feat(backends,conversations,frontends): opus 5.5 progress updates reach telegram, current claude cli in the image

src/beaver_gateway/agents/claude.py

@@ -24,6 +24,9 @@ class ClaudeOptions(BaseModel):
     model_config = ConfigDict(frozen=True, arbitrary_types_allowed=True)
 
     effort: str | None = None
+    thinking: Mapping[str, str] | None = None
+    """Thinking config for the SDK; ``None`` keeps the CLI default."""
+
     include_partial_messages: bool = True
     """Token-level deltas on the wire; off means one delta per finished block."""
 

src/beaver_gateway/backends/claude_sdk.py

@@ -726,6 +726,7 @@ class ClaudeSdkBackend:
         return ClaudeAgentOptions(
             model=agent.model or None,
             effort=cast("Any", opt.effort),
+            thinking=cast("Any", dict(opt.thinking) if opt.thinking else None),
             system_prompt=system_prompt,
             setting_sources=[],
             strict_mcp_config=True,
@@ -933,7 +934,17 @@ async def _disconnect(live: Session) -> None:
         _log.exception("disconnect failed for session %s", live.session_id)
 
 
+CLI_ENV = "BEAVER_CLAUDE_CLI"
+"""Path to a ``claude`` CLI that wins over the one bundled in the wheel."""
+
+
 def _claude_binary() -> str:
+    override = os.environ.get(CLI_ENV)
+    if override:
+        if not Path(override).is_file():
+            msg = f"{CLI_ENV} points at {override!r}, which is not a file"
+            raise FileNotFoundError(msg)
+        return override
     bundled = Path(claude_agent_sdk.__file__).parent / "_bundled" / "claude"
     if bundled.is_file():
         return str(bundled)

src/beaver_gateway/conversations/turns.py

@@ -32,11 +32,11 @@ from beaver_gateway.frontends.accumulate import StreamAccumulator
 from beaver_gateway.storage.models import Conversation, InjectQueueItem, RateLimit
 
 if TYPE_CHECKING:
-    from collections.abc import AsyncIterator, Callable, Sequence
+    from collections.abc import AsyncIterator, Callable, Iterable, Sequence
 
     from beaver_gateway.events.stream import MessageStreamEvent
 
-__all__ = ["Turns", "content_of", "image_block"]
+__all__ = ["Turns", "content_of", "image_block", "reply_text"]
 
 _log = logging.getLogger(__name__)
 
@@ -164,12 +164,7 @@ class Turns(Seeds):
         ):
             acc.feed(event)
         message = acc.finalize(model=agent.model)
-        reply = "\n\n".join(
-            getattr(b, "text", "")
-            for b in message.content
-            if getattr(b, "type", "") == "text"
-        ).strip()
-        return reply, capture
+        return reply_text(message.content), capture
 
     async def before_turn(self, conv: Conversation) -> str | None:  # noqa: ARG002
         return None
@@ -489,6 +484,22 @@ class Turns(Seeds):
 IMAGE_MAX_BYTES = 5 * 1024 * 1024
 
 
+def reply_text(content: Iterable[Any]) -> str:
+    """Text blocks plus non-empty thinking blocks, in block order.
+
+    What the model writes between tool calls rides in thinking blocks; the
+    hidden reasoning in them is empty and drops out on its own.
+    """
+    parts: list[str] = []
+    for block in content:
+        kind = getattr(block, "type", "")
+        if kind == "text":
+            parts.append(getattr(block, "text", "") or "")
+        elif kind == "thinking":
+            parts.append(getattr(block, "thinking", "") or "")
+    return "\n\n".join(p for p in parts if p).strip()
+
+
 def content_of(
     text: str, attachments: Sequence[dict[str, Any]] | None
 ) -> str | list[dict[str, Any]]:

src/beaver_gateway/frontends/telegram/frontend.py

@@ -983,7 +983,12 @@ class TelegramFrontend(Frontend):
                 draft.set_status(self.texts.writing)
                 draft.append(str(delta.get("text") or ""))
             elif delta.get("type") == "thinking_delta":
-                draft.set_status(self.texts.thinking)
+                progress = str(delta.get("thinking") or "")
+                if progress:
+                    draft.set_status(self.texts.writing)
+                    draft.append(progress)
+                else:
+                    draft.set_status(self.texts.thinking)
         elif kind == "content_block_start":
             block = raw.get("content_block") or {}
             if block.get("type") == "tool_use":