Search values MCP tool

app/mcp/v1/tools/ch_reader.py

@@ -8,6 +8,7 @@
     get_statistics_by_type_from_ch,
     get_trend_data_from_ch,
     search_health_records_from_ch,
+    search_values_from_ch,
 )
 
 ch_reader_router = FastMCP(name="CH Reader MCP")
@@ -22,8 +23,15 @@ def get_health_summary_ch() -> dict[str, Any]:
 
     Notes for LLM:
     - IMPORTANT - Do not guess, autofill, or assume any missing data.
-    - When asked for medical advice, ask the user whether he wants to use DuckDB, ClickHouse or
-    Elasticsearch.
+    - If there are multiple databases available (DuckDB, ClickHouse, Elasticsearch):
+      first, ask the user which one he wants to use. DO NOT call any tools before
+      the user specifies his intent.
+    - If the user decides on an option, only use tools from this database,
+      do not switch over to another until the user specifies that he wants
+      to use a different one. You do not have to keep asking whether
+      the user wants to use the same database that he used before.
+    - If there is only one database available (DuckDB, ClickHouse, Elasticsearch):
+      you can use the tools from this database without the user specifying it.
     """
     try:
         return get_health_summary_from_ch()
@@ -44,14 +52,21 @@ def search_health_records_ch(params: HealthRecordSearchParams) -> dict[str, Any]
       matching the search criteria.
     - Each document in the list should represent a single health record as stored in ClickHouse.
     - If an error occurs, the function should return a list with a single dict
-      containing an 'error' key and the error message.
+     containing an 'error' key and the error message.
     - Use this to retrieve structured health data for further analysis, filtering, or display.
     - Example source_name: "Rob’s iPhone", "Polar Flow", "Sync Solver".
     - Example date_from/date_to: "2020-01-01T00:00:00+00:00"
     - Example value_min/value_max: "10", "100.5"
     - IMPORTANT - Do not guess, autofill, or assume any missing data.
-    - When asked for medical advice, ask the user whether he wants to use DuckDB, ClickHouse or
-    Elasticsearch.
+    - If there are multiple databases available (DuckDB, ClickHouse, Elasticsearch):
+      first, ask the user which one he wants to use. DO NOT call any tools before
+      the user specifies his intent.
+    - If the user decides on an option, only use tools from this database,
+      do not switch over to another until the user specifies that he wants
+      to use a different one. You do not have to keep asking whether
+      the user wants to use the same database that he used before.
+    - If there is only one database available (DuckDB, ClickHouse, Elasticsearch):
+      you can use the tools from this database without the user specifying it.
     """
     try:
         return search_health_records_from_ch(params)
@@ -85,16 +100,23 @@ def get_statistics_by_type_ch(record_type: RecordType | str) -> dict[str, Any]:
       max, avg, sum) for the 'value' field.
     - The sources breakdown shows which devices/apps contributed data for this record type.
     - Example types: "HKQuantityTypeIdentifierStepCount",
-      "HKQuantityTypeIdentifierBodyMassIndex", "HKQuantityTypeIdentifierHeartRate", etc.
+     "HKQuantityTypeIdentifierBodyMassIndex", "HKQuantityTypeIdentifierHeartRate", etc.
     - Use this function to understand the distribution, range, and trends of
       specific health metrics.
     - The function is useful for health analysis, identifying outliers, and
       understanding data quality.
     - date_range key for query is commented, since it contained hardcoded from
       date, but you can use it anyway if you replace startDate with your data.
     - IMPORTANT - Do not guess, autofill, or assume any missing data.
-    - When asked for medical advice, ask the user whether he wants to use DuckDB, ClickHouse or
-    Elasticsearch.
+    - If there are multiple databases available (DuckDB, ClickHouse, Elasticsearch):
+      first, ask the user which one he wants to use. DO NOT call any tools before
+      the user specifies his intent.
+    - If the user decides on an option, only use tools from this database,
+      do not switch over to another until the user specifies that he wants
+      to use a different one. You do not have to keep asking whether
+      the user wants to use the same database that he used before.
+    - If there is only one database available (DuckDB, ClickHouse, Elasticsearch):
+      you can use the tools from this database without the user specifying it.
     """
     try:
         return get_statistics_by_type_from_ch(record_type)
@@ -116,7 +138,7 @@ def get_trend_data_ch(
     Parameters:
     - record_type: The type of health record to analyze (e.g., "HKQuantityTypeIdentifierStepCount")
     - interval: Time interval for aggregation.
-    - date_from, date_to: Optional ISO8601 date strings for filtering date range
+    - date_from, date_to: Optional date strings for filtering date range
 
     Returns:
     - record_type: The analyzed record type
@@ -134,10 +156,55 @@ def get_trend_data_ch(
     - IMPORTANT - interval must be one of: "day", "week", "month", or "year".
       Do not use other values.
     - Do not guess, autofill, or assume any missing data.
-    - When asked for medical advice, ask the user whether he wants to use DuckDB, ClickHouse or
-    Elasticsearch.
+    - If there are multiple databases available (DuckDB, ClickHouse, Elasticsearch):
+      first, ask the user which one he wants to use. DO NOT call any tools before
+      the user specifies his intent.
+    - If the user decides on an option, only use tools from this database,
+      do not switch over to another until the user specifies that he wants
+      to use a different one. You do not have to keep asking whether
+      the user wants to use the same database that he used before.
+    - If there is only one database available (DuckDB, ClickHouse, Elasticsearch):
+      you can use the tools from this database without the user specifying it.
     """
     try:
         return get_trend_data_from_ch(record_type, interval, date_from, date_to)
     except Exception as e:
         return {"error": f"Failed to get trend data: {str(e)}"}
+
+
+@ch_reader_router.tool
+def search_values_ch(
+    record_type: RecordType | str | None,
+    value: str,
+    date_from: str | None = None,
+    date_to: str | None = None,
+) -> dict[str, Any]:
+    """
+    Search for records with exactly matching values (including text) using ClickHouse.
+
+    Parameters:
+    - record_type: The type of health record to analyze (e.g., "HKQuantityTypeIdentifierStepCount")
+    - value: Value to search for in the data
+    - date_from, date_to: Optional date strings for filtering date range
+
+    Notes for LLMs:
+    - Use this to search for specific values (for example statistical outliers) in health data
+    - It can also be used for text values: e.g.
+      you can search for "HKCategoryTypeIdentifierSleepAnalysis"
+      records with the value of "HKCategoryValueSleepAnalysisAsleepDeep"
+    - The function automatically handles date filtering if date_from/date_to are provided
+    - Do not guess, autofill, or assume any missing data.
+    - If there are multiple databases available (DuckDB, ClickHouse, Elasticsearch):
+      first, ask the user which one he wants to use. DO NOT call any tools before
+      the user specifies his intent.
+    - If the user decides on an option, only use tools from this database,
+      do not switch over to another until the user specifies that he wants
+      to use a different one. You do not have to keep asking whether
+      the user wants to use the same database that he used before.
+    - If there is only one database available (DuckDB, ClickHouse, Elasticsearch):
+      you can use the tools from this database without the user specifying it.
+    """
+    try:
+        return search_values_from_ch(record_type, value, date_from, date_to)
+    except Exception as e:
+        return {"error": f"Failed to search for values: {str(e)}"}

app/mcp/v1/tools/duckdb_reader.py

@@ -8,6 +8,7 @@
     get_statistics_by_type_from_duckdb,
     get_trend_data_from_duckdb,
     search_health_records_from_duckdb,
+    search_values_from_duckdb,
 )
 
 duckdb_reader_router = FastMCP(name="CH Reader MCP")
@@ -18,12 +19,19 @@ def get_health_summary_duckdb() -> list[dict[str, Any]]:
     """
     Get a summary of Apple Health data from DuckDB.
     The function returns total record count, record type breakdown, and
-    (optionally) a date range aggregation.
+     (optionally) a date range aggregation.
 
     Notes for LLM:
     - IMPORTANT - Do not guess, autofill, or assume any missing data.
-    - When asked for medical advice, ask the user whether he wants to use DuckDB, ClickHouse or
-    Elasticsearch.
+    - If there are multiple databases available (DuckDB, ClickHouse, Elasticsearch):
+      first, ask the user which one he wants to use. DO NOT call any tools before
+      the user specifies his intent.
+    - If the user decides on an option, only use tools from this database,
+      do not switch over to another until the user specifies that he wants
+      to use a different one. You do not have to keep asking whether
+      the user wants to use the same database that he used before.
+    - If there is only one database available (DuckDB, ClickHouse, Elasticsearch):
+      you can use the tools from this database without the user specifying it.
     """
     try:
         return get_health_summary_from_duckdb()
@@ -50,8 +58,15 @@ def search_health_records_duckdb(params: HealthRecordSearchParams) -> list[dict[
     - Example date_from/date_to: "2020-01-01T00:00:00+00:00"
     - Example value_min/value_max: "10", "100.5"
     - IMPORTANT - Do not guess, autofill, or assume any missing data.
-    - When asked for medical advice, ask the user whether he wants to use DuckDB, ClickHouse or
-    Elasticsearch.
+    - If there are multiple databases available (DuckDB, ClickHouse, Elasticsearch):
+      first, ask the user which one he wants to use. DO NOT call any tools before
+      the user specifies his intent.
+    - If the user decides on an option, only use tools from this database,
+      do not switch over to another until the user specifies that he wants
+      to use a different one. You do not have to keep asking whether
+      the user wants to use the same database that he used before.
+    - If there is only one database available (DuckDB, ClickHouse, Elasticsearch):
+      you can use the tools from this database without the user specifying it.
     """
     try:
         return search_health_records_from_duckdb(params)
@@ -87,14 +102,21 @@ def get_statistics_by_type_duckdb(record_type: RecordType | str) -> list[dict[st
     - Example types: "HKQuantityTypeIdentifierStepCount",
       "HKQuantityTypeIdentifierBodyMassIndex", "HKQuantityTypeIdentifierHeartRate", etc.
     - Use this function to understand the distribution, range, and trends of
-      specific health metrics.
+     specific health metrics.
     - The function is useful for health analysis, identifying outliers, and
       understanding data quality.
     - date_range key for query is commented, since it contained hardcoded from
       date, but you can use it anyway if you replace startDate with your data.
     - IMPORTANT - Do not guess, autofill, or assume any missing data.
-    - When asked for medical advice, ask the user whether he wants to use DuckDB, ClickHouse or
-    Elasticsearch.
+    - If there are multiple databases available (DuckDB, ClickHouse, Elasticsearch):
+      first, ask the user which one he wants to use. DO NOT call any tools before
+      the user specifies his intent.
+    - If the user decides on an option, only use tools from this database,
+      do not switch over to another until the user specifies that he wants
+      to use a different one. You do not have to keep asking whether
+      the user wants to use the same database that he used before.
+    - If there is only one database available (DuckDB, ClickHouse, Elasticsearch):
+      you can use the tools from this database without the user specifying it.
     """
     try:
         return get_statistics_by_type_from_duckdb(record_type)
@@ -111,7 +133,7 @@ def get_trend_data_duckdb(
 ) -> list[dict[str, Any]]:
     """
     Get trend data for a specific health record type over time using DuckDB
-    date histogram aggregation.
+     date histogram aggregation.
 
     Parameters:
     - record_type: The type of health record to analyze (e.g., "HKQuantityTypeIdentifierStepCount")
@@ -134,10 +156,55 @@ def get_trend_data_duckdb(
     - IMPORTANT - interval must be one of: "day", "week", "month", or "year".
       Do not use other values.
     - Do not guess, autofill, or assume any missing data.
-    - When asked for medical advice, ask the user whether he wants to use DuckDB, ClickHouse or
-    Elasticsearch.
+    - If there are multiple databases available (DuckDB, ClickHouse, Elasticsearch):
+      first, ask the user which one he wants to use. DO NOT call any tools before
+      the user specifies his intent.
+    - If the user decides on an option, only use tools from this database,
+      do not switch over to another until the user specifies that he wants
+      to use a different one. You do not have to keep asking whether
+      the user wants to use the same database that he used before.
+    - If there is only one database available (DuckDB, ClickHouse, Elasticsearch):
+      you can use the tools from this database without the user specifying it.
     """
     try:
         return get_trend_data_from_duckdb(record_type, interval, date_from, date_to)
     except Exception as e:
         return [{"error": f"Failed to get trend data: {str(e)}"}]
+
+
+@duckdb_reader_router.tool
+def search_values_duckdb(
+    record_type: RecordType | str | None,
+    value: str,
+    date_from: str | None = None,
+    date_to: str | None = None,
+) -> list[dict[str, Any]]:
+    """
+    Search for records with exactly matching values (including text) using DuckDB.
+
+    Parameters:
+    - record_type: The type of health record to analyze (e.g., "HKQuantityTypeIdentifierStepCount")
+    - value: Value to search for in the data
+    - date_from, date_to: Optional ISO8601 date strings for filtering date range
+
+    Notes for LLMs:
+    - Use this to search for specific values (for example statistical outliers) in health data
+    - It can also be used for text values: e.g.
+      you can search for "HKCategoryTypeIdentifierSleepAnalysis"
+      records with the value of "HKCategoryValueSleepAnalysisAsleepDeep"
+    - The function automatically handles date filtering if date_from/date_to are provided
+    - Do not guess, autofill, or assume any missing data.
+    - If there are multiple databases available (DuckDB, ClickHouse, Elasticsearch):
+      first, ask the user which one he wants to use. DO NOT call any tools before
+      the user specifies his intent.
+    - If the user decides on an option, only use tools from this database,
+      do not switch over to another until the user specifies that he wants
+      to use a different one. You do not have to keep asking whether
+      the user wants to use the same database that he used before.
+    - If there is only one database available (DuckDB, ClickHouse, Elasticsearch):
+      you can use the tools from this database without the user specifying it.
+    """
+    try:
+        return search_values_from_duckdb(record_type, value, date_from, date_to)
+    except Exception as e:
+        return [{"error": f"Failed to get trend data: {str(e)}"}]