mirrors/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-06-26 11:12:03 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions 5

feat(memory): batch operations for single-turn memory updates (#48507)

Browse source 
The memory tool was strictly one-op-per-call. With the store running near
its char limit by design, a new add that would overflow gets rejected with
'consolidate now, then retry' -- but the model could not consolidate and add
in one call. It had to remove/replace across several turns, then retry the
add, each turn re-sending the whole conversation context. Expensive thrash.

Add an 'operations' array: a list of add/replace/remove ops applied
atomically against the FINAL char budget. The model frees space and adds new
entries in ONE call, even when an add alone would overflow. All-or-nothing:
any bad op aborts the whole batch, nothing written.

Root-cause note: the two agent-level memory interception sites
(agent_runtime_helpers.py, tool_executor.py) silently dropped any param not
in their explicit kwarg list, so 'operations' never reached the handler and
batch calls failed with 'Unknown action None'. Both now pass it through and
bridge each add/replace op to external memory providers.

Also: success response is now terminal (done=true + 'do not repeat' note,
no full-entries echo that invited re-edits); schema rewritten to lead with
the batch mechanism and an explicit one-shot stop rule (2138 -> 1476 chars).

Live-verified: near-full consolidate-and-add went 7 calls -> 1 call,
stable across 3 reps. 103 memory/approval tests + 398 background-review/
run_agent tests green; 6 new batch tests added.
This commit is contained in:
Teknium 2026-06-18 10:19:33 -07:00 • committed by GitHub
parent 2fa16ec2d2
commit 38c8a9c10f
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
6 changed files with 417 additions and 60 deletions
Download patch file Download diff file Expand all files Collapse all files

40
agent/tool_executor.py
View file

@ -1012,28 +1012,42 @@ def execute_tool_calls_sequential(agent, assistant_message, messages: list, effe
elif function_name == "memory":
def _execute(next_args: dict) -> Any:
target = next_args.get("target", "memory")
operations = next_args.get("operations")
from tools.memory_tool import memory_tool as _memory_tool
result = _memory_tool(
action=next_args.get("action"),
target=target,
content=next_args.get("content"),
old_text=next_args.get("old_text"),
operations=operations,
store=agent._memory_store,
)
# Bridge: notify external memory provider of built-in memory writes
if agent._memory_manager and next_args.get("action") in {"add", "replace"}:
try:
agent._memory_manager.on_memory_write(
next_args.get("action", ""),
target,
next_args.get("content", ""),
metadata=agent._build_memory_write_metadata(
task_id=effective_task_id,
tool_call_id=getattr(tool_call, "id", None),
),
# Bridge: notify external memory provider of built-in memory writes.
# Covers both the single-op shape and each add/replace inside a batch.
if agent._memory_manager:
if operations:
_mem_ops = [
op for op in operations
if isinstance(op, dict) and op.get("action") in {"add", "replace"}
]
else:
_mem_ops = (
[{"action": next_args.get("action"), "content": next_args.get("content")}]
if next_args.get("action") in {"add", "replace"} else []
)
except Exception:
pass
for _op in _mem_ops:
try:
agent._memory_manager.on_memory_write(
_op.get("action", ""),
target,
_op.get("content", "") or "",
metadata=agent._build_memory_write_metadata(
task_id=effective_task_id,
tool_call_id=getattr(tool_call, "id", None),
),
)
except Exception:
pass
return result
function_result, function_args = _run_agent_tool_execution_middleware(
agent,