disinto/supervisor/supervisor-run.sh

334 lines
15 KiB
Bash
Raw Normal View History

#!/usr/bin/env bash
# =============================================================================
# supervisor-run.sh — Polling-loop wrapper: supervisor execution via SDK + formula
#
# Synchronous bash loop using claude -p (one-shot invocation).
# No tmux sessions, no phase files — the bash script IS the state machine.
#
# Flow:
# 1. Guards: run lock, memory check
# 2. Housekeeping: clean up stale crashed worktrees
# 3. Collect pre-flight metrics (supervisor/preflight.sh)
# 4. Load formula (formulas/run-supervisor.toml)
# 5. Context: AGENTS.md, preflight metrics, structural graph
# 6. agent_run(worktree, prompt) → Claude monitors, may clean up
#
# Usage:
# supervisor-run.sh [projects/disinto.toml] # project config (default: disinto)
#
# Called by: entrypoint.sh polling loop (every 20 minutes)
# =============================================================================
set -euo pipefail
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
FACTORY_ROOT="$(dirname "$SCRIPT_DIR")"
# Accept project config from argument; default to disinto
export PROJECT_TOML="${1:-$FACTORY_ROOT/projects/disinto.toml}"
# Set override BEFORE sourcing env.sh so it survives any later re-source of
# env.sh from nested shells / claude -p tools (#762, #747)
export FORGE_TOKEN_OVERRIDE="${FORGE_SUPERVISOR_TOKEN:-}"
# shellcheck source=../lib/env.sh
source "$FACTORY_ROOT/lib/env.sh"
# shellcheck source=../lib/formula-session.sh
source "$FACTORY_ROOT/lib/formula-session.sh"
# shellcheck source=../lib/worktree.sh
source "$FACTORY_ROOT/lib/worktree.sh"
# shellcheck source=../lib/guard.sh
source "$FACTORY_ROOT/lib/guard.sh"
# shellcheck source=../lib/agent-sdk.sh
source "$FACTORY_ROOT/lib/agent-sdk.sh"
LOG_FILE="${DISINTO_LOG_DIR}/supervisor/supervisor.log"
# shellcheck disable=SC2034 # consumed by agent-sdk.sh
LOGFILE="$LOG_FILE"
# shellcheck disable=SC2034 # consumed by agent-sdk.sh
SID_FILE="/tmp/supervisor-session-${PROJECT_NAME}.sid"
SCRATCH_FILE="/tmp/supervisor-${PROJECT_NAME}-scratch.md"
WORKTREE="/tmp/${PROJECT_NAME}-supervisor-run"
# WP agent container name (configurable via env var)
export WP_AGENT_CONTAINER_NAME="${WP_AGENT_CONTAINER_NAME:-disinto-woodpecker-agent}"
# Override LOG_AGENT for consistent agent identification
# shellcheck disable=SC2034 # consumed by agent-sdk.sh and env.sh log()
LOG_AGENT="supervisor"
# ── OPS Repo Detection (Issue #544) ──────────────────────────────────────
# Detect if OPS_REPO_ROOT is available and set degraded mode flag if not.
# This allows the supervisor to run with fallback knowledge files and
# local journal/vault paths when the ops repo is absent.
if [ -z "${OPS_REPO_ROOT:-}" ] || [ ! -d "${OPS_REPO_ROOT}" ]; then
log "WARNING: OPS_REPO_ROOT not set or directory missing — running in degraded mode (no playbooks, no journal continuity, no vault destination)"
export OPS_REPO_DEGRADED=1
# Set fallback paths for degraded mode
export OPS_KNOWLEDGE_ROOT="${FACTORY_ROOT}/knowledge"
export OPS_JOURNAL_ROOT="${FACTORY_ROOT}/state/supervisor-journal"
export OPS_VAULT_ROOT="${PROJECT_REPO_ROOT}/vault/pending"
mkdir -p "$OPS_JOURNAL_ROOT" "$OPS_VAULT_ROOT" 2>/dev/null || true
else
export OPS_REPO_DEGRADED=0
export OPS_KNOWLEDGE_ROOT="${OPS_REPO_ROOT}/knowledge"
export OPS_JOURNAL_ROOT="${OPS_REPO_ROOT}/journal/supervisor"
export OPS_VAULT_ROOT="${OPS_REPO_ROOT}/vault/pending"
mkdir -p "$OPS_JOURNAL_ROOT" "$OPS_VAULT_ROOT" 2>/dev/null || true
fi
# Override log() to append to supervisor-specific log file
# shellcheck disable=SC2034
log() {
local agent="${LOG_AGENT:-supervisor}"
printf '[%s] %s: %s\n' "$(date -u '+%Y-%m-%dT%H:%M:%SZ')" "$agent" "$*" >> "$LOG_FILE"
}
# ── Guards ────────────────────────────────────────────────────────────────
check_active supervisor
acquire_run_lock "/tmp/supervisor-run.lock"
memory_guard 2000
log "--- Supervisor run start ---"
# ── Resolve forge remote for git operations ─────────────────────────────
# Run git operations from the project checkout, not the baked code dir
cd "$PROJECT_REPO_ROOT"
# ── Housekeeping: clean up stale crashed worktrees (>24h) ────────────────
cleanup_stale_crashed_worktrees 24
# ── Resolve agent identity for .profile repo ────────────────────────────
resolve_agent_identity || true
# ── Collect pre-flight metrics ────────────────────────────────────────────
log "Running preflight.sh"
PREFLIGHT_OUTPUT=""
PREFLIGHT_RC=0
if PREFLIGHT_OUTPUT=$(bash "$SCRIPT_DIR/preflight.sh" "$PROJECT_TOML" 2>&1); then
log "Preflight collected ($(echo "$PREFLIGHT_OUTPUT" | wc -l) lines)"
else
PREFLIGHT_RC=$?
log "WARNING: preflight.sh failed (exit code $PREFLIGHT_RC), continuing with partial data"
if [ -n "$PREFLIGHT_OUTPUT" ]; then
log "Preflight error: $(echo "$PREFLIGHT_OUTPUT" | tail -3)"
fi
fi
# ── Load formula + context ───────────────────────────────────────────────
load_formula_or_profile "supervisor" "$FACTORY_ROOT/formulas/run-supervisor.toml" || exit 1
build_context_block AGENTS.md
# ── Prepare .profile context (lessons injection) ─────────────────────────
formula_prepare_profile_context
# ── Read scratch file (compaction survival) ───────────────────────────────
SCRATCH_CONTEXT=$(read_scratch_context "$SCRATCH_FILE")
SCRATCH_INSTRUCTION=$(build_scratch_instruction "$SCRATCH_FILE")
# ── Build prompt ─────────────────────────────────────────────────────────
build_sdk_prompt_footer
export CLAUDE_MODEL="sonnet"
# ── Create worktree (before prompt assembly so trap is set early) ────────
formula_worktree_setup "$WORKTREE"
# Inject OPS repo status into prompt
if [ "${OPS_REPO_DEGRADED:-0}" = "1" ]; then
OPS_STATUS="
## OPS Repo Status
**DEGRADED MODE**: OPS repo is not available. Using bundled knowledge files and local journal/vault paths.
- Knowledge files: ${OPS_KNOWLEDGE_ROOT:-<unset>}
- Journal: ${OPS_JOURNAL_ROOT:-<unset>}
- Vault destination: ${OPS_VAULT_ROOT:-<unset>}
"
else
OPS_STATUS="
## OPS Repo Status
**FULL MODE**: OPS repo available at ${OPS_REPO_ROOT}
- Knowledge files: ${OPS_KNOWLEDGE_ROOT:-<unset>}
- Journal: ${OPS_JOURNAL_ROOT:-<unset>}
- Vault destination: ${OPS_VAULT_ROOT:-<unset>}
"
fi
PROMPT="You are the supervisor agent for ${FORGE_REPO}. Work through the formula below.
You have full shell access and --dangerously-skip-permissions.
Fix what you can. File vault items for what you cannot. Do NOT ask permission — act first, report after.
## Pre-flight metrics (collected $(date -u +%H:%M) UTC)
${PREFLIGHT_OUTPUT}
## Project context
${CONTEXT_BLOCK}$(formula_lessons_block)
${SCRATCH_CONTEXT:+${SCRATCH_CONTEXT}
}
${OPS_STATUS}
Priority order: P0 memory > P1 disk > P2 stopped > P3 degraded > P4 housekeeping
${FORMULA_CONTENT}
${SCRATCH_INSTRUCTION}
${PROMPT_FOOTER}"
# ── WP Agent Health Recovery ──────────────────────────────────────────────
# Check preflight output for WP agent health issues and trigger recovery if needed
_WP_HEALTH_CHECK_FILE="${DISINTO_LOG_DIR}/supervisor/wp-agent-health-check.md"
echo "$PREFLIGHT_OUTPUT" > "$_WP_HEALTH_CHECK_FILE"
# Extract WP agent health status from preflight output
_wp_agent_healthy=$(grep "^WP Agent Health:" "$_WP_HEALTH_CHECK_FILE" 2>/dev/null | grep -q "healthy" && echo "true" || echo "false")
_wp_health_reason=$(grep "^Reason:" "$_WP_HEALTH_CHECK_FILE" 2>/dev/null | sed 's/^Reason: //' || echo "")
if [ "$_wp_agent_healthy" = "false" ] && [ -n "$_wp_health_reason" ]; then
log "WP agent detected as UNHEALTHY: $_wp_health_reason"
# Check for idempotency guard - have we already restarted in this run?
_WP_HEALTH_HISTORY_FILE="${DISINTO_LOG_DIR}/supervisor/wp-agent-health.history"
_wp_last_restart_ts=0
_wp_last_restart="never"
if [ -f "$_WP_HEALTH_HISTORY_FILE" ]; then
_wp_last_restart_ts=$(grep -m1 '^LAST_RESTART_TS=' "$_WP_HEALTH_HISTORY_FILE" 2>/dev/null | cut -d= -f2 || echo "0")
if [ -n "$_wp_last_restart_ts" ] && [ "$_wp_last_restart_ts" != "0" ] 2>/dev/null; then
_wp_last_restart=$(date -d "@$_wp_last_restart_ts" '+%Y-%m-%d %H:%M UTC' 2>/dev/null || echo "$_wp_last_restart_ts")
fi
fi
_current_ts=$(date +%s)
_restart_threshold=300 # 5 minutes between restarts
if [ -z "$_wp_last_restart_ts" ] || [ "$_wp_last_restart_ts" = "0" ] || [ $((_current_ts - _wp_last_restart_ts)) -gt $_restart_threshold ]; then
log "Triggering WP agent restart..."
# Restart the WP agent container
if docker restart "$WP_AGENT_CONTAINER_NAME" >/dev/null 2>&1; then
_restart_time=$(date -u '+%Y-%m-%d %H:%M UTC')
log "Successfully restarted WP agent container: $_wp_agent_healthy"
# Update history file
echo "LAST_RESTART_TS=$_current_ts" > "$_WP_HEALTH_HISTORY_FILE"
echo "LAST_RESTART_TIME=$_restart_time" >> "$_WP_HEALTH_HISTORY_FILE"
# Post recovery notice to journal
_journal_file="${OPS_JOURNAL_ROOT}/$(date -u +%Y-%m-%d).md"
if [ -f "$_journal_file" ]; then
{
echo ""
echo "### WP Agent Recovery - $_restart_time"
echo ""
echo "WP agent was unhealthy: $_wp_health_reason"
echo "Container restarted automatically."
} >> "$_journal_file"
fi
# Scan for issues updated in the last 30 minutes with blocked: ci_exhausted label
log "Scanning for ci_exhausted issues updated in last 30 minutes..."
_now_epoch=$(date +%s)
_thirty_min_ago=$(( _now_epoch - 1800 ))
# Fetch open issues with blocked label
_blocked_issues=$(forge_api GET "/issues?state=open&labels=blocked&type=issues&limit=100" 2>/dev/null || echo "[]")
_blocked_count=$(echo "$_blocked_issues" | jq 'length' 2>/dev/null || echo "0")
_issues_processed=0
_issues_recovered=0
if [ "$_blocked_count" -gt 0 ]; then
# Process each blocked issue
echo "$_blocked_issues" | jq -c '.[]' 2>/dev/null | while IFS= read -r issue_json; do
[ -z "$issue_json" ] && continue
_issue_num=$(echo "$issue_json" | jq -r '.number // empty')
_issue_updated=$(echo "$issue_json" | jq -r '.updated_at // empty')
_issue_labels=$(echo "$issue_json" | jq -r '.labels | map(.name) | join(",")' 2>/dev/null || echo "")
# Check if issue has ci_exhausted label
if ! echo "$_issue_labels" | grep -q "ci_exhausted"; then
continue
fi
# Parse updated_at timestamp
_issue_updated_epoch=$(date -d "$_issue_updated" +%s 2>/dev/null || echo "0")
_time_since_update=$(( _now_epoch - _issue_updated_epoch ))
# Check if updated in last 30 minutes
if [ "$_time_since_update" -lt 1800 ] && [ "$_time_since_update" -ge 0 ]; then
_issues_processed=$(( _issues_processed + 1 ))
# Check for idempotency guard - already swept by supervisor?
_issue_body=$(echo "$issue_json" | jq -r '.body // ""' 2>/dev/null || echo "")
if echo "$_issue_body" | grep -q "<!-- supervisor-swept -->"; then
log "Issue #$_issue_num already swept by supervisor, skipping"
continue
fi
log "Processing ci_exhausted issue #$_issue_num (updated $_time_since_update seconds ago)"
# Get issue assignee
_issue_assignee=$(echo "$issue_json" | jq -r '.assignee.login // empty' 2>/dev/null || echo "")
# Unassign the issue
if [ -n "$_issue_assignee" ]; then
log "Unassigning issue #$_issue_num from $_issue_assignee"
curl -sf -X PATCH \
-H "Authorization: token ${FORGE_SUPERVISOR_TOKEN:-$FORGE_TOKEN}" \
-H "Content-Type: application/json" \
"${FORGE_API}/issues/$_issue_num" \
-d '{"assignees":[]}' >/dev/null 2>&1 || true
fi
# Remove blocked label
_blocked_label_id=$(forge_api GET "/labels" 2>/dev/null | jq -r '.[] | select(.name == "blocked") | .id' 2>/dev/null || echo "")
if [ -n "$_blocked_label_id" ]; then
log "Removing blocked label from issue #$_issue_num"
curl -sf -X DELETE \
-H "Authorization: token ${FORGE_SUPERVISOR_TOKEN:-$FORGE_TOKEN}" \
"${FORGE_API}/issues/$_issue_num/labels/$_blocked_label_id" >/dev/null 2>&1 || true
fi
# Add comment about infra-flake recovery
_recovery_comment=$(cat <<EOF
<!-- supervisor-swept -->
**Automated Recovery — $(date -u '+%Y-%m-%d %H:%M UTC')**
CI agent was unhealthy between $_restart_time and now. The prior retry budget may have been spent on infra flake, not real failures.
**Recovery Actions:**
- Unassigned from pool and returned for fresh attempt
- CI agent container restarted
- Related pipelines will be retriggered automatically
**Next Steps:**
Please re-attempt this issue. The CI environment has been refreshed.
EOF
)
curl -sf -X POST \
-H "Authorization: token ${FORGE_SUPERVISOR_TOKEN:-$FORGE_TOKEN}" \
-H "Content-Type: application/json" \
"${FORGE_API}/issues/$_issue_num/comments" \
-d "{\"body\":$_recovery_comment}" >/dev/null 2>&1 || true
log "Recovered issue #$_issue_num - returned to pool"
fi
done
fi
log "WP agent restart and issue recovery complete"
else
log "ERROR: Failed to restart WP agent container"
fi
else
log "WP agent restart already performed in this run (since $_wp_last_restart), skipping"
fi
fi
# ── Run agent ─────────────────────────────────────────────────────────────
agent_run --worktree "$WORKTREE" "$PROMPT"
log "agent_run complete"
# Write journal entry post-session
profile_write_journal "supervisor-run" "Supervisor run $(date -u +%Y-%m-%d)" "complete" "" || true
rm -f "$SCRATCH_FILE"
log "--- Supervisor run done ---"