docs: mark the serve query service as production ready (#482)

The `serve` query service has been validated internally in production CI,
so the docs should no longer present it as experimental.

- Drop the "(experimental)" qualifier from the Query Service heading and
  add a lead sentence stating it is production ready and supported.
- Reword the `modifiedFilepaths` caveat: the superset requirement is a
  correctness contract on the caller's list, not a marker of instability.
- Retitle "Notes and current limitations" to "Notes and operational
  guidance"; the remaining bullets are operational facts, not caveats.
- Drop the "Experimental:" prefix from the `generate-hashes
  --modified-filepaths` help text, whose semantics the serve POST body
  shares.

README.md and tools/readme_template.md are kept in sync (the README is
generated from the template), and the rendered CLI help block was taken
verbatim from a fresh //tools:cli_help_output build rather than
hand-wrapped.

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>

Author: tinder-maxwellelliott

README.md

@@ -191,7 +191,10 @@ neither ever truncates silently — the full count is always reported:
 * `--maxDepth` (default `-1`, unbounded) stops the search N dependency hops above the queried
   target. When a bound cuts the search short, a warning says so.
 
-## Query Service (experimental)
+## Query Service
+
+`bazel-diff serve` is production ready — it has been validated internally in production CI and is
+supported for production use.
 
 Instead of running `generate-hashes` from scratch on every CI invocation, you can run `bazel-diff` as
 a long-running HTTP service that answers affectedness queries between two git revisions and caches
@@ -242,8 +245,8 @@ curl 'http://localhost:8080/impacted_targets?from=main&to=my-feature-branch'
   as unchanged, turning an O(all source files) content read into O(changed files) — the same
   optimization as `generate-hashes --modified-filepaths`. The list must be a **superset** of what
   actually changed: a truly-changed file left off it is content-skipped on both sides and its
-  impacted targets are missed (hence experimental). Omit it (or send `[]`) for the full-content hash,
-  identical to the GET form. `POST /impacted_targets_with_distances` accepts the same body.
+  impacted targets are missed, so treat the list as a correctness contract. Omit it (or send `[]`)
+  for the full-content hash, identical to the GET form. `POST /impacted_targets_with_distances` accepts the same body.
 
 ```bash
 curl -X POST http://localhost:8080/impacted_targets \
@@ -343,7 +346,7 @@ revision are also harmless — entries are deterministic per key, so last-write-
 content. The `--cacheMax*` pruning flags bound the *local* tier only; bound the bucket with an S3
 lifecycle policy instead.
 
-Notes and current limitations:
+Notes and operational guidance:
 
 * Distance metrics (`/impacted_targets_with_distances`) and the generate-hashes graph
   (`/dependency_edges`) require the dependency-edge graph, which is
@@ -546,12 +549,11 @@ workspace.
                             hashes. Disabling it catches configuration issues
                             by failing loudly. Defaults to `false`
   -m, --modified-filepaths=<modifiedFilepaths>
-                          Experimental: A text file containing a newline
-                            separated list of filepaths (relative to the
-                            workspace) these filepaths should represent the
-                            modified files between the specified revisions and
-                            will be used to scope what files are hashed during
-                            hash generation.
+                          A text file containing a newline separated list of
+                            filepaths (relative to the workspace) these
+                            filepaths should represent the modified files
+                            between the specified revisions and will be used to
+                            scope what files are hashed during hash generation.
   -s, --seed-filepaths=<seedFilepaths>
                           A text file containing a newline separated list of
                             filepaths. Each file in this list will be read and

cli/src/main/kotlin/com/bazel_diff/cli/GenerateHashesCommand.kt

@@ -179,7 +179,7 @@ open class GenerateHashesCommand : Callable<Int> {
       names = ["-m", "--modified-filepaths"],
       description =
           [
-              "Experimental: A text file containing a newline separated list of filepaths (relative to the workspace) these filepaths should represent the modified files between the specified revisions and will be used to scope what files are hashed during hash generation."])
+              "A text file containing a newline separated list of filepaths (relative to the workspace) these filepaths should represent the modified files between the specified revisions and will be used to scope what files are hashed during hash generation."])
   var modifiedFilepaths: File? = null
 
   @CommandLine.Option(

tools/readme_template.md

@@ -191,7 +191,10 @@ neither ever truncates silently — the full count is always reported:
 * `--maxDepth` (default `-1`, unbounded) stops the search N dependency hops above the queried
   target. When a bound cuts the search short, a warning says so.
 
-## Query Service (experimental)
+## Query Service
+
+`bazel-diff serve` is production ready — it has been validated internally in production CI and is
+supported for production use.
 
 Instead of running `generate-hashes` from scratch on every CI invocation, you can run `bazel-diff` as
 a long-running HTTP service that answers affectedness queries between two git revisions and caches
@@ -242,8 +245,8 @@ curl 'http://localhost:8080/impacted_targets?from=main&to=my-feature-branch'
   as unchanged, turning an O(all source files) content read into O(changed files) — the same
   optimization as `generate-hashes --modified-filepaths`. The list must be a **superset** of what
   actually changed: a truly-changed file left off it is content-skipped on both sides and its
-  impacted targets are missed (hence experimental). Omit it (or send `[]`) for the full-content hash,
-  identical to the GET form. `POST /impacted_targets_with_distances` accepts the same body.
+  impacted targets are missed, so treat the list as a correctness contract. Omit it (or send `[]`)
+  for the full-content hash, identical to the GET form. `POST /impacted_targets_with_distances` accepts the same body.
 
 ```bash
 curl -X POST http://localhost:8080/impacted_targets \
@@ -343,7 +346,7 @@ revision are also harmless — entries are deterministic per key, so last-write-
 content. The `--cacheMax*` pruning flags bound the *local* tier only; bound the bucket with an S3
 lifecycle policy instead.
 
-Notes and current limitations:
+Notes and operational guidance:
 
 * Distance metrics (`/impacted_targets_with_distances`) and the generate-hashes graph
   (`/dependency_edges`) require the dependency-edge graph, which is