v0.20.1: free-quota budget — halve auxiliary call volume, keep the daily daily

.github/workflows/cloud-agent.yml

@@ -65,6 +65,8 @@ jobs:
       AI_GATEWAY_FALLBACK_MODELS: ${{ vars.AI_GATEWAY_FALLBACK_MODELS || 'google/gemini-2.5-flash-lite,openai/gpt-5-nano' }}
       AI_GATEWAY_SCREEN_FALLBACK_MODELS: ${{ vars.AI_GATEWAY_SCREEN_FALLBACK_MODELS || 'google/gemini-2.5-flash-lite,openai/gpt-oss-120b' }}
       AI_GATEWAY_MAX_OUTPUT_TOKENS: ${{ vars.AI_GATEWAY_MAX_OUTPUT_TOKENS || '32768' }}
+      # Free-tier quota mode: merge the 4 screening lanes into 2 calls.
+      SCREENING_SHARD_GROUPS: ${{ vars.SCREENING_SHARD_GROUPS || '2' }}
       # Tuned in-repo (repo Actions variables held stale lower values and
       # silently clamped collection breadth; see docs/release-v0.19.0.md).
       MAX_RELEASE_REPOS: '32'

scripts/agent_radar.py

@@ -58,7 +58,7 @@
 INIT_PROTECTED_DIRS = ("prompts", "automation", "docs")
 
 
-__version__ = "0.20.0"
+__version__ = "0.20.1"
 
 CORE_FILES = [
     "README.md",

scripts/cloud_agent_runner.py

@@ -852,10 +852,14 @@ def run_cli(root: Path, command: str, day: dt.date) -> None:
 
 
 def auto_tasks(day: dt.date) -> list[str]:
-    tasks = ["daily", "source-sweep"]
+    tasks = ["daily"]
+    # Free-quota budget (2026-07-22): auxiliary tasks produce slow-moving
+    # assets (research-log candidates, source discovery, promotions) and no
+    # longer run daily. The daily report keeps its daily cadence.
+    if day.weekday() in {0, 3}:  # Mon/Thu
+        tasks.append("source-sweep")
     if day.weekday() == 6:
         tasks.append("weekly")
-    if day.weekday() in {2, 6}:
         tasks.append("promote-candidates")
     # Mid-month refresh keeps the monthly from staying a day-1 seed until month end.
     if day.day == 15 or (day + dt.timedelta(days=1)).month != day.month:
@@ -2149,6 +2153,14 @@ def validate_daily_append_size(rel_path: str, mode: str, content: str) -> None:
 )
 
 
+# Pairwise merges applied when SCREENING_SHARD_GROUPS=2 (free-tier quota mode):
+# discussion keeps first position so merge dedup preserves community framing.
+SCREENING_SHARD_MERGE_2: tuple[tuple[str, tuple[str, ...]], ...] = (
+    ("discussion-official", ("discussion", "official-vendor")),
+    ("oss-packages", ("github-oss", "packages")),
+)
+
+
 def screening_shard_items(items: list[dict[str, str]]) -> list[tuple[str, list[dict[str, str]]]]:
     """Split scored items into per-lane-group shards for separate screening."""
     lane_to_shard: dict[str, str] = {}
@@ -2160,7 +2172,18 @@ def screening_shard_items(items: list[dict[str, str]]) -> list[tuple[str, list[d
         lane = source_lane(item.get("source", ""))
         shard_name = lane_to_shard.get(lane, "official-vendor")
         buckets[shard_name].append(item)
-    return [(name, buckets[name]) for name, _ in SCREENING_SHARD_LANES if buckets[name]]
+    shards = [(name, buckets[name]) for name, _ in SCREENING_SHARD_LANES if buckets[name]]
+    if env_int("SCREENING_SHARD_GROUPS", 4) <= 2:
+        merged: list[tuple[str, list[dict[str, str]]]] = []
+        by_name = dict(shards)
+        for group_name, members in SCREENING_SHARD_MERGE_2:
+            combined: list[dict[str, str]] = []
+            for member in members:
+                combined.extend(by_name.get(member, []))
+            if combined:
+                merged.append((group_name, combined))
+        return merged
+    return shards
 
 
 def candidate_dedupe_key(candidate: dict[str, Any]) -> str:

tests/test_cloud_agent_runner.py

@@ -69,16 +69,13 @@ def test_ai_gateway_default_fallbacks_are_tiered_by_workload(self) -> None:
     def test_auto_tasks_include_candidate_promotion_on_sunday(self) -> None:
         tasks = cloud_agent_runner.auto_tasks(cloud_agent_runner.parse_date("2026-07-05"))
         self.assertIn("daily", tasks)
-        self.assertIn("source-sweep", tasks)
         self.assertIn("weekly", tasks)
         self.assertIn("promote-candidates", tasks)
 
-    def test_auto_tasks_promote_twice_a_week(self) -> None:
+    def test_auto_tasks_weekday_is_daily_only(self) -> None:
+        # v0.20.1 free-quota schedule: a plain Wednesday runs only the daily.
         tasks = cloud_agent_runner.auto_tasks(cloud_agent_runner.parse_date("2026-07-01"))
-        self.assertIn("daily", tasks)
-        self.assertIn("source-sweep", tasks)
-        self.assertIn("promote-candidates", tasks)
-        self.assertNotIn("weekly", tasks)
+        self.assertEqual(tasks, ["daily"])
 
     def test_daily_can_update_source_registry(self) -> None:
         self.assertIn("sources.md", cloud_agent_runner.TASK_CONFIG["daily"]["allowed"])
@@ -2504,6 +2501,39 @@ def test_ai_gateway_accepts_fenced_json_without_fallback(self) -> None:
         self.assertEqual(json.loads(cloud_agent_runner.response_output_text(parsed))["summary"], "ok")
         urlopen_mock.assert_called_once()
 
+    def test_auto_tasks_free_quota_schedule(self) -> None:
+        # v0.20.1: daily every day; source-sweep Mon/Thu; promote Sun; weekly Sun.
+        mon = cloud_agent_runner.parse_date("2026-07-20")
+        thu = cloud_agent_runner.parse_date("2026-07-23")
+        sun = cloud_agent_runner.parse_date("2026-07-26")
+        tue = cloud_agent_runner.parse_date("2026-07-21")
+        self.assertIn("source-sweep", cloud_agent_runner.auto_tasks(mon))
+        self.assertIn("source-sweep", cloud_agent_runner.auto_tasks(thu))
+        self.assertNotIn("source-sweep", cloud_agent_runner.auto_tasks(tue))
+        sun_tasks = cloud_agent_runner.auto_tasks(sun)
+        self.assertIn("weekly", sun_tasks)
+        self.assertIn("promote-candidates", sun_tasks)
+        for d in (mon, tue, thu, sun):
+            self.assertIn("daily", cloud_agent_runner.auto_tasks(d))
+        mid = cloud_agent_runner.parse_date("2026-07-15")
+        self.assertIn("monthly", cloud_agent_runner.auto_tasks(mid))
+
+    def test_screening_shard_groups_merge_to_two(self) -> None:
+        items = [
+            {"source": "reddit-rss:LocalLLaMA", "title": "a", "url": "u1", "note": ""},
+            {"source": "openai-blog", "title": "b", "url": "u2", "note": ""},
+            {"source": "github", "title": "c", "url": "u3", "note": ""},
+            {"source": "npm", "title": "d", "url": "u4", "note": ""},
+        ]
+        with mock.patch.dict(os.environ, {"SCREENING_SHARD_GROUPS": "2"}, clear=False):
+            shards = cloud_agent_runner.screening_shard_items(items)
+        names = [name for name, _ in shards]
+        self.assertEqual(names, ["discussion-official", "oss-packages"])
+        self.assertEqual(sum(len(x) for _, x in shards), 4)
+        with mock.patch.dict(os.environ, {"SCREENING_SHARD_GROUPS": "4"}, clear=False):
+            shards4 = cloud_agent_runner.screening_shard_items(items)
+        self.assertGreaterEqual(len(shards4), 3)
+
     def test_ai_gateway_429_walks_rounds_with_backoff(self) -> None:
         # Issue #76: one pass over a same-pool chain died on free-tier 429s.
         # The loop must walk the chain for several rounds with real backoff.