feat(daily-regen): default to Haiku, add spec_id + model inputs

.github/workflows/bulk-generate.yml

@@ -36,6 +36,15 @@ on:
         description: "Seconds to wait between dispatches (0 = fire all at once, default 120 = 2 min)"
         required: false
         default: '120'
+      model:
+        description: "Claude model to use across generate / review / repair (default sonnet)"
+        required: false
+        type: choice
+        default: 'sonnet'
+        options:
+          - haiku
+          - sonnet
+          - opus
 
 env:
   ALL_LIBRARIES: "matplotlib seaborn plotly bokeh altair plotnine pygal highcharts letsplot"
@@ -168,13 +177,14 @@ jobs:
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           MATRIX: ${{ needs.build-matrix.outputs.matrix }}
           PACE_SECONDS: ${{ inputs.pace_seconds || '120' }}
+          MODEL: ${{ inputs.model || 'sonnet' }}
         run: |
           set -u
 
           pace="${PACE_SECONDS}"
           pairs=$(echo "$MATRIX" | jq -r '.include[] | "\(.specification_id) \(.library)"')
           total=$(echo "$pairs" | wc -l | tr -d ' ')
-          echo "::notice::Dispatching $total item(s) with ${pace}s pacing between each"
+          echo "::notice::Dispatching $total item(s) with ${pace}s pacing between each (model=${MODEL})"
 
           i=0
           failed=0
@@ -203,11 +213,13 @@ jobs:
                 gh workflow run impl-generate.yml --repo "${{ github.repository }}" \
                   -f specification_id="${SPEC_ID}" \
                   -f library="${LIBRARY}" \
-                  -f issue_number="${ISSUE}" && dispatched=1 && break
+                  -f issue_number="${ISSUE}" \
+                  -f model="${MODEL}" && dispatched=1 && break
               else
                 gh workflow run impl-generate.yml --repo "${{ github.repository }}" \
                   -f specification_id="${SPEC_ID}" \
-                  -f library="${LIBRARY}" && dispatched=1 && break
+                  -f library="${LIBRARY}" \
+                  -f model="${MODEL}" && dispatched=1 && break
               fi
               echo "::warning::Dispatch attempt $attempt failed for ${SPEC_ID}/${LIBRARY}, retrying in 10s"
               sleep 10

.github/workflows/daily-regen.yml

@@ -1,5 +1,5 @@
 name: "Scheduled: Regen oldest specs"
-run-name: "Scheduled regen (${{ github.event.inputs.count || '1' }} specs)"
+run-name: "Scheduled regen (${{ github.event.inputs.specification_id || github.event.inputs.count || '1' }} / ${{ github.event.inputs.model || 'haiku' }})"
 
 # Picks the N oldest specs (by most-recent implementation `updated` timestamp)
 # and re-dispatches `bulk-generate.yml` for each. Default N=1 per cron tick.
@@ -9,21 +9,34 @@ run-name: "Scheduled regen (${{ github.event.inputs.count || '1' }} specs)"
 # 02, 04, 06, 08, 10, 12, 14, 16. The 20:00 and 22:00 Berlin slots (UTC 18, 20)
 # are intentionally skipped so runs never start during the user's evening.
 #
-# bulk-generate is serialised via its own concurrency group. With Sonnet +
-# reduced bulk-generate pace, a single spec completes well within the 2h slot,
-# leaving the user window clean.
+# bulk-generate is serialised via its own concurrency group. Default model is
+# Haiku, so a single spec completes well within the 2h slot, leaving the user
+# window clean. The model can be overridden per manual run.
 #
 # Triggers:
-#   - schedule: 10× daily (UTC, every 2h except 18:00 and 20:00 UTC)
-#   - workflow_dispatch: manual, with inputs for count + dry-run
+#   - schedule: 10× daily (UTC, every 2h except 18:00 and 20:00 UTC) → Haiku
+#   - workflow_dispatch: manual, with inputs for spec id, model, count, dry-run
 
 on:
   schedule:
     - cron: '0 0,2,4,6,8,10,12,14,16,22 * * *'
   workflow_dispatch:
     inputs:
+      specification_id:
+        description: "Specific spec id to regen (leave empty to pick the oldest)"
+        required: false
+        default: ''
+      model:
+        description: "Claude model to use across all generate / review / repair steps"
+        required: false
+        type: choice
+        default: 'haiku'
+        options:
+          - haiku
+          - sonnet
+          - opus
       count:
-        description: "How many of the oldest specs to regen (default 1)"
+        description: "How many of the oldest specs to regen (ignored when specification_id is set, default 1)"
         required: false
         default: '1'
       min_age_hours:
@@ -60,14 +73,16 @@ jobs:
       - name: Install PyYAML
         run: pip install pyyaml
 
-      - name: Pick oldest spec(s)
+      - name: Pick oldest spec(s) (or use override)
         id: pick
         env:
           COUNT: ${{ inputs.count || '1' }}
           MIN_AGE_HOURS: ${{ inputs.min_age_hours || '20' }}
+          SPEC_OVERRIDE: ${{ inputs.specification_id }}
         run: |
           python3 <<'PY'
           import os
+          import sys
           from datetime import datetime, timedelta, timezone
           from pathlib import Path
 
@@ -76,9 +91,24 @@ jobs:
 
           COUNT = int(os.environ["COUNT"])
           MIN_AGE = timedelta(hours=int(os.environ["MIN_AGE_HOURS"]))
+          OVERRIDE = (os.environ.get("SPEC_OVERRIDE") or "").strip()
           NOW = datetime.now(timezone.utc)
 
           specs_dir = Path("plots")
+
+          # Manual override: caller specified a spec id — validate and use it.
+          if OVERRIDE:
+              if not (specs_dir / OVERRIDE / "specification.md").is_file():
+                  print(f"::error::Spec '{OVERRIDE}' not found at plots/{OVERRIDE}/specification.md")
+                  sys.exit(1)
+              picks = [OVERRIDE]
+              print(f"::notice::Using override spec: {OVERRIDE}")
+              github_output = os.environ["GITHUB_OUTPUT"]
+              with open(github_output, "a", encoding="utf-8") as f:
+                  f.write(f"specs={OVERRIDE}\n")
+                  f.write(f"count=1\n")
+              sys.exit(0)
+
           candidates: list[tuple[datetime, str]] = []
 
           for spec_dir in sorted(specs_dir.iterdir()):
@@ -142,13 +172,15 @@ jobs:
         env:
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           SPECS: ${{ needs.pick.outputs.specs }}
+          MODEL: ${{ inputs.model || 'haiku' }}
         run: |
           for spec in $SPECS; do
-            echo "::notice::Dispatching bulk-generate for $spec (all 9 libs)"
+            echo "::notice::Dispatching bulk-generate for $spec (all 9 libs, model=$MODEL)"
             gh workflow run bulk-generate.yml \
               --repo "${{ github.repository }}" \
               -f specification_id="$spec" \
-              -f library=all
+              -f library=all \
+              -f model="$MODEL"
             # Small pause between dispatches so GitHub's webhook processing has a moment.
             sleep 5
           done

.github/workflows/impl-generate.yml

@@ -33,6 +33,15 @@ on:
         description: "Issue number (optional, for tracking)"
         required: false
         type: string
+      model:
+        description: "Claude model to use (also threaded into review + repair)"
+        required: false
+        type: choice
+        default: 'sonnet'
+        options:
+          - haiku
+          - sonnet
+          - opus
 
 # Global concurrency: max 3 concurrent implementation workflows
 concurrency:
@@ -72,12 +81,14 @@ jobs:
             SPEC_ID="${{ inputs.specification_id }}"
             LIBRARY="${{ inputs.library }}"
             ISSUE="${{ inputs.issue_number }}"
+            MODEL="${{ inputs.model }}"
           else
             # From label trigger: generate:{library}
             LIBRARY=$(echo "$LABEL_NAME" | sed 's/^generate://')
             # Extract spec ID from issue title: [spec-id] ...
             SPEC_ID=$(echo "$ISSUE_TITLE" | sed -n 's/^\[\([a-z0-9-]*\)\].*/\1/p')
             ISSUE="${{ github.event.issue.number }}"
+            MODEL=""
           fi
 
           if [ -z "$SPEC_ID" ]; then
@@ -90,15 +101,21 @@ jobs:
             exit 1
           fi
 
+          # Default model when caller didn't supply one (label trigger path).
+          if [ -z "$MODEL" ]; then
+            MODEL="sonnet"
+          fi
+
           # Language: only python supported today. Future multi-language work will derive this per spec.
           LANGUAGE="python"
 
           echo "specification_id=$SPEC_ID" >> $GITHUB_OUTPUT
           echo "library=$LIBRARY" >> $GITHUB_OUTPUT
           echo "language=$LANGUAGE" >> $GITHUB_OUTPUT
           echo "issue_number=$ISSUE" >> $GITHUB_OUTPUT
+          echo "model=$MODEL" >> $GITHUB_OUTPUT
 
-          echo "::notice::Generating $LANGUAGE/$LIBRARY for $SPEC_ID (issue: ${ISSUE:-none})"
+          echo "::notice::Generating $LANGUAGE/$LIBRARY for $SPEC_ID (issue: ${ISSUE:-none}, model: ${MODEL})"
 
       - name: Checkout repository
         uses: actions/checkout@de0fac2e4500dabe0009e67214ff5f5447ce83dd  # v6
@@ -308,7 +325,7 @@ jobs:
         uses: anthropics/claude-code-action@ef50f123a3a9be95b60040d042717517407c7256  # v1
         with:
           claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
-          claude_args: "--model sonnet"
+          claude_args: "--model ${{ steps.inputs.outputs.model }}"
           # bulk-generate dispatches us from the github-actions bot; explicitly allow it.
           allowed_bots: '*'
           prompt: |
@@ -327,7 +344,7 @@ jobs:
         uses: anthropics/claude-code-action@ef50f123a3a9be95b60040d042717517407c7256  # v1
         with:
           claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
-          claude_args: "--model sonnet"
+          claude_args: "--model ${{ steps.inputs.outputs.model }}"
           # bulk-generate dispatches us from the github-actions bot; explicitly allow it.
           allowed_bots: '*'
           prompt: |
@@ -350,6 +367,7 @@ jobs:
           LIBRARY: ${{ steps.inputs.outputs.library }}
           ISSUE: ${{ steps.issue.outputs.number }}
           BRANCH: ${{ steps.branch.outputs.branch }}
+          MODEL: ${{ steps.inputs.outputs.model }}
         run: |
           IMPL_DIR="plots/${SPEC_ID}/implementations/${LANGUAGE}"
 
@@ -445,6 +463,7 @@ jobs:
           lib_ver = '$LIBRARY_VERSION'
           has_html = '$HAS_HTML' == 'true'
           metadata_file = '$METADATA_FILE'
+          model = '$MODEL'
           base_url = f'https://storage.googleapis.com/anyplot-images/plots/{spec}/{language}/{lib}'
 
           # Preserve the original 'created' timestamp on regenerations.
@@ -466,11 +485,11 @@ jobs:
               'specification_id': spec,
               'created': created_ts,
               'updated': ts,
-              # Reflects what claude_args=`--model sonnet` actually runs: whatever
-              # Claude Code's current "sonnet" alias resolves to.
+              # Reflects what claude_args=`--model {model}` actually runs: whatever
+              # Claude Code's current alias for the chosen model resolves to.
               # Use the family name instead of a frozen version string so the
               # metadata doesn't go stale every model release.
-              'generated_by': 'claude-sonnet',
+              'generated_by': f'claude-{model}',
               'workflow_run': run_id,
               'issue': issue,
               'python_version': py_ver,
@@ -710,12 +729,14 @@ jobs:
         env:
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           PR_NUMBER: ${{ steps.pr.outputs.pr_number }}
+          MODEL: ${{ steps.inputs.outputs.model }}
         run: |
           # Use repository_dispatch as workaround for workflow_dispatch caching issue
           gh api repos/${{ github.repository }}/dispatches \
             -f event_type=review-pr \
-            -f 'client_payload[pr_number]='"$PR_NUMBER"
-          echo "::notice::Triggered impl-review.yml via repository_dispatch for PR #$PR_NUMBER"
+            -f "client_payload[pr_number]=$PR_NUMBER" \
+            -f "client_payload[model]=$MODEL"
+          echo "::notice::Triggered impl-review.yml via repository_dispatch for PR #$PR_NUMBER (model=$MODEL)"
 
       - name: Determine result
         id: result
@@ -736,6 +757,7 @@ jobs:
           SPEC_ID: ${{ steps.inputs.outputs.specification_id }}
           LIBRARY: ${{ steps.inputs.outputs.library }}
           ISSUE: ${{ steps.issue.outputs.number }}
+          MODEL: ${{ steps.inputs.outputs.model }}
         run: |
           echo "::notice::Handling generation failure for $LIBRARY/$SPEC_ID"
 
@@ -775,7 +797,7 @@ jobs:
 
           To retry manually:
           \`\`\`
-          gh workflow run impl-generate.yml -f specification_id=${SPEC_ID} -f library=${LIBRARY} -f issue_number=${ISSUE}
+          gh workflow run impl-generate.yml -f specification_id=${SPEC_ID} -f library=${LIBRARY} -f issue_number=${ISSUE} -f model=${MODEL}
           \`\`\`
 
           ---
@@ -798,7 +820,8 @@ jobs:
             gh workflow run impl-generate.yml \
               -f specification_id="${SPEC_ID}" \
               -f library="${LIBRARY}" \
-              -f issue_number="${ISSUE}"
+              -f issue_number="${ISSUE}" \
+              -f model="${MODEL}"
 
-            echo "::notice::Triggered automatic retry for ${LIBRARY}/${SPEC_ID} (attempt $((ATTEMPT + 1)))"
+            echo "::notice::Triggered automatic retry for ${LIBRARY}/${SPEC_ID} (attempt $((ATTEMPT + 1)), model=${MODEL})"
           fi

.github/workflows/impl-repair.yml

@@ -23,6 +23,15 @@ on:
         description: "Current attempt number (1, 2, 3, or 4)"
         required: true
         type: string
+      model:
+        description: "Claude model to use (also threaded back into review)"
+        required: false
+        type: choice
+        default: 'sonnet'
+        options:
+          - haiku
+          - sonnet
+          - opus
 
 # Per-library deps now come from `pyproject.toml` `lib-{library}` extras
 # (single source of truth — same change as impl-generate.yml).
@@ -107,7 +116,7 @@ jobs:
         uses: anthropics/claude-code-action@ef50f123a3a9be95b60040d042717517407c7256  # v1
         with:
           claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
-          claude_args: "--model sonnet"
+          claude_args: "--model ${{ inputs.model || 'sonnet' }}"
           allowed_bots: '*'
           prompt: |
             Read `prompts/workflow-prompts/impl-repair-claude.md` and follow those instructions.
@@ -126,7 +135,7 @@ jobs:
         uses: anthropics/claude-code-action@ef50f123a3a9be95b60040d042717517407c7256  # v1
         with:
           claude_code_oauth_token: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
-          claude_args: "--model sonnet"
+          claude_args: "--model ${{ inputs.model || 'sonnet' }}"
           allowed_bots: '*'
           prompt: |
             Read `prompts/workflow-prompts/impl-repair-claude.md` and follow those instructions.
@@ -225,12 +234,14 @@ jobs:
         env:
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
           PR_NUM: ${{ inputs.pr_number }}
+          MODEL: ${{ inputs.model || 'sonnet' }}
         run: |
           # Use repository_dispatch as workaround for workflow_dispatch caching issue
           gh api repos/${{ github.repository }}/dispatches \
             -f event_type=review-pr \
-            -f "client_payload[pr_number]=$PR_NUM"
-          echo "::notice::Triggered impl-review.yml via repository_dispatch for PR #$PR_NUM"
+            -f "client_payload[pr_number]=$PR_NUM" \
+            -f "client_payload[model]=$MODEL"
+          echo "::notice::Triggered impl-review.yml via repository_dispatch for PR #$PR_NUM (model=$MODEL)"
 
       # ========================================================================
       # Failure handling: when the repair workflow itself crashes (e.g. Claude
@@ -245,6 +256,7 @@ jobs:
           SPEC_ID: ${{ inputs.specification_id }}
           LIBRARY: ${{ inputs.library }}
           ATTEMPT: ${{ inputs.attempt }}
+          MODEL: ${{ inputs.model || 'sonnet' }}
         run: |
           # Restore ai-rejected label that was removed at start of repair so the
           # PR state stays consistent (otherwise it looks "stuck approved").
@@ -264,7 +276,7 @@ jobs:
 
           **Manual restart:**
           \`\`\`
-          gh workflow run impl-repair.yml -f pr_number=${PR_NUM} -f specification_id=${SPEC_ID} -f library=${LIBRARY} -f attempt=${ATTEMPT}
+          gh workflow run impl-repair.yml -f pr_number=${PR_NUM} -f specification_id=${SPEC_ID} -f library=${LIBRARY} -f attempt=${ATTEMPT} -f model=${MODEL}
           \`\`\`
 
           ---
@@ -283,5 +295,6 @@ jobs:
               -f pr_number="$PR_NUM" \
               -f specification_id="$SPEC_ID" \
               -f library="$LIBRARY" \
-              -f attempt="$ATTEMPT"
+              -f attempt="$ATTEMPT" \
+              -f model="$MODEL"
           fi