K8s: Eden supported distros update

.claude/skills/tce-examples/reference/testing.md

@@ -130,15 +130,36 @@ The portable pins lag the newest commands. As of DOC-7172 (HIMPORT, Redis 8.10),
 portable pins predated the first release with the command: jedis 7.5.3 (needs 8.0.0),
 lettuce-core 6.5.5 in `pom-lettuce-sync.xml` (needs 7.7.0; the async and reactive poms are
 already on 7.7.0), StackExchange.Redis 3.0.0 in `dotnet/harness.csproj` (needs 3.0.25),
-the cached `npm i redis` 6.1.0 (needs 6.2.0) and ioredis 5.11.1 (needs 6.0.0) in `work/`,
-and go-redis v9.18.0 in `fidelity/go-redis.mod` (needs 9.22.0, which also needs Go 1.24).
+the cached `npm i redis` 6.1.0 (needs 6.2.0), and ioredis 5.11.1 (needs 6.0.0) in `work/`.
+
+The go-redis pin in `fidelity/go-redis.mod` was raised to v9.22.0 (`go 1.24`) in DOC-7197,
+so it now supports HIMPORT. On a machine whose Go toolchain is older than 1.24, `run.sh`
+reports every Go set as SKIP rather than running it.
 
 A run against those pins fails with "method not found" or "unknown command", which looks
 like an example bug but is a pin gap. Until the pins are raised (a separate change, because
 raising them can break other sets), test such a set outside `run.sh`. Give each client a
 scratch project pinned to its first supporting release, with its own throwaway
 `redis:<version>` container on a spare port, and change the port only in the scratch copy.
 
+## Run twice: state that survives a rerun
+
+Some server-side state outlives the example that created it, so one green run proves less
+than it seems. Compact hashes (`HIMPORT`, Redis 8.10) are the known case (DOC-7196):
+
+- A compact hash returns its fields in a different order from the declared one (age, city,
+  name, email for the search examples), so any print or assert of a whole hash or search
+  document changes. Queries that `RETURN` named fields are unaffected.
+- `HSET` on a key that's already a compact hash keeps it compact. An example that
+  re-creates its keys without deleting them first passes on a clean server and fails on the
+  second run.
+
+Run any example that touches compact hashes **twice in a row against the same server**.
+If the second run differs, add a hidden `REMOVE` delete of the keys after the connect step.
+When an example rewrites identical values over existing keys, also run a copy with the
+original write removed. Otherwise a stale index or cached value can produce the expected
+output on its own.
+
 ## Codex evaluation (Phase 4)
 
 An independent reviewer with fresh context. It catches what the harness structurally cannot:

build/example-test-harness/fidelity/go-redis.mod

@@ -1,11 +1,11 @@
 module redis_examples
 
-go 1.21
+go 1.24
 
-require github.com/redis/go-redis/v9 v9.18.0
+require github.com/redis/go-redis/v9 v9.22.0
 
 require (
 	github.com/cespare/xxhash/v2 v2.3.0 // indirect
-	github.com/dgryski/go-rendezvous v0.0.0-20200823014737-9f7001d12a5f // indirect
 	go.uber.org/atomic v1.11.0 // indirect
+	golang.org/x/sys v0.30.0 // indirect
 )

build/example-test-harness/fidelity/pom-jedis.xml

@@ -27,6 +27,21 @@
             <version>5.10.0</version>
             <scope>test</scope>
         </dependency>
+        <dependency>
+            <groupId>ai.djl</groupId>
+            <artifactId>api</artifactId>
+            <version>0.33.0</version>
+        </dependency>
+        <dependency>
+            <groupId>ai.djl.huggingface</groupId>
+            <artifactId>tokenizers</artifactId>
+            <version>0.33.0</version>
+        </dependency>
+        <dependency>
+            <groupId>ai.djl.pytorch</groupId>
+            <artifactId>pytorch-engine</artifactId>
+            <version>0.33.0</version>
+        </dependency>
     </dependencies>
 
     <build>

build/example-test-harness/fidelity/pom-lettuce-async.xml

@@ -19,7 +19,22 @@
         <dependency>
             <groupId>io.lettuce</groupId>
             <artifactId>lettuce-core</artifactId>
-            <version>7.4.0.RELEASE</version>
+            <version>7.7.0.RELEASE</version>
+        </dependency>
+        <dependency>
+            <groupId>ai.djl</groupId>
+            <artifactId>api</artifactId>
+            <version>0.33.0</version>
+        </dependency>
+        <dependency>
+            <groupId>ai.djl.huggingface</groupId>
+            <artifactId>tokenizers</artifactId>
+            <version>0.33.0</version>
+        </dependency>
+        <dependency>
+            <groupId>ai.djl.pytorch</groupId>
+            <artifactId>pytorch-engine</artifactId>
+            <version>0.33.0</version>
         </dependency>
         <dependency>
             <groupId>org.junit.jupiter</groupId>

build/example-test-harness/fidelity/pom-lettuce-reactive.xml

@@ -19,13 +19,28 @@
         <dependency>
             <groupId>io.lettuce</groupId>
             <artifactId>lettuce-core</artifactId>
-            <version>7.4.0.RELEASE</version>
+            <version>7.7.0.RELEASE</version>
         </dependency>
         <dependency>
             <groupId>io.projectreactor</groupId>
             <artifactId>reactor-core</artifactId>
             <version>3.5.0</version>
         </dependency>
+        <dependency>
+            <groupId>ai.djl</groupId>
+            <artifactId>api</artifactId>
+            <version>0.33.0</version>
+        </dependency>
+        <dependency>
+            <groupId>ai.djl.huggingface</groupId>
+            <artifactId>tokenizers</artifactId>
+            <version>0.33.0</version>
+        </dependency>
+        <dependency>
+            <groupId>ai.djl.pytorch</groupId>
+            <artifactId>pytorch-engine</artifactId>
+            <version>0.33.0</version>
+        </dependency>
         <dependency>
             <groupId>org.junit.jupiter</groupId>
             <artifactId>junit-jupiter</artifactId>

build/example-test-harness/pom-jedis.xml

@@ -5,6 +5,9 @@
   <properties><maven.compiler.release>17</maven.compiler.release><project.build.sourceEncoding>UTF-8</project.build.sourceEncoding></properties>
   <dependencies>
     <dependency><groupId>redis.clients</groupId><artifactId>jedis</artifactId><version>7.5.3</version></dependency>
+    <dependency><groupId>ai.djl</groupId><artifactId>api</artifactId><version>0.33.0</version></dependency>
+    <dependency><groupId>ai.djl.huggingface</groupId><artifactId>tokenizers</artifactId><version>0.33.0</version></dependency>
+    <dependency><groupId>ai.djl.pytorch</groupId><artifactId>pytorch-engine</artifactId><version>0.33.0</version></dependency>
     <dependency><groupId>org.junit.jupiter</groupId><artifactId>junit-jupiter</artifactId><version>5.10.2</version><scope>test</scope></dependency>
     <dependency><groupId>org.assertj</groupId><artifactId>assertj-core</artifactId><version>3.25.3</version><scope>test</scope></dependency>
   </dependencies>

build/example-test-harness/pom-lettuce-async.xml

@@ -6,6 +6,9 @@
   <dependencies>
     <dependency><groupId>io.lettuce</groupId><artifactId>lettuce-core</artifactId><version>7.7.0.RELEASE</version></dependency>
     <dependency><groupId>com.fasterxml.jackson.core</groupId><artifactId>jackson-databind</artifactId><version>2.17.1</version></dependency>
+    <dependency><groupId>ai.djl</groupId><artifactId>api</artifactId><version>0.33.0</version></dependency>
+    <dependency><groupId>ai.djl.huggingface</groupId><artifactId>tokenizers</artifactId><version>0.33.0</version></dependency>
+    <dependency><groupId>ai.djl.pytorch</groupId><artifactId>pytorch-engine</artifactId><version>0.33.0</version></dependency>
     <dependency><groupId>org.junit.jupiter</groupId><artifactId>junit-jupiter</artifactId><version>5.10.2</version><scope>test</scope></dependency>
     <dependency><groupId>org.assertj</groupId><artifactId>assertj-core</artifactId><version>3.25.3</version><scope>test</scope></dependency>
   </dependencies>

build/example-test-harness/pom-lettuce-reactive.xml

@@ -6,6 +6,9 @@
   <dependencies>
     <dependency><groupId>io.lettuce</groupId><artifactId>lettuce-core</artifactId><version>7.7.0.RELEASE</version></dependency>
     <dependency><groupId>com.fasterxml.jackson.core</groupId><artifactId>jackson-databind</artifactId><version>2.17.1</version></dependency>
+    <dependency><groupId>ai.djl</groupId><artifactId>api</artifactId><version>0.33.0</version></dependency>
+    <dependency><groupId>ai.djl.huggingface</groupId><artifactId>tokenizers</artifactId><version>0.33.0</version></dependency>
+    <dependency><groupId>ai.djl.pytorch</groupId><artifactId>pytorch-engine</artifactId><version>0.33.0</version></dependency>
     <dependency><groupId>org.junit.jupiter</groupId><artifactId>junit-jupiter</artifactId><version>5.10.2</version><scope>test</scope></dependency>
     <dependency><groupId>org.assertj</groupId><artifactId>assertj-core</artifactId><version>3.25.3</version><scope>test</scope></dependency>
   </dependencies>

build/test_version_archiver.py

@@ -131,6 +131,46 @@ def test_other_products_use_their_own_prefix():
     print("✓ redis-data-integration (integrate prefix) is versioned")
 
 
+def test_links_into_older_versions_untouched():
+    """A link from latest into an older snapshot must keep pointing there.
+
+    Archiving kubernetes 8.2.0 turned /content/operate/kubernetes/7.22/... into
+    /content/operate/kubernetes/8.2.0/7.22/..., which does not exist.
+    """
+    both = ('[m]({{< relref "/operate/kubernetes/7.22/re-clusters/auto-tiering" >}})\n'
+            '[n](/content/operate/kubernetes/7.22/re-clusters/auto-tiering.md)')
+    out = archive("kubernetes", "8.2.0", DEEP, both)
+    assert out == both, out
+    print("✓ links into older version directories are left alone")
+
+
+def test_agent_and_os_files_not_copied():
+    """Agent instruction files, .DS_Store and hidden dirs (.claude/) stay out of a version."""
+    cwd = os.getcwd()
+    with tempfile.TemporaryDirectory() as tmp:
+        os.chdir(tmp)
+        try:
+            archiver = VersionArchiver("kubernetes", "9.9")
+            sub = os.path.join(archiver.latest, "flex")
+            os.makedirs(sub)
+            for name in ("AGENTS.md", "CLAUDE.md", "CLAUDE.local.md", ".DS_Store", "_index.md"):
+                open(os.path.join(archiver.latest, name), "w").close()
+            open(os.path.join(sub, "AGENTS.md"), "w").close()
+            open(os.path.join(sub, "plan.md"), "w").close()
+            hidden = os.path.join(sub, ".claude")
+            os.makedirs(hidden)
+            open(os.path.join(hidden, "state"), "w").close()
+            archiver.archive_version()
+            copied = sorted(
+                os.path.relpath(os.path.join(r, f), archiver.new_directory)
+                for r, _, fs in os.walk(archiver.new_directory) for f in fs
+            )
+        finally:
+            os.chdir(cwd)
+    assert copied == ["_index.md", os.path.join("flex", "plan.md")], copied
+    print("✓ agent instruction and OS files are not copied into the version")
+
+
 def main():
     tests = [
         test_relref_is_versioned,
@@ -141,6 +181,8 @@ def main():
         test_already_versioned_is_idempotent,
         test_other_product_and_external_urls_untouched,
         test_other_products_use_their_own_prefix,
+        test_links_into_older_versions_untouched,
+        test_agent_and_os_files_not_copied,
     ]
     try:
         for t in tests:

build/version_archiver.py

@@ -1,5 +1,15 @@
 import os, shutil, re, argparse
 
+# Files that live in the latest tree but are not documentation. Agent
+# instruction files are loaded by directory, so a frozen copy inside a version
+# would shadow the live rules for anyone editing that version.
+EXCLUDED_FILES = {"AGENTS.md", "CLAUDE.md", "CLAUDE.local.md", ".DS_Store"}
+
+
+def is_version_dir(name):
+    """Same rule archive_version uses to skip other releases: any digit."""
+    return any(char.isdigit() for char in name)
+
 
 class VersionArchiver:
     def __init__(self, product, version):
@@ -24,7 +34,9 @@ def archive_version(self):
             dirs[:] = [
                 d
                 for d in dirs
-                if not any(char.isdigit() for char in d) and d != "release-notes"
+                if not is_version_dir(d)
+                and d != "release-notes"
+                and not d.startswith(".")  # local tool state such as .claude/
             ]
 
             # Create the corresponding destination directory if it does not exist
@@ -34,6 +46,8 @@ def archive_version(self):
                 os.makedirs(dest_dir)
 
             for file in files:
+                if file in EXCLUDED_FILES:
+                    continue
                 src_file = os.path.join(root, file)
                 dest_file = os.path.join(dest_dir, file)
                 shutil.copy2(src_file, dest_file)
@@ -74,11 +88,12 @@ def replace_link(match):
                 full_match = match.group(0)  # The entire match
                 link = match.group(1)  # The actual relref link
 
-                # Check if the link contains 'release-notes' and whether the replacement has already happened
-                if (
-                    "release-notes" not in link
-                    and f"/{self.prefix}/{product}/{version}" not in link
-                ):
+                # Leave release notes alone, and any link that already points
+                # into a version directory: this one (idempotency) or an older
+                # snapshot such as 7.22/, which must not become <version>/7.22/.
+                after_product = link.split(f"/{self.prefix}/{product}/", 1)[-1]
+                first_segment = after_product.split("/", 1)[0]
+                if "release-notes" not in link and not is_version_dir(first_segment):
                     # Otherwise, replace it
                     new_link = link.replace(
                         f"/{self.prefix}/{product}/",

content/commands/bitpos.md

@@ -100,7 +100,7 @@ order to index bytes starting from the end of the string, where -1 is the last
 byte, -2 is the penultimate, and so forth. When `BIT` is specified, -1 is the last
 bit, -2 is the penultimate, and so forth.
 
-Non-existent keys are treated as empty strings.
+If the key does not exist, it is treated as an infinite array of zero bits: searching for a clear bit (0) returns 0, while searching for a set bit (1) returns -1. In contrast, an existing empty string (a key holding "") contains no bits at all, so searching for either 0 or 1 always returns -1.
 
 ## Required arguments
 

content/commands/ft.hybrid.md

@@ -167,6 +167,10 @@ arguments:
   name: sorting
   optional: true
   type: oneof
+- name: explainscore
+  optional: true
+  token: EXPLAINSCORE
+  type: pure-token
 - arguments:
   - name: params
     token: PARAMS
@@ -514,7 +518,7 @@ syntax_fmt: "FT.HYBRID index\n  SEARCH query\n    [SCORER scorer]\n    [YIELD_SC
   \ nargs property [property ...]\n  [GROUPBY nargs property [property ...]\n   \
   \ [REDUCE function nargs arg [arg ...] [AS name]\n    [REDUCE function nargs arg\
   \ [arg ...] [AS name] ...]] ...]]\n  [APPLY expression AS name [APPLY expression\
-  \ AS name ...]]\n  [FILTER filter]\n  PARAMS nargs vector_param vector_blob [name\
+  \ AS name ...]]\n  [FILTER filter]\n  [EXPLAINSCORE]\n  PARAMS nargs vector_param vector_blob [name\
   \ value ...]\n  [TIMEOUT timeout]"
 title: FT.HYBRID
 ---
@@ -657,16 +661,26 @@ defines parameter substitution for the query. Parameters can be referenced in se
 Example: `PARAMS 4 min_price 50 max_price 200`
 </details>
 
-<!--<details open>
+<details open>
 <summary><code>EXPLAINSCORE</code></summary>
 
-includes detailed score explanations in the results, showing how both text search and vector similarity scores were calculated and combined.
-</details>-->
+returns an explanation of how each result's combined score was calculated. Available in Redis 8.10 and later.
+
+Each result gets a `score` field holding a two-element array: the combined score, followed by a nested explanation tree. You don't need `WITHSCORES`. The tree shows:
+
+- The final score formula and the fusion method with its parameters: `RRF` with `window` and `constant`, or `LINEAR` with `alpha`, `beta`, and `window`.
+- The text branch: the scorer name, such as `BM25STD`, and the scorer's own breakdown. With `RRF`, it shows the text rank. With `LINEAR`, it shows the text contribution and the normalized text score.
+- The vector branch: `KNN` or `RANGE` with its radius and epsilon. With `RRF`, it shows the vector rank. With `LINEAR`, it shows the vector contribution and the normalized vector score. `RANGE` also shows whether the document matched within the radius.
+
+A branch that didn't match the document shows `<no match>`.
+
+`EXPLAINSCORE` can't be combined with `GROUPBY`. The command returns the error `EXPLAINSCORE is not supported with GROUPBY`.
+</details>
 
 <details open>
 <summary><code>TIMEOUT timeout</code></summary>
 
-sets a runtime timeout for the query execution in milliseconds.
+sets a runtime timeout for the query execution in milliseconds. When the query times out, the [`search-on-timeout`](/content/develop/ai/search-and-query/administration/configuration.md#search-on-timeout) policy applies. With `RETURN` or `RETURN_STRICT`, the command returns the results accumulated so far and reports the timeout in `warnings`. With `FAIL`, it returns an error.
 </details>
 
 <!--<details open>
@@ -779,6 +793,45 @@ Use parameter substitution for dynamic queries:
 {{< / highlight >}}
 </details>
 
+<details open>
+<summary><b>Explain hybrid scores</b></summary>
+
+Return the top result with an explanation of its RRF score. The document ranked first in both the text and vector branches:
+
+{{< highlight bash >}}
+127.0.0.1:6379> FT.HYBRID products-idx
+  SEARCH "shoes"
+  VSIM @embedding $query_vec
+  KNN 2 K 3
+  EXPLAINSCORE
+  LIMIT 0 1
+  PARAMS 2 query_vec <vector_blob>
+1) total_results
+2) (integer) 3
+3) results
+4) 1) 1) score
+      2) 1) "0.03278688524590164"
+         2) 1) final score: 1 / (constant 60.00 + rank 1) + 1 / (constant 60.00 + rank 1) = 0.032786885245901641
+            2) 1) 1) Hybrid score (RRF: window=20, constant=60.00)
+                  2) 1) 1) text rank = 1
+                        2) 1) 1) Text scorer: BM25STD
+                              2) 1) 1) Final BM25 : words BM25 1.39 * document score 1.00
+                                    2) 1) 1) (Weight 1.00 * children BM25 1.39)
+                                          2) 1) shoes: (0.69 = Weight 1.00 * IDF 0.69 * (F 1.00 * (k1 1.2 + 1)) / (F 1.00 + k1 1.2 * (1 - b 0.75 + b 0.75 * Doc Len 3 / Average Doc Len 3.00)))
+                                             2) +shoe: (0.69 = Weight 1.00 * IDF 0.69 * (F 1.00 * (k1 1.2 + 1)) / (F 1.00 + k1 1.2 * (1 - b 0.75 + b 0.75 * Doc Len 3 / Average Doc Len 3.00)))
+                     2) 1) vector branch (KNN)
+                        2) 1) vector rank = 1
+      3) "__key"
+      4) "product:1"
+      5) "__score"
+      6) "0.0327868852459"
+5) warnings
+6) (empty array)
+7) execution_time
+8) "2.0055"
+{{< / highlight >}}
+</details>
+
 ## Complexity
 
 FT.HYBRID complexity depends on both the text search and vector similarity components: