devforgeai 1.0.5 → 1.0.7

package/src/claude/scripts/plan_file_kb.sh

@@ -0,0 +1,704 @@
+#!/bin/bash
+# Plan File Knowledge Base Functions
+# STORY-222: Extract Plan File Knowledge Base for Decision Archive
+#
+# Provides 4 functions for plan file parsing and decision archive building:
+# - extract_yaml_frontmatter: Parse YAML frontmatter from plan files
+# - extract_story_ids: Extract STORY-NNN patterns with context
+# - build_decision_archive: Build bidirectional story<->plan mapping
+# - query_archive: Query archive for related plan files
+
+set -euo pipefail
+
+# =============================================================================
+# Helper Functions
+# =============================================================================
+
+# Validate that a file exists before processing
+# Usage: validate_file_exists "$file_path"
+# Returns: 0 if exists, 1 if not (also outputs JSON error)
+validate_file_exists() {
+    local file="$1"
+    if [[ ! -f "$file" ]]; then
+        echo '{"error": "File not found"}'
+        return 1
+    fi
+    return 0
+}
+
+# Escape special characters for JSON output
+# Usage: json_escape "$string"
+# Handles: backslash, quotes, newlines, carriage returns, tabs
+json_escape() {
+    local str="$1"
+    str="${str//\\/\\\\}"
+    str="${str//\"/\\\"}"
+    str="${str//$'\n'/\\n}"
+    str="${str//$'\r'/\\r}"
+    str="${str//$'\t'/\\t}"
+    echo "$str"
+}
+
+# Build a JSON array from newline-separated lines
+# Usage: json_array_from_lines < input_lines
+# Example: echo -e "STORY-001\nSTORY-002" | json_array_from_lines
+json_array_from_lines() {
+    awk 'BEGIN{printf "["} NR>1{printf ", "} {printf "\"%s\"", $0} END{printf "]"}'
+}
+
+# Parse a specific YAML field from content
+# Usage: parse_yaml_field "$yaml_content" "status"
+# Returns: field value without quotes
+parse_yaml_field() {
+    local content="$1"
+    local field_name="$2"
+    echo "$content" | grep -E "^${field_name}:" | sed "s/^${field_name}:[[:space:]]*//" | tr -d '"' | tr -d "'" || echo ""
+}
+
+# =============================================================================
+# Function 1: extract_yaml_frontmatter
+# AC#1: Parse YAML frontmatter from plan files
+# SM-010: Parse YAML frontmatter from plan files (Critical)
+# =============================================================================
+extract_yaml_frontmatter() {
+    local plan_file="$1"
+
+    validate_file_exists "$plan_file" || return 1
+
+    # Check if file starts with ---
+    local first_line
+    first_line=$(head -n1 "$plan_file" 2>/dev/null || echo "")
+
+    if [[ "$first_line" != "---" ]]; then
+        # No YAML frontmatter - return empty/default
+        echo '{"status": "", "created": "", "author": "", "related_stories": []}'
+        return 0
+    fi
+
+    # Extract content between first and second ---
+    local yaml_content
+    yaml_content=$(awk 'BEGIN{found=0} /^---$/{found++; if(found==2) exit; next} found==1{print}' "$plan_file" 2>/dev/null || echo "")
+
+    if [[ -z "$yaml_content" ]]; then
+        echo '{"status": "", "created": "", "author": "", "related_stories": []}'
+        return 0
+    fi
+
+    # Parse individual fields from YAML using helper
+    local status
+    status=$(parse_yaml_field "$yaml_content" "status")
+
+    local created
+    created=$(parse_yaml_field "$yaml_content" "created")
+
+    local author
+    author=$(parse_yaml_field "$yaml_content" "author")
+
+    # Extract related_stories array
+    local related_stories
+    related_stories=$(parse_related_stories "$yaml_content")
+
+    # Output JSON
+    printf '{"status": "%s", "created": "%s", "author": "%s", "related_stories": %s}\n' \
+        "$status" "$created" "$author" "$related_stories"
+}
+
+# Helper to parse related_stories field (handles inline and multiline formats)
+parse_related_stories() {
+    local yaml_content="$1"
+    local related_stories_field
+    related_stories_field=$(echo "$yaml_content" | grep -E "^related_stories:" || echo "")
+
+    if [[ -z "$related_stories_field" ]]; then
+        echo "[]"
+        return 0
+    fi
+
+    # Try to extract STORY-XXX patterns from the field
+    local stories_array
+    stories_array=$(echo "$yaml_content" | grep -E "^related_stories:" | \
+        sed 's/^related_stories:[[:space:]]*//' | \
+        grep -oE 'STORY-[0-9]+' | \
+        json_array_from_lines)
+
+    # If no patterns found, return empty array
+    if [[ -z "$stories_array" || "$stories_array" == "[]" ]]; then
+        echo "[]"
+    else
+        echo "$stories_array"
+    fi
+}
+
+# =============================================================================
+# Function 2: extract_story_ids
+# AC#2: Extract STORY-NNN patterns with surrounding context
+# SM-011: Extract STORY-NNN patterns with regex (High)
+# =============================================================================
+extract_story_ids() {
+    local plan_file="$1"
+
+    validate_file_exists "$plan_file" || return 1
+
+    # Extract all STORY-NNN patterns (3+ digits)
+    local story_ids
+    story_ids=$(grep -oE 'STORY-[0-9]{3,}' "$plan_file" 2>/dev/null | sort -u || echo "")
+
+    if [[ -z "$story_ids" ]]; then
+        echo '{"story_ids": [], "contexts": {}}'
+        return 0
+    fi
+
+    # Build JSON array of story IDs using helper
+    local ids_json
+    ids_json=$(echo "$story_ids" | json_array_from_lines)
+
+    # Build contexts object with surrounding text for each story ID
+    local contexts_json="{"
+    local is_first_id_entry=true
+
+    while IFS= read -r story_id; do
+        if [[ -z "$story_id" ]]; then continue; fi
+
+        # Get line with context (extract the line containing the story ID)
+        local context
+        context=$(grep -m1 "$story_id" "$plan_file" 2>/dev/null | head -c 200 | tr '\n' ' ' | tr '"' "'" || echo "")
+
+        if [[ "$is_first_id_entry" == "true" ]]; then
+            is_first_id_entry=false
+        else
+            contexts_json+=", "
+        fi
+        contexts_json+="\"$story_id\": \"$context\""
+    done <<< "$story_ids"
+
+    contexts_json+="}"
+
+    printf '{"story_ids": %s, "contexts": %s}\n' "$ids_json" "$contexts_json"
+}
+
+# =============================================================================
+# Function 3: build_decision_archive
+# AC#3: Build bidirectional story<->decision mapping
+# SM-012: Build story→decision bidirectional mapping (High)
+# NFR-010: Index 350+ plan files within 10 seconds
+# =============================================================================
+# Process a single plan file and update mappings
+process_plan_file() {
+    local plan_file="$1"
+    local -n story_to_plans_ref=$2
+    local -n plan_to_stories_ref=$3
+    local -n plan_metadata_ref=$4
+
+    local plan_name
+    plan_name=$(basename "$plan_file")
+
+    # Extract frontmatter
+    local frontmatter
+    frontmatter=$(extract_yaml_frontmatter "$plan_file")
+    plan_metadata_ref["$plan_name"]="$frontmatter"
+
+    # Extract story IDs
+    local story_ids
+    story_ids=$(grep -oE 'STORY-[0-9]+' "$plan_file" 2>/dev/null | sort -u || echo "")
+
+    # Build plan_to_stories mapping using helper
+    if [[ -n "$story_ids" ]]; then
+        local stories_array
+        stories_array=$(echo "$story_ids" | json_array_from_lines)
+        plan_to_stories_ref["$plan_name"]="$stories_array"
+
+        # Build story_to_plans mapping (bidirectional)
+        while IFS= read -r story_id; do
+            if [[ -n "$story_id" ]]; then
+                if [[ -n "${story_to_plans_ref[$story_id]:-}" ]]; then
+                    story_to_plans_ref["$story_id"]+=", \"$plan_name\""
+                else
+                    story_to_plans_ref["$story_id"]="\"$plan_name\""
+                fi
+            fi
+        done <<< "$story_ids"
+    else
+        plan_to_stories_ref["$plan_name"]="[]"
+    fi
+}
+
+# Build JSON object from associative array
+build_json_object_from_array() {
+    local -n array_ref=$1
+    local json_str="{"
+    local is_first=true
+
+    for key in "${!array_ref[@]}"; do
+        if [[ "$is_first" == "true" ]]; then
+            is_first=false
+        else
+            json_str+=", "
+        fi
+        json_str+="\"$key\": ${array_ref[$key]}"
+    done
+    json_str+="}"
+    echo "$json_str"
+}
+
+# Build JSON array of plan names from story_to_plans mapping
+build_story_to_plans_json() {
+    local -n array_ref=$1
+    local json_str="{"
+    local is_first=true
+
+    for story_id in "${!array_ref[@]}"; do
+        if [[ "$is_first" == "true" ]]; then
+            is_first=false
+        else
+            json_str+=", "
+        fi
+        json_str+="\"$story_id\": [${array_ref[$story_id]}]"
+    done
+    json_str+="}"
+    echo "$json_str"
+}
+
+build_decision_archive() {
+    local plans_dir="$1"
+    local archive_dir="$2"
+
+    if [[ ! -d "$plans_dir" ]]; then
+        echo '{"error": "Plans directory not found"}'
+        return 1
+    fi
+
+    # Create archive directory if it doesn't exist
+    mkdir -p "$archive_dir"
+
+    # Initialize data structures
+    declare -A story_to_plans
+    declare -A plan_to_stories
+    declare -A plan_metadata
+
+    # Process all plan files
+    local plan_count=0
+    while IFS= read -r -d '' plan_file; do
+        process_plan_file "$plan_file" story_to_plans plan_to_stories plan_metadata
+        ((plan_count++))
+    done < <(find "$plans_dir" -name "*.md" -type f -print0 2>/dev/null)
+
+    # Build final JSON using helpers
+    local story_to_plans_json
+    story_to_plans_json=$(build_story_to_plans_json story_to_plans)
+
+    local plan_to_stories_json
+    plan_to_stories_json=$(build_json_object_from_array plan_to_stories)
+
+    local plan_metadata_json
+    plan_metadata_json=$(build_json_object_from_array plan_metadata)
+
+    local archive_json
+    archive_json="{\"story_to_plans\": $story_to_plans_json, \"plan_to_stories\": $plan_to_stories_json, \"metadata\": $plan_metadata_json}"
+
+    # Write archive to file
+    echo "$archive_json" > "$archive_dir/decision_archive.json"
+
+    # Return summary
+    printf '{"status": "success", "plan_count": %d, "archive_path": "%s/decision_archive.json"}\n' \
+        "$plan_count" "$archive_dir"
+}
+
+# =============================================================================
+# Function 4: query_archive
+# AC#4: Query archive for related plan files
+# =============================================================================
+# Extract plans for a story from archive content
+extract_plans_from_archive() {
+    local archive_content="$1"
+    local story_id="$2"
+
+    # Try to extract story_to_plans entry
+    local plans_array
+    plans_array=$(echo "$archive_content" | grep -oE "\"$story_id\": *\[[^]]*\]" | \
+        sed 's/.*\[\([^]]*\)\].*/[\1]/' | head -1 || echo "[]")
+
+    # If not found, try alternate pattern
+    if [[ -z "$plans_array" || "$plans_array" == "[]" ]]; then
+        local story_entry
+        story_entry=$(echo "$archive_content" | grep -o "\"$story_id\": \[[^]]*\]" || echo "")
+        if [[ -n "$story_entry" ]]; then
+            plans_array=$(echo "$story_entry" | sed 's/.*\[\(.*\)\]/[\1]/')
+        fi
+    fi
+
+    echo "$plans_array"
+}
+
+query_archive() {
+    local archive_dir="$1"
+    local story_id="$2"
+
+    local archive_file="$archive_dir/decision_archive.json"
+
+    if [[ ! -f "$archive_file" ]]; then
+        echo '{"error": "Archive not found", "story_id": "'"$story_id"'", "plans": []}'
+        return 1
+    fi
+
+    # Read archive
+    local archive_content
+    archive_content=$(cat "$archive_file")
+
+    # Extract plans using helper
+    local plans_array
+    plans_array=$(extract_plans_from_archive "$archive_content" "$story_id")
+
+    if [[ -z "$plans_array" || "$plans_array" == "[]" ]]; then
+        echo '{"story_id": "'"$story_id"'", "plans": [], "count": 0}'
+        return 0
+    fi
+
+    # Count plans
+    local count
+    count=$(echo "$plans_array" | grep -oE '"[^"]*\.md"' | wc -l || echo "0")
+
+    # Build result with plan details
+    local result='{"story_id": "'"$story_id"'", "plans": '"$plans_array"', "count": '"$count"'}'
+
+    echo "$result"
+}
+
+# =============================================================================
+# Helper: Get archive statistics
+# =============================================================================
+get_archive_stats() {
+    local archive_dir="$1"
+    local archive_file="$archive_dir/decision_archive.json"
+
+    if [[ ! -f "$archive_file" ]]; then
+        echo '{"error": "Archive not found"}'
+        return 1
+    fi
+
+    local story_count
+    story_count=$(grep -oE '"STORY-[0-9]+":' "$archive_file" | sort -u | wc -l || echo "0")
+
+    local plan_count
+    plan_count=$(grep -oE '"[^"]+\.md":' "$archive_file" | sort -u | wc -l || echo "0")
+
+    printf '{"story_count": %d, "plan_count": %d}\n' "$story_count" "$plan_count"
+}
+
+# =============================================================================
+# STORY-232 Helper: Extract markdown section by header name
+# Usage: extract_markdown_section "$file" "Section Name"
+# Extracts content from "## Section Name" to next "##" or EOF
+# =============================================================================
+extract_markdown_section() {
+    local file="$1"
+    local section_name="$2"
+
+    awk -v section="$section_name" '
+        $0 ~ ("^## " section "$") { capture=1; next }
+        /^## / && capture { capture=0 }
+        capture { print }
+    ' "$file" 2>/dev/null | sed '/^$/d' | tr '\n' ' ' | sed 's/  */ /g' | sed 's/^ //;s/ $//' || echo ""
+}
+
+# =============================================================================
+# STORY-232 Helper: Extract JSON field value from JSON string
+# Usage: extract_json_field "$json_string" "field_name"
+# =============================================================================
+extract_json_field() {
+    local json_string="$1"
+    local field_name="$2"
+
+    echo "$json_string" | grep -oE "\"$field_name\": *\"[^\"]*\"" | \
+        sed "s/\"$field_name\": *\"//" | sed 's/"$//' || echo ""
+}
+
+# =============================================================================
+# STORY-232 Helper: Add keyword-to-plan mapping
+# Usage: add_keyword_mapping keyword_array_name "$keyword" "$plan_name"
+# =============================================================================
+add_keyword_mapping() {
+    local -n keyword_map_ref=$1
+    local keyword="$2"
+    local plan_name="$3"
+
+    if [[ -z "$keyword" ]]; then return; fi
+
+    if [[ -n "${keyword_map_ref[$keyword]:-}" ]]; then
+        # Check if plan already in list
+        if ! echo "${keyword_map_ref[$keyword]}" | grep -q "\"$plan_name\""; then
+            keyword_map_ref["$keyword"]="${keyword_map_ref[$keyword]}, \"$plan_name\""
+        fi
+    else
+        keyword_map_ref["$keyword"]="\"$plan_name\""
+    fi
+}
+
+# =============================================================================
+# STORY-232: Function 5: extract_decision_sections
+# AC#2: Extract ## Decision and ## Technical Approach sections from plan file
+# Usage: extract_decision_sections "$plan_file"
+# Returns: JSON with "decision" and "technical_approach" keys
+# =============================================================================
+extract_decision_sections() {
+    local plan_file="${1:-}"
+
+    # Validate argument provided
+    if [[ -z "$plan_file" ]]; then
+        echo '{"error": "plan_file argument required"}'
+        return 1
+    fi
+
+    # Validate path doesn't contain traversal
+    if [[ "$plan_file" == *".."* ]]; then
+        echo '{"error": "Path traversal not allowed"}'
+        return 1
+    fi
+
+    validate_file_exists "$plan_file" || return 1
+
+    # Extract sections using helper
+    local decision_content
+    decision_content=$(extract_markdown_section "$plan_file" "Decision")
+
+    local technical_approach_content
+    technical_approach_content=$(extract_markdown_section "$plan_file" "Technical Approach")
+
+    # Escape content for JSON
+    local escaped_decision
+    escaped_decision=$(json_escape "$decision_content")
+
+    local escaped_technical
+    escaped_technical=$(json_escape "$technical_approach_content")
+
+    # Return JSON
+    printf '{"decision": "%s", "technical_approach": "%s"}\n' "$escaped_decision" "$escaped_technical"
+}
+
+# =============================================================================
+# STORY-232: Function 6: build_searchable_index
+# AC#1: Build searchable index with frontmatter (story ID, status, created date)
+# AC#3: Create searchable_index.json with full-text content
+# Usage: build_searchable_index "$plans_dir" "$index_dir"
+# Returns: JSON with "status", "plan_count", "index_path" keys
+# =============================================================================
+build_searchable_index() {
+    local plans_dir="${1:-}"
+    local index_dir="${2:-}"
+
+    # Validate arguments provided
+    if [[ -z "$plans_dir" ]] || [[ -z "$index_dir" ]]; then
+        echo '{"error": "plans_dir and index_dir arguments required"}'
+        return 1
+    fi
+
+    # Validate paths don't contain traversal
+    if [[ "$plans_dir" == *".."* ]] || [[ "$index_dir" == *".."* ]]; then
+        echo '{"error": "Path traversal not allowed"}'
+        return 1
+    fi
+
+    if [[ ! -d "$plans_dir" ]]; then
+        echo '{"error": "Plans directory not found"}'
+        return 1
+    fi
+
+    # Create index directory if it doesn't exist
+    mkdir -p "$index_dir"
+
+    # Start building JSON
+    local plans_json="{"
+    local keywords_json="{"
+    local is_first_plan=true
+    local plan_count=0
+
+    # Associative array for keyword tracking
+    declare -A keyword_to_plans
+
+    # Process all plan files - OPTIMIZED: Single file read per plan
+    while IFS= read -r -d '' plan_file; do
+        local plan_name
+        plan_name=$(basename "$plan_file")
+
+        # Extract story ID from filename (STORY-NNN pattern) - no subshell
+        local story_id=""
+        [[ "$plan_name" =~ STORY-([0-9]+) ]] && story_id="STORY-${BASH_REMATCH[1]}"
+
+        # OPTIMIZATION: Read file content ONCE into memory
+        local file_content
+        file_content=$(cat "$plan_file" 2>/dev/null) || file_content=""
+
+        # Extract frontmatter from cached content
+        local status="" created=""
+        if [[ "$file_content" == "---"* ]]; then
+            local yaml_block
+            yaml_block=$(echo "$file_content" | awk 'BEGIN{f=0} /^---$/{f++; if(f==2) exit; next} f==1{print}')
+            status=$(echo "$yaml_block" | grep -E "^status:" | sed 's/^status:[[:space:]]*//' | tr -d '"' | tr -d "'" || echo "")
+            created=$(echo "$yaml_block" | grep -E "^created:" | sed 's/^created:[[:space:]]*//' | tr -d '"' | tr -d "'" || echo "")
+        fi
+
+        # Extract decision sections from cached content
+        local decision="" technical_approach=""
+        decision=$(echo "$file_content" | awk '/^## Decision$/{capture=1; next} /^## /{capture=0} capture{print}' | tr '\n' ' ' | sed 's/  */ /g;s/^ //;s/ $//' || echo "")
+        technical_approach=$(echo "$file_content" | awk '/^## Technical Approach$/{capture=1; next} /^## /{capture=0} capture{print}' | tr '\n' ' ' | sed 's/  */ /g;s/^ //;s/ $//' || echo "")
+
+        # Build full text from cached content
+        local full_text
+        full_text=$(echo "$file_content" | tr '\n' ' ' | sed 's/  */ /g' | head -c 500 || echo "")
+        local escaped_full_text escaped_decision escaped_technical
+        escaped_full_text=$(json_escape "$full_text")
+        escaped_decision=$(json_escape "$decision")
+        escaped_technical=$(json_escape "$technical_approach")
+
+        # Build keywords from cached content - use word list directly, skip common words
+        local content_words
+        content_words=$(echo "$file_content" | tr '[:upper:]' '[:lower:]' | grep -oE '[a-z]{4,}' | \
+            grep -vE '^(that|this|with|from|have|been|were|will|would|could|should|their|there|which|about)$' | \
+            sort -u | head -50 || echo "")
+
+        # Track keywords for this plan using helper
+        while IFS= read -r word; do
+            [[ -n "$word" ]] && add_keyword_mapping keyword_to_plans "$word" "$plan_name"
+        done <<< "$content_words"
+
+        # Add to plans JSON
+        if [[ "$is_first_plan" == "true" ]]; then
+            is_first_plan=false
+        else
+            plans_json+=", "
+        fi
+
+        plans_json+="\"$plan_name\": {\"story_id\": \"$story_id\", \"status\": \"$status\", \"created\": \"$created\", \"decision\": \"$escaped_decision\", \"technical_approach\": \"$escaped_technical\", \"full_text\": \"$escaped_full_text\"}"
+
+        ((plan_count++))
+    done < <(find "$plans_dir" -name "*.md" -type f -print0 2>/dev/null)
+
+    plans_json+="}"
+
+    # Build keywords JSON
+    local is_first_keyword=true
+    for keyword in "${!keyword_to_plans[@]}"; do
+        if [[ "$is_first_keyword" == "true" ]]; then
+            is_first_keyword=false
+        else
+            keywords_json+=", "
+        fi
+        keywords_json+="\"$keyword\": [${keyword_to_plans[$keyword]}]"
+    done
+    keywords_json+="}"
+
+    # Combine into final index
+    local index_json="{\"plans\": $plans_json, \"keywords\": $keywords_json}"
+
+    # Write to file
+    echo "$index_json" > "$index_dir/searchable_index.json"
+
+    # Return success
+    printf '{"status": "success", "plan_count": %d, "index_path": "%s/searchable_index.json"}\n' \
+        "$plan_count" "$index_dir"
+}
+
+# =============================================================================
+# STORY-232: Function 7: search_index
+# AC#3: Search index for keyword matches in decision, technical approach, full text
+# Usage: search_index "$index_dir" "$keyword"
+# Returns: JSON with "query", "matches", "count" keys
+# =============================================================================
+search_index() {
+    local index_dir="${1:-}"
+    local keyword="${2:-}"
+
+    # Validate index_dir provided
+    if [[ -z "$index_dir" ]]; then
+        echo '{"error": "index_dir argument required", "query": "", "matches": [], "count": 0}'
+        return 1
+    fi
+
+    # Validate path doesn't contain traversal
+    if [[ "$index_dir" == *".."* ]]; then
+        echo '{"error": "Path traversal not allowed", "query": "", "matches": [], "count": 0}'
+        return 1
+    fi
+
+    local index_file="$index_dir/searchable_index.json"
+
+    if [[ ! -f "$index_file" ]]; then
+        echo '{"error": "Index not found", "query": "'"$keyword"'", "matches": [], "count": 0}'
+        return 1
+    fi
+
+    # Handle empty keyword
+    if [[ -z "$keyword" ]]; then
+        echo '{"error": "Empty query", "query": "", "matches": [], "count": 0}'
+        return 1
+    fi
+
+    local index_content
+    index_content=$(cat "$index_file")
+
+    # Convert keyword to lowercase for case-insensitive search
+    local keyword_lower
+    keyword_lower=$(echo "$keyword" | tr '[:upper:]' '[:lower:]')
+
+    # Sanitize regex metacharacters to prevent command injection
+    local sanitized_keyword
+    sanitized_keyword=$(printf '%s\n' "$keyword_lower" | sed 's/[.[\*^$()+?{|\\]/\\&/g')
+
+    # Collect matching plan files
+    declare -A matches_set
+
+    # Method 1: Check keyword index (pre-built keywords section)
+    # Look for "keyword": ["plan1.md", "plan2.md"] pattern
+    local keyword_matches
+    keyword_matches=$(echo "$index_content" | grep -oE "\"$sanitized_keyword\": *\[[^\]]*\]" | \
+        grep -oE '"STORY-[^"]+\.md"' | tr -d '"' || echo "")
+
+    while IFS= read -r match; do
+        if [[ -n "$match" ]]; then
+            matches_set["$match"]=1
+        fi
+    done <<< "$keyword_matches"
+
+    # Method 2: Search in full text content (case insensitive grep through entire file)
+    # Find plan names that contain the keyword in their content
+    local plan_names
+    plan_names=$(echo "$index_content" | grep -oE '"STORY-[^"]+\.md":' | tr -d '":' || echo "")
+
+    while IFS= read -r plan_name; do
+        if [[ -z "$plan_name" ]]; then continue; fi
+
+        # Use awk to extract the block for this plan and search for keyword
+        local found
+        found=$(echo "$index_content" | awk -v plan="$plan_name" -v kw="$keyword_lower" '
+            BEGIN { IGNORECASE=1; in_plan=0; content="" }
+            $0 ~ ("\"" plan "\":") { in_plan=1 }
+            in_plan { content = content $0 }
+            in_plan && /\}[,]?$/ && !/\{/ {
+                if (content ~ kw) print plan
+                in_plan=0; content=""
+            }
+        ' || echo "")
+
+        if [[ -n "$found" ]]; then
+            matches_set["$plan_name"]=1
+        fi
+    done <<< "$plan_names"
+
+    # Build matches array
+    local matches_json="["
+    local is_first=true
+    local count=0
+
+    for match in "${!matches_set[@]}"; do
+        if [[ "$is_first" == "true" ]]; then
+            is_first=false
+        else
+            matches_json+=", "
+        fi
+        matches_json+="\"$match\""
+        ((count++))
+    done
+    matches_json+="]"
+
+    # Return result
+    printf '{"query": "%s", "matches": %s, "count": %d}\n' "$keyword" "$matches_json" "$count"
+}