feat: add support for CJK angle bracket and trailing attribute separator drift in DSML tool parsing

This commit is contained in:
CJACK
2026-05-10 01:54:31 +08:00
parent 77b6d83266
commit 61d42f8b72
12 changed files with 222 additions and 35 deletions

View File

@@ -40,7 +40,7 @@ Docs: [Overview](README.en.md) / [Architecture](docs/ARCHITECTURE.en.md) / [Depl
- OpenAI / Claude / Gemini protocols are now mounted on one shared `chi` router tree assembled in `internal/server/router.go`.
- Adapter responsibilities are streamlined to: **request normalization → DeepSeek invocation → protocol-shaped rendering**, reducing legacy split-logic paths.
- Tool-calling semantics are aligned between Go and Node runtime: models should output the fullwidth-separator DSML shell `<DSMLtool_calls>``<DSMLinvoke name="...">``<DSMLparameter name="...">`; DS2API also accepts the halfwidth DSML wrapper `<|DSML|tool_calls>`, DSML wrapper aliases such as `<dsml|tool_calls>`, `<|tool_calls>`, `<tool_calls>`, common DSML separator drift such as `<|DSML tool_calls>`, collapsed DSML local names such as `<DSMLtool_calls>`, control-separator drift such as `<DSML␂tool_calls>` / raw STX `\x02`, arbitrary protocol prefixes such as `<proto💥tool_calls>`, and legacy canonical XML `<tool_calls>``<invoke name="...">``<parameter name="...">`. The scanner normalizes fixed local names (`tool_calls` / `invoke` / `parameter`) back to XML before parsing; only wrapped tool blocks or the narrow missing-opening-wrapper repair path enter the tool path, while bare `<invoke>` does not count as supported syntax. JSON literal parameter bodies are preserved as structured values, explicit empty or whitespace-only parameters are preserved as empty strings, malformed complete wrappers are released as plain text, and loose CDATA is narrowly repaired at final parse/flush when it can preserve a complete outer tool call.
- Tool-calling semantics are aligned between Go and Node runtime: models should output the fullwidth-separator DSML shell `<DSMLtool_calls>``<DSMLinvoke name="...">``<DSMLparameter name="...">`; DS2API also accepts the halfwidth DSML wrapper `<|DSML|tool_calls>`, DSML wrapper aliases such as `<dsml|tool_calls>`, `<|tool_calls>`, `<tool_calls>`, common DSML separator drift such as `<|DSML tool_calls>`, collapsed DSML local names such as `<DSMLtool_calls>`, control-separator drift such as `<DSML␂tool_calls>` / raw STX `\x02`, CJK angle bracket and trailing attribute separator drift such as `<DSMparameter name="command">...〈/DSMparameter〉`, arbitrary protocol prefixes such as `<proto💥tool_calls>`, and legacy canonical XML `<tool_calls>``<invoke name="...">``<parameter name="...">`. The scanner normalizes fixed local names (`tool_calls` / `invoke` / `parameter`) back to XML before parsing; only wrapped tool blocks or the narrow missing-opening-wrapper repair path enter the tool path, while bare `<invoke>` does not count as supported syntax. JSON literal parameter bodies are preserved as structured values, explicit empty or whitespace-only parameters are preserved as empty strings, malformed complete wrappers are released as plain text, and loose CDATA is narrowly repaired at final parse/flush when it can preserve a complete outer tool call.
- `Admin API` separates static config from runtime policy: `/admin/config*` for configuration state, `/admin/settings*` for runtime behavior.
- When upstream returns a thinking-only response with no visible text, the Go main path for both streaming and non-streaming completions retries once in the same DeepSeek session: it appends the prompt suffix `"Previous reply had no visible output. Please regenerate the visible final answer or tool call now."` and sets `parent_message_id`. If that same-account retry would still end as `429 upstream_empty_output`, managed-account mode switches to the next available account, creates a fresh session, and retries the original payload once before returning 429.
- Citation/reference marker boundary: streaming output hides upstream `[citation:N]` / `[reference:N]` placeholders by default; non-stream output converts DeepSeek search reference markers into Markdown links.
@@ -355,7 +355,7 @@ When `tools` is present, DS2API performs anti-leak handling:
Additional notes:
- The parser treats the recommended DSML shell tool blocks (`<DSMLtool_calls>` / `<DSMLinvoke name="...">` / `<DSMLparameter name="...">`), halfwidth DSML shell blocks (`<|DSML|tool_calls>` / `<|DSML|invoke name="...">` / `<|DSML|parameter name="...">`), DSML wrapper aliases (`<dsml|tool_calls>`, `<|tool_calls>`, `<tool_calls>`), common DSML separator drift (`<|DSML tool_calls>` / `<|DSML invoke>` / `<|DSML parameter>`), collapsed DSML local names (`<DSMLtool_calls>` / `<DSMLinvoke>` / `<DSMLparameter>`), control-separator drift (`<DSML␂tool_calls>` / raw STX `\x02`), arbitrary protocol prefixes (`<proto💥tool_calls>`), and legacy canonical XML tool blocks (`<tool_calls>` / `<invoke name="...">` / `<parameter name="...">`) as executable tool calls. These shells normalize back to XML first, while internal parsing remains XML-based. Legacy `<tools>`, `<tool_call>`, `<tool_name>`, `<param>`, `<function_call>`, `tool_use`, antml variants, and standalone JSON `tool_calls` payloads are treated as plain text; complete but malformed wrappers are also released as plain text.
- The parser treats the recommended DSML shell tool blocks (`<DSMLtool_calls>` / `<DSMLinvoke name="...">` / `<DSMLparameter name="...">`), halfwidth DSML shell blocks (`<|DSML|tool_calls>` / `<|DSML|invoke name="...">` / `<|DSML|parameter name="...">`), DSML wrapper aliases (`<dsml|tool_calls>`, `<|tool_calls>`, `<tool_calls>`), common DSML separator drift (`<|DSML tool_calls>` / `<|DSML invoke>` / `<|DSML parameter>`), collapsed DSML local names (`<DSMLtool_calls>` / `<DSMLinvoke>` / `<DSMLparameter>`), control-separator drift (`<DSML␂tool_calls>` / raw STX `\x02`), CJK angle bracket and trailing attribute separator drift (`<DSMparameter name="command">...〈/DSMparameter〉`), arbitrary protocol prefixes (`<proto💥tool_calls>`), and legacy canonical XML tool blocks (`<tool_calls>` / `<invoke name="...">` / `<parameter name="...">`) as executable tool calls. These shells normalize back to XML first, while internal parsing remains XML-based. Legacy `<tools>`, `<tool_call>`, `<tool_name>`, `<param>`, `<function_call>`, `tool_use`, antml variants, and standalone JSON `tool_calls` payloads are treated as plain text; complete but malformed wrappers are also released as plain text.
- The parser no longer drops tool calls solely because parameter values are empty; explicit empty strings or whitespace-only parameters become empty strings in structured `tool_calls`. Prompting still tells the model not to emit blank parameters, and missing/empty argument rejection belongs in the tool executor or client schema validation.
- If the final visible response text is empty but the reasoning stream contains an executable tool call, Chat / Responses emits a standard OpenAI `tool_calls` / `function_call` output during finalization. If thinking/reasoning was not enabled by the client, that reasoning text is used only for detection and is not exposed as visible text or `reasoning_content`.
- `tool_calls` shown inside fenced markdown code blocks (for example, ```json ... ```) are treated as examples, not executable calls.

4
API.md
View File

@@ -40,7 +40,7 @@
- OpenAI / Claude / Gemini 三套协议已统一挂在同一 `chi` 路由树上,由 `internal/server/router.go` 负责装配。
- 适配器层职责收敛为:**请求归一化 → DeepSeek 调用 → 协议形态渲染**,减少历史版本中“同能力多处实现”的分叉。
- Tool Calling 的解析策略在 Go 与 Node Runtime 间保持一致:推荐模型输出全角分隔符 DSML 外壳 `<DSMLtool_calls>``<DSMLinvoke name="...">``<DSMLparameter name="...">`;兼容层也接受半角 DSML wrapper `<|DSML|tool_calls>`、DSML wrapper 别名 `<dsml|tool_calls>``<|tool_calls>``<tool_calls>`、常见 DSML 分隔符漏写形态(如 `<|DSML tool_calls>`)、`DSML` 与工具标签名黏连的常见 typo`<DSMLtool_calls>`)、控制分隔符漂移(如 `<DSML␂tool_calls>` / 原始 STX `\x02`)、任意协议前缀壳(如 `<proto💥tool_calls>`),以及旧式 canonical XML `<tool_calls>``<invoke name="...">``<parameter name="...">`。实现上采用结构扫描:只要固定本地标签名是 `tool_calls` / `invoke` / `parameter`,前缀壳会在解析入口归一化;只有 `tool_calls` wrapper 或可修复的缺失 opening wrapper 会进入工具路径,裸 `<invoke>` 不计为已支持语法;流式场景继续执行防泄漏筛分。若参数体本身是合法 JSON 字面量(如 `123``true``null`、数组或对象),会按结构化值输出,不再一律当作字符串;显式空字符串和纯空白参数会结构化保留为空字符串,是否拒绝缺参由工具执行侧决定;完整但 malformed 的 wrapper 会作为普通文本释放,不会吞掉或伪造成工具调用;若 CDATA 偶发漏闭合,则会在最终 parse / flush 恢复阶段做窄修复,尽量保住已完整包裹的外层工具调用。
- Tool Calling 的解析策略在 Go 与 Node Runtime 间保持一致:推荐模型输出全角分隔符 DSML 外壳 `<DSMLtool_calls>``<DSMLinvoke name="...">``<DSMLparameter name="...">`;兼容层也接受半角 DSML wrapper `<|DSML|tool_calls>`、DSML wrapper 别名 `<dsml|tool_calls>``<|tool_calls>``<tool_calls>`、常见 DSML 分隔符漏写形态(如 `<|DSML tool_calls>`)、`DSML` 与工具标签名黏连的常见 typo`<DSMLtool_calls>`)、控制分隔符漂移(如 `<DSML␂tool_calls>` / 原始 STX `\x02`)、CJK 尖括号与属性尾部分隔符漂移(如 `<DSMparameter name="command">...〈/DSMparameter〉`)、任意协议前缀壳(如 `<proto💥tool_calls>`),以及旧式 canonical XML `<tool_calls>``<invoke name="...">``<parameter name="...">`。实现上采用结构扫描:只要固定本地标签名是 `tool_calls` / `invoke` / `parameter`,前缀壳会在解析入口归一化;只有 `tool_calls` wrapper 或可修复的缺失 opening wrapper 会进入工具路径,裸 `<invoke>` 不计为已支持语法;流式场景继续执行防泄漏筛分。若参数体本身是合法 JSON 字面量(如 `123``true``null`、数组或对象),会按结构化值输出,不再一律当作字符串;显式空字符串和纯空白参数会结构化保留为空字符串,是否拒绝缺参由工具执行侧决定;完整但 malformed 的 wrapper 会作为普通文本释放,不会吞掉或伪造成工具调用;若 CDATA 偶发漏闭合,则会在最终 parse / flush 恢复阶段做窄修复,尽量保住已完整包裹的外层工具调用。
- `Admin API` 将配置与运行时策略分开:`/admin/config*` 管静态配置,`/admin/settings*` 管运行时行为。
- 当上游返回 thinking-only 响应模型输出了推理链但无可见文本Go 主路径的流式与非流式补全都会先自动重试一次:以多轮对话 follow-up 方式追加 prompt 后缀 `"Previous reply had no visible output. Please regenerate the visible final answer or tool call now."` 并设置 `parent_message_id` 在同一 DeepSeek session 内让模型重新输出;同账号重试最大 1 次。若同账号重试后仍即将返回 `429 upstream_empty_output`,托管账号模式会在返回 429 前自动切换到下一个可用账号,新建 session用原始 payload 再 fresh retry 一次。
- 引用标记处理边界:流式输出默认隐藏 `[citation:N]` / `[reference:N]` 这类上游内部占位符;非流式输出默认把 DeepSeek 搜索引用标记转换为 Markdown 引用链接。
@@ -357,7 +357,7 @@ data: [DONE]
补充说明:
- **非代码块上下文**下,工具负载即使与普通文本混合,也会按特征识别并产出可执行 tool call前后普通文本仍可透传
- 解析器当前把推荐 DSML 外壳(`<DSMLtool_calls>` / `<DSMLinvoke name="...">` / `<DSMLparameter name="...">`)、半角 DSML 外壳(`<|DSML|tool_calls>` / `<|DSML|invoke name="...">` / `<|DSML|parameter name="...">`、DSML wrapper 别名(`<dsml|tool_calls>``<|tool_calls>``<tool_calls>`)、常见 DSML 分隔符漏写形态(如 `<|DSML tool_calls>` / `<|DSML invoke>` / `<|DSML parameter>`)、`DSML` 与工具标签名黏连的常见 typo`<DSMLtool_calls>` / `<DSMLinvoke>` / `<DSMLparameter>`)、控制分隔符漂移(如 `<DSML␂tool_calls>` / 原始 STX `\x02`)、任意协议前缀壳(如 `<proto💥tool_calls>`)和旧式 canonical XML 工具块(`<tool_calls>` / `<invoke name="...">` / `<parameter name="...">`)作为可执行调用解析;这些前缀壳会先归一化回 XML内部仍以 XML 解析语义为准。旧式 `<tools>``<tool_call>``<tool_name>``<param>``<function_call>``tool_use`、antml 风格与纯 JSON `tool_calls` 片段默认都会按普通文本处理;完整但 malformed 的 wrapper 同样会作为普通文本释放。
- 解析器当前把推荐 DSML 外壳(`<DSMLtool_calls>` / `<DSMLinvoke name="...">` / `<DSMLparameter name="...">`)、半角 DSML 外壳(`<|DSML|tool_calls>` / `<|DSML|invoke name="...">` / `<|DSML|parameter name="...">`、DSML wrapper 别名(`<dsml|tool_calls>``<|tool_calls>``<tool_calls>`)、常见 DSML 分隔符漏写形态(如 `<|DSML tool_calls>` / `<|DSML invoke>` / `<|DSML parameter>`)、`DSML` 与工具标签名黏连的常见 typo`<DSMLtool_calls>` / `<DSMLinvoke>` / `<DSMLparameter>`)、控制分隔符漂移(如 `<DSML␂tool_calls>` / 原始 STX `\x02`)、CJK 尖括号与属性尾部分隔符漂移(如 `<DSMparameter name="command">...〈/DSMparameter〉`)、任意协议前缀壳(如 `<proto💥tool_calls>`)和旧式 canonical XML 工具块(`<tool_calls>` / `<invoke name="...">` / `<parameter name="...">`)作为可执行调用解析;这些前缀壳会先归一化回 XML内部仍以 XML 解析语义为准。旧式 `<tools>``<tool_call>``<tool_name>``<param>``<function_call>``tool_use`、antml 风格与纯 JSON `tool_calls` 片段默认都会按普通文本处理;完整但 malformed 的 wrapper 同样会作为普通文本释放。
- 解析层不会因为参数值为空而丢弃工具调用;显式空字符串或纯空白参数会按空字符串进入结构化 `tool_calls`。Prompt 会要求模型不要主动输出空参数,缺参/空命令的拒绝应由工具执行侧或客户端 schema 校验负责。
- 当最终可见正文为空但思维链里包含可执行工具调用时Chat / Responses 会在收尾阶段补发标准 OpenAI `tool_calls` / `function_call` 输出;如果客户端未开启 thinking / reasoning该思维链只用于检测不会作为可见正文或 `reasoning_content` 暴露。
- Markdown fenced code block例如 ```json ... ```)中的 `tool_calls` 仅视为示例文本,不会被执行。

View File

@@ -168,7 +168,7 @@ OpenAI Chat / Responses 在标准化后、current input file 之前,会默认
4. 把这整段内容并入 system prompt。
工具调用正例现在优先示范全角分隔符 DSML 风格:`<DSMLtool_calls>``<DSMLinvoke name="...">``<DSMLparameter name="...">`
兼容层仍接受旧式纯 `<tool_calls>` wrapper并会容错若干 DSML 标签变体,包括短横线形式 `<dsml-tool-calls>` / `<dsml-invoke>` / `<dsml-parameter>`、下划线形式 `<dsml_tool_calls>` / `<dsml_invoke>` / `<dsml_parameter>`,以及其他前缀分隔形态如 `<vendor|tool_calls>` / `<vendor_tool_calls>` / `<vendor - tool_calls>`;标签壳扫描还会把全角 ASCII 漂移归一化,例如 `<tool_calls>` 与全角 `` 结束符。更一般地Go / Node tag 扫描以固定本地标签名 `tool_calls` / `invoke` / `parameter` 为准,标签名前任意协议前缀壳都会在解析入口剥离,例如 `<DSML␂tool_calls>``<proto💥tool_calls>` 这类控制符或非 ASCII 分隔符漂移也会归一化回现有 XML 标签后继续走同一套 parser。但提示词会优先要求模型输出官方 DSML 标签,并强调不能只输出 closing wrapper 而漏掉 opening tag。需要注意这是“兼容 DSML 外壳,内部仍以 XML 解析语义为准”,不是原生 DSML 全链路实现。解析器会先截获非代码块中的疑似工具 wrapper完整解析失败或工具语义无效时再按普通文本放行。
兼容层仍接受旧式纯 `<tool_calls>` wrapper并会容错若干 DSML 标签变体,包括短横线形式 `<dsml-tool-calls>` / `<dsml-invoke>` / `<dsml-parameter>`、下划线形式 `<dsml_tool_calls>` / `<dsml_invoke>` / `<dsml_parameter>`,以及其他前缀分隔形态如 `<vendor|tool_calls>` / `<vendor_tool_calls>` / `<vendor - tool_calls>`;标签壳扫描还会把全角 ASCII 漂移归一化,例如 `<tool_calls>` 与全角 `` 结束符,也会容错 CJK 尖括号和属性尾部分隔符漂移,例如 `<DSMparameter name="command">...〈/DSMparameter〉`。更一般地Go / Node tag 扫描以固定本地标签名 `tool_calls` / `invoke` / `parameter` 为准,标签名前任意协议前缀壳都会在解析入口剥离,例如 `<DSML␂tool_calls>``<proto💥tool_calls>` 这类控制符或非 ASCII 分隔符漂移也会归一化回现有 XML 标签后继续走同一套 parser。但提示词会优先要求模型输出官方 DSML 标签,并强调不能只输出 closing wrapper 而漏掉 opening tag。需要注意这是“兼容 DSML 外壳,内部仍以 XML 解析语义为准”,不是原生 DSML 全链路实现。解析器会先截获非代码块中的疑似工具 wrapper完整解析失败或工具语义无效时再按普通文本放行。
数组参数使用 `<item>...</item>` 子节点表示;当某个参数体只包含 item 子节点时Go / Node 解析器会把它还原成数组,避免 `questions` / `options` 这类 schema 中要求 array 的参数被误解析成 `{ "item": ... }` 对象。除此之外,解析器还会回收一些更松散的列表写法,例如 JSON array 字面量或逗号分隔的 JSON 项序列,只要它们足够明确;但 `<item>` 仍然是首选形态。若模型把完整结构化 XML fragment 误包进 CDATA兼容层会在保护 `content` / `command` 等原文字段的前提下,尝试把非原文字段中的 CDATA XML fragment 还原成 object / array。不过如果 CDATA 只是单个平面的 XML/HTML 标签,例如 `<b>urgent</b>` 这种行内标记,兼容层会保留原始字符串,不会强行升成 object / array只有明显表示结构的 CDATA 片段,例如多兄弟节点、嵌套子节点或 `item` 列表,才会触发结构化恢复。对 `command` / `content` 等长文本参数CDATA 内部的 Markdown fenced DSML / XML 示例会作为原文保护;示例里的 `]]></parameter>``</tool_calls>` 不会截断外层工具调用,解析器会继续等待围栏外真正的参数 / wrapper 结束标签。
Go 侧读取 DeepSeek SSE 时不再依赖 `bufio.Scanner` 的固定 2MiB 单行上限;当写文件类工具把很长的 `content` 放在单个 `data:` 行里返回时,非流式收集、流式解析和 auto-continue 透传都会保留完整行,再进入同一套工具解析与序列化流程。
在 assistant 最终回包阶段,如果某个 tool 参数在声明 schema 中明确是 `string`,兼容层会在把解析后的 `tool_calls` / `function_call` 重新序列化成 OpenAI / Responses / Claude 可见参数前,递归把该路径上的 number / bool / object / array 统一转成字符串;其中 object / array 会压成紧凑 JSON 字符串。这个保护只对 schema 明确声明为 string 的路径生效,不会改写本来就是 `number` / `boolean` / `object` / `array` 的参数。这样可以兼容 DeepSeek 输出了结构化片段、但上游客户端工具 schema 又严格要求字符串参数的场景(例如 `content``prompt``path``taskId` 等)。

View File

@@ -39,8 +39,8 @@
兼容修复:
- 如果模型漏掉 opening wrapper但后面仍输出了一个或多个 invoke 并以 closing wrapper 收尾Go 解析链路会在解析前补回缺失的 opening wrapper。
- Go / Node 解析层不再枚举每一种 DSML typo。它以固定本地标签名 `tool_calls` / `invoke` / `parameter` 为准,把标签名前的任意协议前缀壳视为可容忍噪声,并继续兼容管道符 `|` / ``、空白、重复 leading `<`、可视控制符 `␂`、原始 STX `\x02`、非 ASCII 分隔符等漂移。例如 `<DSML|tool_calls>``<<|DSML|tool_calls>``<|DSML tool_calls>``<DSMLtool_calls>``<<DSML|DSML|tool_calls>``<DSML␂tool_calls>``<proto💥tool_calls>` 都会归一化;相似但非固定标签名(如 `tool_calls_extra`)仍按普通文本处理。
- 如果模型在固定工具标签名后多输出一个尾部管道符,例如 `<|DSML|tool_calls|` / `<|DSML|invoke|` / `<|DSML|parameter|`,兼容层会把这个尾部 `|` 当作异常标签终止符并补齐缺失的 `>`;如果后面已经有 `>`,也会消费这个多余 `|` 后再归一化。
- Go / Node 解析层不再枚举每一种 DSML typo。它以固定本地标签名 `tool_calls` / `invoke` / `parameter` 为准,把标签名前的任意协议前缀壳视为可容忍噪声,并继续兼容管道符 `|` / ``、空白、重复 leading `<`、可视控制符 `␂`、原始 STX `\x02`、非 ASCII 分隔符、CJK 尖括号 `〈` / `〉` 等漂移。例如 `<DSML|tool_calls>``<<|DSML|tool_calls>``<|DSML tool_calls>``<DSMLtool_calls>``<<DSML|DSML|tool_calls>``<DSML␂tool_calls>``<proto💥tool_calls>``<DSMtool_calls>...〈/DSMtool_calls〉` 都会归一化;相似但非固定标签名(如 `tool_calls_extra`)仍按普通文本处理。
- 如果模型在固定工具标签名后多输出一个尾部管道符,例如 `<|DSML|tool_calls|` / `<|DSML|invoke|` / `<|DSML|parameter|`或在带属性标签的结束符前多输出一个尾部管道符(如 `<DSMparameter name="command">`兼容层会把这个尾部 `|` / `` 当作异常标签终止符并补齐或归一化;如果后面已经有 `>` / `〉`,也会消费这个多余分隔符后再归一化。
- 这是一个针对常见模型失误的窄修复不改变推荐输出格式prompt 仍要求模型直接输出完整 DSML 外壳。
-`<invoke ...>` / `<parameter ...>` 不会被当成“已支持的工具语法”;只有 `tool_calls` wrapper 或可修复的缺失 opening wrapper 才会进入工具调用路径。
@@ -102,7 +102,7 @@ go test -v -run 'TestParseToolCalls|TestProcessToolSieve' ./internal/toolcall ./
- DSML `<DSMLtool_calls>` wrapper 正常解析
- legacy canonical `<tool_calls>` wrapper 正常解析
- 固定本地标签名的 DSML 噪声容错形态(如 `<DSML|tool_calls>``<<|DSML|tool_calls>``<|DSML tool_calls>``<DSMLtool_calls>``<<DSML|DSML|tool_calls>`)正常解析
- 固定本地标签名的 DSML 噪声容错形态(如 `<DSML|tool_calls>``<<|DSML|tool_calls>``<|DSML tool_calls>``<DSMLtool_calls>``<<DSML|DSML|tool_calls>``<DSMtool_calls>...〈/DSMtool_calls〉`)正常解析
- 混搭标签DSML wrapper + canonical inner归一化后正常解析
- 波浪线围栏 `~~~` 内的示例不执行
- 嵌套围栏4 反引号嵌套 3 反引号)内的示例不执行

View File

@@ -1,6 +1,6 @@
'use strict';
const CDATA_PATTERN = /^<!\[CDATA\[([\s\S]*?)]](?:>|)$/i;
const CDATA_PATTERN = /^(?:<|〈)!\[CDATA\[([\s\S]*?)]](?:>||〉)$/i;
const XML_ATTR_PATTERN = /\b([a-z0-9_:-]+)\s*=\s*("([^"]*)"|'([^']*)')/gi;
const TOOL_MARKUP_NAMES = [
{ raw: 'tool_calls', canonical: 'tool_calls' },
@@ -431,26 +431,31 @@ function skipXmlIgnoredSection(lower, i) {
function findCDATAEnd(text, from) {
const ascii = text.indexOf(']]>', from);
const fullwidth = text.indexOf(']]', from);
if (ascii < 0 && fullwidth < 0) {
const cjk = text.indexOf(']]〉', from);
if (ascii < 0 && fullwidth < 0 && cjk < 0) {
return { index: -1, len: 0 };
}
if (ascii < 0) {
return { index: fullwidth, len: ']]'.length };
let best = { index: -1, len: 0 };
for (const candidate of [
{ index: ascii, len: ']]>'.length },
{ index: fullwidth, len: ']]'.length },
{ index: cjk, len: ']]〉'.length },
]) {
if (candidate.index >= 0 && (best.index < 0 || candidate.index < best.index)) {
best = candidate;
}
}
if (fullwidth < 0 || ascii < fullwidth) {
return { index: ascii, len: ']]>'.length };
}
return { index: fullwidth, len: ']]'.length };
return best;
}
function scanToolMarkupTagAt(text, start) {
const raw = toStringSafe(text);
if (!raw || start < 0 || start >= raw.length || raw[start] !== '<') {
if (!raw || start < 0 || start >= raw.length || normalizeFullwidthASCIIChar(raw[start]) !== '<') {
return null;
}
const lower = raw.toLowerCase();
let i = start + 1;
while (i < raw.length && raw[i] === '<') {
while (i < raw.length && normalizeFullwidthASCIIChar(raw[i]) === '<') {
i += 1;
}
const closing = raw[i] === '/';
@@ -822,6 +827,12 @@ function normalizeFullwidthASCIIChar(ch) {
if (!ch) {
return ch;
}
if (ch === '〈') {
return '<';
}
if (ch === '〉') {
return '>';
}
const code = ch.charCodeAt(0);
if (code >= 0xff01 && code <= 0xff5e) {
return String.fromCharCode(code - 0xfee0);
@@ -858,9 +869,28 @@ function matchNormalizedASCII(raw, start, expected) {
function normalizeToolMarkupTagTailForXML(tail) {
let out = '';
for (const ch of typeof tail === 'string' ? tail : String(tail || '')) {
const raw = typeof tail === 'string' ? tail : String(tail || '');
let quote = '';
for (let i = 0; i < raw.length; i += 1) {
const ch = raw[i];
const normalized = normalizeFullwidthASCIIChar(ch);
if (['>', '/', '=', '"', "'"].includes(normalized)) {
if (quote) {
out += normalized;
if (normalized === quote) {
quote = '';
}
} else if (normalized === '"' || normalized === "'") {
quote = normalized;
out += normalized;
} else if (normalized === '|') {
let j = i + 1;
while (j < raw.length && [' ', '\t', '\r', '\n'].includes(raw[j])) {
j += 1;
}
if (normalizeFullwidthASCIIChar(raw[j] || '') !== '>') {
out += normalized;
}
} else if (['>', '/', '='].includes(normalized)) {
out += normalized;
} else {
out += ch;

View File

@@ -65,6 +65,7 @@ func normalizeToolMarkupTagTailForXML(tail string) string {
}
var b strings.Builder
b.Grow(len(tail))
quote := rune(0)
for i := 0; i < len(tail); {
r, size := utf8.DecodeRuneInString(tail[i:])
if r == utf8.RuneError && size == 1 {
@@ -72,9 +73,38 @@ func normalizeToolMarkupTagTailForXML(tail string) string {
i++
continue
}
switch normalizeFullwidthASCII(r) {
case '>', '/', '=', '"', '\'':
b.WriteRune(normalizeFullwidthASCII(r))
ch := normalizeFullwidthASCII(r)
if quote != 0 {
b.WriteRune(ch)
if ch == quote {
quote = 0
}
i += size
continue
}
switch ch {
case '"', '\'':
quote = ch
b.WriteRune(ch)
case '|':
j := i + size
for j < len(tail) {
next, nextSize := utf8.DecodeRuneInString(tail[j:])
if nextSize <= 0 {
break
}
if next == ' ' || next == '\t' || next == '\r' || next == '\n' {
j += nextSize
continue
}
break
}
next, _ := normalizedASCIIAt(tail, j)
if next != '>' {
b.WriteRune(ch)
}
case '>', '/', '=':
b.WriteRune(ch)
default:
b.WriteString(tail[i : i+size])
}

View File

@@ -10,7 +10,7 @@ import (
var toolCallMarkupKVPattern = regexp.MustCompile(`(?is)<(?:[a-z0-9_:-]+:)?([a-z0-9_\-.]+)\b[^>]*>(.*?)</(?:[a-z0-9_:-]+:)?([a-z0-9_\-.]+)>`)
// cdataPattern matches a standalone CDATA section.
var cdataPattern = regexp.MustCompile(`(?is)^<!\[CDATA\[(.*?)]](?:>|)$`)
var cdataPattern = regexp.MustCompile(`(?is)^(?:<|〈)!\[CDATA\[(.*?)]](?:>||〉)$`)
func parseMarkupKVObject(text string) map[string]any {
matches := toolCallMarkupKVPattern.FindAllStringSubmatch(strings.TrimSpace(text), -1)

View File

@@ -305,19 +305,23 @@ func indexToolCDATAClose(text string, from int) int {
}
asciiIdx := strings.Index(text[from:], "]]>")
fullIdx := strings.Index(text[from:], "]]")
if asciiIdx < 0 && fullIdx < 0 {
cjkIdx := strings.Index(text[from:], "]]〉")
if asciiIdx < 0 && fullIdx < 0 && cjkIdx < 0 {
return -1
}
if asciiIdx < 0 {
return from + fullIdx
best := -1
for _, idx := range []int{asciiIdx, fullIdx, cjkIdx} {
if idx >= 0 && (best < 0 || idx < best) {
best = idx
}
}
if fullIdx < 0 || asciiIdx < fullIdx {
return from + asciiIdx
}
return from + fullIdx
return from + best
}
func toolCDATACloseLenAt(text string, idx int) int {
if strings.HasPrefix(text[idx:], "]]〉") {
return len("]]〉")
}
if strings.HasPrefix(text[idx:], "]]") {
return len("]]")
}

View File

@@ -134,12 +134,17 @@ func FindMatchingToolMarkupClose(text string, open ToolMarkupTag) (ToolMarkupTag
}
func scanToolMarkupTagAt(text string, start int) (ToolMarkupTag, bool) {
if start < 0 || start >= len(text) || text[start] != '<' {
next, ok := consumeToolMarkupLessThan(text, start)
if !ok {
return ToolMarkupTag{}, false
}
i := start + 1
for i < len(text) && text[i] == '<' {
i++
i := next
for {
next, ok := consumeToolMarkupLessThan(text, i)
if !ok {
break
}
i = next
}
closing := false
if i < len(text) && text[i] == '/' {
@@ -459,6 +464,14 @@ func consumeToolMarkupPipe(text string, idx int) (int, bool) {
return idx, false
}
func consumeToolMarkupLessThan(text string, idx int) (int, bool) {
ch, size := normalizedASCIIAt(text, idx)
if size <= 0 || ch != '<' {
return idx, false
}
return idx + size, true
}
func hasToolMarkupBoundary(text string, idx int) bool {
if idx >= len(text) {
return true
@@ -488,6 +501,12 @@ func normalizedASCIIAt(text string, idx int) (byte, int) {
}
func normalizeFullwidthASCII(r rune) rune {
switch r {
case '〈':
return '<'
case '〉':
return '>'
}
if r >= '' && r <= '' {
return r - 0xFEE0
}

View File

@@ -132,6 +132,37 @@ func TestParseToolCallsSupportsFullwidthDSMLShell(t *testing.T) {
}
}
func TestParseToolCallsSupportsCJKAngleDSMDrift(t *testing.T) {
text := `<DSMtool_calls>
<DSMinvoke name="Bash">
<DSMparameter name="description">〈![CDATA[Show commits on local dev not on origin/dev]]〉〈/DSMparameter〉
<DSMparameter name="command">〈![CDATA[git log --oneline origin/dev..dev]]〉〈/DSMparameter〉
〈/DSMinvoke〉
<DSMinvoke name="Bash">
<DSMparameter name="description">〈![CDATA[Show commits on origin/dev not on local dev]]〉〈/DSMparameter〉
<DSMparameter name="command">〈![CDATA[git log --oneline dev..origin/dev]]〉〈/DSMparameter〉
〈/DSMinvoke〉
<DSMinvoke name="Bash">
<DSMparameter name="description">〈![CDATA[Check tracking branch status]]〉〈/DSMparameter〉
<DSMparameter name="command">〈![CDATA[git status -b --short]]〉〈/DSMparameter〉
〈/DSMinvoke〉
〈/DSMtool_calls〉`
calls := ParseToolCalls(text, []string{"Bash"})
if len(calls) != 3 {
t.Fatalf("expected three CJK-angle DSM drift calls, got %#v", calls)
}
if calls[0].Name != "Bash" || calls[0].Input["command"] != "git log --oneline origin/dev..dev" {
t.Fatalf("unexpected first CJK-angle DSM drift call: %#v", calls[0])
}
if calls[1].Name != "Bash" || calls[1].Input["description"] != "Show commits on origin/dev not on local dev" {
t.Fatalf("unexpected second CJK-angle DSM drift call: %#v", calls[1])
}
if calls[2].Name != "Bash" || calls[2].Input["command"] != "git status -b --short" {
t.Fatalf("unexpected third CJK-angle DSM drift call: %#v", calls[2])
}
}
func TestParseToolCallsIgnoresBareHyphenatedToolCallsLookalike(t *testing.T) {
text := `<tool-calls><invoke name="Bash"><parameter name="command">pwd</parameter></invoke></tool-calls>`
calls := ParseToolCalls(text, []string{"Bash"})

View File

@@ -1228,3 +1228,37 @@ func TestProcessToolSieveDSMLBarePrefixVariantDoesNotLeak(t *testing.T) {
t.Fatalf("expected one tool call from DSML bare prefix variant, got %d events=%#v", toolCalls, events)
}
}
func TestProcessToolSieveCJKAngleDSMDriftDoesNotLeak(t *testing.T) {
var state State
chunks := []string{
"<DSMtool_calls>\n",
"<DSMinvoke name=\"Bash\">\n",
"<DSMparameter name=\"description\">〈![CDATA[Check tracking branch status]]〉〈/DSMparameter〉\n",
"<DSMparameter name=\"command\">〈![CDATA[git status -b --short]]〉〈/DSMparameter〉\n",
"〈/DSMinvoke〉\n",
"〈/DSMtool_calls〉",
}
var events []Event
for _, c := range chunks {
events = append(events, ProcessChunk(&state, c, []string{"Bash"})...)
}
events = append(events, Flush(&state, []string{"Bash"})...)
var textContent string
var calls []toolcall.ParsedToolCall
for _, evt := range events {
textContent += evt.Content
calls = append(calls, evt.ToolCalls...)
}
if strings.Contains(textContent, "DSM") || strings.Contains(textContent, "git status") {
t.Fatalf("CJK-angle DSM drift leaked to text: %q events=%#v", textContent, events)
}
if len(calls) != 1 {
t.Fatalf("expected one CJK-angle DSM drift tool call, got %d events=%#v", len(calls), events)
}
if calls[0].Name != "Bash" || calls[0].Input["command"] != "git status -b --short" {
t.Fatalf("unexpected CJK-angle DSM drift call: %#v", calls[0])
}
}

View File

@@ -129,6 +129,29 @@ test('parseToolCalls parses fullwidth DSML shell drift', () => {
assert.deepEqual(calls[1].input, { file_path: '/Users/aq/Desktop/myproject/Personal_Blog/index.html' });
});
test('parseToolCalls parses CJK-angle DSM drift', () => {
const payload = `<DSMtool_calls>
<DSMinvoke name="Bash">
<DSMparameter name="description">〈![CDATA[Show commits on local dev not on origin/dev]]〉〈/DSMparameter〉
<DSMparameter name="command">〈![CDATA[git log --oneline origin/dev..dev]]〉〈/DSMparameter〉
〈/DSMinvoke〉
<DSMinvoke name="Bash">
<DSMparameter name="description">〈![CDATA[Show commits on origin/dev not on local dev]]〉〈/DSMparameter〉
<DSMparameter name="command">〈![CDATA[git log --oneline dev..origin/dev]]〉〈/DSMparameter〉
〈/DSMinvoke〉
<DSMinvoke name="Bash">
<DSMparameter name="description">〈![CDATA[Check tracking branch status]]〉〈/DSMparameter〉
<DSMparameter name="command">〈![CDATA[git status -b --short]]〉〈/DSMparameter〉
〈/DSMinvoke〉
〈/DSMtool_calls〉`;
const calls = parseToolCalls(payload, ['Bash']);
assert.equal(calls.length, 3);
assert.equal(calls[0].name, 'Bash');
assert.equal(calls[0].input.command, 'git log --oneline origin/dev..dev');
assert.equal(calls[1].input.description, 'Show commits on origin/dev not on local dev');
assert.equal(calls[2].input.command, 'git status -b --short');
});
test('parseToolCalls parses DSML control separator drift', () => {
for (const sep of ['␂', '\x02']) {
const payload = `<DSML${sep}tool_calls>
@@ -523,6 +546,22 @@ test('sieve emits tool_calls for arbitrary-prefixed tool tags', () => {
assert.equal(text.includes('💥'), false);
});
test('sieve emits tool_calls for CJK-angle DSM drift', () => {
const events = runSieve([
'<DSMtool_calls>\n',
'<DSMinvoke name="Bash">\n',
'<DSMparameter name="description">〈![CDATA[Check tracking branch status]]〉〈/DSMparameter〉\n',
'<DSMparameter name="command">〈![CDATA[git status -b --short]]〉〈/DSMparameter〉\n',
'〈/DSMinvoke〉\n',
'〈/DSMtool_calls〉',
], ['Bash']);
const finalCalls = events.flatMap((evt) => (evt.type === 'tool_calls' ? evt.calls : []));
assert.equal(finalCalls.length, 1);
assert.equal(finalCalls[0].name, 'Bash');
assert.equal(finalCalls[0].input.command, 'git status -b --short');
assert.equal(collectText(events), '');
});
test('sieve emits all-empty arbitrary-prefixed tool tags without leaking text', () => {
const payload = [
'<TDSMLtool_calls>\n',