[v1.38] Release v1.38 documentation to main

REMOVED_CONTENT_CLEANUP.md

@@ -0,0 +1,27 @@
+# Removed content — cleanup tracking
+
+This file tracks documentation content that describes **removed** Weaviate features,
+environment variables, or configuration fields. Rather than deleting such content the
+moment a feature is removed, we keep it in place with a short "Removed in `vX.Y`" note so
+that users upgrading from an older version can still find the entry and understand what
+happened to it. Once enough releases have passed that few users are upgrading across the
+removal boundary, the noted content should be deleted.
+
+## Policy
+
+- When a feature/env var/config field is removed from Weaviate, **mark** the corresponding
+  docs entry as `Removed in vX.Y` instead of deleting it immediately.
+- Add a row to the table below so the kept-but-stale content can be found and cleaned later.
+- **Suggested cleanup window:** keep the note for roughly three minor releases after the
+  removal, then delete the entry and remove its row here. (Adjust per the supported-version
+  policy at cleanup time — these are guidelines, not hard commitments.)
+
+## Tracked entries
+
+| Page / file | Removed item | Removed in | Suggested cleanup | Notes |
+| --- | --- | --- | --- | --- |
+| `docs/deploy/configuration/env-vars/index.md` | `ASYNC_REPLICATION_CLUSTER_MAX_WORKERS` (table row) | `v1.38` | `v1.41`+ | Replaced by `ASYNC_REPLICATION_SCHEDULER_WORKERS`. |
+| `docs/deploy/configuration/env-vars/index.md` | `ASYNC_REPLICATION_ALIVE_NODES_CHECKING_FREQUENCY` (table row) | `v1.38` | `v1.41`+ | Scheduler no longer polls alive nodes separately. |
+| `docs/deploy/configuration/env-vars/runtime-config.md` | `async_replication_cluster_max_workers` (override mapping row) | `v1.38` | `v1.41`+ | Runtime override removed alongside the env var. |
+| `docs/deploy/configuration/async-rep.md` | "Removed environment variables (v1.38)" `<details>` block (both vars above) | `v1.38` | `v1.41`+ | Delete the whole block at cleanup. |
+| `docs/deploy/configuration/replication.md` | "Removed in `v1.38`" `:::note` admonition | `v1.38` | `v1.41`+ | Mentions both removed env vars. |

_includes/async-replication-per-collection-config.mdx

@@ -1,3 +1,3 @@
 :::info Collection-level configuration — Added in `v1.36`
-Async replication parameters can also be set per-collection via the `asyncConfig` object in `replicationConfig`. Per-collection settings override the cluster-wide environment variable defaults. See [Collection `asyncConfig` parameters](/weaviate/config-refs/collections#async-config) for details.
+Async replication runs by default for any collection with a replication factor greater than `1` (as of `v1.38`). To fine-tune its behavior for a specific collection, set the `asyncConfig` object in `replicationConfig`. Cluster-wide environment variable settings override per-collection settings. See [Collection `asyncConfig` parameters](/weaviate/config-refs/collections#async-config) for details.
 :::

_includes/code/config-refs/reference.collections.py

@@ -57,7 +57,6 @@
     ),
     replication_config=Configure.replication(
         factor=1,
-        async_enabled=False,
         deletion_strategy=ReplicationDeletionStrategy.TIME_BASED_RESOLUTION,
     ),
 )
@@ -733,7 +732,6 @@
     # highlight-start
     replication_config=Configure.replication(
         factor=3,
-        async_enabled=True,
         deletion_strategy=ReplicationDeletionStrategy.TIME_BASED_RESOLUTION,
     ),
     # highlight-end
@@ -743,7 +741,6 @@
 # Test
 collection = client.collections.use("Article")
 config = collection.config.get()
-assert config.replication_config.async_enabled == True
 assert (
     config.replication_config.deletion_strategy
     == ReplicationDeletionStrategy.TIME_BASED_RESOLUTION

_includes/code/configuration/replication-consistency.mdx

@@ -43,8 +43,7 @@ curl \
         }
     ],
     "replicationConfig": {
-      "factor": 3,
-      "asyncEnabled": true
+      "factor": 3
     }
 }' \
 http://localhost:8080/v1/schema

_includes/code/howto/go/docs/manage-data.classes_test.go

@@ -585,8 +585,8 @@ func Test_ManageDataClasses(t *testing.T) {
 		articleClass := &models.Class{
 			Class:       "Article",
 			Description: "Collection of articles",
+			// Async replication runs by default when the replication factor is greater than 1
 			ReplicationConfig: &models.ReplicationConfig{
-				AsyncEnabled:     true,
 				Factor:           3,
 				DeletionStrategy: models.ReplicationConfigDeletionStrategyTimeBasedResolution,
 			},

_includes/code/howto/manage-data.collections.py

@@ -786,9 +786,9 @@
 client.collections.create(
     "Article",
     # highlight-start
+    # Async replication runs by default when the replication factor is greater than 1
     replication_config=Configure.replication(
         factor=3,
-        async_enabled=True,
     ),
     # highlight-end
 )
@@ -819,10 +819,8 @@
     # highlight-start
     replication_config=Configure.replication(
         factor=3,
-        async_enabled=True,
         deletion_strategy=ReplicationDeletionStrategy.TIME_BASED_RESOLUTION,
         async_config=Configure.Replication.async_config(
-            max_workers=5,
             hashtree_height=16,
             frequency=30,
         ),
@@ -840,7 +838,6 @@
 collection.config.update(
     replication_config=Reconfigure.replication(
         async_config=Reconfigure.Replication.async_config(
-            max_workers=10,
             frequency=60,
         ),
     ),
@@ -851,7 +848,6 @@
 # Test
 collection = client.collections.use("Article")
 config = collection.config.get()
-assert config.replication_config.async_enabled == True
 assert (
     config.replication_config.deletion_strategy
     == ReplicationDeletionStrategy.TIME_BASED_RESOLUTION

_includes/code/howto/manage-data.collections.ts

@@ -677,9 +677,9 @@ import { configure } from 'weaviate-client';
 await client.collections.create({
   name: 'Article',
   // highlight-start
+  // Async replication runs by default when the replication factor is greater than 1
   replication: configure.replication({
     factor: 1,
-    asyncEnabled: true,
   }),
   // highlight-end
 })
@@ -710,10 +710,8 @@ await replicationClient.collections.create({
   // highlight-start
   replication: configure.replication({
     factor: 3,
-    asyncEnabled: true,
     deletionStrategy: 'TimeBasedResolution',
     asyncConfig: {
-      maxWorkers: 5,
       hashtreeHeight: 16,
       frequency: 30,
     },
@@ -729,7 +727,6 @@ const articleReplication = replicationClient.collections.use('Article')
 await articleReplication.config.update({
   replication: reconfigure.replication({
     asyncConfig: {
-      maxWorkers: 10,
       frequency: 60,
     },
   }),

_includes/code/howto/search.boost.py

@@ -0,0 +1,240 @@
+# How-to: Search > Boost results — Python examples.
+#
+# Requires Weaviate v1.38+ and the Python client release that adds Boost
+# support (PR weaviate/weaviate-python-client#2030). Boost is gRPC-only —
+# REST/curl is not supported.
+#
+# Uses the text2vec-transformers vectorizer. Run against the local stack
+# in tests/docker-compose-anon.yml (Weaviate + transformers inference).
+
+import time
+from datetime import datetime, timedelta, timezone
+
+import weaviate
+from weaviate.classes.config import Configure, DataType, Property, Tokenization
+from weaviate.classes.query import Boost, Filter
+
+client = weaviate.connect_to_local()
+
+# ---- Fixture: an Articles collection with date + numeric properties ----
+client.collections.delete("Articles")
+client.collections.create(
+    name="Articles",
+    vector_config=Configure.Vectors.text2vec_transformers(),
+    properties=[
+        Property(name="title", data_type=DataType.TEXT),
+        Property(name="category", data_type=DataType.TEXT, tokenization=Tokenization.FIELD),
+        Property(name="published", data_type=DataType.DATE),
+        Property(name="likes", data_type=DataType.INT),
+        Property(name="price", data_type=DataType.NUMBER),
+        Property(name="draft", data_type=DataType.BOOL),
+    ],
+)
+
+now = datetime.now(timezone.utc)
+articles = client.collections.use("Articles")
+articles.data.insert_many([
+    {"title": "Transformers explained",        "category": "research", "published": now - timedelta(days=2),   "likes": 100,     "price": 49.99, "draft": False},
+    {"title": "Old transformer survey",        "category": "research", "published": now - timedelta(days=400), "likes": 5000,    "price": 49.99, "draft": False},
+    {"title": "How to fine-tune a model",      "category": "tutorial", "published": now - timedelta(days=1),   "likes": 30,      "price": 9.99,  "draft": False},
+    {"title": "Pricing transformers",          "category": "tutorial", "published": now - timedelta(days=10),  "likes": 5000000, "price": 199.0, "draft": False},
+    {"title": "Draft: transformer architecture","category": "research", "published": now - timedelta(days=3),   "likes": 200,     "price": 9.99,  "draft": True},
+])
+
+# Wait briefly for the vectorizer to finish indexing the new objects.
+time.sleep(3)
+
+
+# ==========================================
+# ===== Filter boost (soft WHERE) =====
+# ==========================================
+
+# START BoostFilter
+# Promote articles in the "research" category without filtering others out.
+response = articles.query.near_text(
+    query="transformer architectures",
+    limit=5,
+    # highlight-start
+    boost=Boost.filter(
+        Filter.by_property("category").equal("research"),
+        weight=0.5,
+    ),
+    # highlight-end
+    return_properties=["title", "category"],
+)
+
+for o in response.objects:
+    print(o.properties["category"], "-", o.properties["title"])
+# END BoostFilter
+
+assert response.objects[0].properties["category"] == "research"
+
+
+# ==========================================
+# ===== Property boost (numeric value) =====
+# ==========================================
+
+# START BoostProperty
+# Bias toward articles with more `likes`. LOG1P dampens the long tail so a
+# single 5-million-likes outlier doesn't dominate.
+response = articles.query.near_text(
+    query="transformer architectures",
+    limit=5,
+    # highlight-start
+    boost=Boost.property(
+        "likes",
+        modifier=Boost.Modifier.LOG1P,
+        weight=0.7,
+    ),
+    # highlight-end
+    return_properties=["title", "likes"],
+)
+
+for o in response.objects:
+    print(o.properties["likes"], "-", o.properties["title"])
+# END BoostProperty
+
+
+# ==========================================
+# ===== Time decay (boost recent docs) =====
+# ==========================================
+
+# START BoostTimeDecay
+# Score decays exponentially over time. "30d scale" + decay=0.5 means an
+# article that's 30 days old gets half the score of one published "now".
+response = articles.query.near_text(
+    query="transformer architectures",
+    limit=5,
+    # highlight-start
+    boost=Boost.time_decay(
+        "published",
+        origin="now",
+        scale=timedelta(days=30),
+        curve=Boost.Curve.EXPONENTIAL,
+        decay=0.5,
+        weight=0.6,
+    ),
+    # highlight-end
+    return_properties=["title", "published"],
+)
+# END BoostTimeDecay
+
+# The 400-day-old "Old transformer survey" should be demoted vs the 2-day-old article.
+top_titles = [o.properties["title"] for o in response.objects[:2]]
+assert "Old transformer survey" not in top_titles
+
+
+# ==========================================
+# ===== Numeric decay (closest to a value) =====
+# ==========================================
+
+# START BoostNumericDecay
+# Score peaks at a target price and falls off symmetrically. Gauss gives a
+# bell-shaped falloff: items within `offset` of $49.99 score 1.0, items at
+# $59.99 (one scale away) score `decay`.
+response = articles.query.near_text(
+    query="transformer architectures",
+    limit=5,
+    # highlight-start
+    boost=Boost.numeric_decay(
+        "price",
+        origin=49.99,
+        scale=10.0,
+        curve=Boost.Curve.GAUSSIAN,
+        decay=0.5,
+        weight=0.5,
+    ),
+    # highlight-end
+    return_properties=["title", "price"],
+)
+# END BoostNumericDecay
+
+# Both top-2 results have the target price; the $199 outlier is pushed down.
+assert all(o.properties["price"] == 49.99 for o in response.objects[:2])
+
+
+# ==========================================
+# ===== Blend multiple conditions =====
+# ==========================================
+
+# START BoostBlend
+# Combine two soft signals: recency (weight 2) + popularity (weight 1).
+# The outer weight=0.4 controls how much the blended rank affects the
+# final score; the inner weights are *per-condition* and balance each
+# other.
+response = articles.query.near_text(
+    query="transformer architectures",
+    limit=5,
+    # highlight-start
+    boost=Boost.blend(
+        Boost.time_decay("published", origin="now", scale=timedelta(days=30), weight=2.0),
+        Boost.property("likes", modifier=Boost.Modifier.LOG1P, weight=1.0),
+        weight=0.4,
+        depth=200,  # rescore the top 200 vector matches
+    ),
+    # highlight-end
+    return_properties=["title", "likes", "published"],
+)
+# END BoostBlend
+
+# Recency dominates, but the LOG1P-dampened 5M-likes article still surfaces
+# in the top 3 — popularity is helping, not invisible.
+assert any(o.properties["likes"] == 5_000_000 for o in response.objects[:3])
+
+
+# ==========================================
+# ===== Negative weights demote =====
+# ==========================================
+
+# START BoostNegativeWeight
+# A negative per-condition weight pushes matching documents DOWN — they
+# stay in the result set but lose ground against everything else. Use
+# this to deprioritize drafts without filtering them out entirely.
+response = articles.query.bm25(
+    query="transformer",
+    limit=5,
+    # highlight-start
+    boost=Boost.blend(
+        Boost.filter(Filter.by_property("draft").equal(True), weight=-2.0),
+        weight=0.5,
+    ),
+    # highlight-end
+    return_properties=["title", "draft"],
+)
+
+# The draft article is still in results, just no longer first.
+all_titles = [o.properties["title"] for o in response.objects]
+assert any("Draft" in t for t in all_titles)
+assert response.objects[0].properties["draft"] is False
+# END BoostNegativeWeight
+
+
+# ==========================================
+# ===== Boost on hybrid search =====
+# ==========================================
+
+# START BoostOnHybrid
+# Hybrid keeps its own alpha-blend of BM25 + vector. The boost runs once
+# over the fused hybrid result — the sub-search legs don't see it.
+response = articles.query.hybrid(
+    query="transformer architectures",
+    alpha=0.75,
+    limit=5,
+    # highlight-start
+    boost=Boost.blend(
+        Boost.filter(Filter.by_property("category").equal("research"), weight=1.0),
+        Boost.filter(Filter.by_property("draft").equal(True), weight=-2.0),
+        weight=0.3,
+    ),
+    # highlight-end
+    return_properties=["title", "category", "draft"],
+)
+# END BoostOnHybrid
+
+# Top result is research and not a draft — both boost legs worked.
+assert response.objects[0].properties["category"] == "research"
+assert response.objects[0].properties["draft"] is False
+
+
+client.collections.delete("Articles")
+client.close()