From 2977be33c639c4bfc0f06413a6758129876cfe8f Mon Sep 17 00:00:00 2001 From: Liang Jiaqing Date: Thu, 9 Apr 2026 18:43:20 +0800 Subject: [PATCH] Update llmcore truncation, adb_ui parsing, stapp regex, and subagent docs --- frontends/stapp.py | 6 +++--- llmcore.py | 49 +++++++++++++++++++++++++--------------------- memory/adb_ui.py | 14 ++++++++----- memory/subagent.md | 1 + 4 files changed, 40 insertions(+), 30 deletions(-) diff --git a/frontends/stapp.py b/frontends/stapp.py index 10525b4..17e0ffc 100644 --- a/frontends/stapp.py +++ b/frontends/stapp.py @@ -79,9 +79,9 @@ def fold_turns(text): turns.append((marker, content)) for idx, (marker, content) in enumerate(turns): if idx < len(turns) - 1: - m = re.search(r'\s*(.*?)\s*', content, re.DOTALL) - if m: - title = m.group(1).strip() + matches = re.findall(r'\s*((?:(?!).)*?)\s*', content, re.DOTALL) + if matches: + title = matches[-1].strip() title = title.split('\n')[0] if len(title) > 50: title = title[:50] + '...' else: title = marker.strip('*') diff --git a/llmcore.py b/llmcore.py index dc53603..9549afa 100644 --- a/llmcore.py +++ b/llmcore.py @@ -22,18 +22,28 @@ def compress_history_tags(messages, keep_recent=10, max_len=800, force=False): _before = sum(len(json.dumps(m, ensure_ascii=False)) for m in messages) _pats = {tag: re.compile(rf'(<{tag}>)([\s\S]*?)()') for tag in ('thinking', 'think', 'tool_use', 'tool_result')} _hist_pat = re.compile(r'<(history|key_info)>[\s\S]*?') + def _trunc_str(s): return s[:max_len//2] + '\n...[Truncated]...\n' + s[-max_len//2:] if isinstance(s, str) and len(s) > max_len else s def _trunc(text): text = _hist_pat.sub(lambda m: f'<{m.group(1)}>[...]', text) - for pat in _pats.values(): text = pat.sub(lambda m: m.group(1) + m.group(2)[:max_len] + '...' + m.group(3) if len(m.group(2)) > max_len else m.group(0), text) + for pat in _pats.values(): text = pat.sub(lambda m: m.group(1) + _trunc_str(m.group(2)) + m.group(3), text) return text for i, msg in enumerate(messages): if i >= len(messages) - keep_recent: break c = msg['content'] if isinstance(c, str): msg['content'] = _trunc(c) elif isinstance(c, list): - for block in c: - if isinstance(block, dict) and block.get('type') == 'text' and isinstance(block.get('text'), str): - block['text'] = _trunc(block['text']) + for b in c: + if not isinstance(b, dict): continue + t = b.get('type') + if t == 'text' and isinstance(b.get('text'), str): b['text'] = _trunc(b['text']) + elif t == 'tool_result': + tc = b.get('content') + if isinstance(tc, str): b['content'] = _trunc_str(tc) + elif isinstance(tc, list): + for sub in tc: + if isinstance(sub, dict) and sub.get('type') == 'text': sub['text'] = _trunc_str(sub.get('text')) + elif t == 'tool_use' and isinstance(b.get('input'), dict): + for k, v in b['input'].items(): b['input'][k] = _trunc_str(v) print(f"[Cut] {_before} -> {sum(len(json.dumps(m, ensure_ascii=False)) for m in messages)}") return messages @@ -420,8 +430,8 @@ class BaseSession: proxy = cfg.get('proxy') self.proxies = {"http": proxy, "https": proxy} if proxy else None self.max_retries = max(0, int(cfg.get('max_retries', 2))) - self.connect_timeout = max(1, int(cfg.get('connect_timeout', 10))) - self.read_timeout = max(5, int(cfg.get('read_timeout', 60))) + self.connect_timeout = max(1, int(cfg.get('timeout', 5))) + self.read_timeout = max(5, int(cfg.get('read_timeout', 30))) effort = cfg.get('reasoning_effort') effort = None if effort is None else str(effort).strip().lower() self.reasoning_effort = effort if effort in ('none', 'minimal', 'low', 'medium', 'high', 'xhigh') else None @@ -457,13 +467,13 @@ class ClaudeSession(BaseSession): headers = {"x-api-key": self.api_key, "Content-Type": "application/json", "anthropic-version": "2023-06-01", "anthropic-beta": "prompt-caching-2024-07-31"} payload = {"model": model, "messages": messages, "temperature": temperature, "max_tokens": max_tokens, "stream": True} if self.system: payload["system"] = [{"type": "text", "text": self.system, "cache_control": {"type": "persistent"}}] - content_blocks = [] try: - with requests.post(auto_make_url(self.api_base, "messages"), headers=headers, json=payload, stream=True, timeout=(5,30)) as r: - r.raise_for_status() - content_blocks = yield from _parse_claude_sse(r.iter_lines()) - except Exception as e: yield f"Error: {str(e)}" - return content_blocks or [] + with requests.post(auto_make_url(self.api_base, "messages"), headers=headers, json=payload, stream=True, timeout=(self.connect_timeout, self.read_timeout)) as r: + if r.status_code != 200: raise Exception(f"HTTP {r.status_code} {r.text[:500]}") + return (yield from _parse_claude_sse(r.iter_lines())) or [] + except Exception as e: + yield (err := f"Error: {e}") + return [{"type": "text", "text": err}] def make_messages(self, raw_list): msgs = [{"role": m['role'], "content": list(m['content'])} for m in raw_list] c = msgs[-1]["content"] @@ -508,17 +518,12 @@ class NativeClaudeSession(BaseSession): messages[-1] = {**messages[-1], "content": list(messages[-1]["content"])} messages[-1]["content"][-1] = dict(messages[-1]["content"][-1], cache_control={"type": "ephemeral"}) try: - resp = requests.post(auto_make_url(self.api_base, "messages"), headers=headers, json=payload, stream=True, timeout=60) - if resp.status_code != 200: - error_msg = f"Error: HTTP {resp.status_code} {resp.text[:500]}" - yield error_msg - return [{"type": "text", "text": error_msg}] + resp = requests.post(auto_make_url(self.api_base, "messages"), headers=headers, json=payload, stream=True, timeout=(self.connect_timeout, self.read_timeout)) + if resp.status_code != 200: raise Exception(f"HTTP {resp.status_code} {resp.text[:500]}") + return (yield from _parse_claude_sse(resp.iter_lines())) or [] except Exception as e: - error_msg = f"Error: {e}" - yield error_msg - return [{"type": "text", "text": error_msg}] - content_blocks = yield from _parse_claude_sse(resp.iter_lines()) - return content_blocks or [] + yield (err := f"Error: {e}") + return [{"type": "text", "text": err}] def ask(self, msg, model=None): assert type(msg) is dict diff --git a/memory/adb_ui.py b/memory/adb_ui.py index b2a5fc1..e69cf74 100644 --- a/memory/adb_ui.py +++ b/memory/adb_ui.py @@ -42,7 +42,7 @@ def _parse_xml(xml_str, keyword=None, clickable_only=False, raw=False): cls = n.get("class", "").split(".")[-1] rid = n.get("resource-id", "") label = text or desc - if not label and not raw: continue + if not label and not click and not raw: continue if clickable_only and not click: continue if keyword and keyword.lower() not in label.lower(): continue cx, cy = 0, 0 @@ -51,8 +51,8 @@ def _parse_xml(xml_str, keyword=None, clickable_only=False, raw=False): if len(m) == 2: cx = (int(m[0][0]) + int(m[1][0])) // 2 cy = (int(m[0][1]) + int(m[1][1])) // 2 - nodes.append({"text": text or desc, "click": click, - "bounds": bounds, "cx": cx, "cy": cy, "class": cls, "id": rid}) + edit = cls == "EditText" + nodes.append({"text": text or desc, "click": click, "edit": edit, "cx": cx, "cy": cy, "cls": cls, "rid": rid}) return nodes def ui(keyword=None, clickable_only=False, raw=False): @@ -66,9 +66,13 @@ def ui(keyword=None, clickable_only=False, raw=False): nodes = _parse_xml(xml_str, keyword, clickable_only, raw) if not raw: for n in nodes: - flag = "Y" if n["click"] else " " + flag = "E" if n.get("edit") else ("Y" if n["click"] else " ") coord = f"({n['cx']},{n['cy']})" if n['cx'] else "" - print(f"[{flag}] {n['text']} {coord} {n['bounds']}") + display_text = n['text'] + if not display_text: + hint = n.get('rid', '').split('/')[-1] or n.get('cls', 'icon') + display_text = f"<{hint}>" + print(f"[{flag}] {display_text} {coord}") print(f"\ntotal: {len(nodes)} nodes") return nodes diff --git a/memory/subagent.md b/memory/subagent.md index 6526b1e..5a8acb9 100644 --- a/memory/subagent.md +++ b/memory/subagent.md @@ -9,6 +9,7 @@ - 流程：启动 → 轮询 output.txt（`[ROUND END]`=该轮完成）→ 写 reply.txt 继续 → 不写则5min退出 - output1/2/3.txt：reply后各轮输出（递增编号），同样append+`[ROUND END]` - input.txt：目标+约束，可指定SOP名。禁写具体步骤（除非已读SOP确认）。大量数据给路径禁塞入 +- ⚠ `--input`走命令行，长文本/含引号会超时。超过一句话时：先写input.txt，启动时不带`--input` - --bg启动瞬间返回，可同一code_run内sleep后poll；非--bg方式禁止合并启动+轮询 ```python