feat: format progress updates per exec spec

codex_telegram_bridge/src/codex_telegram_bridge/exec_bridge.py

@@ -60,29 +60,34 @@ class ProgressEditor:
         self.edit_every_s = edit_every_s
 
         self._lock = threading.Lock()
-        self._pending: Optional[str] = None
-        self._last_sent: Optional[str] = None
+        self._pending: Optional[tuple[str, Optional[list[dict[str, Any]]]]] = None
+        self._last_sent: Optional[tuple[str, Optional[list[dict[str, Any]]]]] = None
         self._last_edit_at = 0.0
 
         self._stop = threading.Event()
         self._thread = threading.Thread(target=self._run, daemon=True)
         self._thread.start()
 
-    def set(self, text: str) -> None:
+    def set(self, text: str, entities: Optional[list[dict[str, Any]]] = None) -> None:
         text = _clamp_tg_text(text)
         with self._lock:
-            self._pending = text
+            self._pending = (text, entities)
+
+    def set_markdown(self, text: str) -> None:
+        rendered_text, entities = render_markdown(text)
+        self.set(rendered_text, entities or None)
 
     def stop(self) -> None:
         self._stop.set()
         self._thread.join(timeout=1.0)
 
-    def _edit(self, text: str) -> None:
+    def _edit(self, text: str, entities: Optional[list[dict[str, Any]]]) -> None:
         try:
             self.bot.edit_message_text(
                 chat_id=self.chat_id,
                 message_id=self.message_id,
                 text=text,
+                entities=entities,
             )
         except Exception as e:
             log(
@@ -92,7 +97,7 @@ class ProgressEditor:
 
     def _run(self) -> None:
         while not self._stop.is_set():
-            to_send: Optional[str] = None
+            to_send: Optional[tuple[str, Optional[list[dict[str, Any]]]]] = None
             now = time.monotonic()
             with self._lock:
                 if self._pending is not None and (now - self._last_edit_at) >= self.edit_every_s:
@@ -103,7 +108,7 @@ class ProgressEditor:
                     self._pending = None
 
             if to_send is not None:
-                self._edit(to_send)
+                self._edit(to_send[0], to_send[1])
 
             self._stop.wait(0.2)
 
@@ -358,12 +363,19 @@ def run(
         )
         typing_thread.start()
 
+        started_at = time.monotonic()
+        session_box: dict[str, Optional[str]] = {"id": resume_session}
+        progress_renderer = ExecProgressRenderer(max_actions=5)
+
         progress_id: Optional[int] = None
         progress: Optional[ProgressEditor] = None
         try:
+            initial_text = progress_renderer.render_progress(0.0)
+            initial_rendered, initial_entities = render_markdown(initial_text)
             progress_msg = bot.send_message(
                 chat_id=chat_id,
-                text="Working...",
+                text=initial_rendered,
+                entities=initial_entities or None,
                 reply_to_message_id=user_msg_id,
                 disable_notification=silent_progress,
             )
@@ -374,10 +386,6 @@ def run(
         if progress_id is not None:
             progress = ProgressEditor(bot, chat_id, progress_id, edit_every_s)
 
-        started_at = time.monotonic()
-        session_box: dict[str, Optional[str]] = {"id": resume_session}
-        progress_renderer = ExecProgressRenderer(max_lines=4)
-
         def on_event(evt: Dict[str, Any]) -> None:
             event_type = evt.get("type")
             if event_type == "thread.started":
@@ -392,15 +400,10 @@ def run(
                             thread_id,
                             meta={"workspace": workspace},
                         )
-            line = progress_renderer.note_event(evt)
-            if line and progress is not None:
-                elapsed = int(time.monotonic() - started_at)
-                session_id = session_box["id"]
-                header = f"Working... ({elapsed}s)"
-                if session_id:
-                    header += f"\nSession: {session_id}"
-                msg = progress_renderer.render(header)
-                progress.set(msg)
+            if progress_renderer.note_event(evt) and progress is not None:
+                elapsed = time.monotonic() - started_at
+                msg = progress_renderer.render_progress(elapsed)
+                progress.set_markdown(msg)
 
         def _stop_background() -> None:
             typing_stop.set()
@@ -443,13 +446,15 @@ def run(
         _stop_background()
 
         answer = answer or "(No agent_message captured from JSON stream.)"
-        final_text, final_entities = render_markdown(answer)
+        elapsed = time.monotonic() - started_at
+        final_md = progress_renderer.render_final(elapsed, answer)
+        final_text, final_entities = render_markdown(final_md)
         can_edit_final = progress_id is not None and len(final_text) <= TELEGRAM_TEXT_LIMIT
 
         if loud_final or not can_edit_final:
             sent_msgs = bot.send_message_markdown_chunked(
                 chat_id=chat_id,
-                text=answer,
+                text=final_md,
                 reply_to_message_id=user_msg_id,
             )
             for m in sent_msgs:

codex_telegram_bridge/src/codex_telegram_bridge/exec_render.py

@@ -5,7 +5,93 @@ import re
 from collections import deque
 from dataclasses import dataclass, field
 from textwrap import indent
-from typing import Any, Dict, List, Optional
+from typing import Any, Optional
+
+ELLIPSIS = "…"
+STATUS_RUNNING = "▸"
+STATUS_DONE = "✓"
+HEADER_SEP = " · "
+
+MAX_CMD_LEN = 40
+MAX_REASON_LEN = 80
+MAX_QUERY_LEN = 60
+MAX_PATH_LEN = 40
+MAX_PROGRESS_CHARS = 300
+
+
+def _one_line(text: str) -> str:
+    return " ".join((text or "").split())
+
+
+def _truncate(text: str, max_len: int) -> str:
+    text = _one_line(text)
+    if len(text) <= max_len:
+        return text
+    if max_len <= len(ELLIPSIS):
+        return text[:max_len]
+    return text[: max_len - len(ELLIPSIS)] + ELLIPSIS
+
+
+def _inline_code(text: str) -> str:
+    return f"`{text}`"
+
+
+def _format_elapsed(elapsed_s: float) -> str:
+    total = max(0, int(elapsed_s))
+    minutes, seconds = divmod(total, 60)
+    hours, minutes = divmod(minutes, 60)
+    if hours:
+        return f"{hours}h {minutes:02d}m"
+    if minutes:
+        return f"{minutes}m {seconds:02d}s"
+    return f"{seconds}s"
+
+
+def _format_header(elapsed_s: float, turn: Optional[int], done: bool) -> str:
+    label = "Done" if done else "codex"
+    elapsed = _format_elapsed(elapsed_s)
+    if turn is not None:
+        return f"{label}{HEADER_SEP}{elapsed}{HEADER_SEP}turn {turn}"
+    return f"{label}{HEADER_SEP}{elapsed}"
+
+
+def _format_reasoning(text: str) -> str:
+    if not text:
+        return ""
+    return f"_{_truncate(text, MAX_REASON_LEN)}_"
+
+
+def _format_command(command: str) -> str:
+    command = _truncate(command, MAX_CMD_LEN)
+    if not command:
+        command = "(empty)"
+    return _inline_code(command)
+
+
+def _format_query(query: str) -> str:
+    return _truncate(query, MAX_QUERY_LEN)
+
+
+def _format_paths(paths: list[str]) -> str:
+    rendered = []
+    for path in paths:
+        rendered.append(_inline_code(_truncate(path, MAX_PATH_LEN)))
+    return ", ".join(rendered)
+
+
+def _format_file_change(changes: list[dict[str, Any]]) -> str:
+    paths = [change.get("path") for change in changes if change.get("path")]
+    if not paths:
+        total = len(changes)
+        return "updated files" if total == 0 else f"updated {total} files"
+    if len(paths) <= 3:
+        return f"updated {_format_paths(paths)}"
+    return f"updated {len(paths)} files"
+
+
+def _format_tool_call(server: str, tool: str) -> str:
+    name = ".".join(part for part in (server, tool) if part)
+    return name or "tool"
 
 
 def _truncate_output(text: str, max_lines: int = 20, max_chars: int = 4000) -> str:
@@ -19,23 +105,6 @@ def _truncate_output(text: str, max_lines: int = 20, max_chars: int = 4000) -> s
     return "\n".join(lines)
 
 
-def _format_todo(items: list[dict[str, Any]]) -> str:
-    rendered: list[str] = []
-    for item in items:
-        status = "done" if item.get("completed") else "todo"
-        rendered.append(f"- [{status}] {item.get('text', '')}")
-    return "\n".join(rendered)
-
-
-def _format_todo_summary(items: list[dict[str, Any]]) -> str:
-    total = len(items)
-    done = sum(1 for item in items if item.get("completed"))
-    next_text = next((item.get("text", "") for item in items if not item.get("completed")), "")
-    if next_text:
-        return f"plan: {done}/{total} done, next: {_shorten(next_text, 80)}"
-    return f"plan: {done}/{total} done"
-
-
 def _maybe_parse_json(text: str) -> Optional[Any]:
     try:
         return json.loads(text)
@@ -43,16 +112,14 @@ def _maybe_parse_json(text: str) -> Optional[Any]:
         return None
 
 
-def _shorten(text: str, max_len: int = 140) -> str:
-    if len(text) <= max_len:
-        return text
-    return text[: max_len - 3] + "..."
-
-
 @dataclass
 class ExecRenderState:
     items: dict[str, dict[str, Any]] = field(default_factory=dict)
-    recent: deque[str] = field(default_factory=lambda: deque(maxlen=4))
+    recent_actions: deque[str] = field(default_factory=lambda: deque(maxlen=5))
+    current_action: Optional[str] = None
+    current_action_id: Optional[str] = None
+    pending_reasoning: Optional[str] = None
+    current_reasoning: Optional[str] = None
     last_turn: Optional[int] = None
 
 
@@ -65,6 +132,31 @@ def _record_item(state: ExecRenderState, item: dict[str, Any]) -> None:
             state.last_turn = int(match.group(1))
 
 
+def _set_current_action(state: ExecRenderState, item_id: Optional[str], line: str) -> bool:
+    changed = False
+    if state.current_action != line or state.current_action_id != item_id:
+        state.current_action = line
+        state.current_action_id = item_id
+        if state.pending_reasoning:
+            state.current_reasoning = state.pending_reasoning
+            state.pending_reasoning = None
+        changed = True
+    return changed
+
+
+def _complete_action(state: ExecRenderState, item_id: Optional[str], line: str) -> bool:
+    changed = False
+    if line:
+        state.recent_actions.append(line)
+        changed = True
+    if item_id and state.current_action_id == item_id:
+        state.current_action = None
+        state.current_action_id = None
+        state.current_reasoning = None
+        changed = True
+    return changed
+
+
 def render_event_cli(
     event: dict[str, Any],
     state: ExecRenderState,
@@ -75,26 +167,13 @@ def render_event_cli(
     lines: list[str] = []
 
     if etype == "thread.started":
-        thread_id = event.get("thread_id", "")
-        if thread_id:
-            lines.append(f"thread started: {thread_id}")
-            lines.append(f"resume with: codex exec resume {thread_id}")
-        else:
-            lines.append("thread started")
-        return lines
+        return ["thread started"]
 
     if etype == "turn.started":
         return ["turn started"]
 
     if etype == "turn.completed":
-        usage = event.get("usage", {})
-        lines.append(
-            "turn completed "
-            f"(in={usage.get('input_tokens', 0)} "
-            f"cached={usage.get('cached_input_tokens', 0)} "
-            f"out={usage.get('output_tokens', 0)})"
-        )
-        return lines
+        return ["turn completed"]
 
     if etype == "turn.failed":
         error = event.get("error", {}).get("message", "")
@@ -108,9 +187,8 @@ def render_event_cli(
         _record_item(state, item)
 
         itype = item.get("type")
-        status = item.get("status")
 
-        if itype == "agent_message":
+        if itype == "agent_message" and etype == "item.completed":
             text = item.get("text", "")
             parsed = _maybe_parse_json(text)
             if parsed is not None:
@@ -121,177 +199,155 @@ def render_event_cli(
                 lines.extend(indent(text, "  ").splitlines() if text else ["  (empty)"])
 
         elif itype == "reasoning" and show_reasoning:
-            lines.append(f"reasoning: {item.get('text', '')}")
+            reasoning = _format_reasoning(item.get("text", ""))
+            if reasoning:
+                lines.append(reasoning)
 
         elif itype == "command_execution":
-            command = item.get("command", "")
+            command = _format_command(item.get("command", ""))
             if etype == "item.started":
-                lines.append(f"run: {command}")
-            else:
+                lines.append(f"{STATUS_RUNNING} running: {command}")
+            elif etype == "item.completed":
                 exit_code = item.get("exit_code")
-                outcome = "ok" if status == "completed" else status or "unknown"
-                lines.append(f"command {outcome} (exit={exit_code}): {command}")
+                exit_part = f" (exit {exit_code})" if exit_code is not None else ""
+                lines.append(f"{STATUS_DONE} ran: {command}{exit_part}")
                 output = _truncate_output(item.get("aggregated_output", ""))
                 if output:
                     lines.extend(indent(output, "  ").splitlines())
 
-        elif itype == "file_change":
-            changes = item.get("changes", [])
-            counts = {"add": 0, "update": 0, "delete": 0}
-            for change in changes:
-                kind = change.get("kind")
-                if kind in counts:
-                    counts[kind] += 1
-            lines.append(
-                "file changes "
-                f"(status={status}) add={counts['add']} "
-                f"update={counts['update']} delete={counts['delete']}"
-            )
-            for change in changes:
-                lines.append(f"  {change.get('kind')} {change.get('path')}")
+        elif itype == "file_change" and etype == "item.completed":
+            line = _format_file_change(item.get("changes", []))
+            lines.append(f"{STATUS_DONE} {line}")
 
         elif itype == "mcp_tool_call":
-            server = item.get("server", "")
-            tool = item.get("tool", "")
+            name = _format_tool_call(item.get("server", ""), item.get("tool", ""))
             if etype == "item.started":
-                lines.append(f"tool call: {server}.{tool}")
-            else:
-                outcome = "ok" if status == "completed" else status or "unknown"
-                lines.append(f"tool {outcome}: {server}.{tool}")
-                if item.get("error"):
-                    lines.append(f"  error: {item['error'].get('message', '')}")
-                result = item.get("result") or {}
-                if result.get("structured_content") is not None:
-                    lines.append("  result:")
-                    lines.extend(
-                        indent(json.dumps(result["structured_content"], indent=2), "    ").splitlines()
-                    )
+                lines.append(f"{STATUS_RUNNING} tool: {name}")
+            elif etype == "item.completed":
+                lines.append(f"{STATUS_DONE} tool: {name}")
 
-        elif itype == "web_search":
-            lines.append(f"web search: {item.get('query', '')}")
+        elif itype == "web_search" and etype == "item.completed":
+            query = _format_query(item.get("query", ""))
+            lines.append(f"{STATUS_DONE} searched: {query}")
 
-        elif itype == "todo_list":
-            todo = _format_todo(item.get("items", []))
-            lines.append(f"plan ({etype}):")
-            lines.extend(indent(todo, "  ").splitlines())
-
-        elif itype == "error":
-            lines.append(f"warning: {item.get('message', '')}")
-
-        else:
-            lines.append(f"{etype}: {item}")
-
-    else:
-        if etype:
-            lines.append(f"event: {etype}")
-        else:
-            lines.append(f"event: {event}")
+        elif itype == "error" and etype == "item.completed":
+            warning = _truncate(item.get("message", ""), 120)
+            lines.append(f"{STATUS_DONE} warning: {warning}")
 
     return lines
 
 
-def render_event_progress(event: dict[str, Any], state: ExecRenderState) -> Optional[str]:
+class ExecProgressRenderer:
+    def __init__(self, max_actions: int = 5, max_chars: int = MAX_PROGRESS_CHARS) -> None:
+        self.state = ExecRenderState(recent_actions=deque(maxlen=max_actions))
+        self.max_chars = max_chars
+
+    def note_event(self, event: dict[str, Any]) -> bool:
         etype = event.get("type")
+        changed = False
 
-    if etype == "thread.started":
-        thread_id = event.get("thread_id", "")
-        return f"thread started: {thread_id}" if thread_id else "thread started"
-
-    if etype == "turn.started":
-        return "turn started"
-
-    if etype == "turn.completed":
-        usage = event.get("usage", {})
-        return (
-            "turn completed "
-            f"(in={usage.get('input_tokens', 0)} "
-            f"cached={usage.get('cached_input_tokens', 0)} "
-            f"out={usage.get('output_tokens', 0)})"
-        )
-
-    if etype == "turn.failed":
-        error = event.get("error", {}).get("message", "")
-        return f"turn failed: {error}"
-
-    if etype == "error":
-        return f"stream error: {event.get('message', '')}"
+        if etype in {"thread.started", "turn.started"}:
+            return True
 
         if etype in {"item.started", "item.updated", "item.completed"}:
             item = event.get("item", {}) or {}
-        _record_item(state, item)
-
+            _record_item(self.state, item)
             itype = item.get("type")
-        status = item.get("status")
+            item_id = item.get("id") if isinstance(item.get("id"), str) else None
 
-        if itype == "agent_message" and etype == "item.completed":
-            text = item.get("text", "")
-            snippet = text.splitlines()[0] if text else ""
-            if snippet:
-                return f"assistant: {_shorten(snippet, 120)}"
-            return "assistant response ready"
+            if itype == "reasoning":
+                reasoning = _format_reasoning(item.get("text", ""))
+                if reasoning:
+                    if self.state.current_action and not self.state.current_reasoning:
+                        self.state.current_reasoning = reasoning
+                        changed = True
+                    else:
+                        self.state.pending_reasoning = reasoning
+                return changed
 
             if itype == "command_execution":
-            command = item.get("command", "")
+                command = _format_command(item.get("command", ""))
                 if etype == "item.started":
-                return f"run: {_shorten(command, 160)}"
+                    line = f"{STATUS_RUNNING} running: {command}"
+                    changed = _set_current_action(self.state, item_id, line) or changed
+                elif etype == "item.completed":
                     exit_code = item.get("exit_code")
-            outcome = "ok" if status == "completed" else status or "unknown"
-            return f"command {outcome} (exit={exit_code}): {_shorten(command, 120)}"
-
-        if itype == "file_change":
-            changes = item.get("changes", [])
-            counts = {"add": 0, "update": 0, "delete": 0}
-            for change in changes:
-                kind = change.get("kind")
-                if kind in counts:
-                    counts[kind] += 1
-            return (
-                "file changes "
-                f"+{counts['add']} ~{counts['update']} -{counts['delete']}"
-            )
+                    exit_part = f" (exit {exit_code})" if exit_code is not None else ""
+                    line = f"{STATUS_DONE} ran: {command}{exit_part}"
+                    changed = _complete_action(self.state, item_id, line) or changed
+                return changed
 
             if itype == "mcp_tool_call":
-            server = item.get("server", "")
-            tool = item.get("tool", "")
+                name = _format_tool_call(item.get("server", ""), item.get("tool", ""))
                 if etype == "item.started":
-                return f"tool call: {server}.{tool}"
-            outcome = "ok" if status == "completed" else status or "unknown"
-            return f"tool {outcome}: {server}.{tool}"
+                    line = f"{STATUS_RUNNING} tool: {name}"
+                    changed = _set_current_action(self.state, item_id, line) or changed
+                elif etype == "item.completed":
+                    line = f"{STATUS_DONE} tool: {name}"
+                    changed = _complete_action(self.state, item_id, line) or changed
+                return changed
 
-        if itype == "web_search":
-            return f"web search: {_shorten(item.get('query', ''), 120)}"
+            if itype == "web_search" and etype == "item.completed":
+                query = _format_query(item.get("query", ""))
+                line = f"{STATUS_DONE} searched: {query}"
+                return _complete_action(self.state, item_id, line) or changed
 
-        if itype == "todo_list":
-            return _format_todo_summary(item.get("items", []))
+            if itype == "file_change" and etype == "item.completed":
+                line = f"{STATUS_DONE} {_format_file_change(item.get('changes', []))}"
+                return _complete_action(self.state, item_id, line) or changed
 
-        if itype == "error":
-            return f"warning: {_shorten(item.get('message', ''), 120)}"
+            if itype == "error" and etype == "item.completed":
+                warning = _truncate(item.get("message", ""), 120)
+                line = f"{STATUS_DONE} warning: {warning}"
+                return _complete_action(self.state, item_id, line) or changed
 
-    return None
+        return changed
 
+    def render_progress(self, elapsed_s: float) -> str:
+        header = _format_header(elapsed_s, self.state.last_turn, done=False)
+        actions = list(self.state.recent_actions)
+        current_reasoning = self.state.current_reasoning
+        current_action = self.state.current_action
 
-class ExecProgressRenderer:
-    def __init__(self, max_lines: int = 4) -> None:
-        self.state = ExecRenderState(recent=deque(maxlen=max_lines))
-
-    def note_event(self, event: dict[str, Any]) -> Optional[str]:
-        line = render_event_progress(event, self.state)
-        if not line:
-            return None
-        line = line.strip()
-        if not line:
-            return None
-        if self.state.recent and self.state.recent[-1] == line:
-            return line
-        self.state.recent.append(line)
-        return line
-
-    def render(self, header: str) -> str:
         lines: list[str] = []
-        if self.state.last_turn is not None:
-            lines.append(f"Turn: {self.state.last_turn}")
-        if self.state.recent:
-            lines.extend(self.state.recent)
+        if actions:
+            lines.extend(actions)
+        if current_reasoning and current_action:
+            lines.append(current_reasoning)
+        if current_action:
+            lines.append(current_action)
+
+        message = self._assemble(header, lines)
+        if len(message) <= self.max_chars:
+            return message
+
+        trimmed_actions = list(actions)
+        while len(message) > self.max_chars and trimmed_actions:
+            trimmed_actions.pop(0)
+            lines = []
+            if trimmed_actions:
+                lines.extend(trimmed_actions)
+            if current_reasoning and current_action:
+                lines.append(current_reasoning)
+            if current_action:
+                lines.append(current_action)
+            message = self._assemble(header, lines)
+
+        return message
+
+    def render_final(self, elapsed_s: float, answer: str) -> str:
+        header = _format_header(elapsed_s, self.state.last_turn, done=True)
+        lines: list[str] = []
+        if self.state.recent_actions:
+            lines.extend(self.state.recent_actions)
+        body = self._assemble(header, lines)
+        answer = (answer or "").strip()
+        if answer:
+            body = body + "\n\n" + answer
+        return body
+
+    @staticmethod
+    def _assemble(header: str, lines: list[str]) -> str:
         if not lines:
             return header
         return header + "\n\n" + "\n".join(lines)