Add translation

chart/values-prod.yaml.example

@@ -98,7 +98,7 @@ media:
 
 # --- Ingestion ----------------------------------------------------------------
 # A deployment loads every sentence the filters admit, not the development
-# sample. For the Thai 25.0 release that is 21,429 exercises and as many clips,
+# sample. For the Thai 27.0 release that is 21,442 exercises and as many clips,
 # about 400 MB on the media volume, and a Job that runs for minutes. The Job
 # still only runs when rendered with ingest.enabled=true; see docs/deployment.md.
 ingest:

chart/values.yaml

@@ -223,7 +223,7 @@ ingest:
   enabled: false
   # How much of the corpus to load: a positive integer selects that many of the
   # best candidates, `all` selects every sentence the filters admit (about 21k
-  # for the Thai 25.0 release). Small by default, because a development cluster
+  # for the Thai 27.0 release). Small by default, because a development cluster
   # needs a catalog, not the catalog; a deployment asks for `all`.
   count: 100
   # The release directory as the pod sees it. In development kind mounts the

docs/deployment.md

@@ -106,14 +106,14 @@ onto the media volume. The corpus has to be on a node, mounted from
 `ingest.count` decides how much of it is loaded. A positive integer selects that
 many of the best candidates, and `all` selects every sentence the quality and
 beginner filters admit. The chart default is `100`, a development-sized sample;
-`values-prod.yaml.example` sets `all`. For the Thai 25.0 release, `all` is 21,429
+`values-prod.yaml.example` sets `all`. For the Thai 27.0 release, `all` is 21,442
 exercises and about 400 MB of clips, and the Job runs for minutes, logging its
 progress every thousand exercises. Any other value fails at render time.
 
 ```bash
 kubectl delete job ingest -n typelearn --ignore-not-found
 helm template typelearn ./chart --namespace typelearn --values my-values.yaml \
-  --set ingest.enabled=true --set ingest.corpusHostPath=/srv/cv-corpus-25.0-2026-03-09 \
+  --set ingest.enabled=true --set ingest.corpusHostPath=/srv/cv-corpus-27.0-2026-09-11 \
   --show-only templates/ingest-job.yaml | kubectl apply -n typelearn -f -
 kubectl logs -n typelearn -f job/ingest
 ```

docs/development.md

@@ -90,11 +90,11 @@ bootstrap.
 ### 3. Load exercises
 
 Ingestion needs a local Common Voice release directory
-(`cv-corpus-25.0-2026-03-09` for Thai). It stays outside the repository and is
+(`cv-corpus-27.0-2026-09-11` for Thai). It stays outside the repository and is
 never committed — point the cluster at it before creating the cluster:
 
 ```bash
-export TYPELEARN_CORPUS_DIR=/path/to/cv-corpus-25.0-2026-03-09
+export TYPELEARN_CORPUS_DIR=/path/to/cv-corpus-27.0-2026-09-11
 ```
 
 Then press **Ingest the corpus** in the Tilt UI. It selects 100 exercises from

openspec/config.yaml

@@ -24,8 +24,8 @@ context: |
     configuration at all. Keep it that way.
 
   Corpus:
-  - Mozilla Common Voice Thai 25.0, on disk in the release directory
-    cv-corpus-25.0-2026-03-09/th — outside the repository, read in place,
+  - Mozilla Common Voice Thai 27.0, on disk in the release directory
+    cv-corpus-27.0-2026-09-11/th — outside the repository, read in place,
     never committed. Its path is a parameter, never hardcoded.
   - Select exercises from validated.tsv (has path + up_votes). Never from
     validated_sentences.tsv, which has neither.

openspec/specs/corpus-ingestion/spec.md

@@ -38,7 +38,7 @@ later rows for an already-selected sentence are skipped. A count selects a prefi
 of the "all" selection.
 
 #### Scenario: Selecting the MVP subset
-- **WHEN** the ingestion script runs against the Thai corpus at the supplied corpus root (release directory `cv-corpus-25.0-2026-03-09`) with no selection size given
+- **WHEN** the ingestion script runs against the Thai corpus at the supplied corpus root (release directory `cv-corpus-27.0-2026-09-11`) with no selection size given
 - **THEN** it emits exactly 100 exercises, each with a unique `sentence`, drawn from the candidate pool matching the filters above
 
 #### Scenario: Selecting an explicit count

specs/tech-stack.md

@@ -100,9 +100,9 @@ Tailwind 3 and its `init -p` command was removed in v4.
 
 | Layer | Choice | Notes |
 |---|---|---|
-| Corpus source | Mozilla Common Voice | CC-0 license, 150+ languages, Thai release 25.0 |
-| Corpus location | Release directory `cv-corpus-25.0-2026-03-09/th`, path supplied to the ingestion script | 9.6 GB, **outside the repository**, read in place and never committed |
-| Selection source | `validated.tsv` | 148,765 rows; the only manifest carrying both a clip `path` and vote counts |
+| Corpus source | Mozilla Common Voice | CC-0 license, 150+ languages, Thai release 27.0 |
+| Corpus location | Release directory `cv-corpus-27.0-2026-09-11/th`, path supplied to the ingestion script | 9.7 GB, **outside the repository**, read in place and never committed |
+| Selection source | `validated.tsv` | 148,823 rows; the only manifest carrying both a clip `path` and vote counts |
 | Clip durations | `clip_durations.tsv` | 366,508 rows mapping clip filename → duration in ms |
 | Audio | `clips/*.mp3` | 366,508 files; only the ~100 selected clips are copied into `MEDIA_ROOT` |
 | MVP dataset | JSON file | `data/processed/mvp_dataset.json`, produced by the ingestion script |
@@ -206,7 +206,7 @@ TypeLearn/
 ```
 
 The Common Voice corpus is **not** part of this tree. It lives outside the
-repository in the release directory `cv-corpus-25.0-2026-03-09/`, is
+repository in the release directory `cv-corpus-27.0-2026-09-11/`, is
 read in place by the ingestion script, and only the selected clips are copied into
 `data/media/`. The corpus path is a parameter of the ingestion script, not a
 hardcoded constant.