Sprint 0+1: Package restructure, source tiers, verdicts, multi-variant editions

- New oracle/ package (11 modules) with unified CLI (python -m oracle)
- Source tiers: Tier 1 (arxiv/github/hf), Tier 2 (rss/hn), Tier 3 (reddit)
- Composite verdicts: PUBLISH/WATCH/ARCHIVE/DROP based on signal score + age
- Content-hash dedup: SHA-256[:16] normalized, atomic at insert time
- Multi-variant editions: 4 YAML configs (default/research/devops/brief)
- Variant engine: filter → rank → render (HTML + JSON, themed)
- Per-adapter timeout (10s) + threading fallback
- Consolidated 12 root scripts → thin wrappers + oracle/ package
- Archived stale scripts (_engagement, _live_compare, reddit_proof)
- Updated .gitignore, README.md, schema.sql

oracle/config.py

@@ -0,0 +1,68 @@
+"""Centralized configuration for the AI Research Oracle.
+
+Single source of truth for all paths, defaults, and constants.
+"""
+import os
+from pathlib import Path
+
+# ── Paths ──────────────────────────────────────────────────────────────────
+ROOT = Path(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+DB_PATH = ROOT / "oracle.db"
+SCHEMA_PATH = ROOT / "schema.sql"
+
+# ── Web output ─────────────────────────────────────────────────────────────
+WEBROOT = "/var/www/preprod2"
+FALLBACK_WEBROOT = ROOT / "site"
+SEEN_JSON = ROOT.parent / "ai-oracle-site" / "seen_urls.json"
+
+# ── Pipeline defaults ──────────────────────────────────────────────────────
+ENABLED_SOURCES = ["github", "arxiv", "reddit", "hackernews", "huggingface", "rss"]
+DEFAULT_LIMIT = 20
+
+# ── Source tiers (World Monitor pattern) ───────────────────────────────────
+# Tier 1: Primary trusted sources (official releases, peer-reviewed)
+# Tier 2: Secondary credible sources (curated communities, major outlets)
+# Tier 3: Tertiary noise sources (user-generated, unverified)
+SOURCE_TIERS = {
+    "arxiv": {"tier": 1, "label": "PRIMARY", "description": "Peer-reviewed research"},
+    "github": {"tier": 1, "label": "PRIMARY", "description": "Official code releases"},
+    "huggingface": {"tier": 1, "label": "PRIMARY", "description": "Model registry"},
+    "rss": {"tier": 2, "label": "SECONDARY", "description": "Curated tech media"},
+    "hackernews": {"tier": 2, "label": "SECONDARY", "description": "Curated community"},
+    "reddit": {"tier": 3, "label": "TERTIARY", "description": "User-generated discussion"},
+}
+
+# Tier-based signal score bonus/penalty (applied to final_score)
+TIER_BONUS = {1: 0.05, 2: 0.0, 3: -0.05}
+
+# Freshness SLA per tier (hours after which a source is flagged stale)
+FRESHNESS_SLA_H = {1: 48, 2: 24, 3: 12}
+
+# ── Composite verdict thresholds ───────────────────────────────────────────
+# PUBLISH: High score + fresh, goes to top
+# WATCH:   Medium score, monitor for follow-ups
+# ARCHIVE: Low score or aged out, move to archive
+# DROP:    Junk score, ignore
+VERDICT_THRESHOLDS = {
+    "PUBLISH": {"min_score": 6.0, "max_age_h": 48},
+    "WATCH":   {"min_score": 4.0, "max_age_h": 168},   # 7 days
+    "ARCHIVE": {"min_score": 2.0, "max_age_h": 720},    # 30 days
+    "DROP":    {"min_score": 0.0, "max_age_h": 999999}, # catch-all
+}
+
+# ── Content hash dedup ─────────────────────────────────────────────────────
+CONTENT_HASH_PREFIX = "sha256"
+HASH_LENGTH = 16  # characters
+
+# ── Clickability weights ───────────────────────────────────────────────────
+VEL_W = 0.50
+ENG_W = 0.50
+SIG_W = 0.0
+HALF_LIFE_H = 18.0
+
+# ── Render ─────────────────────────────────────────────────────────────────
+TOP_N = 8
+
+# ── Archive ────────────────────────────────────────────────────────────────
+ARCHIVE_DAYS = 30
+ARCHIVE_CAP = 5000