally illuminate the same topic — often produces the most valuable insights and can reframe or deepen the interpretation of the original result.
+
+**Always ask**: *What else is related to this that would be worth knowing?* Then include it.
+
+If the user asks about Monday morning opening price trends, also plot order flow imbalance, session volatility, and volume — these directly affect how the price trend should be interpreted. If the user asks about RSI divergences, also show the distribution of returns following each divergence type. If asked about a specific symbol's correlation with BTC, also show correlation stability over time and during high-volatility regimes.
+
+Concretely:
+- **Add subplots** for related metrics (volume, volatility, spread, order flow) alongside the primary chart
+- **Include summary statistics** the user didn't ask for but that contextualize the result (e.g. sample size, statistical significance, base rates, regime breakdowns)
+- **Surface anomalies or surprises** you notice in the data, even if tangential
+- **Stratify results** by relevant dimensions (time of day, day of week, bull/bear regime, high/low volatility) when the sample is large enough
+
+Keep it focused — adjacent analysis should feel like natural extensions of the same question, not a data dump. Two or three well-chosen additions are better than ten loosely related ones.
+
 ## Data Selection: Resolution and Time Window

 > **Rule**: Every research script must fetch the maximum useful history — target 100,000–200,000 bars, hard cap at 5 years. **Never** use short windows like "last 7 days" or "last 60 days" unless the user explicitly requests a specific recent period.
@@ -49,36 +66,11 @@ Quick reference — approximate bars per resolution at various windows:

 **When to shorten the window**: only if 5 years at the chosen resolution would far exceed 200,000 bars (e.g., 5m over 5 years ≈ 525k → shorten to ~2 years). Otherwise always use the full 5 years.

-## Available Tools
+## Tool Behavior Notes

-You have direct access to these MCP tools:
-
- **PythonWrite**: Create a new script (research, strategy, or indicator category)
-  - Required: category, name, description, details, code
-  - Optional: metadata (category-specific fields — see below)
-  - **For research**: fully executes the script and returns all output (stdout, stderr) and captured chart images. The response IS the execution result — **do not call `ExecuteResearch` afterward**.
-  - **For indicator/strategy**: runs against synthetic test data to catch compile/runtime errors; no chart images are generated.
-  - Returns validation results and execution output (text + images for research)
-
- **PythonEdit**: Update an existing script
-  - Required: category, name
-  - Optional: code, patches, description, details (full replacement), detail_patches (targeted text replacements in details), metadata
-  - **For research**: re-executes the script when code is changed and returns all output and images. **Do not call `ExecuteResearch` afterward**.
-  - **For indicator/strategy**: re-runs the validation test only.
-  - Returns validation results and execution output
-
- **PythonRead**: Read an existing research script
-  - Returns: code, metadata
-
- **PythonList**: List all research scripts
-  - Returns: array of {name, description, metadata}
-
- **ExecuteResearch**: Run a research script that already exists on disk
-  - Use this **only** when the user explicitly asks to re-run a script, or to run a script that was written in a previous session and already exists
-  - **Do not call this after `PythonWrite` or `PythonEdit`** — those tools already executed the script and returned its output
-  - Returns: text output and images
-
- **WebSearch**, **FetchPage**, **ArxivSearch**: Search the web or fetch pages for reference information when researching methodologies or indicators
+- **`PythonWrite` / `PythonEdit` for research**: auto-executes the script and returns all output (stdout, stderr) and captured images. **Do not call `ExecuteResearch` afterward** — the script has already run.
+- **`PythonWrite` / `PythonEdit` for indicator/strategy**: runs against synthetic test data only; no chart images are generated.
+- **`ExecuteResearch`**: use **only** when the user explicitly asks to re-run a script, or to run one written in a previous session. Never call it after `PythonWrite` or `PythonEdit`.

 ## Research Script API

@@ -109,6 +101,10 @@ When a user requests analysis:

 2. **Use the provided name**: The instruction will begin with `Research script name: "<name>"`. Always use that exact name when calling `PythonWrite` or `PythonEdit`. Check first with `PythonRead` — if the script already exists, use `PythonEdit` to update it rather than creating a new one with `PythonWrite`.

+   **One script per analysis idea**: If the name matches an existing script, the user is iterating on that idea — update it in place rather than creating a variant with a different name. Old versions are preserved in git history; there is no need to keep multiple scripts for variations of the same analysis.
+
+   **Duplicate detection**: Also review the **Existing Research Scripts** list above. If a script already exists there that appears to cover the same analysis as your current instruction — even under a different name — note this in your response after completing the task, so the user can decide whether to consolidate.
+
 3. **Write the script**: Use `PythonWrite` (new) or `PythonEdit` (existing)
   - Write clean, well-commented Python code
   - Include proper error handling
@@ -127,7 +123,17 @@ When a user requests analysis:
   - Use `PythonEdit` to fix the script
   - The script will auto-execute again

-6. **Return results**: Once successful, summarize what was done
+6. **Summarize findings**: After successful execution, update the research summary entry
+   using `ResearchSummaryPatch`:
+   - Replace the `**Findings:**` line(s) with 3–5 concise bullet points of key results
+   - Include only **statistically significant or practically notable** findings —
+     p-values, effect sizes, actionable patterns
+   - If nothing notable emerged: a single bullet `No significant patterns found`
+   - Keep the entire findings block under ~100 words; full output is always readable via
+     `PythonReadOutput(category="research", name="<script-name>")`
+   - This applies after `PythonWrite`, `PythonEdit`, and `ExecuteResearch` runs
+
+7. **Return results**: Once successful, summarize what was done
   - The user will receive both your text response AND the chart images
   - Don't try to describe the images in detail - the user can see them

--- a/gateway/prompt/agent-web-explore.md
+++ b/gateway/prompt/agent-web-explore.md
@@ -6,14 +6,6 @@ recursionLimit: 15

 You are a research assistant that searches the web and academic databases to answer questions or gather information according to the given instructions.

-## Tools
-
-You have three tools:
-
- **`WebSearch`** — Search the web broadly (Tavily). Returns titles, URLs, and content summaries. Best for general information, news, documentation, proprietary/niche topics, trading indicators, software papers, and anything not likely to be on arXiv.
- **`ArxivSearch`** — Search arXiv for academic preprints. Returns titles, authors, abstracts, and PDF links. Use this **only** for peer-reviewed or academic research (e.g. machine learning, statistics, finance theory). Most trading indicators, technical analysis tools, and proprietary methods are NOT on arXiv.
- **`FetchPage`** — Fetch the full content of a URL (web page or PDF). PDFs are automatically converted to text. Use this after searching to read the complete content of a promising result.
-
 ## Strategy

 1. **Choose the right search tool first:**
--- a/gateway/src/clients/duckdb-client.ts
+++ b/gateway/src/clients/duckdb-client.ts
@@ -44,6 +44,7 @@ export class DuckDBClient {
  private conversationsBucket?: string;
  private logger: FastifyBaseLogger;
  private initialized = false;
+  private initPromise: Promise<void> | null = null;

  constructor(config: DuckDBConfig, logger: FastifyBaseLogger) {
    this.logger = logger;
@@ -63,10 +64,16 @@ export class DuckDBClient {
   * Initialize DuckDB connection and configure S3/Iceberg extensions
   */
  async initialize(): Promise<void> {
-    if (this.initialized) {
-      return;
+    if (this.initialized) return;
+    if (!this.initPromise) {
+      this.initPromise = this._initialize().finally(() => {
+        this.initPromise = null;
+      });
    }
+    await this.initPromise;
+  }

+  private async _initialize(): Promise<void> {
    try {
      this.db = new Database(':memory:');
      this.conn = this.db.connect();
@@ -409,10 +416,12 @@ export class DuckDBClient {
      // duplicate parquet files (e.g. from repeated Flink job runs on the same key
      // range) never produce more than one row per (ticker, period_seconds, timestamp).
      const sql = `
-        SELECT timestamp, ticker, period_seconds, open, high, low, close, volume
+        SELECT timestamp, ticker, period_seconds, open, high, low, close,
+               volume, buy_vol, sell_vol, open_time, close_time, num_trades, quote_volume
        FROM (
          SELECT
-            timestamp, ticker, period_seconds, open, high, low, close, volume, ingested_at,
+            timestamp, ticker, period_seconds, open, high, low, close,
+            volume, buy_vol, sell_vol, open_time, close_time, num_trades, quote_volume, ingested_at,
            ROW_NUMBER() OVER (
              PARTITION BY timestamp
              ORDER BY ingested_at DESC
--- a/gateway/src/clients/zmq-relay-client.ts
+++ b/gateway/src/clients/zmq-relay-client.ts
@@ -37,7 +37,7 @@ export interface ZMQRelayConfig {
  relayRequestEndpoint: string;       // e.g., "tcp://relay:5559"
  relayNotificationEndpoint: string;  // e.g., "tcp://relay:5558"
  clientId?: string;                  // Optional client ID, will generate if not provided
-  requestTimeout?: number;            // Request timeout in ms (default: 60000)
+  requestTimeout?: number;            // Request timeout in ms (default: 120000)
  onMetadataUpdate?: () => Promise<void>; // Callback when symbol metadata updates
 }

@@ -77,7 +77,7 @@ export class ZMQRelayClient {
      relayRequestEndpoint: config.relayRequestEndpoint,
      relayNotificationEndpoint: config.relayNotificationEndpoint,
      clientId: config.clientId || `gateway-${randomUUID().slice(0, 8)}`,
-      requestTimeout: config.requestTimeout || 60000,
+      requestTimeout: config.requestTimeout || 120000,
      onMetadataUpdate: config.onMetadataUpdate || (async () => {}),
    };
    this.logger = logger;
--- a/gateway/src/harness/agent-harness.ts
+++ b/gateway/src/harness/agent-harness.ts
@@ -138,14 +138,29 @@ export class AgentHarness {
      }
    });

-    // Register the user-preferences virtual wiki page (loaded fresh each turn)
-    this.wikiLoader.registerVirtual('user-preferences', async (ctx) => {
+    // Register existing research scripts as a virtual wiki page
+    this.wikiLoader.registerVirtual('research-scripts', async (ctx) => {
+      if (!ctx.mcpClient) return '';
+      return this.fetchResearchScriptsSection(ctx.mcpClient);
+    });
+
+    this.registerFileDocument('user-preferences', 'PreferencesRead', 'User Preferences');
+    this.registerFileDocument('research-summary', 'ResearchSummaryRead', 'Research Summary');
+  }
+
+  /**
+   * Register a virtual wiki page backed by a sandbox file-document tool.
+   * The tool must return { content: string, exists: boolean }.
+   * Pages are loaded fresh every turn (never cached).
+   */
+  private registerFileDocument(virtualName: string, toolName: string, heading: string): void {
+    this.wikiLoader.registerVirtual(virtualName, async (ctx) => {
      if (!ctx.mcpClient) return '';
      try {
-        const result = await ctx.mcpClient.callTool('PreferencesRead', {});
+        const result = await ctx.mcpClient.callTool(toolName, {});
        const parsed = JSON.parse(String(result));
        if (!parsed.exists || !parsed.content?.trim()) return '';
-        return `## User Preferences\n\n${parsed.content}`;
+        return `## ${heading}\n\n${parsed.content}`;
      } catch {
        return '';
      }
@@ -235,6 +250,35 @@ export class AgentHarness {
    }
  }

+  /**
+   * Fetch existing research scripts and return a formatted markdown section.
+   * Used as the virtual wiki page 'research-scripts'.
+   */
+  private async fetchResearchScriptsSection(mcpClient: MCPClientConnector): Promise<string> {
+    try {
+      const raw = await mcpClient.callTool('PythonList', { category: 'research' });
+      const r = raw as any;
+      const text = r?.content?.[0]?.text ?? r?.[0]?.text;
+      const parsed = typeof text === 'string' ? JSON.parse(text) : raw;
+      const items: any[] = parsed?.items ?? [];
+      if (items.length === 0) return '';
+
+      const lines: string[] = ['## Existing Research Scripts', ''];
+      lines.push('The following research scripts already exist. Before creating a new one, check whether any of these already covers the requested analysis.\n');
+      lines.push('| Name | Description |');
+      lines.push('|------|-------------|');
+      for (const item of items) {
+        const name: string = item.name ?? 'unknown';
+        const desc: string = item.description ?? '';
+        lines.push(`| ${name} | ${desc} |`);
+      }
+      return lines.join('\n');
+    } catch (err) {
+      this.config.logger.warn({ err }, 'Failed to fetch research scripts for wiki page');
+      return '';
+    }
+  }
+
  /**
   * Fetch custom indicators from the sandbox and return a formatted markdown section.
   * Used as the virtual wiki page 'custom-indicators'.
--- a/gateway/src/harness/spawn/wiki-loader.ts
+++ b/gateway/src/harness/spawn/wiki-loader.ts
@@ -1,5 +1,5 @@
 import { readFile, readdir } from 'fs/promises';
-import { join, dirname } from 'path';
+import { join, dirname, relative } from 'path';
 import { fileURLToPath } from 'url';
 import yaml from 'js-yaml';
 import type { MCPClientConnector } from '../mcp-client.js';
@@ -16,6 +16,7 @@ const KNOWLEDGE_DIR = join(__dirname, '..', '..', '..', 'knowledge');
 const PROMPT_DIR = join(__dirname, '..', '..', '..', 'prompt');

 export interface WikiFrontmatter {
+  description?: string;
  maxTokens?: number;
  recursionLimit?: number;
  spawnsImages?: boolean;
@@ -157,36 +158,74 @@ export class WikiLoader {
  }

  /**
-   * Return the base prompt text: prompt/index.md body + prompt/tools.md body, concatenated.
+   * Return the base prompt text: prompt/index.md body + prompt/tools.md body + KB catalog.
   * Result is cached for the lifetime of this WikiLoader instance.
   */
  async getBasePrompt(): Promise<string> {
    if (this.basePromptCache !== null) return this.basePromptCache;

-    const [index, tools] = await Promise.all([
+    const [index, tools, catalog] = await Promise.all([
      this.loadPromptPage('index'),
      this.loadPromptPage('tools'),
+      this.getKBCatalog(),
    ]);

    const parts: string[] = [];
    if (index) parts.push(index.body);
    if (tools) parts.push(tools.body);
+    parts.push(catalog);
    this.basePromptCache = parts.join('\n\n');
    return this.basePromptCache;
  }

  /**
   * List all .md file names available in the knowledge directory (without extension).
+   * Returns paths relative to KNOWLEDGE_DIR, e.g. "api-reference", "platform/workspace".
   */
  async listPages(): Promise<string[]> {
-    try {
-      const files = await readdir(KNOWLEDGE_DIR);
-      return files
-        .filter(f => f.endsWith('.md'))
-        .map(f => f.slice(0, -3));
-    } catch {
-      return [];
+    const results: string[] = [];
+    const scan = async (dir: string): Promise<void> => {
+      let entries;
+      try {
+        entries = await readdir(dir, { withFileTypes: true });
+      } catch {
+        return;
+      }
+      for (const entry of entries) {
+        const full = join(dir, entry.name);
+        if (entry.isDirectory()) {
+          await scan(full);
+        } else if (entry.isFile() && entry.name.endsWith('.md')) {
+          const rel = relative(KNOWLEDGE_DIR, full);
+          results.push(rel.slice(0, -3)); // strip .md
+        }
+      }
+    };
+    await scan(KNOWLEDGE_DIR);
+    return results.sort();
+  }
+
+  /**
+   * Build a markdown catalog table of all KB pages with their descriptions.
+   * Used by getBasePrompt() to inject into the cached system prompt.
+   */
+  async getKBCatalog(): Promise<string> {
+    const pages = await this.listPages();
+    const rows: string[] = [];
+    for (const name of pages) {
+      const page = await this.loadPage(name);
+      const desc = page?.frontmatter.description ?? '';
+      rows.push(`| ${name} | ${desc} |`);
    }
+    return [
+      '## Knowledge Base',
+      '',
+      'The following reference articles are available via `MemoryLookup({page: "..."})`. Read an article when you need detailed reference information.',
+      '',
+      '| Page | Description |',
+      '|------|-------------|',
+      ...rows,
+    ].join('\n');
  }

  /**
--- a/gateway/src/services/ohlc-service.ts
+++ b/gateway/src/services/ohlc-service.ts
@@ -277,7 +277,7 @@ export class OHLCService {
        pricescale: 100,
        has_intraday: true,
        has_daily: true,
-        has_weekly_and_monthly: false,
+        has_weekly_and_monthly: true,
        supported_resolutions: DEFAULT_SUPPORTED_RESOLUTIONS,
        data_status: 'streaming',
      };
--- a/gateway/src/services/symbol-index-service.ts
+++ b/gateway/src/services/symbol-index-service.ts
@@ -25,6 +25,7 @@ export class SymbolIndexService {
  private logger: FastifyBaseLogger;
  private symbols: Map<string, SymbolMetadata> = new Map(); // key: "MARKET_ID.EXCHANGE" (Nautilus format)
  private initialized: boolean = false;
+  private initPromise: Promise<void> | null = null;

  constructor(config: SymbolIndexServiceConfig) {
    this.icebergClient = config.icebergClient;
@@ -57,7 +58,9 @@ export class SymbolIndexService {
        this.symbols.set(key, symbol);
      }

-      this.initialized = true;
+      if (symbols.length > 0) {
+        this.initialized = true;
+      }
      this.logger.info({
        count: this.symbols.size,
        totalRows: symbols.length,
@@ -74,12 +77,14 @@ export class SymbolIndexService {
   * Ensure index is initialized (with retry on failure)
   */
  private async ensureInitialized(): Promise<void> {
-    if (this.initialized) {
-      return;
+    if (this.initialized) return;
+    if (!this.initPromise) {
+      this.logger.info('Lazy-loading symbol index');
+      this.initPromise = this.initialize().finally(() => {
+        this.initPromise = null;
+      });
    }
-
-    this.logger.info('Lazy-loading symbol index');
-    await this.initialize();
+    await this.initPromise;
  }

  /**
@@ -213,7 +218,7 @@ export class SymbolIndexService {
      supported_resolutions: supportedResolutions.length > 0 ? supportedResolutions : DEFAULT_SUPPORTED_RESOLUTIONS,
      has_intraday: true,
      has_daily: true,
-      has_weekly_and_monthly: false,
+      has_weekly_and_monthly: true,
      pricescale,
      minmov: 1,
      base_currency: metadata.base_asset,
--- a/gateway/src/tools/mcp/mcp-tool-wrapper.ts
+++ b/gateway/src/tools/mcp/mcp-tool-wrapper.ts
@@ -125,6 +125,14 @@ export function createMCPToolWrapper(
  });
 }

+/** Silently parse a JSON string; pass non-strings through unchanged. */
+function tryParseJson(val: unknown): unknown {
+  if (typeof val === 'string') {
+    try { return JSON.parse(val); } catch { /* fall through */ }
+  }
+  return val;
+}
+
 /**
 * Convert MCP input schema to Zod schema
 */
@@ -156,17 +164,13 @@ function mcpInputSchemaToZod(inputSchema?: MCPToolInfo['inputSchema']): z.ZodObj
      case 'boolean':
        zodType = z.boolean().describe(prop.description || '');
        break;
-      case 'array':
-        // Handle array items
-        if (prop.items) {
-          const itemType = getZodTypeForProperty(prop.items);
-          zodType = z.array(itemType).describe(prop.description || '');
-        } else {
-          zodType = z.array(z.any()).describe(prop.description || '');
-        }
+      case 'array': {
+        const itemType = prop.items ? getZodTypeForProperty(prop.items) : z.any();
+        zodType = z.preprocess(tryParseJson, z.array(itemType)).describe(prop.description || '');
        break;
+      }
      case 'object':
-        zodType = z.object({}).passthrough().describe(prop.description || '');
+        zodType = z.preprocess(tryParseJson, z.object({}).passthrough()).describe(prop.description || '');
        break;
      default:
        zodType = z.any().describe(prop.description || '');
--- a/gateway/src/tools/platform/get-chart-data.tool.ts
+++ b/gateway/src/tools/platform/get-chart-data.tool.ts
@@ -43,7 +43,10 @@ Parameters:
      from_time: z.union([z.number(), z.string()]).optional().describe('Start time: Unix seconds OR date string (defaults to workspace chartState.start_time)'),
      to_time: z.union([z.number(), z.string()]).optional().describe('End time: Unix seconds OR date string (defaults to workspace chartState.end_time)'),
      countback: z.number().optional().describe('Limit number of bars returned (max 500)'),
-      columns: z.array(z.enum(['volume', 'buy_vol', 'sell_vol', 'open_time', 'high_time', 'low_time', 'close_time', 'open_interest'])).optional().describe('Extra columns beyond OHLC'),
+      columns: z.preprocess(
+        (val) => { if (typeof val === 'string') { try { return JSON.parse(val); } catch { /* fall through */ } } return val; },
+        z.array(z.enum(['volume', 'buy_vol', 'sell_vol', 'open_time', 'high_time', 'low_time', 'close_time', 'open_interest'])).optional()
+      ).describe('Extra columns beyond OHLC'),
    }),
    func: async ({ ticker, period, from_time, to_time, countback, columns }) => {
      const MAX_BARS = 500;
--- a/gateway/src/tools/platform/memory-lookup.tool.ts
+++ b/gateway/src/tools/platform/memory-lookup.tool.ts
@@ -19,31 +19,18 @@ export function createMemoryLookupTool(config: MemoryLookupToolConfig): DynamicS

  return new DynamicStructuredTool({
    name: 'MemoryLookup',
-    description: `Read a knowledge wiki page by name to get detailed reference information.
-
-Pass "index" to list all available pages.
-
-Example pages:
- "api-reference" — DataAPI and ChartingAPI reference for research scripts
- "usage-examples" — Example research scripts
- "pandas-ta-reference" — Full pandas-ta indicator catalog`,
+    description: `Read a knowledge wiki page by name to get detailed reference information. Available pages are listed in the Knowledge Base section of your system prompt.`,
    schema: z.object({
      page: z.string().describe(
-        'Wiki page name to read (without .md extension). Pass "index" to list all pages.'
+        'Wiki page name to read (without .md extension), e.g. "api-reference" or "platform/workspace".'
      ),
    }),
    func: async ({ page }: { page: string }): Promise<string> => {
      logger.info({ page }, 'memory_lookup: reading page');

-      if (page === 'index') {
-        const pages = await wikiLoader.listPages();
-        return `Available wiki pages:\n${pages.map(p => `- ${p}`).join('\n')}`;
-      }
-
      const wikiPage = await wikiLoader.loadPage(page);
      if (!wikiPage) {
-        const pages = await wikiLoader.listPages();
-        return `Page "${page}" not found.\n\nAvailable pages:\n${pages.map(p => `- ${p}`).join('\n')}`;
+        return `Page "${page}" not found. Refer to the Knowledge Base table in your system prompt for available page names.`;
      }

      return wikiPage.body;
--- a/ingestor/src/ccxt-fetcher.js
+++ b/ingestor/src/ccxt-fetcher.js
@@ -156,12 +156,32 @@ export class CCXTFetcher {
    const FETCH_RETRIES = 3;
    const FETCH_RETRY_DELAY_MS = 5000;

+    // Binance provides extended kline data (buy/sell volume split, quote volume, trade count).
+    // We use the raw klines endpoint directly to capture all available fields.
+    const isBinance = exchangeName === 'binance';
+    let binanceMarketId = null;
+    if (isBinance) {
+      if (!exchange.markets || Object.keys(exchange.markets).length === 0) {
+        await exchange.loadMarkets();
+      }
+      binanceMarketId = exchange.market(symbol).id;
+    }
+
    while (since < endMs) {
      let candles;
      let lastError;
      for (let attempt = 1; attempt <= FETCH_RETRIES; attempt++) {
        try {
-          candles = await exchange.fetchOHLCV(symbol, timeframe, since, PAGE_SIZE);
+          if (isBinance) {
+            candles = await exchange.publicGetKlines({
+              symbol: binanceMarketId,
+              interval: timeframe,
+              startTime: since,
+              limit: PAGE_SIZE
+            });
+          } else {
+            candles = await exchange.fetchOHLCV(symbol, timeframe, since, PAGE_SIZE);
+          }
          lastError = null;
          break;
        } catch (error) {
@@ -267,7 +287,7 @@ export class CCXTFetcher {
      } else if (prevClose !== null) {
        // Interior gap — forward-fill with previous close, zero volume
        gapCount++;
-        allCandles.push({
+        const gapBar = {
          ticker,
          timestamp:  (ts * 1_000_000).toString(),
          open:  prevClose,
@@ -277,7 +297,14 @@ export class CCXTFetcher {
          volume: '0',
          open_time:  (ts * 1_000_000).toString(),
          close_time: ((ts + periodSeconds * 1000) * 1_000_000).toString()
-        });
+        };
+        if (isBinance) {
+          gapBar.buy_vol = '0';
+          gapBar.sell_vol = '0';
+          gapBar.num_trades = '0';
+          gapBar.quote_volume = '0';
+        }
+        allCandles.push(gapBar);
      }
    }

@@ -332,27 +359,54 @@ export class CCXTFetcher {
  }

  /**
-   * Convert CCXT OHLCV array to our OHLC format
-   * CCXT format: [timestamp, open, high, low, close, volume]
-   * Uses precision fields from market metadata for proper integer representation
+   * Convert OHLCV array to our OHLC format.
+   *
+   * Accepts two formats:
+   *   - Standard CCXT (6 elements): [timestamp, open, high, low, close, volume]
+   *   - Binance raw klines (12 elements): [openTime, open, high, low, close, baseVolume,
+   *       closeTime, quoteVolume, numTrades, takerBuyBaseVol, takerBuyQuoteVol, ignore]
+   *
+   * Prices/volumes use integer representation scaled by market metadata precision.
   */
  convertToOHLC(candle, ticker, periodSeconds, metadata) {
-    const [timestamp, open, high, low, close, volume] = candle;
+    const timestamp = Number(candle[0]);
+    const open      = parseFloat(candle[1]);
+    const high      = parseFloat(candle[2]);
+    const low       = parseFloat(candle[3]);
+    const close     = parseFloat(candle[4]);
+    const volume    = parseFloat(candle[5]);

    const priceMult = Math.pow(10, metadata.pricePrecision ?? 2);
-    const sizeMult = Math.pow(10, metadata.sizePrecision ?? 8);
+    const sizeMult  = Math.pow(10, metadata.sizePrecision ?? 8);

-    return {
+    const result = {
      ticker,
-      timestamp: (timestamp * 1_000_000).toString(), // Convert ms to nanoseconds
-      open: Math.round(open * priceMult).toString(),
-      high: Math.round(high * priceMult).toString(),
-      low: Math.round(low * priceMult).toString(),
+      timestamp:  (timestamp * 1_000_000).toString(),
+      open:  Math.round(open  * priceMult).toString(),
+      high:  Math.round(high  * priceMult).toString(),
+      low:   Math.round(low   * priceMult).toString(),
      close: Math.round(close * priceMult).toString(),
      volume: Math.round(volume * sizeMult).toString(),
      open_time: (timestamp * 1_000_000).toString(),
-      close_time: ((timestamp + periodSeconds * 1000) * 1_000_000).toString()
    };
+
+    if (candle.length >= 10) {
+      // Binance extended klines format
+      const closeTimeMs    = Number(candle[6]);
+      const quoteVolRaw    = parseFloat(candle[7]);
+      const numTrades      = Number(candle[8]);
+      const takerBuyBase   = parseFloat(candle[9]);
+
+      result.close_time   = (closeTimeMs * 1_000_000).toString();
+      result.quote_volume = Math.round(quoteVolRaw  * priceMult).toString();
+      result.num_trades   = numTrades.toString();
+      result.buy_vol      = Math.round(takerBuyBase  * sizeMult).toString();
+      result.sell_vol     = Math.round((volume - takerBuyBase) * sizeMult).toString();
+    } else {
+      result.close_time = ((timestamp + periodSeconds * 1000) * 1_000_000).toString();
+    }
+
+    return result;
  }

  /**
--- a/ingestor/src/kafka-producer.js
+++ b/ingestor/src/kafka-producer.js
@@ -147,6 +147,12 @@ export class KafkaProducer {
          close: candle.close,
          volume: candle.volume
        };
+        if (candle.buy_vol     != null) protoCandle.buy_vol     = candle.buy_vol;
+        if (candle.sell_vol    != null) protoCandle.sell_vol    = candle.sell_vol;
+        if (candle.open_time   != null) protoCandle.open_time   = candle.open_time;
+        if (candle.close_time  != null) protoCandle.close_time  = candle.close_time;
+        if (candle.num_trades  != null) protoCandle.num_trades  = candle.num_trades;
+        if (candle.quote_volume != null) protoCandle.quote_volume = candle.quote_volume;

        const [frame1, frame2] = encodeMessage(MessageTypeId.OHLC, protoCandle, OHLC);
        const value = Buffer.concat([frame1, frame2]);
@@ -188,7 +194,13 @@ export class KafkaProducer {
          low:       candle.low,
          close:     candle.close,
        };
-        if (candle.volume != null) row.volume = candle.volume;
+        if (candle.volume      != null) row.volume      = candle.volume;
+        if (candle.buy_vol     != null) row.buy_vol     = candle.buy_vol;
+        if (candle.sell_vol    != null) row.sell_vol    = candle.sell_vol;
+        if (candle.open_time   != null) row.open_time   = candle.open_time;
+        if (candle.close_time  != null) row.close_time  = candle.close_time;
+        if (candle.num_trades  != null) row.num_trades  = candle.num_trades;
+        if (candle.quote_volume != null) row.quote_volume = candle.quote_volume;
        return row;
      })
    };
--- a/protobuf/ohlc.proto
+++ b/protobuf/ohlc.proto
@@ -23,6 +23,8 @@ message OHLC {
  optional int64 close_time = 12;
  optional int64 open_interest = 13;
  string ticker = 14;  // Nautilus format: "BTC/USDT.BINANCE"
+  optional int64 num_trades = 15;   // Number of trades in the candle
+  optional int64 quote_volume = 16; // Total quote asset volume (scaled by price precision)
 }

 // Batch of OHLC rows with metadata for historical request tracking
--- a/sandbox/dexorder/impl/data_api_impl.py
+++ b/sandbox/dexorder/impl/data_api_impl.py
@@ -20,6 +20,7 @@ OHLC_OPTIONAL_COLUMNS = [
    "volume", "buy_vol", "sell_vol",
    "open_time", "high_time", "low_time", "close_time",
    "open_interest",
+    "num_trades", "quote_volume",
 ]

 # All valid extra columns available in the Iceberg schema
@@ -27,6 +28,7 @@ VALID_EXTRA_COLUMNS = {
    "volume", "buy_vol", "sell_vol",
    "open_time", "high_time", "low_time", "close_time",
    "open_interest",
+    "num_trades", "quote_volume",
    "ticker", "period_seconds"
 }

@@ -51,7 +53,7 @@ class DataAPIImpl(DataAPI):
        s3_access_key: Optional[str] = None,
        s3_secret_key: Optional[str] = None,
        s3_region: Optional[str] = None,
-        request_timeout: float = 30.0,
+        request_timeout: float = 120.0,
    ):
        """
        Initialize DataAPI implementation.
@@ -65,7 +67,7 @@ class DataAPIImpl(DataAPI):
            s3_access_key: S3/MinIO access key
            s3_secret_key: S3/MinIO secret key
            s3_region: S3/MinIO region (e.g., "us-east-1")
-            request_timeout: Default timeout for historical data requests in seconds (default: 30)
+            request_timeout: Default timeout for historical data requests in seconds (default: 120)
        """
        self.ohlc_client = OHLCClient(
            iceberg_catalog_uri=iceberg_catalog_uri,
--- a/sandbox/dexorder/ohlc_client.py
+++ b/sandbox/dexorder/ohlc_client.py
@@ -90,7 +90,7 @@ class OHLCClient:
        period_seconds: int,
        start_time: int,
        end_time: int,
-        request_timeout: float = 30.0
+        request_timeout: float = 120.0
    ) -> pd.DataFrame:
        """
        Fetch OHLC data with smart caching.
@@ -108,7 +108,7 @@ class OHLCClient:
            period_seconds: OHLC period in seconds (60, 300, 3600, etc.)
            start_time: Start timestamp in nanoseconds
            end_time: End timestamp in nanoseconds
-            request_timeout: Timeout for historical data requests (default: 30s)
+            request_timeout: Timeout for historical data requests (default: 120s)

        Returns:
            DataFrame with OHLC data sorted by timestamp
@@ -180,6 +180,12 @@ class OHLCClient:
                if col in df.columns:
                    df[col] = df[col] / size_divisor

+        if price_precision is not None and price_precision > 0:
+            price_divisor = 10 ** price_precision
+            for col in ("quote_volume",):
+                if col in df.columns:
+                    df[col] = df[col] / price_divisor
+
        return df

    async def __aenter__(self):
--- a/sandbox/main.py
+++ b/sandbox/main.py
@@ -197,6 +197,73 @@ def _get_env_yml() -> Optional[Path]:
    return p if p.exists() else None


+def _coerce_json_arg(val, expected_type: str):
+    """Coerce a possibly-stringified JSON argument to its expected Python type.
+    Handles LLMs that serialize structured arguments as JSON strings.
+    expected_type: 'object' → dict | 'array' → list
+    Returns None if val is None or coercion is not possible.
+    """
+    if val is None:
+        return None
+    target = dict if expected_type == "object" else list
+    if isinstance(val, target):
+        return val
+    if isinstance(val, str):
+        try:
+            parsed = json.loads(val)
+            return parsed if isinstance(parsed, target) else None
+        except (ValueError, TypeError):
+            return None
+    return None
+
+
+def _update_research_summary(data_dir: Path, script_name: str, description: str, text_output: str) -> None:
+    """
+    Upsert the research-summary.md entry for the given script name.
+    Uses HTML comment anchors (<!-- BEGIN:name --> / <!-- END:name -->) to locate entries.
+    New entries get a stub with a findings placeholder; existing entries only have their
+    Last Run (and optionally Description) updated — agent-written findings are preserved.
+    """
+    import re
+    from datetime import datetime, timezone
+    summary_path = data_dir / "research-summary.md"
+    timestamp = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
+
+    begin_marker = f"<!-- BEGIN:{script_name} -->"
+    end_marker   = f"<!-- END:{script_name} -->"
+
+    stub_parts = [begin_marker, f"## {script_name}"]
+    if description:
+        stub_parts.append(f"**Description:** {description}")
+    stub_parts.append(f"**Last Run:** {timestamp}")
+    stub_parts.append("")
+    stub_parts.append("**Findings:** *(awaiting agent summary)*")
+    stub_parts.append(end_marker)
+    stub_entry = "\n".join(stub_parts)
+
+    if not summary_path.exists():
+        summary_path.write_text(f"# Research Summary\n\n{stub_entry}\n", encoding="utf-8")
+        return
+
+    content = summary_path.read_text(encoding="utf-8")
+    begin_idx = content.find(begin_marker)
+    end_idx   = content.find(end_marker)
+
+    if begin_idx != -1 and end_idx != -1:
+        # Entry exists — update only Last Run (and Description if provided), preserve findings
+        existing = content[begin_idx : end_idx + len(end_marker)]
+        updated = re.sub(r'\*\*Last Run:\*\* [^\n]*', f'**Last Run:** {timestamp}', existing)
+        if description:
+            if '**Description:**' in updated:
+                updated = re.sub(r'\*\*Description:\*\* [^\n]*', f'**Description:** {description}', updated)
+            else:
+                updated = re.sub(r'(## [^\n]*\n)', f'\\1**Description:** {description}\n', updated, count=1)
+        new_content = content[:begin_idx] + updated + content[end_idx + len(end_marker):]
+        summary_path.write_text(new_content, encoding="utf-8")
+    else:
+        summary_path.write_text(content.rstrip() + "\n\n---\n\n" + stub_entry + "\n", encoding="utf-8")
+
+

 # =============================================================================
 # Configuration
@@ -398,7 +465,6 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                            "description": "Name of the store"
                        },
                        "patch": {
-                            "type": "array",
                            "description": "JSON Patch operations (RFC 6902)",
                            "items": {
                                "type": "object",
@@ -454,6 +520,46 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                    "required": ["old_str", "new_str"]
                }
            ),
+            Tool(
+                name="ResearchSummaryRead",
+                description="Read the research summary markdown file. Returns the full content of research-summary.md from the user's sandbox data directory.",
+                inputSchema={
+                    "type": "object",
+                    "properties": {}
+                }
+            ),
+            Tool(
+                name="ResearchSummaryWrite",
+                description="Write (fully replace) the research summary markdown file.",
+                inputSchema={
+                    "type": "object",
+                    "properties": {
+                        "content": {
+                            "type": "string",
+                            "description": "Full markdown content for the research summary file"
+                        }
+                    },
+                    "required": ["content"]
+                }
+            ),
+            Tool(
+                name="ResearchSummaryPatch",
+                description="Surgically update a section of the research summary markdown file by finding and replacing text. Fails if old_str is not found.",
+                inputSchema={
+                    "type": "object",
+                    "properties": {
+                        "old_str": {
+                            "type": "string",
+                            "description": "Exact text to find in the research summary file"
+                        },
+                        "new_str": {
+                            "type": "string",
+                            "description": "Replacement text"
+                        }
+                    },
+                    "required": ["old_str", "new_str"]
+                }
+            ),
            Tool(
                name="PythonWrite",
                description="Write a new strategy, indicator, or research script with validation",
@@ -488,7 +594,6 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                            "description": "Python implementation code"
                        },
                        "metadata": {
-                            "type": "object",
                            "description": (
                                "Optional category-specific metadata. "
                                "For strategy: include 'data_feeds' (list of {symbol, period_seconds, description}) "
@@ -527,7 +632,6 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                            "description": "Full replacement Python code. Use only when rewriting the entire implementation; prefer 'patches' for targeted edits."
                        },
                        "patches": {
-                            "type": "array",
                            "description": (
                                "Targeted code edits as old/new string pairs. Preferred over 'code' for small changes. "
                                "Each patch: {\"old_string\": \"exact text to find\", \"new_string\": \"replacement text\"}. "
@@ -552,7 +656,6 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                            "description": "Full replacement for the details field. Use only when rewriting the entire description; prefer 'detail_patches' for targeted edits."
                        },
                        "detail_patches": {
-                            "type": "array",
                            "description": (
                                "Targeted edits to the details field as old/new string pairs. Preferred over 'details' for small changes. "
                                "Each patch: {\"old_string\": \"exact text to find\", \"new_string\": \"replacement text\"}. "
@@ -569,7 +672,6 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                            }
                        },
                        "metadata": {
-                            "type": "object",
                            "description": (
                                "Updated metadata fields (optional). "
                                "For strategy: 'data_feeds' (list of {symbol, period_seconds, description}) "
@@ -621,8 +723,6 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                            "description": "Display name of the item"
                        },
                        "files": {
-                            "type": "array",
-                            "items": {"type": "string"},
                            "description": (
                                "Specific filenames under output/ to return (e.g. [\"analysis.md\", \"img1.png\"]). "
                                "If omitted, returns all output files listed in metadata."
@@ -729,8 +829,6 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                    "type": "object",
                    "properties": {
                        "packages": {
-                            "type": "array",
-                            "items": {"type": "string"},
                            "description": "List of conda package names to install"
                        }
                    },
@@ -781,7 +879,6 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                            "description": "Lowercase pandas-ta function name, e.g. 'rsi', 'macd', 'bbands'"
                        },
                        "parameters": {
-                            "type": "object",
                            "description": "pandas-ta keyword arguments, e.g. {\"length\": 14} or {\"fast\": 12, \"slow\": 26, \"signal\": 9}",
                            "default": {}
                        }
@@ -806,7 +903,6 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                            "description": "Display name of the strategy as saved via PythonWrite"
                        },
                        "feeds": {
-                            "type": "array",
                            "description": "Data feeds to backtest against",
                            "items": {
                                "type": "object",
@@ -861,7 +957,6 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                            "description": "Display name of the strategy as saved via PythonWrite"
                        },
                        "feeds": {
-                            "type": "array",
                            "description": "Data feeds for the strategy",
                            "items": {
                                "type": "object",
@@ -1004,7 +1099,7 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
        elif name == "WorkspacePatch":
            return workspace_store.patch(
                arguments.get("store_name", ""),
-                arguments.get("patch", [])
+                _coerce_json_arg(arguments.get("patch"), "array") or []
            )
        elif name == "PreferencesRead":
            prefs_path = DATA_DIR / "preferences.md"
@@ -1025,6 +1120,25 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                return {"success": False, "error": "old_str not found in preferences file"}
            prefs_path.write_text(content.replace(old_str, new_str, 1), encoding="utf-8")
            return {"success": True}
+        elif name == "ResearchSummaryRead":
+            summary_path = DATA_DIR / "research-summary.md"
+            if not summary_path.exists():
+                return {"content": "", "exists": False}
+            content = summary_path.read_text(encoding="utf-8")
+            return {"content": content, "exists": True}
+        elif name == "ResearchSummaryWrite":
+            summary_path = DATA_DIR / "research-summary.md"
+            summary_path.write_text(arguments.get("content", ""), encoding="utf-8")
+            return {"success": True}
+        elif name == "ResearchSummaryPatch":
+            summary_path = DATA_DIR / "research-summary.md"
+            old_str = arguments.get("old_str", "")
+            new_str = arguments.get("new_str", "")
+            content = summary_path.read_text(encoding="utf-8") if summary_path.exists() else ""
+            if old_str not in content:
+                return {"success": False, "error": "old_str not found in research summary file"}
+            summary_path.write_text(content.replace(old_str, new_str, 1), encoding="utf-8")
+            return {"success": True}
        elif name == "PythonWrite":
            result = await category_manager.write(
                category=arguments.get("category", ""),
@@ -1032,7 +1146,7 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                description=arguments.get("description", ""),
                details=arguments.get("details", ""),
                code=arguments.get("code", ""),
-                metadata=arguments.get("metadata")
+                metadata=_coerce_json_arg(arguments.get("metadata"), "object")
            )
            content = []
            meta_parts = [f"success: {result['success']}"]
@@ -1043,7 +1157,11 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
            if result.get("revision"):
                meta_parts.append(f"revision: {result['revision']}")
            if result.get("validation") and not result["validation"].get("success"):
-                meta_parts.append(f"validation errors: {result['validation'].get('errors', [])}")
+                val = result["validation"]
+                error_detail = val.get('error') or ''
+                if val.get('output'):
+                    error_detail = f"{error_detail}\n{val['output']}" if error_detail else val['output']
+                meta_parts.append(f"validation error: {error_detail.strip()}")
            content.append(TextContent(type="text", text="\n".join(meta_parts)))
            if result.get("execution"):
                exec_content = result["execution"].get("content", [])
@@ -1058,17 +1176,29 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                sync = _workspace_sync_content(workspace_store, arguments.get("category", ""))
                if sync:
                    content.append(sync)
+                if arguments.get("category") == "research" and result.get("execution"):
+                    exec_text = "\n".join(
+                        item.text for item in result["execution"].get("content", [])
+                        if getattr(item, "type", "") == "text"
+                    )
+                    if exec_text.strip():
+                        _update_research_summary(
+                            DATA_DIR,
+                            arguments.get("name", ""),
+                            arguments.get("description", "") or "",
+                            exec_text,
+                        )
            return content
        elif name == "PythonEdit":
            result = await category_manager.edit(
                category=arguments.get("category", ""),
                name=arguments.get("name", ""),
                code=arguments.get("code"),
-                patches=arguments.get("patches"),
+                patches=_coerce_json_arg(arguments.get("patches"), "array"),
                description=arguments.get("description"),
                details=arguments.get("details"),
-                detail_patches=arguments.get("detail_patches"),
-                metadata=arguments.get("metadata")
+                detail_patches=_coerce_json_arg(arguments.get("detail_patches"), "array"),
+                metadata=_coerce_json_arg(arguments.get("metadata"), "object")
            )
            content = []
            meta_parts = [f"success: {result['success']}"]
@@ -1079,7 +1209,11 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
            if result.get("revision"):
                meta_parts.append(f"revision: {result['revision']}")
            if result.get("validation") and not result["validation"].get("success"):
-                meta_parts.append(f"validation errors: {result['validation'].get('errors', [])}")
+                val = result["validation"]
+                error_detail = val.get('error') or ''
+                if val.get('output'):
+                    error_detail = f"{error_detail}\n{val['output']}" if error_detail else val['output']
+                meta_parts.append(f"validation error: {error_detail.strip()}")
            content.append(TextContent(type="text", text="\n".join(meta_parts)))
            if result.get("execution"):
                exec_content = result["execution"].get("content", [])
@@ -1094,6 +1228,18 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                sync = _workspace_sync_content(workspace_store, arguments.get("category", ""))
                if sync:
                    content.append(sync)
+                if arguments.get("category") == "research" and result.get("execution"):
+                    exec_text = "\n".join(
+                        item.text for item in result["execution"].get("content", [])
+                        if getattr(item, "type", "") == "text"
+                    )
+                    if exec_text.strip():
+                        _update_research_summary(
+                            DATA_DIR,
+                            arguments.get("name", ""),
+                            arguments.get("description", "") or "",
+                            exec_text,
+                        )
            return content
        elif name == "PythonRead":
            return category_manager.read(
@@ -1104,7 +1250,7 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
            result = category_manager.read_output(
                category=arguments.get("category", ""),
                name=arguments.get("name", ""),
-                files=arguments.get("files"),
+                files=_coerce_json_arg(arguments.get("files"), "array"),
            )
            if "error" in result:
                return [TextContent(type="text", text=f"Error: {result['error']}")]
@@ -1140,7 +1286,11 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
            if result.get("error"):
                meta_parts.append(f"error: {result['error']}")
            if result.get("validation") and not result["validation"].get("success"):
-                meta_parts.append(f"validation errors: {result['validation'].get('errors', [])}")
+                val = result["validation"]
+                error_detail = val.get('error') or ''
+                if val.get('output'):
+                    error_detail = f"{error_detail}\n{val['output']}" if error_detail else val['output']
+                meta_parts.append(f"validation error: {error_detail.strip()}")
            if result.get("success"):
                _upsert_type(workspace_store, category_manager, arguments.get("category", ""), arguments.get("name", ""))
                sync = _workspace_sync_content(workspace_store, arguments.get("category", ""))
@@ -1175,7 +1325,7 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                environment_yml=_get_env_yml()
            )
        elif name == "CondaInstall":
-            return await install_packages_async(arguments.get("packages", []))
+            return await install_packages_async(_coerce_json_arg(arguments.get("packages"), "array") or [])
        elif name == "ExecuteResearch":
            result = await category_manager.execute_research(name=arguments.get("name", ""))
            if "error" in result:
@@ -1184,6 +1334,12 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
            content = result.get("content", [TextContent(type="text", text="No output")])
            image_count = sum(1 for item in content if item.type == "image")
            logging.info(f"ExecuteResearch '{arguments.get('name')}': returning {len(content)} items, {image_count} images")
+            exec_text = "\n".join(
+                item.text for item in content
+                if getattr(item, "type", "") == "text"
+            )
+            if exec_text.strip():
+                _update_research_summary(DATA_DIR, arguments.get("name", ""), "", exec_text)
            return content
        elif name == "EvaluateIndicator":
            return await evaluate_indicator(
@@ -1192,12 +1348,12 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                to_time=arguments.get("to_time"),
                period_seconds=int(arguments.get("period_seconds", 3600)),
                pandas_ta_name=arguments.get("pandas_ta_name", ""),
-                parameters=arguments.get("parameters") or {},
+                parameters=_coerce_json_arg(arguments.get("parameters"), "object") or {},
            )
        elif name == "BacktestStrategy":
            result = await backtest_strategy(
                strategy_name=arguments.get("strategy_name", ""),
-                feeds=arguments.get("feeds", []),
+                feeds=_coerce_json_arg(arguments.get("feeds"), "array") or [],
                from_time=arguments.get("from_time"),
                to_time=arguments.get("to_time"),
                initial_capital=float(arguments.get("initial_capital", 10_000.0)),
@@ -1214,7 +1370,7 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
                        from_time=arguments.get("from_time"),
                        to_time=arguments.get("to_time"),
                        initial_capital=float(arguments.get("initial_capital", 10_000.0)),
-                        feeds=arguments.get("feeds", []),
+                        feeds=_coerce_json_arg(arguments.get("feeds"), "array") or [],
                        summary=payload.get("summary", {}),
                        statistics=payload.get("statistics", {}),
                        trades=payload.get("trades", []),
@@ -1226,7 +1382,7 @@ def create_mcp_server(config: Config, event_publisher: EventPublisher) -> Server
        elif name == "ActivateStrategy":
            return await activate_strategy(
                strategy_name=arguments.get("strategy_name", ""),
-                feeds=arguments.get("feeds", []),
+                feeds=_coerce_json_arg(arguments.get("feeds"), "array") or [],
                allocation=float(arguments.get("allocation", 0.0)),
                paper=bool(arguments.get("paper", True)),
            )
--- a/web/src/components/ChatPanel.vue
+++ b/web/src/components/ChatPanel.vue
@@ -634,6 +634,23 @@ const injectShadowStyles = () => {
  }
 }

+const updateShadowStopOverride = (enabled: boolean) => {
+  const chatEl = document.querySelector('vue-advanced-chat')
+  if (!chatEl?.shadowRoot) return
+  chatEl.shadowRoot.querySelector('#vac-stop-override')?.remove()
+  if (enabled) {
+    const style = document.createElement('style')
+    style.id = 'vac-stop-override'
+    style.textContent = `
+      .vac-send-disabled {
+        pointer-events: auto !important;
+        cursor: pointer !important;
+      }
+    `
+    chatEl.shadowRoot.appendChild(style)
+  }
+}
+
 watch(() => channelStore.isReady, async (ready) => {
  if (!ready) return
  await nextTick()
@@ -644,6 +661,10 @@ watch(() => channelStore.isReady, async (ready) => {
  }, 100)
 })

+watch(isAgentProcessing, (val) => {
+  updateShadowStopOverride(val)
+})
+
 onUnmounted(() => {
  wsManager.removeHandler(handleMessage)
 })
@@ -711,6 +732,7 @@ onUnmounted(() => {
        @mousedown="isStopPressed = true"
        @mouseup="isStopPressed = false"
        :style="{
+          pointerEvents: 'auto',
          display: 'flex',
          alignItems: 'center',
          justifyContent: 'center',