@m13v/s4l 1.6.197-rc.10

package/skill/run-github.sh

@@ -0,0 +1,102 @@
+#!/bin/bash
+# Social Autoposter - GitHub Issues momentum-gated posting.
+#
+# Delegates to scripts/post_github.py which implements:
+#   - Phase 1: gh search across project topics, snapshot T0 comment + reaction counts
+#   - Sleep (T0 -> T1 momentum window, default 600s, owned by post_github --sleep)
+#   - Phase 2a: re-fetch same issues, compute delta_score
+#   - Phase 2b: adaptive cap (default 1, bump to 3 when >=3 candidates show momentum),
+#              Claude drafts (one-shot, no Bash tools), Python posts via `gh issue comment`
+#              and persists search_topic, language, engagement_style to the posts table.
+#
+# Reduction levers baked in:
+#   (1) Historical (project, style) engagement block in drafter prompt.
+#   (2) top_search_topics feedback so high-scoring seeds get preferred.
+#   (3) Adaptive cap gated by per-cycle momentum.
+#   (4) T0 -> T1 delta filter: stale issues drop out before Claude sees them.
+#   (5) Pre-filter eliminates Claude's tool budget; one JSON in one shot.
+#
+# Called by launchd. Cadence is owned by the .plist, not this script.
+
+set -euo pipefail
+
+[ -f "$HOME/social-autoposter/.env" ] && source "$HOME/social-autoposter/.env"
+
+REPO_DIR="$HOME/social-autoposter"
+LOG_DIR="$REPO_DIR/skill/logs"
+mkdir -p "$LOG_DIR"
+LOG_FILE="$LOG_DIR/github-$(date +%Y-%m-%d_%H%M%S).log"
+RUN_START=$(date +%s)
+
+# Per-cycle batch id stamped onto every claude_sessions row spawned by this
+# run (via SA_CYCLE_ID env -> log_claude_session.py). Lets the dashboard /
+# get_run_cost.py --cycle-id report exact per-cycle cost instead of the
+# legacy script+since query which bleeds costs across concurrent runs.
+# 2026-05-10 cycle_id rollout.
+BATCH_ID="ghcycle-$(date +%Y%m%d-%H%M%S)-$$"
+export SA_CYCLE_ID="$BATCH_ID"
+
+# Idempotent run_monitor.log emitter wired to EXIT/INT/TERM/HUP. Without this,
+# a SIGTERM landing during the post_github.py loop (after `gh issue comment`
+# committed a comment but before the script's own log_run.py call at the
+# bottom of main()) silently drops the run from run_monitor.log. The
+# dashboard reads run_monitor.log, so the operator-visible "last post_github
+# cycle" stays stuck on a stale entry while real comments continue landing
+# unrecorded. Mirrors the same fix shipped to run-reddit-search.sh,
+# run-twitter-cycle.sh, and run-linkedin.sh.
+#
+# Detection mechanism: post_github.py writes a "=== SUMMARY: elapsed=" line
+# to stdout (teed into $LOG_FILE) IMMEDIATELY before its own log_run.py
+# call. Presence of that marker in $LOG_FILE means Python wrote its own
+# summary, so the trap no-ops. Absence means SIGTERM landed before Python
+# could finish, and the trap emits a "sigterm:1" placeholder so the cycle
+# is at least visible on the dashboard.
+#
+# We can't re-derive accurate POSTED counts from the DB at trap-time the way
+# linkedin/twitter do — github posts go through `gh issue comment` and only
+# get a posts-table row from inside post_github.py's log_post call, so a
+# SIGTERM mid-loop may leave 1-2 comments live on github.com with no posts
+# row yet. The placeholder marks the cycle as failed=1 + sigterm:1 so the
+# operator notices; manual reconciliation if the count matters.
+_SA_RUN_SUMMARY_EMITTED=0
+_sa_emit_run_summary_oneshot() {
+    [ "${_SA_RUN_SUMMARY_EMITTED:-0}" = "1" ] && return 0
+    _SA_RUN_SUMMARY_EMITTED=1
+    # Python wrote its own summary? Trust it.
+    if [ -n "${LOG_FILE:-}" ] && [ -f "${LOG_FILE:-}" ] \
+        && grep -q "=== SUMMARY: elapsed=" "$LOG_FILE" 2>/dev/null; then
+        return 0
+    fi
+    local elapsed cost
+    elapsed=$(( $(date +%s) - ${RUN_START:-$(date +%s)} ))
+    cost=$(timeout 10 python3 "$REPO_DIR/scripts/get_run_cost.py" \
+                --since "${RUN_START:-0}" \
+                --scripts "post_github" \
+                2>/dev/null || echo "0.0000")
+    # Classify the most likely cause from the cycle log before falling back to
+    # the generic "sigterm:1" placeholder. Picks up Anthropic-side failures
+    # (stream_idle_timeout, monthly_limit, api_overloaded, context_overflow,
+    # credit_balance) that would otherwise read as a silent sigterm row on the
+    # dashboard. Falls back to sigterm:1 when no API marker is present (true
+    # external SIGTERM, e.g. cycle budget exceeded by watchdog_hung_runs).
+    local trap_reason
+    trap_reason=""
+    if [ -n "${LOG_FILE:-}" ] && [ -f "${LOG_FILE:-}" ]; then
+        trap_reason=$(python3 "$REPO_DIR/scripts/classify_run_error.py" "$LOG_FILE" 2>/dev/null)
+    fi
+    [ -z "$trap_reason" ] && trap_reason="sigterm"
+    python3 "$REPO_DIR/scripts/log_run.py" \
+        --script post_github \
+        --posted 0 --skipped 0 --failed 1 \
+        --cost "$cost" --elapsed "$elapsed" \
+        --failure-reasons "${trap_reason}:1" 2>/dev/null || true
+}
+trap _sa_emit_run_summary_oneshot EXIT INT TERM HUP
+
+echo "=== GitHub Issues Run: $(date) ===" | tee "$LOG_FILE"
+
+python3 "$REPO_DIR/scripts/post_github.py" 2>&1 | tee -a "$LOG_FILE"
+
+echo "=== Run complete: $(date) ===" | tee -a "$LOG_FILE"
+
+find "$LOG_DIR" -name "github-*.log" -mtime +7 -delete 2>/dev/null || true

package/skill/run-instagram-daily.sh

@@ -0,0 +1,149 @@
+#!/bin/bash
+# run-instagram-daily.sh — Pick + post one Instagram Reel per fire.
+#
+# Cadence (set by launchd, com.m13v.social-instagram-daily.plist):
+#   5 fires/day at 09:00, 12:00, 15:00, 18:00, 21:00 local time.
+#
+# Per-fire logic:
+#   1. acquire_lock instagram-poster (file lock; IG is HTTP-only, no browser)
+#   2. pick_ig_account.py -> chooses target IG account (inverse-recent-share
+#      over enabled accounts in config.json instagram.accounts).
+#   3. ig_post_type_picker.py --account <name> -> JSON {post_type, video_path, ...}
+#      scoped to the chosen account's draft pool.
+#   4. call mixer/post_to_ig.py --file <path> --post-type <type> --account <name>
+#      which: uploads to GCS, posts via IG Graph API, writes posted.json,
+#      updates media_posts (status=posted, post_type + target_account coalesced).
+#
+# Exit behavior:
+#   0  - posted, OR queue exhausted (logged), OR another run holds the lock
+#   1  - real failure (DB error, IG API error, picker crash, etc.)
+#
+# Logs: skill/logs/instagram-daily-YYYY-MM-DD_HHMMSS.log
+
+set -uo pipefail
+
+REPO_DIR="$HOME/social-autoposter"
+LOG_DIR="$REPO_DIR/skill/logs"
+mkdir -p "$LOG_DIR"
+
+LOG_FILE="$LOG_DIR/instagram-daily-$(date +%Y-%m-%d_%H%M%S).log"
+PICK_FILE="/tmp/ig_pick_$(date +%s)_$$.json"
+
+if [ -f "$REPO_DIR/.env" ]; then
+  set -a
+  # shellcheck disable=SC1091
+  source "$REPO_DIR/.env"
+  set +a
+fi
+
+log() { echo "[$(date +%H:%M:%S)] $*" | tee -a "$LOG_FILE"; }
+
+# Run accounting for dashboard Job History (Post Threads · Instagram).
+# Each exit site updates POSTED_CT / SKIPPED_CT / FAILED_CT; the EXIT trap
+# always emits one log_run.py line so the run shows up under
+# thread_instagram, matching how thread_twitter / thread_reddit log.
+RUN_START_EPOCH=$(date +%s)
+POSTED_CT=0
+SKIPPED_CT=0
+FAILED_CT=0
+
+cleanup() {
+  local rc=$?
+  rm -f "$PICK_FILE"
+  if [ "$POSTED_CT" -eq 0 ] && [ "$SKIPPED_CT" -eq 0 ] && [ "$FAILED_CT" -eq 0 ]; then
+    if [ "$rc" -eq 0 ]; then SKIPPED_CT=1; else FAILED_CT=1; fi
+  fi
+  local elapsed=$(( $(date +%s) - RUN_START_EPOCH ))
+  local cost
+  cost=$(/usr/bin/python3 "$REPO_DIR/scripts/get_run_cost.py" --since "$RUN_START_EPOCH" --scripts "run-instagram-daily" 2>/dev/null || echo "0.0000")
+  /usr/bin/python3 "$REPO_DIR/scripts/log_run.py" \
+      --script "thread_instagram" \
+      --posted "$POSTED_CT" --skipped "$SKIPPED_CT" --failed "$FAILED_CT" \
+      --cost "$cost" --elapsed "$elapsed" >/dev/null 2>&1 || true
+}
+trap cleanup EXIT INT TERM HUP
+
+log "=== instagram-daily fire: $(date) ==="
+
+# Lock against overlapping fires (pure HTTP pipeline, no browser, but the
+# picker reads-then-posts so two concurrent fires could pick the same row).
+# 30s timeout — if a prior fire is still uploading, skip cleanly.
+# shellcheck source=lock.sh
+source "$REPO_DIR/skill/lock.sh"
+acquire_lock instagram-poster 30
+
+# Step 1: pick target IG account. Per-account plists set FORCE_ACCOUNT in
+# their EnvironmentVariables block so the slot is hard-pinned to one account;
+# fall back to inverse-recent-share picker only when no FORCE_ACCOUNT is set
+# (manual / legacy invocation).
+if [ -n "${FORCE_ACCOUNT:-}" ]; then
+  log "step 1: FORCE_ACCOUNT=$FORCE_ACCOUNT honored from env"
+  TARGET_ACCOUNT=$(/opt/homebrew/bin/python3.11 "$REPO_DIR/scripts/pick_ig_account.py" --account "$FORCE_ACCOUNT" 2>>"$LOG_FILE")
+else
+  log "step 1: pick_ig_account (no FORCE_ACCOUNT in env)"
+  TARGET_ACCOUNT=$(/opt/homebrew/bin/python3.11 "$REPO_DIR/scripts/pick_ig_account.py" 2>>"$LOG_FILE")
+fi
+if [ -z "$TARGET_ACCOUNT" ]; then
+  log "pick_ig_account.py produced no account — exiting non-zero"
+  FAILED_CT=1
+  exit 1
+fi
+log "picker chose account: $TARGET_ACCOUNT"
+
+# Step 2: pick post type + video, scoped to that account
+log "step 2: ig_post_type_picker --account $TARGET_ACCOUNT"
+if ! /opt/homebrew/bin/python3.11 "$REPO_DIR/scripts/ig_post_type_picker.py" \
+        --account "$TARGET_ACCOUNT" > "$PICK_FILE" 2>>"$LOG_FILE"; then
+  rc=$?
+  if [ "$rc" -eq 2 ]; then
+    log "queue exhausted for account=$TARGET_ACCOUNT (no drafts of either type) — exiting cleanly"
+    SKIPPED_CT=1
+    exit 0
+  fi
+  log "picker failed rc=$rc — exiting non-zero"
+  FAILED_CT=1
+  exit 1
+fi
+
+POST_TYPE=$(/opt/homebrew/bin/python3.11 -c "import json; print(json.load(open('$PICK_FILE'))['post_type'])")
+VIDEO_PATH=$(/opt/homebrew/bin/python3.11 -c "import json; print(json.load(open('$PICK_FILE'))['video_path'])")
+POST_NUMBER=$(/opt/homebrew/bin/python3.11 -c "import json; print(json.load(open('$PICK_FILE'))['post_number'])")
+REASON=$(/opt/homebrew/bin/python3.11 -c "import json; print(json.load(open('$PICK_FILE'))['reason'])")
+FALLBACK=$(/opt/homebrew/bin/python3.11 -c "import json; print(json.load(open('$PICK_FILE'))['fallback'])")
+
+log "picker chose: account=${TARGET_ACCOUNT} post-${POST_NUMBER} type=${POST_TYPE} fallback=${FALLBACK}"
+log "picker reason: ${REASON}"
+
+if [ ! -f "$VIDEO_PATH" ]; then
+  log "ERROR: picker pointed at $VIDEO_PATH but file missing on disk"
+  FAILED_CT=1
+  exit 1
+fi
+
+# Step 2: post
+DRY_FLAG=""
+if [ "${IG_DRY_RUN:-0}" = "1" ]; then
+  DRY_FLAG="--dry-run"
+  log "IG_DRY_RUN=1 — passing --dry-run to post_to_ig.py"
+fi
+
+log "step 3: post_to_ig.py --file $(basename "$VIDEO_PATH") --post-type $POST_TYPE --account $TARGET_ACCOUNT $DRY_FLAG"
+if ! /opt/homebrew/bin/python3.11 "$REPO_DIR/mixer/post_to_ig.py" \
+        --file "$VIDEO_PATH" --post-type "$POST_TYPE" --account "$TARGET_ACCOUNT" $DRY_FLAG >>"$LOG_FILE" 2>&1; then
+  log "post_to_ig.py failed — exiting non-zero"
+  FAILED_CT=1
+  exit 1
+fi
+
+POSTED_CT=1
+log "=== finished post-${POST_NUMBER} (${POST_TYPE}) on ${TARGET_ACCOUNT} successfully ==="
+
+# Step 4: mirror the new media_posts row into the cross-platform `posts` table
+# so it surfaces in the dashboard (Trends, Top, Activity, Cohort, Stats by
+# Style) alongside Reddit/Twitter/LinkedIn rows. Idempotent.
+log "step 4: sync_ig_to_posts"
+if ! /opt/homebrew/bin/python3.11 "$REPO_DIR/scripts/sync_ig_to_posts.py" --quiet >>"$LOG_FILE" 2>&1; then
+    log "sync_ig_to_posts failed (post is already on IG; will retry on next stats fire)"
+fi
+
+exit 0