mirrors/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-05-18 04:41:56 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 7

fix(acp): replay assistant reasoning as agent_thought_chunk on session/load (#12285) (#26943)

Browse source 
Persisted assistant `reasoning_content` / `reasoning` fields are now emitted
as ACP `agent_thought_chunk` notifications during `_replay_session_history`,
so editor clients (Zed, etc.) rebuild collapsed Thinking panes when the user
re-opens a session that used a thinking model.

Ordering matches live streaming: thought precedes message text within the
same assistant turn, mirroring how `reasoning_callback` deltas arrive before
`stream_delta_callback` deltas in `events.py::make_thinking_cb` /
`make_message_cb`.

Behavior on non-reasoning histories is unchanged; the replay loop's existing
text / tool_call / tool_call_update / plan emission is preserved bit-for-bit.

Closes #12285.

Credit:
- @Yukipukii1 (#14691) — original thought-replay design via
  `acp.update_agent_thought_text`; the tool-call portion of that PR has
  since landed via #19139, but the reasoning replay is theirs.
- @HenkDz (#17652 / #18578) — established the `_replay_session_history` and
  `_history_*` helper conventions this builds on.
- @D1zzyDwarf (#16531) — also closed by this work.
This commit is contained in:
kshitij 2026-05-16 06:45:29 -07:00 • committed by GitHub
parent a91a57fa5a
commit f3a4af9cf2
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 277 additions and 22 deletions
Download patch file Download diff file Expand all files Collapse all files

93
acp_adapter/server.py
View file

@ -18,6 +18,7 @@ import acp
from acp.schema import (
AgentCapabilities,
AgentMessageChunk,
AgentThoughtChunk,
AuthenticateResponse,
AvailableCommand,
AvailableCommandsUpdate,
@ -788,14 +789,20 @@ class HermesACPAgent(acp.Agent):
# ---- Session management -------------------------------------------------
@staticmethod
def _history_message_text(message: dict[str, Any]) -> str:
"""Extract displayable text from a persisted OpenAI-style message."""
content = message.get("content")
if isinstance(content, str):
return content.strip()
if isinstance(content, list):
def _flatten_history_text(value: Any) -> str:
"""Normalize a persisted text-or-text-parts value into a single string.
OpenAI-style assistant content (and provider reasoning fields) can arrive
as either a scalar string or a list of ``{"text": ...}`` /
``{"type": "text", "content": ...}`` parts. Whitespace-only inputs
collapse to an empty string so callers can treat ``""`` as "nothing to
emit".
"""
if isinstance(value, str):
return value.strip()
if isinstance(value, list):
parts: list[str] = []
for item in content:
for item in value:
if isinstance(item, dict):
text = item.get("text")
if isinstance(text, str):
@ -807,6 +814,29 @@ class HermesACPAgent(acp.Agent):
return "\n".join(part.strip() for part in parts if part and part.strip()).strip()
return ""
@classmethod
def _history_message_text(cls, message: dict[str, Any]) -> str:
"""Extract displayable text from a persisted OpenAI-style message."""
return cls._flatten_history_text(message.get("content"))
@classmethod
def _history_reasoning_text(cls, message: dict[str, Any]) -> str:
"""Extract displayable reasoning/thought text from a persisted assistant message.
Returns the first non-empty value among ``reasoning_content`` (the
canonical field used by DeepSeek / Moonshot and the post-#16892
chat-completions normalizer) and ``reasoning`` (used by the codex
event projector and several other transports). Both keys are
actively written by live code paths, so neither branch is
deprecated — they cover different transports rather than old vs.
new sessions.
"""
for key in ("reasoning_content", "reasoning"):
text = cls._flatten_history_text(message.get(key))
if text:
return text
return ""
@staticmethod
def _history_message_update(
*,
@ -827,6 +857,11 @@ class HermesACPAgent(acp.Agent):
)
return None
@staticmethod
def _history_thought_update(text: str) -> AgentThoughtChunk:
"""Build an ACP history replay update for an assistant thought."""
return acp.update_agent_thought_text(text)
@staticmethod
def _history_tool_call_name_args(tool_call: dict[str, Any]) -> tuple[str, dict[str, Any]]:
"""Extract function name/arguments from an OpenAI-style tool_call."""
@ -858,9 +893,10 @@ class HermesACPAgent(acp.Agent):
Zed's ACP history UI calls ``session/load`` after the user picks an item
from the Agents sidebar. The agent must then replay the full conversation
as user/assistant chunks plus reconstructed tool-call start/completion
notifications; merely restoring server-side state makes Hermes remember
context, but leaves the editor looking like a clean thread.
as user/assistant chunks, thinking-mode thought chunks, plus reconstructed
tool-call start/completion notifications; merely restoring server-side
state makes Hermes remember context, but leaves the editor looking like a
clean thread.
"""
if not self._conn or not state.history:
return
@ -882,24 +918,37 @@ class HermesACPAgent(acp.Agent):
for message in state.history:
role = str(message.get("role") or "")
if role in {"user", "assistant"}:
if role == "user":
text = self._history_message_text(message)
if text:
update = self._history_message_update(role=role, text=text)
if update is not None and not await _send(update):
return
continue
if role == "assistant":
thought = self._history_reasoning_text(message)
if thought and not await _send(self._history_thought_update(thought)):
return
text = self._history_message_text(message)
if text:
update = self._history_message_update(role=role, text=text)
if update is not None and not await _send(update):
return
if role == "assistant" and isinstance(message.get("tool_calls"), list):
for tool_call in message["tool_calls"]:
if not isinstance(tool_call, dict):
continue
tool_call_id = self._history_tool_call_id(tool_call)
if not tool_call_id:
continue
tool_name, args = self._history_tool_call_name_args(tool_call)
active_tool_calls[tool_call_id] = (tool_name, args)
if not await _send(build_tool_start(tool_call_id, tool_name, args)):
return
tool_calls = message.get("tool_calls")
if isinstance(tool_calls, list):
for tool_call in tool_calls:
if not isinstance(tool_call, dict):
continue
tool_call_id = self._history_tool_call_id(tool_call)
if not tool_call_id:
continue
tool_name, args = self._history_tool_call_name_args(tool_call)
active_tool_calls[tool_call_id] = (tool_name, args)
if not await _send(build_tool_start(tool_call_id, tool_name, args)):
return
continue
if role == "tool":

206
tests/acp/test_server.py
View file

@ -13,6 +13,7 @@ from acp.schema import (
AgentCapabilities,
AgentMessageChunk,
AgentPlanUpdate,
AgentThoughtChunk,
AuthenticateResponse,
AvailableCommandsUpdate,
Implementation,
@ -466,6 +467,211 @@ class TestSessionOps:
for update in updates
)
@pytest.mark.asyncio
async def test_load_session_replays_reasoning_thought_before_message(self, agent):
"""Thinking-model thoughts must be replayed via ``agent_thought_chunk``.
Regression for #12285 — when a session is loaded, persisted assistant
``reasoning_content`` / ``reasoning`` fields must surface as ACP
``AgentThoughtChunk`` notifications in the same relative position they
had live (thought streams before the assistant message text), so Zed's
collapsed Thinking pane rebuilds instead of vanishing on reconnect.
"""
mock_conn = MagicMock(spec=acp.Client)
mock_conn.session_update = AsyncMock()
agent._conn = mock_conn
new_resp = await agent.new_session(cwd="/tmp")
state = agent.session_manager.get_session(new_resp.session_id)
state.history = [
{"role": "user", "content": "Walk me through it."},
{
"role": "assistant",
"reasoning_content": "Let me think step by step about the request.",
"content": "Here is the plan.",
},
{"role": "user", "content": "And the legacy case?"},
{
"role": "assistant",
# No reasoning_content — exercise the legacy "reasoning" fallback
# path so sessions persisted before #16892 still replay thoughts.
"reasoning": "Older sessions stored the trace under the internal key.",
"content": "Same idea, older field name.",
},
]
mock_conn.session_update.reset_mock()
resp = await agent.load_session(cwd="/tmp", session_id=new_resp.session_id)
await asyncio.sleep(0)
await asyncio.sleep(0)
assert isinstance(resp, LoadSessionResponse)
replay_kinds = [
getattr(call.kwargs.get("update"), "session_update", None)
for call in mock_conn.session_update.await_args_list
if getattr(call.kwargs.get("update"), "session_update", None)
in {"user_message_chunk", "agent_message_chunk", "agent_thought_chunk"}
]
assert replay_kinds == [
"user_message_chunk",
"agent_thought_chunk",
"agent_message_chunk",
"user_message_chunk",
"agent_thought_chunk",
"agent_message_chunk",
]
thought_updates = [
call.kwargs["update"]
for call in mock_conn.session_update.await_args_list
if isinstance(call.kwargs.get("update"), AgentThoughtChunk)
]
assert len(thought_updates) == 2
assert thought_updates[0].content.text == "Let me think step by step about the request."
assert thought_updates[1].content.text == "Older sessions stored the trace under the internal key."
@pytest.mark.asyncio
async def test_load_session_replays_reasoning_only_turn(self, agent):
"""Assistant turns with reasoning but no content should still emit a thought.
Pure reasoning-only assistant entries (e.g. a thinking step before a
tool-call turn) commonly carry ``reasoning_content`` with empty
``content``. The replay must still surface the thought so the editor's
Thinking pane rebuilds, even when there is no message text to follow.
"""
mock_conn = MagicMock(spec=acp.Client)
mock_conn.session_update = AsyncMock()
agent._conn = mock_conn
new_resp = await agent.new_session(cwd="/tmp")
state = agent.session_manager.get_session(new_resp.session_id)
state.history = [
{
"role": "assistant",
"reasoning_content": "I should call the search tool next.",
"content": "",
},
]
mock_conn.session_update.reset_mock()
await agent.load_session(cwd="/tmp", session_id=new_resp.session_id)
await asyncio.sleep(0)
await asyncio.sleep(0)
thought_updates = [
call.kwargs["update"]
for call in mock_conn.session_update.await_args_list
if isinstance(call.kwargs.get("update"), AgentThoughtChunk)
]
message_updates = [
call.kwargs["update"]
for call in mock_conn.session_update.await_args_list
if isinstance(call.kwargs.get("update"), AgentMessageChunk)
]
assert len(thought_updates) == 1
assert thought_updates[0].content.text == "I should call the search tool next."
assert message_updates == []
@pytest.mark.asyncio
async def test_load_session_skips_empty_reasoning_fields(self, agent):
"""Empty/whitespace reasoning fields must not produce notifications."""
mock_conn = MagicMock(spec=acp.Client)
mock_conn.session_update = AsyncMock()
agent._conn = mock_conn
new_resp = await agent.new_session(cwd="/tmp")
state = agent.session_manager.get_session(new_resp.session_id)
state.history = [
{
"role": "assistant",
"reasoning_content": "",
"reasoning": " \n\t",
"content": "Just a regular answer.",
},
]
mock_conn.session_update.reset_mock()
await agent.load_session(cwd="/tmp", session_id=new_resp.session_id)
await asyncio.sleep(0)
await asyncio.sleep(0)
thought_updates = [
call.kwargs["update"]
for call in mock_conn.session_update.await_args_list
if isinstance(call.kwargs.get("update"), AgentThoughtChunk)
]
assert thought_updates == []
@pytest.mark.asyncio
async def test_load_session_replays_thought_then_tool_call_without_message(self, agent):
"""Canonical thinking-model shape: reasoning + tool_call + no body text.
Thinking models commonly emit a pre-tool thought followed by a
tool_calls turn with empty ``content``. Replay must emit:
``agent_thought_chunk`` then ``tool_call`` then ``tool_call_update``
for the matching tool result — and crucially, NO ``agent_message_chunk``
for the empty-text assistant body. Regression for the canonical
thinking-then-tool flow on #12285.
"""
mock_conn = MagicMock(spec=acp.Client)
mock_conn.session_update = AsyncMock()
agent._conn = mock_conn
new_resp = await agent.new_session(cwd="/tmp")
state = agent.session_manager.get_session(new_resp.session_id)
state.history = [
{"role": "user", "content": "Find the bug."},
{
"role": "assistant",
"reasoning_content": "I should grep for the function name first.",
"content": "",
"tool_calls": [
{
"id": "call_grep_1",
"type": "function",
"function": {
"name": "search_files",
"arguments": '{"pattern":"foo","path":"."}',
},
}
],
},
{
"role": "tool",
"tool_call_id": "call_grep_1",
"content": '{"total_count":1,"matches":[{"path":"x.py","line":1,"content":"foo"}]}',
},
]
mock_conn.session_update.reset_mock()
await agent.load_session(cwd="/tmp", session_id=new_resp.session_id)
await asyncio.sleep(0)
await asyncio.sleep(0)
kinds = [
getattr(call.kwargs.get("update"), "session_update", None)
for call in mock_conn.session_update.await_args_list
if getattr(call.kwargs.get("update"), "session_update", None)
in {
"user_message_chunk",
"agent_thought_chunk",
"agent_message_chunk",
"tool_call",
"tool_call_update",
}
]
# No agent_message_chunk for the empty-content assistant turn.
assert "agent_message_chunk" not in kinds
# Thought must precede the tool_call_start within the assistant turn,
# and the tool result follows.
assert kinds == [
"user_message_chunk",
"agent_thought_chunk",
"tool_call",
"tool_call_update",
]
@pytest.mark.asyncio
async def test_load_session_schedules_history_replay_after_response(self, agent):
"""Zed only attaches replayed updates after session/load has completed."""