diff --git a/crates/bin/ampctl/src/cmd/manifest/generate.rs b/crates/bin/ampctl/src/cmd/manifest/generate.rs
index 6ae90a070..aeae12b9c 100644
--- a/crates/bin/ampctl/src/cmd/manifest/generate.rs
+++ b/crates/bin/ampctl/src/cmd/manifest/generate.rs
@@ -174,6 +174,7 @@ fn generate_evm_rpc_manifest(
             let manifest_table = ManifestTable {
                 schema,
                 network: network.clone(),
+                bloom_filter_columns: Vec::new(),
             };
             (table.name().to_string(), manifest_table)
         })
@@ -204,6 +205,7 @@ fn generate_solana_manifest(
             let manifest_table = ManifestTable {
                 schema,
                 network: network.clone(),
+                bloom_filter_columns: Vec::new(),
             };
             (table.name().to_string(), manifest_table)
         })
@@ -234,6 +236,7 @@ fn generate_firehose_manifest(
             let manifest_table = ManifestTable {
                 schema,
                 network: network.clone(),
+                bloom_filter_columns: Vec::new(),
             };
             (table.name().to_string(), manifest_table)
         })
@@ -264,6 +267,7 @@ fn generate_tempo_manifest(
             let manifest_table = ManifestTable {
                 schema,
                 network: network.clone(),
+                bloom_filter_columns: Vec::new(),
             };
             (table.name().to_string(), manifest_table)
         })
diff --git a/crates/config/src/worker_core.rs b/crates/config/src/worker_core.rs
index d7c030fd6..9566a6a3a 100644
--- a/crates/config/src/worker_core.rs
+++ b/crates/config/src/worker_core.rs
@@ -16,9 +16,6 @@ pub struct ParquetConfig {
     /// Compression algorithm (default: `zstd(1)`).
     #[serde(default)]
     pub compression: Compression,
-    /// Enable Parquet bloom filters (default: false).
-    #[serde(default)]
-    pub bloom_filters: bool,
     /// Parquet metadata cache size in MB (default: 1024).
     #[serde(default = "default_cache_size_mb")]
     pub cache_size_mb: u64,
@@ -48,7 +45,6 @@ impl Default for ParquetConfig {
     fn default() -> Self {
         Self {
             compression: Compression::default(),
-            bloom_filters: false,
             cache_size_mb: default_cache_size_mb(),
             max_row_group_mb: default_max_row_group_mb(),
             target_size: SizeLimitConfig::default(),
@@ -71,7 +67,6 @@ impl From<&ParquetConfig> for amp_worker_core::ParquetConfig {
     fn from(config: &ParquetConfig) -> Self {
         Self {
             compression: (&config.compression).into(),
-            bloom_filters: config.bloom_filters,
             cache_size_mb: config.cache_size_mb,
             max_row_group_mb: config.max_row_group_mb,
             target_size: (&config.target_size).into(),
diff --git a/crates/core/datasets-common/src/dataset.rs b/crates/core/datasets-common/src/dataset.rs
index 9a74c9353..99a278ee0 100644
--- a/crates/core/datasets-common/src/dataset.rs
+++ b/crates/core/datasets-common/src/dataset.rs
@@ -18,7 +18,7 @@ use downcast_rs::{DowncastSync, impl_downcast};
 
 use crate::{
     block_num::BlockNum, dataset_kind_str::DatasetKindStr, hash_reference::HashReference,
-    network_id::NetworkId, table_name::TableName,
+    manifest::BloomFilterColumnConfig, network_id::NetworkId, table_name::TableName,
 };
 
 /// Core trait representing a dataset definition.
@@ -95,6 +95,13 @@ pub trait Table: DowncastSync + std::fmt::Debug {
 
     /// Returns column names by which this table is naturally sorted. Always includes `_block_num`.
     fn sorted_by(&self) -> &BTreeSet<String>;
+
+    /// Returns the bloom filter column configurations for this table.
+    ///
+    /// Empty by default — tables without bloom filter config get no bloom filters.
+    fn bloom_filter_columns(&self) -> &[BloomFilterColumnConfig] {
+        &[]
+    }
 }
 
 // Implement downcasting for `Table`.
diff --git a/crates/core/datasets-common/src/manifest.rs b/crates/core/datasets-common/src/manifest.rs
index 81c6e11c0..5f2019e21 100644
--- a/crates/core/datasets-common/src/manifest.rs
+++ b/crates/core/datasets-common/src/manifest.rs
@@ -61,6 +61,25 @@ impl Default for Schema {
     }
 }
 
+/// Configuration for a bloom filter on a specific column.
+///
+/// Bloom filters enable efficient row-group pruning during queries by quickly
+/// determining whether a row group could contain a specific value. The `ndv`
+/// (number of distinct values) parameter tunes the filter's false-positive rate.
+#[derive(Debug, Clone, serde::Serialize, serde::Deserialize)]
+#[cfg_attr(feature = "schemars", derive(schemars::JsonSchema))]
+pub struct BloomFilterColumnConfig {
+    /// Column name to apply the bloom filter to.
+    pub column: String,
+    /// Estimated number of distinct values for this column (default: 10,000).
+    #[serde(default = "default_bloom_filter_ndv")]
+    pub ndv: u64,
+}
+
+fn default_bloom_filter_ndv() -> u64 {
+    10_000
+}
+
 /// Apache Arrow data _new-type_ wrapper with JSON schema support.
 ///
 /// This wrapper provides serialization and JSON schema generation capabilities
diff --git a/crates/core/datasets-raw/src/dataset.rs b/crates/core/datasets-raw/src/dataset.rs
index 6e63d59f5..ad5fad750 100644
--- a/crates/core/datasets-raw/src/dataset.rs
+++ b/crates/core/datasets-raw/src/dataset.rs
@@ -1,6 +1,9 @@
 //! Concrete raw dataset and table types.
 
-use std::{collections::BTreeSet, sync::Arc};
+use std::{
+    collections::{BTreeMap, BTreeSet},
+    sync::Arc,
+};
 
 use arrow::datatypes::SchemaRef;
 use datasets_common::{
@@ -8,6 +11,7 @@ use datasets_common::{
     dataset::Table as TableTrait,
     dataset_kind_str::DatasetKindStr,
     hash_reference::HashReference,
+    manifest::BloomFilterColumnConfig,
     network_id::NetworkId,
     table_name::TableName,
 };
@@ -37,7 +41,8 @@ impl Dataset {
         reference: HashReference,
         kind: DatasetKindStr,
         network: NetworkId,
-        tables: Vec<Table>,
+        mut tables: Vec<Table>,
+        manifest_tables: &BTreeMap<String, crate::manifest::Table>,
         start_block: Option<BlockNum>,
         finalized_blocks_only: bool,
     ) -> Self {
@@ -50,6 +55,13 @@ impl Dataset {
             "all tables must belong to the same network as the dataset"
         );
 
+        // Apply per-table bloom filter config from the manifest.
+        for table in &mut tables {
+            if let Some(mt) = manifest_tables.get(table.name().as_str()) {
+                table.set_bloom_filter_columns(mt.bloom_filter_columns.clone());
+            }
+        }
+
         let tables: Vec<Arc<dyn TableTrait>> = tables
             .into_iter()
             .map(|t| Arc::new(t) as Arc<dyn TableTrait>)
@@ -111,6 +123,7 @@ pub struct Table {
     schema: SchemaRef,
     network: NetworkId,
     sorted_by: BTreeSet<String>,
+    bloom_filter_columns: Vec<BloomFilterColumnConfig>,
 }
 
 impl Table {
@@ -128,6 +141,7 @@ impl Table {
             schema,
             network,
             sorted_by,
+            bloom_filter_columns: Vec::new(),
         }
     }
 
@@ -135,6 +149,11 @@ impl Table {
     pub fn network_ref(&self) -> &NetworkId {
         &self.network
     }
+
+    /// Sets the bloom filter column configurations for this table.
+    pub fn set_bloom_filter_columns(&mut self, columns: Vec<BloomFilterColumnConfig>) {
+        self.bloom_filter_columns = columns;
+    }
 }
 
 impl TableTrait for Table {
@@ -153,4 +172,8 @@ impl TableTrait for Table {
     fn sorted_by(&self) -> &BTreeSet<String> {
         &self.sorted_by
     }
+
+    fn bloom_filter_columns(&self) -> &[BloomFilterColumnConfig] {
+        &self.bloom_filter_columns
+    }
 }
diff --git a/crates/core/datasets-raw/src/manifest.rs b/crates/core/datasets-raw/src/manifest.rs
index 93de50aea..c741fbf05 100644
--- a/crates/core/datasets-raw/src/manifest.rs
+++ b/crates/core/datasets-raw/src/manifest.rs
@@ -2,7 +2,11 @@
 
 use std::collections::BTreeMap;
 
-use datasets_common::{block_num::BlockNum, manifest::TableSchema, network_id::NetworkId};
+use datasets_common::{
+    block_num::BlockNum,
+    manifest::{BloomFilterColumnConfig, TableSchema},
+    network_id::NetworkId,
+};
 
 use crate::dataset_kind::{
     EvmRpcDatasetKind, FirehoseDatasetKind, SolanaDatasetKind, TempoDatasetKind,
@@ -57,6 +61,10 @@ pub struct Table {
     pub schema: TableSchema,
     /// Network for this table.
     pub network: NetworkId,
+    /// Columns to apply bloom filters to for row-group pruning.
+    /// Omit to disable bloom filters for this table.
+    #[serde(default)]
+    pub bloom_filter_columns: Vec<BloomFilterColumnConfig>,
 }
 
 /// Schema generation types, gated behind the `schemars` feature.
diff --git a/crates/core/worker-core/src/compaction/config.rs b/crates/core/worker-core/src/compaction/config.rs
index 72bcd2929..87a8996ff 100644
--- a/crates/core/worker-core/src/compaction/config.rs
+++ b/crates/core/worker-core/src/compaction/config.rs
@@ -8,8 +8,6 @@ use super::algorithm::Overflow;
 pub struct ParquetConfig {
     /// Compression algorithm: zstd, lz4, gzip, brotli, snappy, uncompressed (default: zstd(1))
     pub compression: Compression,
-    /// Enable bloom filters (default: false)
-    pub bloom_filters: bool,
     /// Parquet metadata cache size in MB (default: 1024)
     pub cache_size_mb: u64,
     /// Max row group size in MB (default: 512)
@@ -26,7 +24,6 @@ impl Default for ParquetConfig {
     fn default() -> Self {
         Self {
             compression: Compression::ZSTD(ZstdLevel::default()),
-            bloom_filters: false,
             cache_size_mb: 1024,   // 1 GB
             max_row_group_mb: 512, // 512 MB
             target_size: SizeLimitConfig::default(),
diff --git a/crates/core/worker-core/src/lib.rs b/crates/core/worker-core/src/lib.rs
index dbe9b6b45..7b790b866 100644
--- a/crates/core/worker-core/src/lib.rs
+++ b/crates/core/worker-core/src/lib.rs
@@ -3,6 +3,7 @@
 use std::{sync::Arc, time::Duration};
 
 use datafusion::parquet::file::properties::WriterProperties as ParquetWriterProperties;
+use datasets_common::manifest::BloomFilterColumnConfig;
 
 pub mod block_ranges;
 pub mod check;
@@ -51,11 +52,6 @@ pub struct WriterProperties {
 
 pub fn parquet_opts(config: impl Into<ParquetConfig>) -> Arc<WriterProperties> {
     let config = config.into();
-    // We have not done our own benchmarking, but the default 1_000_000 value for this adds about a
-    // megabyte of storage per column, per row group. This analysis by InfluxData suggests that
-    // smaller NDV values may be equally effective:
-    // https://www.influxdata.com/blog/using-parquets-bloom-filters/
-    let bloom_filter_ndv = 10_000;
 
     // For DataFusion defaults, see `ParquetOptions` here:
     // https://github.com/apache/arrow-datafusion/blob/main/datafusion/common/src/config.rs
@@ -63,11 +59,7 @@ pub fn parquet_opts(config: impl Into<ParquetConfig>) -> Arc<WriterProperties> {
     // Note: We could set `sorting_columns` for columns like `block_num` and `ordinal`. However,
     // Datafusion doesn't actually read that metadata info anywhere and just reiles on the
     // `file_sort_order` set on the reader configuration.
-    let parquet = ParquetWriterProperties::builder()
-        .set_compression(config.compression)
-        .set_bloom_filter_ndv(bloom_filter_ndv)
-        .set_bloom_filter_enabled(config.bloom_filters)
-        .build();
+    let parquet = build_parquet_writer_properties(config.compression, &[]);
 
     let collector = CollectorProperties::from(&config);
     let compactor = CompactorProperties::from(&config);
@@ -88,3 +80,28 @@ pub fn parquet_opts(config: impl Into<ParquetConfig>) -> Arc<WriterProperties> {
     }
     .into()
 }
+
+/// Builds `ParquetWriterProperties` with per-column bloom filter configuration.
+///
+/// When `bloom_filter_columns` is non-empty, bloom filters are enabled only on the
+/// specified columns with their configured NDV values. When empty, no bloom filters
+/// are written.
+pub fn build_parquet_writer_properties(
+    compression: datafusion::parquet::basic::Compression,
+    bloom_filter_columns: &[BloomFilterColumnConfig],
+) -> ParquetWriterProperties {
+    let mut builder = ParquetWriterProperties::builder().set_compression(compression);
+
+    if bloom_filter_columns.is_empty() {
+        builder.build()
+    } else {
+        // Disable bloom filters globally, then enable per-column
+        builder = builder.set_bloom_filter_enabled(false);
+        for col in bloom_filter_columns {
+            builder = builder
+                .set_column_bloom_filter_enabled(col.column.clone().into(), true)
+                .set_column_bloom_filter_ndv(col.column.clone().into(), col.ndv);
+        }
+        builder.build()
+    }
+}
diff --git a/crates/core/worker-datasets-raw/src/job_impl.rs b/crates/core/worker-datasets-raw/src/job_impl.rs
index 1edc0a3a8..a06ee6dfc 100644
--- a/crates/core/worker-datasets-raw/src/job_impl.rs
+++ b/crates/core/worker-datasets-raw/src/job_impl.rs
@@ -83,7 +83,7 @@ use std::{collections::BTreeMap, ops::RangeInclusive, sync::Arc, time::Instant};
 use amp_data_store::retryable::RetryableErrorExt as _;
 use amp_providers_registry::retryable::RetryableErrorExt as _;
 use amp_worker_core::{
-    ResolvedEndBlock,
+    ResolvedEndBlock, WriterProperties,
     block_ranges::resolve_end_block,
     check::consistency_check,
     compaction::AmpCompactor,
@@ -142,11 +142,28 @@ pub async fn execute(
     let dataset_reference = dataset.reference();
 
     let materialize_start_time = Instant::now();
-    let parquet_opts = amp_worker_core::parquet_opts(ctx.config.parquet_writer.clone());
+    let global_parquet_opts = amp_worker_core::parquet_opts(ctx.config.parquet_writer.clone());
 
     // Initialize physical tables and compactors
     let mut tables: Vec<(Arc<PhysicalTable>, Arc<AmpCompactor>)> = vec![];
+    let mut per_table_opts: BTreeMap<TableName, Arc<WriterProperties>> = BTreeMap::new();
     for table_def in dataset.tables() {
+        // Build per-table WriterProperties with table-specific bloom filter columns
+        let bloom_filter_columns = table_def.bloom_filter_columns();
+        let table_opts = if bloom_filter_columns.is_empty() {
+            global_parquet_opts.clone()
+        } else {
+            let table_parquet = amp_worker_core::build_parquet_writer_properties(
+                ctx.config.parquet_writer.compression,
+                bloom_filter_columns,
+            );
+            Arc::new(WriterProperties {
+                parquet: table_parquet,
+                ..(*global_parquet_opts).clone()
+            })
+        };
+        per_table_opts.insert(table_def.name().clone(), table_opts.clone());
+
         // Try to get existing active physical table (handles retry case)
         let revision = match ctx
             .data_store
@@ -176,7 +193,7 @@ pub async fn execute(
         let compactor = AmpCompactor::start(
             ctx.metadata_db.clone(),
             ctx.data_store.clone(),
-            parquet_opts.clone(),
+            table_opts,
             physical_table.clone(),
             ctx.metrics.clone(),
         )
@@ -367,7 +384,7 @@ pub async fn execute(
                 &client,
                 &ctx,
                 &catalog,
-                parquet_opts.clone(),
+                per_table_opts.clone(),
                 missing_ranges_by_table,
                 compactors_by_table,
                 &tables,
diff --git a/crates/core/worker-datasets-raw/src/job_impl/ranges.rs b/crates/core/worker-datasets-raw/src/job_impl/ranges.rs
index d643135ec..14cde47fa 100644
--- a/crates/core/worker-datasets-raw/src/job_impl/ranges.rs
+++ b/crates/core/worker-datasets-raw/src/job_impl/ranges.rs
@@ -51,7 +51,7 @@ pub(super) async fn materialize_ranges<S: BlockStreamer + Send + Sync>(
     client: &S,
     ctx: &Context,
     catalog: &Catalog,
-    parquet_opts: Arc<WriterProperties>,
+    per_table_opts: BTreeMap<TableName, Arc<WriterProperties>>,
     missing_ranges_by_table: BTreeMap<TableName, Vec<RangeInclusive<BlockNum>>>,
     compactors_by_table: BTreeMap<TableName, Arc<AmpCompactor>>,
     tables: &[(Arc<PhysicalTable>, Arc<AmpCompactor>)],
@@ -113,7 +113,7 @@ pub(super) async fn materialize_ranges<S: BlockStreamer + Send + Sync>(
             data_store: ctx.data_store.clone(),
             catalog: catalog.clone(),
             ranges,
-            parquet_opts: parquet_opts.clone(),
+            per_table_opts: per_table_opts.clone(),
             missing_ranges_by_table: missing_ranges_by_table.clone(),
             compactors_by_table: compactors_by_table.clone(),
             id: i as u32,
@@ -418,8 +418,8 @@ struct MaterializePartition<S: BlockStreamer> {
     catalog: Catalog,
     /// The block ranges to scan
     ranges: Vec<RangeInclusive<BlockNum>>,
-    /// The Parquet writer properties
-    parquet_opts: Arc<WriterProperties>,
+    /// Per-table Parquet writer properties
+    per_table_opts: BTreeMap<TableName, Arc<WriterProperties>>,
     /// The missing block ranges by table
     missing_ranges_by_table: BTreeMap<TableName, Vec<RangeInclusive<BlockNum>>>,
     /// The compactors for each table
@@ -505,7 +505,7 @@ impl<S: BlockStreamer> MaterializePartition<S> {
             self.catalog.clone(),
             self.metadata_db.clone(),
             self.data_store.clone(),
-            self.parquet_opts.clone(),
+            self.per_table_opts.clone(),
             missing_ranges_by_table,
             self.compactors_by_table.clone(),
             self.metrics.clone(),
diff --git a/crates/core/worker-datasets-raw/src/job_impl/writer.rs b/crates/core/worker-datasets-raw/src/job_impl/writer.rs
index 286b31206..56ad5cffd 100644
--- a/crates/core/worker-datasets-raw/src/job_impl/writer.rs
+++ b/crates/core/worker-datasets-raw/src/job_impl/writer.rs
@@ -34,22 +34,23 @@ impl RawDatasetWriter {
         catalog: Catalog,
         metadata_db: MetadataDb,
         store: DataStore,
-        opts: Arc<WriterProperties>,
+        per_table_opts: BTreeMap<TableName, Arc<WriterProperties>>,
         missing_ranges_by_table: BTreeMap<TableName, Vec<RangeInclusive<BlockNum>>>,
         compactors_by_table: BTreeMap<TableName, Arc<AmpCompactor>>,
         metrics: Option<Arc<metrics::MetricsRegistry>>,
     ) -> Result<Self, ParquetError> {
         let mut writers = BTreeMap::new();
         for table in catalog.physical_tables() {
-            // Unwrap: `missing_ranges_by_table` contains an entry for each table.
+            // Unwrap: `missing_ranges_by_table` and `per_table_opts` contain an entry for each table.
             let table_name = table.table_name();
             let ranges = missing_ranges_by_table.get(table_name).unwrap().clone();
             let compactor = Arc::clone(compactors_by_table.get(table_name).unwrap());
+            let opts = Arc::clone(per_table_opts.get(table_name).unwrap());
             let writer = RawTableWriter::new(
                 table.clone(),
                 store.clone(),
                 compactor,
-                opts.clone(),
+                opts,
                 ranges,
                 metrics.clone(),
             )?;
diff --git a/crates/extractors/evm-rpc/src/lib.rs b/crates/extractors/evm-rpc/src/lib.rs
index 8a54f8bf4..1ab795c3e 100644
--- a/crates/extractors/evm-rpc/src/lib.rs
+++ b/crates/extractors/evm-rpc/src/lib.rs
@@ -21,6 +21,7 @@ pub fn dataset(reference: HashReference, manifest: Manifest) -> RawDataset {
         manifest.kind.into(),
         network.clone(),
         tables::all(&network),
+        &manifest.tables,
         Some(manifest.start_block),
         manifest.finalized_blocks_only,
     )
diff --git a/crates/extractors/firehose/src/lib.rs b/crates/extractors/firehose/src/lib.rs
index b0fa2bff8..e11fb825f 100644
--- a/crates/extractors/firehose/src/lib.rs
+++ b/crates/extractors/firehose/src/lib.rs
@@ -25,6 +25,7 @@ pub fn dataset(reference: HashReference, manifest: Manifest) -> RawDataset {
         manifest.kind.into(),
         network.clone(),
         tables::all(&network),
+        &manifest.tables,
         Some(manifest.start_block),
         manifest.finalized_blocks_only,
     )
diff --git a/crates/extractors/solana/src/lib.rs b/crates/extractors/solana/src/lib.rs
index 01bdf4fc9..4f7ae542e 100644
--- a/crates/extractors/solana/src/lib.rs
+++ b/crates/extractors/solana/src/lib.rs
@@ -21,6 +21,7 @@ pub fn dataset(reference: HashReference, manifest: Manifest) -> RawDataset {
         manifest.kind.into(),
         network.clone(),
         tables::all(&network),
+        &manifest.tables,
         Some(manifest.start_block),
         manifest.finalized_blocks_only,
     )
diff --git a/crates/extractors/tempo/src/lib.rs b/crates/extractors/tempo/src/lib.rs
index 337d37684..970aad2af 100644
--- a/crates/extractors/tempo/src/lib.rs
+++ b/crates/extractors/tempo/src/lib.rs
@@ -18,6 +18,7 @@ pub fn dataset(reference: HashReference, manifest: Manifest) -> RawDataset {
         manifest.kind.into(),
         network.clone(),
         tables::all(&network),
+        &manifest.tables,
         Some(manifest.start_block),
         manifest.finalized_blocks_only,
     )
diff --git a/docs/config.sample.toml b/docs/config.sample.toml
index b403214bb..6990b3013 100644
--- a/docs/config.sample.toml
+++ b/docs/config.sample.toml
@@ -70,7 +70,6 @@ url = "postgres://<pg_url>"
 # Writer/Parquet configuration
 [writer]
 # compression = "zstd(1)"    # Compression algorithm: zstd, lz4, gzip, brotli, snappy, uncompressed (default: zstd(1))
-# bloom_filters = false      # Enable bloom filters (default: false)
 # cache_size_mb = 1024       # Parquet metadata cache size in MB (default: 1024)
 # max_row_group_mb = 512     # Max row group size in MB (default: 512)
 # segment_flush_interval_secs = 600.0  # Max wall-clock seconds before closing a segment (default: 600 = 10 min)
diff --git a/docs/schemas/config/ampd.spec.json b/docs/schemas/config/ampd.spec.json
index 528a468c5..a29d04062 100644
--- a/docs/schemas/config/ampd.spec.json
+++ b/docs/schemas/config/ampd.spec.json
@@ -318,11 +318,6 @@
       "description": "Parquet writer and file configuration.\n\nControls compression, caching, segment sizing, compaction, and garbage collection\nfor Parquet files produced by the worker.",
       "type": "object",
       "properties": {
-        "bloom_filters": {
-          "description": "Enable Parquet bloom filters (default: false).",
-          "type": "boolean",
-          "default": false
-        },
         "bytes": {
           "description": "Target bytes per file (default: 2147483648 = 2 GB for target_size, 0 for eager limits).",
           "type": "integer",
diff --git a/docs/schemas/manifest/raw.spec.json b/docs/schemas/manifest/raw.spec.json
index 59b6b0017..8ed07954f 100644
--- a/docs/schemas/manifest/raw.spec.json
+++ b/docs/schemas/manifest/raw.spec.json
@@ -60,6 +60,26 @@
         "fields"
       ]
     },
+    "BloomFilterColumnConfig": {
+      "description": "Configuration for a bloom filter on a specific column.\n\nBloom filters enable efficient row-group pruning during queries by quickly\ndetermining whether a row group could contain a specific value. The `ndv`\n(number of distinct values) parameter tunes the filter's false-positive rate.",
+      "type": "object",
+      "properties": {
+        "column": {
+          "description": "Column name to apply the bloom filter to.",
+          "type": "string"
+        },
+        "ndv": {
+          "description": "Estimated number of distinct values for this column (default: 10,000).",
+          "type": "integer",
+          "format": "uint64",
+          "default": 10000,
+          "minimum": 0
+        }
+      },
+      "required": [
+        "column"
+      ]
+    },
     "DataType": {
       "description": "Arrow data type, e.g. `Int32`, `Utf8`, etc.",
       "type": "string"
@@ -106,6 +126,14 @@
       "description": "Table definition for raw dataset manifests.",
       "type": "object",
       "properties": {
+        "bloom_filter_columns": {
+          "description": "Columns to apply bloom filters to for row-group pruning.\nOmit to disable bloom filters for this table.",
+          "type": "array",
+          "default": [],
+          "items": {
+            "$ref": "#/$defs/BloomFilterColumnConfig"
+          }
+        },
         "network": {
           "description": "Network for this table.",
           "$ref": "#/$defs/NetworkId"
diff --git a/tests/config/manifests/eth_rpc.json b/tests/config/manifests/eth_rpc.json
index fa0782b22..0048ab80d 100644
--- a/tests/config/manifests/eth_rpc.json
+++ b/tests/config/manifests/eth_rpc.json
@@ -173,7 +173,8 @@
           ]
         }
       },
-      "network": "mainnet"
+      "network": "mainnet",
+      "bloom_filter_columns": []
     },
     "logs": {
       "schema": {
@@ -266,7 +267,8 @@
           ]
         }
       },
-      "network": "mainnet"
+      "network": "mainnet",
+      "bloom_filter_columns": []
     },
     "transactions": {
       "schema": {
@@ -567,7 +569,8 @@
           ]
         }
       },
-      "network": "mainnet"
+      "network": "mainnet",
+      "bloom_filter_columns": []
     }
   }
 }
diff --git a/tests/config/manifests/solana.json b/tests/config/manifests/solana.json
index 73830cfe7..4b2b0c817 100644
--- a/tests/config/manifests/solana.json
+++ b/tests/config/manifests/solana.json
@@ -46,7 +46,8 @@
           ]
         }
       },
-      "network": "mainnet"
+      "network": "mainnet",
+      "bloom_filter_columns": []
     },
     "block_rewards": {
       "schema": {
@@ -90,7 +91,8 @@
           ]
         }
       },
-      "network": "mainnet"
+      "network": "mainnet",
+      "bloom_filter_columns": []
     },
     "instructions": {
       "schema": {
@@ -157,7 +159,8 @@
           ]
         }
       },
-      "network": "mainnet"
+      "network": "mainnet",
+      "bloom_filter_columns": []
     },
     "messages": {
       "schema": {
@@ -291,7 +294,8 @@
           ]
         }
       },
-      "network": "mainnet"
+      "network": "mainnet",
+      "bloom_filter_columns": []
     },
     "transactions": {
       "schema": {
@@ -668,7 +672,8 @@
           ]
         }
       },
-      "network": "mainnet"
+      "network": "mainnet",
+      "bloom_filter_columns": []
     }
   }
-}
\ No newline at end of file
+}
diff --git a/tests/config/manifests/tempo.json b/tests/config/manifests/tempo.json
index 13c764d95..901b9bf30 100644
--- a/tests/config/manifests/tempo.json
+++ b/tests/config/manifests/tempo.json
@@ -188,7 +188,8 @@
           ]
         }
       },
-      "network": "testnet"
+      "network": "testnet",
+      "bloom_filter_columns": []
     },
     "logs": {
       "schema": {
@@ -281,7 +282,8 @@
           ]
         }
       },
-      "network": "testnet"
+      "network": "testnet",
+      "bloom_filter_columns": []
     },
     "transactions": {
       "schema": {
@@ -992,7 +994,8 @@
           ]
         }
       },
-      "network": "testnet"
+      "network": "testnet",
+      "bloom_filter_columns": []
     }
   }
-}
\ No newline at end of file
+}