Undefined-bot 3.6.7__tar.gz → 3.6.8__tar.gz

{undefined_bot-3.6.7 → undefined_bot-3.6.8}/CHANGELOG.md

@@ -1,3 +1,15 @@
+## v3.6.8 生图响应自动识别与 WebUI request_params 搜索修复
+
+本版本修复 OpenAI 兼容生图链路对 `response_format` 的错误默认与解析依赖，并加固 base64 解码；同时修正 WebUI 配置搜索会把 `request_params` 结构化编辑器里的 key/type 控件单独隐藏的问题。
+
+- 生图响应按 payload **自动识别**格式：`http(s) url` / data URL / `url` 内原始 base64 / `b64_json` / `base64`，不再依赖请求或配置中的 `response_format` 做解析分支。
+- 停止默认写入 `response_format=base64`：仅当工具参数显式传入时覆盖；否则保留 `request_params` 中的值（若有），避免上游网关因默认格式不兼容而失败。
+- 加固 base64 解码：`validate=True` 校验字母表，并用 PNG/JPEG/GIF/BMP/WEBP 魔数确认结果为图片，拒绝把错误文本当成功生图。
+- 修复 WebUI 配置搜索：索引与过滤仅针对带 `data-path` 的表单项，嵌套 key/type 表单组不再被单独隐藏；`request_params` 编辑器占满网格宽度，搜索匹配时 key/type 仍可见。
+- 同步配置文档与回归测试：覆盖 payload 解析回落、edit-form 默认值、base64 校验，以及 config search 对 request_params 嵌套控件的可见性。
+
+---
+
 ## v3.6.7 Naga 会话级黑白名单
 
 本版本将 Naga 启用粒度从进程全局细化到群聊/私聊会话：在总闸打开后，可按 `mode` 名单控制具体群与私聊是否启用 NagaAgent 提示词、相关 Agent 与外部网关能力，并保证提示词与工具暴露/执行结论一致。

{undefined_bot-3.6.7 → undefined_bot-3.6.8}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: Undefined-bot
-Version: 3.6.7
+Version: 3.6.8
 Summary: QQ bot platform with cognitive memory architecture and multi-agent Skills, via OneBot V11.
 Project-URL: Repository, https://github.com/69gg/Undefined
 Project-URL: Issues, https://github.com/69gg/Undefined/issues

{undefined_bot-3.6.7 → undefined_bot-3.6.8}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "Undefined-bot"
-version = "3.6.7"
+version = "3.6.8"
 description = "QQ bot platform with cognitive memory architecture and multi-agent Skills, via OneBot V11."
 readme = "README.md"
 authors = [

{undefined_bot-3.6.7 → undefined_bot-3.6.8}/src/Undefined/__init__.py

@@ -24,7 +24,7 @@ if TYPE_CHECKING:
     from .skills.registry import BaseRegistry as BaseRegistry
     from .skills.tools import ToolRegistry as ToolRegistry
 
-__version__: str = "3.6.7"
+__version__: str = "3.6.8"
 
 # symbol -> (module_path, attribute_name)；首次访问时才 importlib 加载
 _LAZY_IMPORTS: dict[str, tuple[str, str]] = {

{undefined_bot-3.6.7 → undefined_bot-3.6.8}/src/Undefined/skills/agents/entertainment_agent/tools/ai_draw_one/config.json

@@ -24,7 +24,7 @@
                 },
                 "response_format": {
                     "type": "string",
-                    "description": "图片响应格式（仅 models 模式生效）",
+                    "description": "请求偏好的图片响应格式（仅 models 模式生效，影响上游请求体；返回解析始终按 payload 自动识别 url / b64_json / data URL，不依赖此字段）",
                     "enum": ["url", "b64_json", "base64"]
                 },
                 "delivery": {

{undefined_bot-3.6.7 → undefined_bot-3.6.8}/src/Undefined/skills/agents/entertainment_agent/tools/ai_draw_one/handler.py

@@ -41,10 +41,21 @@ _MAX_REFERENCE_IMAGE_UIDS = 16
 _IMAGE_GEN_MODERATION_PROMPT: str | None = None
 
 
+def _is_http_url(value: str) -> bool:
+    """判断字符串是否为 HTTP(S) URL（技能内本地实现，避免跨包耦合）。"""
+    return value.startswith("http://") or value.startswith("https://")
+
+
+def _is_data_url(value: str) -> bool:
+    """判断字符串是否为 ``data:`` URL（技能内本地实现，避免跨包耦合）。"""
+    return value.startswith("data:")
+
+
 @dataclass
 class _GeneratedImagePayload:
     image_url: str | None = None
     image_bytes: bytes | None = None
+    detected_format: str | None = None
 
 
 def _record_image_gen_usage(
@@ -79,24 +90,128 @@ def _iso_now() -> str:
     return datetime.now().isoformat()
 
 
-def _parse_image_url(data: dict[str, Any]) -> str | None:
-    """从 API 响应中提取图片 URL"""
+def _extract_first_image_item(data: dict[str, Any]) -> dict[str, Any] | None:
+    """从 OpenAI 形 ``{"data":[{...}]}`` 取出首个图片 item。"""
     try:
-        return str(data["data"][0]["url"])
+        image_item = data["data"][0]
     except (KeyError, IndexError, TypeError):
         return None
+    if not isinstance(image_item, dict):
+        return None
+    return image_item
+
+
+def _is_likely_image_bytes(image_bytes: bytes) -> bool:
+    """用魔数判断解码结果是否像常见图片，避免把任意 base64 垃圾当图。"""
+    if not image_bytes:
+        return False
+    if image_bytes.startswith(b"\x89PNG\r\n\x1a\n"):
+        return True
+    if image_bytes.startswith(b"\xff\xd8\xff"):
+        return True
+    if image_bytes.startswith((b"GIF87a", b"GIF89a")):
+        return True
+    if image_bytes.startswith(b"BM"):
+        return True
+    if (
+        len(image_bytes) >= 12
+        and image_bytes.startswith(b"RIFF")
+        and image_bytes[8:12] == b"WEBP"
+    ):
+        return True
+    return False
+
+
+def _decode_image_base64(text: str) -> bytes | None:
+    """解码纯 base64 或 ``data:image/...;base64,...`` 图片内容。
+
+    使用 ``validate=True`` 拒绝非字母表字符，并用魔数确认结果是图片，
+    避免错误/文本 payload 被当作成功生图。
+    """
+    payload = text.strip()
+    if not payload:
+        return None
 
+    if _is_data_url(payload):
+        header, _, encoded = payload.partition(",")
+        if not encoded or ";base64" not in header.lower():
+            return None
+        payload = encoded.strip()
+
+    # 部分网关会在长 base64 中插入空白；去掉后再严格解码
+    compact = "".join(payload.split())
+    if not compact:
+        return None
+    padding = (-len(compact)) % 4
+    if padding:
+        compact = f"{compact}{'=' * padding}"
 
-def _parse_image_bytes(data: dict[str, Any]) -> bytes | None:
-    """从 API 响应中提取 Base64 图片内容。"""
     try:
-        image_item = data["data"][0]
-    except (KeyError, IndexError, TypeError):
+        decoded = base64.b64decode(compact, validate=True)
+    except (binascii.Error, ValueError):
         return None
 
-    if not isinstance(image_item, dict):
+    if not _is_likely_image_bytes(decoded):
+        return None
+    return decoded
+
+
+def _parse_image_url(data: dict[str, Any]) -> str | None:
+    """从 API 响应中提取可下载的 http(s) 图片 URL（星之阁等路径复用）。"""
+    image_item = _extract_first_image_item(data)
+    if image_item is None:
+        return None
+    raw_url = image_item.get("url")
+    if raw_url is None:
+        return None
+    text = str(raw_url).strip()
+    if not text or not _is_http_url(text):
+        return None
+    return text
+
+
+def _parse_generated_image(data: dict[str, Any]) -> _GeneratedImagePayload | None:
+    """按响应 payload 自动检测图片格式，不依赖请求/配置中的 response_format。
+
+    优先级：
+    1. ``url`` 为 http(s) → 远程 URL
+    2. ``url`` 为 data URL → 解码为字节
+    3. ``url`` 为原始 base64（部分网关）→ 解码为字节
+    4. ``b64_json`` / ``base64``（可含 data URL 前缀）→ 解码为字节
+    """
+    image_item = _extract_first_image_item(data)
+    if image_item is None:
         return None
 
+    raw_url = image_item.get("url")
+    if raw_url is not None:
+        url_text = str(raw_url).strip()
+        if url_text:
+            if _is_http_url(url_text):
+                return _GeneratedImagePayload(
+                    image_url=url_text,
+                    detected_format="url",
+                )
+            if _is_data_url(url_text):
+                image_bytes = _decode_image_base64(url_text)
+                if image_bytes is not None:
+                    return _GeneratedImagePayload(
+                        image_bytes=image_bytes,
+                        detected_format="data_url",
+                    )
+                logger.warning("图片 data URL 解码失败，继续尝试其它字段")
+            else:
+                # 非 http(s)/data URL：部分网关把原始 base64 塞进 url
+                image_bytes = _decode_image_base64(url_text)
+                if image_bytes is not None:
+                    return _GeneratedImagePayload(
+                        image_bytes=image_bytes,
+                        detected_format="url_base64",
+                    )
+                logger.warning(
+                    "url 字段既非可下载链接也非有效图片 base64，继续尝试其它字段"
+                )
+
     for key in ("b64_json", "base64"):
         raw_value = image_item.get(key)
         if raw_value is None:
@@ -104,22 +219,13 @@ def _parse_image_bytes(data: dict[str, Any]) -> bytes | None:
         text = str(raw_value).strip()
         if not text:
             continue
-        try:
-            return base64.b64decode(text)
-        except (binascii.Error, ValueError):
-            logger.error("图片 Base64 解码失败: key=%s", key)
-            return None
-    return None
-
-
-def _parse_generated_image(data: dict[str, Any]) -> _GeneratedImagePayload | None:
-    image_url = _parse_image_url(data)
-    if image_url:
-        return _GeneratedImagePayload(image_url=image_url)
-
-    image_bytes = _parse_image_bytes(data)
-    if image_bytes is not None:
-        return _GeneratedImagePayload(image_bytes=image_bytes)
+        image_bytes = _decode_image_base64(text)
+        if image_bytes is not None:
+            return _GeneratedImagePayload(
+                image_bytes=image_bytes,
+                detected_format=key,
+            )
+        logger.warning("图片 Base64 解码失败: key=%s", key)
 
     return None
 
@@ -271,10 +377,9 @@ def _build_openai_models_request_body(
         body["quality"] = quality
     if style:
         body["style"] = style
+    # 仅在工具参数显式传入时覆盖；否则保留 request_params 中的值（若有），不默认塞 response_format
     if response_format:
         body["response_format"] = response_format
-    else:
-        body.setdefault("response_format", "base64")
     return body
 
 
@@ -334,10 +439,9 @@ def _build_openai_models_edit_form(
         body["quality"] = quality
     if style:
         body["style"] = style
+    # 仅在工具参数显式传入时覆盖；否则保留 request_params 中的值（若有），不默认塞 response_format
     if response_format:
         body["response_format"] = response_format
-    else:
-        body.setdefault("response_format", "base64")
     return {key: _stringify_multipart_value(value) for key, value in body.items()}
 
 
@@ -508,7 +612,10 @@ async def _call_openai_models(
         logger.error(f"图片生成 API 返回 (未找到图片链接): {data}")
         return f"API 返回原文 (错误：未找到图片内容): {data}"
 
-    logger.info(f"图片生成 API 返回: {data}")
+    logger.info(
+        "图片生成 API 解析成功: detected_format=%s",
+        generated_image.detected_format or "unknown",
+    )
     if generated_image.image_url:
         logger.info(f"提取图片链接: {generated_image.image_url}")
     elif generated_image.image_bytes is not None:
@@ -639,7 +746,10 @@ async def _call_openai_models_edit(
         logger.error(f"参考图生图 API 返回 (未找到图片内容): {data}")
         return f"API 返回原文 (错误：未找到图片内容): {data}"
 
-    logger.info(f"参考图生图 API 返回: {data}")
+    logger.info(
+        "参考图生图 API 解析成功: detected_format=%s",
+        generated_image.detected_format or "unknown",
+    )
     if generated_image.image_url:
         logger.info(f"提取图片链接: {generated_image.image_url}")
     elif generated_image.image_bytes is not None:

{undefined_bot-3.6.7 → undefined_bot-3.6.8}/src/Undefined/webui/static/css/components.css

@@ -35,6 +35,8 @@
 .form-group { margin-bottom: 18px; }
 .form-group.is-hidden { display: none; }
 .form-group.is-match { border: 1px solid rgba(217, 119, 87, 0.4); border-radius: var(--radius-sm); padding: 10px; background: rgba(217, 119, 87, 0.05); }
+/* request_params structured editor needs full row width in form-fields grid */
+.form-group.config-request-params { grid-column: 1 / -1; }
 .form-label { display: block; font-size: 12px; font-weight: 500; color: var(--text-secondary); margin-bottom: 8px; text-transform: capitalize; }
 .form-hint { font-size: 12px; color: var(--text-tertiary); margin-bottom: 8px; }
 .form-control { width: 100%; padding: 10px 14px; border-radius: var(--radius-sm); border: 1px solid var(--border-color); background: var(--bg-input); color: var(--text-primary); font-size: 14px; outline: none; }

{undefined_bot-3.6.7 → undefined_bot-3.6.8}/src/Undefined/webui/static/js/config-form.js

@@ -34,9 +34,12 @@ function updateCommentTexts() {
 }
 
 function updateConfigSearchIndex() {
-    document.querySelectorAll(".form-group").forEach((group) => {
-        const label = group.querySelector(".form-label");
-        const hint = group.querySelector(".form-hint");
+    // Only index real config fields (data-path). Nested structural
+    // form-groups inside request_params editors (key/type) must not
+    // participate, or search will hide them independently of the parent.
+    document.querySelectorAll(".form-group[data-path]").forEach((group) => {
+        const label = group.querySelector(":scope > .form-label");
+        const hint = group.querySelector(":scope > .form-hint");
         const path = group.dataset.path || "";
         group.dataset.searchText =
             `${path} ${label ? label.innerText : ""} ${hint ? hint.innerText : ""}`.toLowerCase();
@@ -184,7 +187,10 @@ function applyConfigFilter() {
     let matchCount = 0;
     document.querySelectorAll(".config-card").forEach((card) => {
         let cardMatches = 0;
-        card.querySelectorAll(".form-group").forEach((group) => {
+        // Only filter top-level config fields. Nested form-groups inside
+        // structured editors (request_params key/type) have no data-path
+        // and must stay visible whenever their parent field is shown.
+        card.querySelectorAll(".form-group[data-path]").forEach((group) => {
             const isMatch =
                 !query || (group.dataset.searchText || "").includes(query);
             group.classList.toggle("is-hidden", !isMatch);
@@ -193,7 +199,7 @@ function applyConfigFilter() {
         });
         card.querySelectorAll(".form-subsection").forEach((section) => {
             section.style.display = section.querySelector(
-                ".form-group:not(.is-hidden)",
+                ".form-group[data-path]:not(.is-hidden)",
             )
                 ? ""
                 : "none";
@@ -516,7 +522,7 @@ function inferScalarType(value) {
 
 function createRequestParamsWidget(path, value) {
     const group = document.createElement("div");
-    group.className = "form-group";
+    group.className = "form-group config-request-params";
     group.dataset.path = path;
 
     const label = document.createElement("label");