Fix native tool-calling follow-up round on Gemini and Ollama (#867)

The agent's multi-round (tool-result) follow-up request was rejected with
HTTP 400 on two providers, so tools ran but the agent never produced an answer:

- OpenAI-compatible streaming (Gemini 3) dropped the per-call thought_signature
  and collided parallel tool calls, which arrive with index=None: they all
  landed in slot 0, overwriting the first call's name and corrupting its
  arguments by concatenation, so the follow-up request 400'd. Capture and replay
  each call's extra_content (thought_signature), and give every parallel call
  its own accumulator slot (allocated above the max key, so sparse or mixed
  indices can't collide).
- Native Ollama /api/chat expects object tool-call arguments, but Odysseus
  carries them as a JSON string, which Ollama rejected ("Value looks like
  object, but can't find closing '}' symbol"). Convert them to objects in the
  Ollama payload builder.

Both compose with the no-prose null-content sanitize fix from #862.

Tested: python -m pytest tests/test_llm_core_streaming.py
tests/test_llm_core_ollama.py tests/test_agent_loop.py (53 pass), and
python -m py_compile src/llm_core.py src/agent_loop.py.
This commit is contained in:
James Arslan
2026-06-02 04:39:40 +02:00
committed by GitHub
parent 54ac4a74fb
commit a327df6936
5 changed files with 334 additions and 3 deletions

View File

@@ -301,3 +301,39 @@ class TestAppendToolResultsNativeContent:
assert messages[0]["content"] == "thinking..."
assert messages[1]["role"] == "user"
assert "tool output" in messages[1]["content"]
class TestAppendToolResultsThoughtSignature:
"""Gemini 3 returns an opaque thought_signature (in extra_content) with each
function call and rejects the follow-up turn with HTTP 400 unless it is
echoed back on the assistant tool_call. _append_tool_results must replay it
when present, and omit the field entirely otherwise (other providers never
send it)."""
def test_extra_content_is_replayed_when_present(self):
native = [{
"id": "call_g",
"name": "app_api",
"arguments": '{"action": "get_memory"}',
"extra_content": {"google": {"thought_signature": "EuIDCt8DAQ=="}},
}]
messages = []
_append_tool_results(
messages, "", native, [{}], ["mem"],
used_native=True, round_num=1,
)
tc = messages[0]["tool_calls"][0]
assert tc["extra_content"] == {"google": {"thought_signature": "EuIDCt8DAQ=="}}
# function payload is still well-formed alongside it
assert tc["function"]["name"] == "app_api"
assert tc["id"] == "call_g"
def test_no_extra_content_key_when_absent(self):
native = [{"id": "call_o", "name": "app_api", "arguments": "{}"}]
messages = []
_append_tool_results(
messages, "", native, [{}], ["r"],
used_native=True, round_num=1,
)
# No empty/None extra_content leaks onto non-Gemini tool calls.
assert "extra_content" not in messages[0]["tool_calls"][0]