mirrors/hermes-agent
1
0
Fork 0
mirror of https://github.com/NousResearch/hermes-agent.git synced 2026-04-25 00:51:20 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

fix: prevent tool name duplication in streaming accumulator (MiniMax/NVIDIA NIM)

Browse source 
Based on #11984 by @maxchernin.  Fixes #8259.

Some providers (MiniMax M2.7 via NVIDIA NIM) resend the full function
name in every streaming chunk instead of only the first.  The old
accumulator used += which concatenated them into 'read_fileread_file'.

Changed to simple assignment (=), matching the OpenAI Node SDK, LiteLLM,
and Vercel AI SDK patterns.  Function names are atomic identifiers
delivered complete — no provider splits them across chunks, so
concatenation was never correct semantics.
This commit is contained in:
jarvischer 2026-04-18 22:46:36 +05:30 • committed by kshitijk4poor
parent b0bde98b0f
commit aa24177340
3 changed files with 54 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

10
run_agent.py
View file

@ -5868,7 +5868,15 @@ class AIAgent:
entry["id"] = tc_delta.id
if tc_delta.function:
if tc_delta.function.name:
entry["function"]["name"] += tc_delta.function.name
# Use assignment, not +=. Function names are
# atomic identifiers delivered complete in the
# first chunk (OpenAI spec). Some providers
# (MiniMax M2.7 via NVIDIA NIM) resend the full
# name in every chunk; concatenation would
# produce "read_fileread_file". Assignment
# (matching the OpenAI Node SDK / LiteLLM /
# Vercel AI patterns) is immune to this.
entry["function"]["name"] = tc_delta.function.name
if tc_delta.function.arguments:
entry["function"]["arguments"] += tc_delta.function.arguments
extra = getattr(tc_delta, "extra_content", None)

1
scripts/release.py
View file

@ -264,6 +264,7 @@ AUTHOR_MAP = {
"asurla@nvidia.com": "anniesurla",
"limkuan24@gmail.com": "WideLee",
"aviralarora002@gmail.com": "AviArora02-commits",
"jarvischer@gmail.com": "maxchernin",
}

44
tests/run_agent/test_streaming.py
View file

@ -141,6 +141,50 @@ class TestStreamingAccumulator:
assert tc[0].function.name == "terminal"
assert tc[0].function.arguments == '{"command": "ls"}'
@patch("run_agent.AIAgent._create_request_openai_client")
@patch("run_agent.AIAgent._close_request_openai_client")
def test_tool_name_not_duplicated_when_resent_per_chunk(self, mock_close, mock_create):
"""MiniMax M2.7 via NVIDIA NIM resends the full name in every chunk.
Bug #8259: the old += accumulation produced "read_fileread_file".
Assignment (matching OpenAI Node SDK / LiteLLM) prevents this.
"""
from run_agent import AIAgent
chunks = [
_make_stream_chunk(tool_calls=[
_make_tool_call_delta(index=0, tc_id="call_nim", name="read_file")
]),
_make_stream_chunk(tool_calls=[
_make_tool_call_delta(index=0, tc_id="call_nim", name="read_file", arguments='{"path":')
]),
_make_stream_chunk(tool_calls=[
_make_tool_call_delta(index=0, tc_id="call_nim", name="read_file", arguments=' "x.py"}')
]),
_make_stream_chunk(finish_reason="tool_calls"),
]
mock_client = MagicMock()
mock_client.chat.completions.create.return_value = iter(chunks)
mock_create.return_value = mock_client
agent = AIAgent(
model="test/model",
quiet_mode=True,
skip_context_files=True,
skip_memory=True,
)
agent.api_mode = "chat_completions"
agent._interrupt_requested = False
response = agent._interruptible_streaming_api_call({})
tc = response.choices[0].message.tool_calls
assert tc is not None
assert len(tc) == 1
assert tc[0].function.name == "read_file"
assert tc[0].function.arguments == '{"path": "x.py"}'
@patch("run_agent.AIAgent._create_request_openai_client")
@patch("run_agent.AIAgent._close_request_openai_client")
def test_tool_call_extra_content_preserved(self, mock_close, mock_create):