@ozzylabs/feedradar 0.1.5 → 0.1.7

package/dist/templates/agents/AGENTS.md

@@ -56,12 +56,13 @@ radar source test <id> [--limit N] [--show-content]   # state/items を書き換
 radar source remove <id>
 
 # JSON API recipe を pagination 付きで追加（ADR-0012）
+# `facets:` (年・カテゴリ単位の sweep) は flag では設定できず recipe のみ (ADR-0017)
 radar source add aws-whats-new --kind json-api \
-  --url "https://aws.amazon.com/api/dirs/items/search?item.directoryId=whats-new&size=100&page=0" \
+  --url "https://aws.amazon.com/api/dirs/items/search?item.directoryId=whats-new-v2&size=100&page=0" \
   --keywords "Bedrock,Claude" \
-  --pagination-strategy page --page-size 100 --max-pages 200
+  --pagination-strategy page --page-size 100 --max-pages 30
 
-# 同じことを bundled recipe で 1 行で
+# 同じことを bundled recipe で 1 行で (year facet sweep 付き、全 21,834 件カバー)
 radar source add aws-watch --recipe aws-whats-new --keywords "Bedrock,Claude"
 
 # JSON Feed (1.0 / 1.1) は URL のみで動く zero-config
@@ -73,7 +74,8 @@ radar source add example-microblog --kind json-feed \
 radar watch run
 
 # 過去全履歴の一括取り込み (kind: json-api / github-releases / npm-registry)
-radar watch run --source aws-whats-new --backfill --max-pages 200
+# AWS は recipe の facets.year + per-facet maxPages=30 で 21,834 件を完全カバー (ADR-0017)
+radar watch run --source aws-whats-new --backfill
 
 # 検出済み item に対する操作
 radar research <item-id> --agent <agent> [--verbose]   # 調査レポートを生成 (status: detected -> researched)。--verbose で agent stdout を直接見る (ADR-0015)
@@ -142,6 +144,36 @@ radar research <item-id>      # 自動 triage は推奨しない (ユーザー
 
 `watch run` は cron / GitHub Actions / Claude Routines から呼ぶことを想定しています。`research` / `review` / `update` / `dismiss` は人間の判断が伴うため、interactive session 経由を推奨します。
 
+### scheduled triage workflow 例 (ADR-0018 §W5)
+
+`triagePolicy:` を持つ source を登録済みの場合、scheduled GHA cron で `watch → triage → research → review` を**無人実行**できます。雛形は `radar workflow generate combined-with-triage` で生成:
+
+```bash
+radar workflow generate combined-with-triage \
+  --watch-cron "0 6 * * *" \
+  --triage-agent gemini-cli \
+  --research-agent claude-code \
+  --review-agent codex-cli \
+  --max-items 10
+# → .github/workflows/feedradar-daily.yaml
+```
+
+生成される workflow が 1 cron tick で走らせる 5 ステップ:
+
+```text
+1. radar watch run                                            # 新着 → detected
+2. radar triage --apply --triage-agent gemini-cli             # detected → triaged_research / triaged_digest / triaged_unsure / dismissed
+3. radar research --batch --status triaged_research \
+     --max-items 10 --agent claude-code                       # triaged_research → researched (1 item 1 report)
+4. radar research --digest <ids per triage.group> \
+     --agent claude-code                                      # triaged_digest → researched (group ごとに 1 report 集約)
+5. radar review --batch --status researched --agent codex-cli # researched → reviewed (cross-agent)
+```
+
+末尾には `triaged_unsure` キュー深度を Slack 通知する `if: always()` step と、`peter-evans/create-pull-request@v6` で `items/ state/ research/` を 1 PR にまとめる step が付く。**triage の cost は research に比べて 1-2 桁安い** (cheap-model channel、`gemini-2.5-flash-lite` 想定で月数千 item でも \$0.10 未満) ため、cost gating の主防御は引き続き `--max-items` (ADR-0014 D3a)。
+
+詳細・secrets setup・policy 書き方・cost 試算・troubleshooting は `radar` リポジトリの [`docs/user-guide.md` §triage workflow](https://github.com/ozzy-labs/feedradar/blob/main/docs/user-guide.md#triage-workflow) を参照。
+
 ## エージェント選択ガイド (cross-agent review)
 
 [ADR-0001](https://github.com/ozzy-labs/feedradar/blob/main/docs/adr/0001-agent-adapter-interface.md) に基づき、`research` と `review` は **別の agent** で実行することを推奨します:

package/dist/templates/workflows/combined-with-triage.template.yaml.tmpl

@@ -0,0 +1,133 @@
+# Generated by `radar workflow generate combined-with-triage` (ADR-0018 §W5).
+# Chains watch run -> LLM triage -> capped research --batch -> per-group
+# research --digest -> review --batch in one GitHub Actions job. The triage
+# layer slashes detection volume to research-worthy items only, so the
+# LoC-heavy research/review steps see ~5-15% of the raw detected count
+# instead of the full firehose. Template placeholders / hard-cap rationale
+# are documented in `src/cli/workflow/generate-combined-with-triage.ts`.
+
+name: feedradar-daily
+
+on:
+  schedule:
+    - cron: "{{watchCron}}"
+  workflow_dispatch: {}
+
+permissions:
+  contents: write
+  pull-requests: write
+
+concurrency:
+  # Distinct from feedradar-watch / feedradar-combined so this can co-exist
+  # with the simpler workflows on the same branch without serializing.
+  group: feedradar-daily-${{ github.ref }}
+  cancel-in-progress: false
+
+jobs:
+  daily:
+    runs-on: ubuntu-latest
+    timeout-minutes: 45
+    env:
+{{envBlock}}
+    steps:
+      - name: Checkout workspace
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+          persist-credentials: true
+
+      - name: Set up Node.js
+        uses: actions/setup-node@v4
+        with:
+          node-version: "22.21"
+
+      - name: Install FeedRadar
+        run: npm install -g @ozzylabs/feedradar
+
+      - name: Run watch (detect new items)
+        run: radar watch run
+
+      - name: Triage detected items (agent={{triageAgent}})
+        # ADR-0018 §W3: per-source `triagePolicy:` drives the cheap-model
+        # classification. Items without a policy are skipped and remain in
+        # `detected` so the existing untriaged-feeds workflow still picks
+        # them up.
+        run: radar triage --apply --triage-agent {{triageAgent}}
+
+      - name: Research triaged_research items (capped at {{maxItems}}, agent={{researchAgent}})
+        # `--batch --status triaged_research` consumes only items the triage
+        # step promoted, leaving `triaged_digest` for the digest step and
+        # `triaged_unsure` for the human-review queue. The hard cap protects
+        # against a triage misfire that promotes everything.
+        run: radar research --batch --status triaged_research --max-items {{maxItems}} --agent {{researchAgent}}
+
+      - name: Research digest groups (one report per triage.group)
+        # `radar items list --triage-group <g>` returns every item the triage
+        # adapter tagged with the same group hint, which `radar research
+        # --digest` then collapses into a single combined report. We walk
+        # the unique groups one at a time so each digest stays self-contained.
+        # `sort -u` dedupes since `items list --field` emits one row per
+        # matching item.
+        run: |
+          set -euo pipefail
+          GROUPS=$(radar items list --status triaged_digest --field triage.group | sort -u | grep -v '^-$' || true)
+          if [ -z "$GROUPS" ]; then
+            echo "no triaged_digest groups to process"
+            exit 0
+          fi
+          while IFS= read -r GROUP; do
+            [ -z "$GROUP" ] && continue
+            echo "::group::digest for triage.group=$GROUP"
+            IDS=$(radar items list --triage-group "$GROUP" --status triaged_digest --field id | tr '\n' ' ')
+            # shellcheck disable=SC2086
+            radar research --digest $IDS --agent {{researchAgent}}
+            echo "::endgroup::"
+          done <<< "$GROUPS"
+
+      - name: Review researched items (agent={{reviewAgent}})
+        # Cross-agent review (ADR-0001): the review step intentionally uses
+        # a different agent than the research step so the reviewer does not
+        # inherit the same model's blind spots.
+        run: radar review --batch --status researched --agent {{reviewAgent}}
+
+      - name: Notify unsure queue
+        # `if: always()` so a triage spike still alerts even when a
+        # downstream research step failed. The Slack notification is
+        # opt-in via `--slack-webhook secrets.<NAME>`; an empty webhook
+        # collapses the curl branch into a no-op so the step still succeeds
+        # without leaking the alert.
+        if: always()
+        env:
+          SLACK_WEBHOOK_URL: {{slackWebhookExpr}}
+        run: |
+          set -euo pipefail
+          UNSURE=$(radar items list --status triaged_unsure --json | jq length)
+          echo "triaged_unsure queue depth: $UNSURE"
+          if [ "$UNSURE" -gt 0 ] && [ -n "${SLACK_WEBHOOK_URL:-}" ]; then
+            curl -X POST "$SLACK_WEBHOOK_URL" \
+              -H 'Content-Type: application/json' \
+              -d "{\"text\":\"feedradar: $UNSURE items need human triage review\"}"
+          fi
+
+      - name: Create PR with research output
+        # `peter-evans/create-pull-request@v6` stages items/ state/ research/
+        # into a single PR per cron tick. Human reviews the PR before
+        # research/ lands on main, giving an explicit gate on auto-generated
+        # content (ADR-0014 §X5 / ADR-0018 §W5).
+        uses: peter-evans/create-pull-request@v6
+        with:
+          commit-message: "chore(feedradar): daily watch + triage + research"
+          # peter-evans/create-pull-request does not run shell on these
+          # fields, so `$(date ...)` would land literally in the PR title.
+          # Use the `github.run_id` expression to keep titles unique per run.
+          title: "feedradar: daily triage + research (run ${{ github.run_id }})"
+          body: |
+            Automated feedradar pipeline output. Review the research/ Markdown
+            before merging — generated content is untrusted (ADR-0009).
+          branch: feedradar/daily
+          base: ${{ github.ref_name }}
+          delete-branch: true
+          add-paths: |
+            items/
+            state/
+            research/

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ozzylabs/feedradar",
-  "version": "0.1.5",
+  "version": "0.1.7",
   "description": "Multi-agent CLI that watches blogs and release feeds, then turns keyword hits into Markdown research reports",
   "type": "module",
   "publishConfig": {