@joekytc/dsh-swarm 0.1.2 → 0.2.0

package/skills/llm-wiki-skill/scripts/validate-step1.sh

@@ -0,0 +1,144 @@
+#!/bin/bash
+# 验证 ingest Step 1 的 JSON 输出格式
+# 用法：bash validate-step1.sh <json_file>
+# 返回：0 = 格式正确，1 = 格式有问题（触发回退）
+
+SCRIPT_DIR="${BASH_SOURCE[0]%/*}"
+[ "$SCRIPT_DIR" = "${BASH_SOURCE[0]}" ] && SCRIPT_DIR="."
+SCRIPT_DIR="$(cd "$SCRIPT_DIR" && pwd)"
+# shellcheck disable=SC1091
+source "$SCRIPT_DIR/shared-config.sh"
+
+JSON_FILE="$1"
+
+# 参数检查
+[ -z "$1" ] && { echo "ERROR: usage: validate-step1.sh <json_file>"; exit 1; }
+
+# 检查 jq 是否可用（必需依赖）
+command -v jq >/dev/null 2>&1 || {
+  echo "ERROR: jq is not installed. Install it via:" >&2
+  print_install_hint jq
+  exit 1
+}
+
+# 检查文件是否存在
+[ -f "$JSON_FILE" ] || { echo "ERROR: file not found: $JSON_FILE"; exit 1; }
+
+# 检查是否是有效 JSON
+jq empty "$JSON_FILE" 2>/dev/null || { echo "ERROR: invalid JSON format"; exit 1; }
+
+# 检查必需字段存在且类型正确
+jq -e '.entities | type == "array"' "$JSON_FILE" >/dev/null 2>&1 || { echo "ERROR: 'entities' must be an array"; exit 1; }
+jq -e '.topics | type == "array"' "$JSON_FILE" >/dev/null 2>&1 || { echo "ERROR: 'topics' must be an array"; exit 1; }
+jq -e '.connections | type == "array"' "$JSON_FILE" >/dev/null 2>&1 || { echo "ERROR: 'connections' must be an array"; exit 1; }
+jq -e '.contradictions | type == "array"' "$JSON_FILE" >/dev/null 2>&1 || { echo "ERROR: 'contradictions' must be an array"; exit 1; }
+jq -e '.new_vs_existing | type == "object"' "$JSON_FILE" >/dev/null 2>&1 || { echo "ERROR: 'new_vs_existing' must be an object"; exit 1; }
+
+# 检查每个 entity 的必需子字段
+VALID_CONFIDENCE="EXTRACTED|INFERRED|AMBIGUOUS|UNVERIFIED"
+
+ENTITY_COUNT=$(jq '.entities | length' "$JSON_FILE" 2>/dev/null)
+if [ "$ENTITY_COUNT" -gt 0 ] 2>/dev/null; then
+    NON_OBJECT_ENTITY_COUNT=$(jq '[.entities[] | select(type != "object")] | length' "$JSON_FILE" 2>/dev/null)
+    if [ "$NON_OBJECT_ENTITY_COUNT" -gt 0 ] 2>/dev/null; then
+        echo "ERROR: $NON_OBJECT_ENTITY_COUNT entity/entities must be objects"
+        exit 1
+    fi
+
+    # name, type, confidence 必须存在且非空
+    BAD_ENTITY_COUNT=$(jq '
+        [.entities[] | select(
+            (.name // "" | length) == 0 or
+            (.type // "" | length) == 0 or
+            (.confidence // "" | length) == 0
+        )] | length
+    ' "$JSON_FILE" 2>/dev/null)
+    if [ "$BAD_ENTITY_COUNT" -gt 0 ] 2>/dev/null; then
+        echo "ERROR: $BAD_ENTITY_COUNT entity/entities missing required fields (name/type/confidence)"
+        exit 1
+    fi
+
+    # confidence 值必须是四个有效值之一
+    INVALID=$(jq -r '.entities[]? | (.confidence // "MISSING")' "$JSON_FILE" 2>/dev/null | \
+        grep -v -E "^($VALID_CONFIDENCE)$" | head -3)
+    if [ -n "$INVALID" ]; then
+        echo "ERROR: invalid entity confidence value(s): $INVALID"
+        echo "       Valid values: EXTRACTED | INFERRED | AMBIGUOUS | UNVERIFIED"
+        exit 1
+    fi
+
+    # EXTRACTED 和 INFERRED 必须提供 evidence 字段
+    NO_EVIDENCE_COUNT=$(jq '
+        [.entities[] | select(
+            (.confidence == "EXTRACTED" or .confidence == "INFERRED") and
+            ((.evidence // "" | length) == 0)
+        )] | length
+    ' "$JSON_FILE" 2>/dev/null)
+    if [ "$NO_EVIDENCE_COUNT" -gt 0 ] 2>/dev/null; then
+        echo "WARN: $NO_EVIDENCE_COUNT entity/entities with EXTRACTED/INFERRED confidence missing 'evidence' field"
+    fi
+fi
+
+# 检查每个 topic 的必需子字段
+TOPIC_COUNT=$(jq '.topics | length' "$JSON_FILE" 2>/dev/null)
+if [ "$TOPIC_COUNT" -gt 0 ] 2>/dev/null; then
+    NON_OBJECT_TOPIC_COUNT=$(jq '[.topics[] | select(type != "object")] | length' "$JSON_FILE" 2>/dev/null)
+    if [ "$NON_OBJECT_TOPIC_COUNT" -gt 0 ] 2>/dev/null; then
+        echo "ERROR: $NON_OBJECT_TOPIC_COUNT topic(s) must be objects"
+        exit 1
+    fi
+
+    BAD_TOPIC_COUNT=$(jq '
+        [.topics[] | select(
+            (.name // "" | length) == 0
+        )] | length
+    ' "$JSON_FILE" 2>/dev/null)
+    if [ "$BAD_TOPIC_COUNT" -gt 0 ] 2>/dev/null; then
+        echo "ERROR: $BAD_TOPIC_COUNT topic(s) missing required 'name' field"
+        exit 1
+    fi
+fi
+
+# 检查每个 connection 的必需子字段（from, to, confidence）
+CONN_COUNT=$(jq '.connections | length' "$JSON_FILE" 2>/dev/null)
+if [ "$CONN_COUNT" -gt 0 ] 2>/dev/null; then
+    NON_OBJECT_CONN_COUNT=$(jq '[.connections[] | select(type != "object")] | length' "$JSON_FILE" 2>/dev/null)
+    if [ "$NON_OBJECT_CONN_COUNT" -gt 0 ] 2>/dev/null; then
+        echo "ERROR: $NON_OBJECT_CONN_COUNT connection(s) must be objects"
+        exit 1
+    fi
+
+    BAD_CONN_COUNT=$(jq '
+        [.connections[] | select(
+            (.from // "" | length) == 0 or
+            (.to // "" | length) == 0 or
+            (.confidence // "" | length) == 0
+        )] | length
+    ' "$JSON_FILE" 2>/dev/null)
+    if [ "$BAD_CONN_COUNT" -gt 0 ] 2>/dev/null; then
+        echo "ERROR: $BAD_CONN_COUNT connection(s) missing required fields (from/to/confidence)"
+        exit 1
+    fi
+
+    INVALID_CONN_CONF=$(jq -r '.connections[]? | (.confidence // "MISSING")' "$JSON_FILE" 2>/dev/null | \
+        grep -v -E "^($VALID_CONFIDENCE)$" | head -3)
+    if [ -n "$INVALID_CONN_CONF" ]; then
+        echo "ERROR: invalid connection confidence value(s): $INVALID_CONN_CONF"
+        echo "       Valid values: EXTRACTED | INFERRED | AMBIGUOUS | UNVERIFIED"
+        exit 1
+    fi
+
+    # EXTRACTED 和 INFERRED connections 必须提供 evidence
+    NO_CONN_EVIDENCE=$(jq '
+        [.connections[] | select(
+            (.confidence == "EXTRACTED" or .confidence == "INFERRED") and
+            ((.evidence // "" | length) == 0)
+        )] | length
+    ' "$JSON_FILE" 2>/dev/null)
+    if [ "$NO_CONN_EVIDENCE" -gt 0 ] 2>/dev/null; then
+        echo "WARN: $NO_CONN_EVIDENCE connection(s) with EXTRACTED/INFERRED confidence missing 'evidence' field"
+    fi
+fi
+
+echo "OK: Step 1 JSON validation passed"
+exit 0

package/skills/llm-wiki-skill/scripts/wiki-compat.sh

@@ -0,0 +1,267 @@
+#!/bin/bash
+# 旧知识库兼容脚本：惰性默认、目录检查、按需创建
+# 原则：migration_required=no，只有确实无法兼容时才引入显式迁移
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+SOURCE_REGISTRY_SCRIPT="$SCRIPT_DIR/source-registry.sh"
+
+LEGACY_REQUIRED_RAW_DIRS=(
+  "raw/articles"
+  "raw/tweets"
+  "raw/wechat"
+  "raw/pdfs"
+  "raw/notes"
+  "raw/assets"
+)
+
+REQUIRED_PATHS=(
+  ".wiki-schema.md"
+  "index.md"
+  "log.md"
+  "raw"
+  "wiki"
+  "wiki/entities"
+  "wiki/topics"
+  "wiki/sources"
+  "wiki/comparisons"
+  "wiki/synthesis"
+  "wiki/overview.md"
+)
+
+usage() {
+  cat <<'EOF'
+用法：
+  bash scripts/wiki-compat.sh inspect <wiki_root>
+  bash scripts/wiki-compat.sh validate <wiki_root>
+  bash scripts/wiki-compat.sh ensure-source-dir <wiki_root> <source_id>
+EOF
+}
+
+trim() {
+  printf '%s' "$1" | awk '{ gsub(/^[[:space:]]+|[[:space:]]+$/, "", $0); printf "%s", $0 }'
+}
+
+require_wiki_root() {
+  local wiki_root="$1"
+
+  [ -n "$wiki_root" ] || {
+    usage
+    exit 1
+  }
+
+  [ -d "$wiki_root" ] || {
+    echo "知识库不存在：$wiki_root" >&2
+    exit 1
+  }
+}
+
+schema_field_value() {
+  local wiki_root="$1"
+  local field_name="$2"
+  local default_value="$3"
+  local schema_path value
+
+  schema_path="$wiki_root/.wiki-schema.md"
+
+  if [ ! -f "$schema_path" ]; then
+    printf '%s\n' "$default_value"
+    return 0
+  fi
+
+  value="$(
+    awk -v field_name="$field_name" '
+      $0 ~ "^-[[:space:]]*" field_name "[：:]" {
+        line = $0
+        sub("^-[[:space:]]*" field_name "[：:][[:space:]]*", "", line)
+        print line
+        exit
+      }
+    ' "$schema_path"
+  )"
+
+  value="$(trim "$value")"
+
+  if [ -n "$value" ]; then
+    printf '%s\n' "$value"
+  else
+    printf '%s\n' "$default_value"
+  fi
+}
+
+resolved_language() {
+  local wiki_root="$1"
+  local raw_value
+
+  raw_value="$(schema_field_value "$wiki_root" "语言" "")"
+
+  case "$raw_value" in
+    English|english|EN|en)
+      printf 'en\n'
+      ;;
+    *)
+      printf 'zh\n'
+      ;;
+  esac
+}
+
+resolved_schema_version() {
+  local wiki_root="$1"
+
+  schema_field_value "$wiki_root" "版本" "1.0"
+}
+
+is_legacy_required_raw_dir() {
+  case "$1" in
+    raw/articles|raw/tweets|raw/wechat|raw/pdfs|raw/notes|raw/assets)
+      return 0
+      ;;
+    *)
+      return 1
+      ;;
+  esac
+}
+
+missing_optional_raw_dirs() {
+  local wiki_root="$1"
+  local raw_dir
+  local missing=()
+
+  while IFS= read -r raw_dir; do
+    [ -n "$raw_dir" ] || continue
+
+    if is_legacy_required_raw_dir "$raw_dir"; then
+      continue
+    fi
+
+    if [ ! -d "$wiki_root/$raw_dir" ]; then
+      missing+=("$raw_dir")
+    fi
+  done < <(
+    bash "$SOURCE_REGISTRY_SCRIPT" list | awk -F '\t' 'NR > 1 { print $6 }' | LC_ALL=C sort -u
+  )
+
+  if [ "${#missing[@]}" -eq 0 ]; then
+    printf '%s\n' '-'
+  else
+    local IFS=,
+    printf '%s\n' "${missing[*]}"
+  fi
+}
+
+file_presence() {
+  local wiki_root="$1"
+  local relative_path="$2"
+
+  if [ -e "$wiki_root/$relative_path" ]; then
+    printf 'present\n'
+  else
+    printf 'missing\n'
+  fi
+}
+
+validate_layout() {
+  local wiki_root="$1"
+  local failed=0
+  local path
+
+  require_wiki_root "$wiki_root"
+
+  for path in "${REQUIRED_PATHS[@]}"; do
+    if [ ! -e "$wiki_root/$path" ]; then
+      echo "缺少必要路径：$path" >&2
+      failed=1
+    fi
+  done
+
+  for path in "${LEGACY_REQUIRED_RAW_DIRS[@]}"; do
+    if [ ! -d "$wiki_root/$path" ]; then
+      echo "缺少必要旧目录：$path" >&2
+      failed=1
+    fi
+  done
+
+  if [ "$failed" -ne 0 ]; then
+    exit 1
+  fi
+}
+
+source_raw_dir() {
+  local source_id="$1"
+  local record raw_dir
+
+  record="$(
+    bash "$SOURCE_REGISTRY_SCRIPT" get "$source_id" 2>/dev/null
+  )" || {
+    echo "未知来源：$source_id" >&2
+    exit 1
+  }
+
+  IFS=$'\t' read -r _ _ _ _ _ raw_dir _ _ _ _ <<EOF
+$record
+EOF
+
+  printf '%s\n' "$raw_dir"
+}
+
+print_inspect() {
+  local wiki_root="$1"
+  local schema_version language optional_dirs legacy_mode purpose_file cache_file
+
+  validate_layout "$wiki_root"
+
+  schema_version="$(resolved_schema_version "$wiki_root")"
+  language="$(resolved_language "$wiki_root")"
+  optional_dirs="$(missing_optional_raw_dirs "$wiki_root")"
+  purpose_file="$(file_presence "$wiki_root" "purpose.md")"
+  cache_file="$(file_presence "$wiki_root" ".wiki-cache.json")"
+
+  if [ "$schema_version" = "1.0" ] || [ "$optional_dirs" != "-" ] || [ "$purpose_file" = "missing" ] || [ "$cache_file" = "missing" ]; then
+    legacy_mode="yes"
+  else
+    legacy_mode="no"
+  fi
+
+  printf 'wiki_root=%s\n' "$wiki_root"
+  printf 'schema_version=%s\n' "$schema_version"
+  printf 'language=%s\n' "$language"
+  printf 'legacy_mode=%s\n' "$legacy_mode"
+  printf 'migration_required=no\n'
+  printf 'missing_optional_raw_dirs=%s\n' "$optional_dirs"
+  printf 'purpose_file=%s\n' "$purpose_file"
+  printf 'cache_file=%s\n' "$cache_file"
+}
+
+ensure_source_dir() {
+  local wiki_root="$1"
+  local source_id="$2"
+  local raw_dir
+
+  validate_layout "$wiki_root"
+  raw_dir="$(source_raw_dir "$source_id")"
+
+  mkdir -p "$wiki_root/$raw_dir"
+  printf '%s\n' "$wiki_root/$raw_dir"
+}
+
+command_name="${1:-}"
+
+case "$command_name" in
+  inspect)
+    [ "$#" -eq 2 ] || { usage; exit 1; }
+    print_inspect "$2"
+    ;;
+  validate)
+    [ "$#" -eq 2 ] || { usage; exit 1; }
+    print_inspect "$2" > /dev/null
+    ;;
+  ensure-source-dir)
+    [ "$#" -eq 3 ] || { usage; exit 1; }
+    ensure_source_dir "$2" "$3"
+    ;;
+  *)
+    usage
+    exit 1
+    ;;
+esac

package/skills/llm-wiki-skill/setup.sh

@@ -0,0 +1,8 @@
+# 已废弃：请使用 bash install.sh --platform claude
+#!/bin/bash
+# Claude 旧入口兼容包装
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
+
+exec bash "$SCRIPT_DIR/install.sh" --platform claude "$@"

package/skills/llm-wiki-skill/templates/entity-template.md

@@ -0,0 +1,30 @@
+---
+tags: [实体]
+created: {{DATE}}
+updated: {{DATE}}
+sources: []
+---
+
+# {{ENTITY_NAME}}
+
+> 一句话描述这个实体是什么
+
+## 简介
+
+（这个实体的基本介绍）
+
+## 关键信息
+
+- **类型**：（人物 / 组织 / 概念 / 工具 / 事件）
+- **领域**：（所属领域）
+- **相关概念**：（关联的其他实体）
+
+## 详细内容
+
+（从素材中提取的关于这个实体的详细信息）
+
+## 不同素材中的观点
+
+（不同素材对同一个实体的不同描述或评价，标注来源）
+
+## 相关页面

package/skills/llm-wiki-skill/templates/graph-styles/wash/footer.html

@@ -0,0 +1,27 @@
+</script>
+
+<!-- DEPRECATED: Legacy wash template retained for one release cycle after Stage 4 engine HTML migration. Do not extend; use packages/graph-engine instead. -->
+
+<!-- TWEAKS DEFAULTS (editable by host) -->
+<script>
+window.__TWEAKS = /*EDITMODE-BEGIN*/{
+  "variant": "wash",
+  "sizeMode": "uniform",
+  "bubbleMode": "wash",
+  "nodeStyle": "card"
+}/*EDITMODE-END*/;
+</script>
+
+<!-- Libraries -->
+<script src="d3.min.js" defer></script>
+<script src="rough.min.js" defer></script>
+<script src="marked.min.js" defer></script>
+<script src="purify.min.js" defer></script>
+
+<!-- Sample graph data -->
+<!-- Main app -->
+<script src="graph-wash-helpers.js" defer></script>
+<script src="graph-wash.js" defer></script>
+
+</body>
+</html>