eagle-mem 3.0.1 → 3.1.0

package/lib/hooks-posttool.sh

@@ -0,0 +1,138 @@
+#!/usr/bin/env bash
+# ═══════════════════════════════════════════════════════════
+# Eagle Mem — PostToolUse extracted responsibilities
+# Source from hooks/post-tool-use.sh after common.sh + db.sh
+# ═══════════════════════════════════════════════════════════
+[ -n "${_EAGLE_HOOKS_POSTTOOL_LOADED:-}" ] && return 0
+_EAGLE_HOOKS_POSTTOOL_LOADED=1
+
+eagle_posttool_mirror_writes() {
+    local tool_name="$1" fp="$2" session_id="$3" project="$4"
+
+    case "$tool_name" in
+        Write|Edit)
+            if [ -n "$fp" ]; then
+                case "$fp" in
+                    *..*) ;; # path traversal — skip
+                    "$EAGLE_CLAUDE_PROJECTS_DIR"/*/memory/*.md)
+                        local mem_base
+                        mem_base=$(basename "$fp")
+                        if [ "$mem_base" != "MEMORY.md" ] && [ -f "$fp" ]; then
+                            eagle_capture_claude_memory "$fp" "$session_id" "$project"
+                        fi
+                        ;;
+                    "$EAGLE_CLAUDE_PLANS_DIR/"*.md)
+                        if [ -f "$fp" ]; then
+                            eagle_capture_claude_plan "$fp" "$session_id" "$project"
+                        fi
+                        ;;
+                esac
+            fi
+            ;;
+    esac
+}
+
+eagle_posttool_mirror_tasks() {
+    local tool_name="$1" session_id="$2" project="$3" input="$4"
+
+    case "$tool_name" in
+        TaskCreate|TaskUpdate)
+            if eagle_validate_session_id "$session_id"; then
+                local task_dir="$EAGLE_CLAUDE_TASKS_DIR/$session_id"
+                if [ -d "$task_dir" ]; then
+                    local task_id
+                    task_id=$(echo "$input" | jq -r '.tool_input.id // empty')
+                    if [ -z "$task_id" ]; then
+                        local newest
+                        newest=$(ls -t "$task_dir"/*.json 2>/dev/null | head -1)
+                        [ -n "$newest" ] && [ -f "$newest" ] && eagle_capture_claude_task "$newest" "$session_id" "$project"
+                    elif eagle_validate_session_id "$task_id"; then
+                        local task_json="$task_dir/$task_id.json"
+                        [ -f "$task_json" ] && eagle_capture_claude_task "$task_json" "$session_id" "$project"
+                    fi
+                fi
+            fi
+            ;;
+    esac
+}
+
+eagle_posttool_stale_hint() {
+    local tool_name="$1" fp="$2" project="$3"
+
+    case "$tool_name" in
+        Write|Edit)
+            if [ -n "$fp" ]; then
+                local fname fname_stem
+                fname=$(basename "$fp")
+                fname_stem="${fname%.*}"
+                case "$fp" in
+                    "$HOME/.claude/"*) ;; # skip Claude config files
+                    *)
+                        if [ ${#fname_stem} -ge 3 ]; then
+                            local fts_query
+                            fts_query=$(eagle_fts_sanitize "$fname_stem")
+                            if [ -n "$fts_query" ]; then
+                                local stale_hit
+                                stale_hit=$(eagle_search_stale_memories "$project" "$fts_query")
+                                if [ -n "$stale_hit" ]; then
+                                    local stale_msg="Eagle Mem: Memory '${stale_hit}' may reference '${fname}'. If your edit contradicts it, update the memory."
+                                    jq -nc --arg ctx "$stale_msg" '{"hookSpecificOutput":{"hookEventName":"PostToolUse","additionalContext":$ctx}}'
+                                fi
+                            fi
+                        fi
+                        ;;
+                esac
+            fi
+            ;;
+    esac
+}
+
+eagle_posttool_decision_surface() {
+    local tool_name="$1" fp="$2" project="$3"
+
+    case "$tool_name" in
+        Read)
+            if [ -n "$fp" ]; then
+                local fname fname_stem read_context=""
+                fname=$(basename "$fp")
+                fname_stem="${fname%.*}"
+                case "$fp" in
+                    "$HOME/.claude/"*) ;; # skip Claude config files
+                    *)
+                        if [ ${#fname_stem} -ge 3 ]; then
+                            local fts_query
+                            fts_query=$(eagle_fts_sanitize "$fname_stem")
+                            if [ -n "$fts_query" ]; then
+                                local decision_hit
+                                decision_hit=$(eagle_search_decisions_for_file "$project" "$fts_query")
+                                if [ -n "$decision_hit" ]; then
+                                    read_context+="Eagle Mem decision history for '${fname}': ${decision_hit} — Do not revert without explicit user request. "
+                                fi
+                            fi
+                        fi
+
+                        local feature_hit
+                        feature_hit=$(eagle_find_features_for_file "$project" "$fp")
+                        if [ -n "$feature_hit" ]; then
+                            while IFS='|' read -r feat_name feat_desc feat_verified _role feat_deps feat_other_files feat_smoke; do
+                                [ -z "$feat_name" ] && continue
+                                read_context+="Eagle Mem: '${fname}' is part of feature '${feat_name}'"
+                                [ -n "$feat_desc" ] && read_context+=" ($feat_desc)"
+                                read_context+="."
+                                [ -n "$feat_verified" ] && read_context+=" Last verified: ${feat_verified}."
+                                [ -n "$feat_deps" ] && read_context+=" Dependencies: ${feat_deps}."
+                                [ -n "$feat_other_files" ] && read_context+=" Other files in pipeline: ${feat_other_files}."
+                                [ -n "$feat_smoke" ] && read_context+=" Smoke tests: ${feat_smoke}."
+                                read_context+=" Changes require re-testing after deploy. "
+                            done <<< "$feature_hit"
+                        fi
+
+                        if [ -n "$read_context" ]; then
+                            jq -nc --arg ctx "$read_context" '{"hookSpecificOutput":{"hookEventName":"PostToolUse","additionalContext":$ctx}}'
+                        fi
+                        ;;
+                esac
+            fi
+            ;;
+    esac
+}

package/lib/provider.sh

@@ -151,8 +151,9 @@ model = "claude-haiku-4-5-20251001"
 model = "gpt-4o-mini"
 
 [curator]
-# How often the curator runs: "manual", "daily", "weekly"
+# "auto" = triggers at session end after min_sessions; "manual" = CLI only
 schedule = "manual"
+min_sessions = 5
 
 [redaction]
 # Additional secret patterns (regex) beyond built-in defaults

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "eagle-mem",
-  "version": "3.0.1",
+  "version": "3.1.0",
   "description": "Persistent memory for Claude Code — SQLite + FTS5, no daemon, no bloat",
   "bin": {
     "eagle-mem": "bin/eagle-mem"

package/scripts/curate.sh

@@ -361,5 +361,6 @@ echo ""
 if [ "$DRY_RUN" -eq 1 ]; then
     eagle_footer "Dry run complete. Run without --dry-run to apply changes."
 else
+    eagle_meta_set "last_curated_at" "$(date -u +%Y-%m-%dT%H:%M:%SZ)" "$project"
     eagle_footer "Curation complete for '$project'."
 fi

package/scripts/health.sh

@@ -0,0 +1,218 @@
+#!/usr/bin/env bash
+# ═══════════════════════════════════════════════════════════
+# Eagle Mem — Health Check
+# Diagnoses how well the self-learning pipeline is working
+# ═══════════════════════════════════════════════════════════
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+LIB_DIR="$SCRIPT_DIR/../lib"
+
+. "$LIB_DIR/common.sh"
+. "$SCRIPT_DIR/style.sh"
+. "$LIB_DIR/db.sh"
+. "$LIB_DIR/provider.sh"
+
+eagle_header "Health Check"
+
+project=""
+JSON_OUT=0
+
+while [ $# -gt 0 ]; do
+    case "$1" in
+        -p|--project) project="$2"; shift 2 ;;
+        -j|--json) JSON_OUT=1; shift ;;
+        *) shift ;;
+    esac
+done
+
+if [ -z "$project" ]; then
+    project=$(eagle_project_from_cwd "$(pwd)")
+fi
+
+if [ -z "$project" ]; then
+    eagle_err "Cannot determine project (ephemeral directory). Use -p <project>."
+    exit 1
+fi
+
+p_esc=$(eagle_sql_escape "$project")
+
+eagle_info "Project: ${BOLD}$project${RESET}"
+echo ""
+
+score=0
+max_score=0
+issues=()
+
+# ─── 1. Summary enrichment rate ──────────────────────────
+
+max_score=$((max_score + 30))
+
+total_summaries=$(eagle_db "SELECT COUNT(*) FROM summaries WHERE project = '$p_esc';")
+enriched_summaries=$(eagle_db "SELECT COUNT(*) FROM summaries WHERE project = '$p_esc' AND (decisions IS NOT NULL AND decisions != '' OR gotchas IS NOT NULL AND gotchas != '' OR key_files IS NOT NULL AND key_files != '');")
+
+if [ "${total_summaries:-0}" -eq 0 ]; then
+    enrich_pct=0
+else
+    enrich_pct=$((enriched_summaries * 100 / total_summaries))
+fi
+
+if [ "$enrich_pct" -ge 50 ]; then
+    eagle_ok "Enriched summaries: ${enriched_summaries}/${total_summaries} (${enrich_pct}%)"
+    score=$((score + 30))
+elif [ "$enrich_pct" -ge 20 ]; then
+    eagle_warn "Enriched summaries: ${enriched_summaries}/${total_summaries} (${enrich_pct}%) — aim for 50%+"
+    score=$((score + 15))
+    issues+=("Low enrichment rate (${enrich_pct}%). Eagle-summary blocks aren't being emitted reliably.")
+elif [ "${total_summaries:-0}" -gt 0 ]; then
+    eagle_fail "Enriched summaries: ${enriched_summaries}/${total_summaries} (${enrich_pct}%) — self-learning not working"
+    issues+=("Critical: ${enrich_pct}% enrichment. Decisions/gotchas/key_files are not being captured.")
+else
+    eagle_dim "  No summaries yet"
+fi
+
+# ─── 2. Feature discovery ────────────────────────────────
+
+max_score=$((max_score + 20))
+
+feature_count=$(eagle_db "SELECT COUNT(*) FROM features WHERE project = '$p_esc' AND status = 'active';")
+feature_file_count=$(eagle_db "SELECT COUNT(*) FROM feature_files ff JOIN features f ON ff.feature_id = f.id WHERE f.project = '$p_esc' AND f.status = 'active';")
+
+if [ "${feature_count:-0}" -ge 3 ]; then
+    eagle_ok "Features tracked: ${feature_count} (${feature_file_count} files mapped)"
+    score=$((score + 20))
+elif [ "${feature_count:-0}" -ge 1 ]; then
+    eagle_warn "Features tracked: ${feature_count} — curator needs more sessions"
+    score=$((score + 10))
+    issues+=("Only ${feature_count} features discovered. Run curator more often.")
+else
+    eagle_fail "Features tracked: 0 — feature graph is empty"
+    issues+=("No features discovered. Run: eagle-mem curate")
+fi
+
+# ─── 3. Command intelligence ─────────────────────────────
+
+max_score=$((max_score + 15))
+
+rule_count=$(eagle_db "SELECT COUNT(*) FROM command_rules WHERE (project = '$p_esc' OR project IS NULL) AND enabled = 1;")
+obs_with_metrics=$(eagle_db "SELECT COUNT(*) FROM observations WHERE project = '$p_esc' AND tool_name = 'Bash' AND output_bytes IS NOT NULL AND output_bytes > 0;")
+
+if [ "${rule_count:-0}" -ge 2 ]; then
+    eagle_ok "Command rules: ${rule_count} active (${obs_with_metrics} observations with metrics)"
+    score=$((score + 15))
+elif [ "${rule_count:-0}" -ge 1 ]; then
+    eagle_warn "Command rules: ${rule_count} — learning in progress"
+    score=$((score + 8))
+elif [ "${obs_with_metrics:-0}" -gt 20 ]; then
+    eagle_fail "Command rules: 0 (but ${obs_with_metrics} observations available — run curator)"
+    issues+=("Command metrics collected but no rules generated yet.")
+else
+    eagle_dim "  Command metrics: ${obs_with_metrics} observations (need more data)"
+    score=$((score + 5))
+fi
+
+# ─── 4. Provider configured ──────────────────────────────
+
+max_score=$((max_score + 15))
+
+provider=$(eagle_config_get "provider" "type" "none")
+if [ "$provider" != "none" ]; then
+    model=$(eagle_config_get "$provider" "model" "default")
+    eagle_ok "LLM provider: ${provider} (${model})"
+    score=$((score + 15))
+else
+    eagle_fail "No LLM provider — curator and enrichment extraction disabled"
+    issues+=("Configure a provider: eagle-mem config init")
+fi
+
+# ─── 5. Project data quality ─────────────────────────────
+
+max_score=$((max_score + 10))
+
+tmp_sessions=$(eagle_db "SELECT COUNT(*) FROM sessions WHERE project IN ('tmp', 'private', '');")
+total_sessions=$(eagle_db "SELECT COUNT(*) FROM sessions;")
+
+if [ "${total_sessions:-0}" -eq 0 ]; then
+    noise_pct=0
+else
+    noise_pct=$((tmp_sessions * 100 / total_sessions))
+fi
+
+if [ "$noise_pct" -le 5 ]; then
+    eagle_ok "Data quality: ${tmp_sessions} ephemeral sessions (${noise_pct}% noise)"
+    score=$((score + 10))
+elif [ "$noise_pct" -le 20 ]; then
+    eagle_warn "Data quality: ${tmp_sessions} ephemeral sessions (${noise_pct}% noise)"
+    score=$((score + 5))
+    issues+=("${noise_pct}% of sessions from ephemeral dirs. Skiplist should prevent new ones.")
+else
+    eagle_fail "Data quality: ${noise_pct}% noise — ${tmp_sessions}/${total_sessions} sessions from ephemeral dirs"
+    issues+=("Heavy ephemeral pollution. Update Eagle Mem to get skiplist protection.")
+fi
+
+# ─── 6. Curator activity ─────────────────────────────────
+
+max_score=$((max_score + 10))
+
+curator_schedule=$(eagle_config_get "curator" "schedule" "manual")
+last_curated=$(eagle_db "SELECT value FROM eagle_meta WHERE key = 'last_curated_at' AND (project = '$p_esc' OR project IS NULL) ORDER BY CASE WHEN project IS NOT NULL THEN 0 ELSE 1 END LIMIT 1;" 2>/dev/null || echo "")
+
+if [ -n "$last_curated" ]; then
+    eagle_ok "Curator: last run ${last_curated} (schedule: ${curator_schedule})"
+    score=$((score + 10))
+elif [ "$curator_schedule" = "auto" ]; then
+    eagle_warn "Curator: auto-scheduled but hasn't run yet"
+    score=$((score + 5))
+    issues+=("Auto-curate is configured but hasn't run. It triggers at session end.")
+else
+    eagle_fail "Curator: never run (schedule: ${curator_schedule})"
+    issues+=("Curator has never run. Try: eagle-mem curate --dry-run")
+fi
+
+# ─── Score ────────────────────────────────────────────────
+
+echo ""
+echo -e "  ${DIM}─────────────────────────────────────${RESET}"
+
+pct=$((score * 100 / max_score))
+if [ "$pct" -ge 80 ]; then
+    color="$GREEN"
+    grade="Healthy"
+elif [ "$pct" -ge 50 ]; then
+    color="$YELLOW"
+    grade="Needs attention"
+else
+    color="$RED"
+    grade="Unhealthy"
+fi
+
+echo -e "  ${BOLD}Score: ${color}${score}/${max_score} (${pct}%)${RESET}  ${color}${grade}${RESET}"
+
+if [ ${#issues[@]} -gt 0 ]; then
+    echo ""
+    echo -e "  ${BOLD}Issues:${RESET}"
+    for issue in "${issues[@]}"; do
+        echo -e "    ${YELLOW}!${RESET} $issue"
+    done
+fi
+
+eagle_footer "Health check complete."
+
+if [ "$JSON_OUT" -eq 1 ]; then
+    jq -nc \
+        --arg project "$project" \
+        --argjson score "$score" \
+        --argjson max_score "$max_score" \
+        --argjson pct "$pct" \
+        --arg grade "$grade" \
+        --argjson total_summaries "${total_summaries:-0}" \
+        --argjson enriched_summaries "${enriched_summaries:-0}" \
+        --argjson features "${feature_count:-0}" \
+        --argjson command_rules "${rule_count:-0}" \
+        --arg provider "$provider" \
+        --argjson noise_pct "$noise_pct" \
+        '{project:$project, score:$score, max:$max_score, pct:$pct, grade:$grade,
+          enrichment:{total:$total_summaries, enriched:$enriched_summaries},
+          features:$features, command_rules:$command_rules,
+          provider:$provider, noise_pct:$noise_pct}'
+fi

package/scripts/help.sh

@@ -27,6 +27,7 @@ echo -e "    ${CYAN}prune${RESET}       Remove old observations and orphaned chu
 echo -e "    ${CYAN}config${RESET}      View or change LLM provider settings"
 echo -e "    ${CYAN}curate${RESET}      Run the self-learning curator (LLM-powered analysis)"
 echo -e "    ${CYAN}feature${RESET}     Manage feature graph (list/show/verify/add)"
+echo -e "    ${CYAN}health${RESET}      Diagnose self-learning pipeline health"
 echo -e "    ${CYAN}install${RESET}     First-time setup: hooks, database, skills"
 echo -e "    ${CYAN}update${RESET}      Re-deploy hooks and run migrations"
 echo -e "    ${CYAN}uninstall${RESET}   Remove hooks and optionally delete data"

package/scripts/memories.sh

@@ -438,7 +438,7 @@ memories_sync() {
     eagle_info "Scanning for Claude Code auto-memory files..."
     echo ""
 
-    local claude_mem_root="$HOME/.claude/projects"
+    local claude_mem_root="$EAGLE_CLAUDE_PROJECTS_DIR"
     local mem_synced=0
     local mem_skipped=0
 
@@ -471,7 +471,7 @@ memories_sync() {
     eagle_info "Scanning for Claude Code plan files..."
     echo ""
 
-    local plans_dir="$HOME/.claude/plans"
+    local plans_dir="$EAGLE_CLAUDE_PLANS_DIR"
     local plan_synced=0
     local plan_skipped=0
 
@@ -504,7 +504,7 @@ memories_sync() {
     eagle_info "Scanning for Claude Code task files..."
     echo ""
 
-    local tasks_dir="$HOME/.claude/tasks"
+    local tasks_dir="$EAGLE_CLAUDE_TASKS_DIR"
     local task_synced=0
     local task_skipped=0