Use Claude Code, Codex, VSCode, Pi, and OpenCode (and 6 other harnesses) for free (1.3B+ free tokens) from your terminal, app, IDE, or phone, and now from the browser with native browser sessions (multi-harness + multi-model) like OpenClaw (voice supported + ToS friendly)
Python
import json
import pytest
from free_claude_code.core.anthropic import ReasoningReplayMode
from free_claude_code.core.anthropic.stream_contracts import parse_sse_text
from free_claude_code.core.openai_responses import (
OpenAIResponsesRequest,
build_responses_chat_request,
)
from free_claude_code.providers.openai_chat.stream_output import (
AnthropicChatStreamOutput,
ChatStreamOutput,
ChatStreamUsage,
ResponsesChatStreamOutput,
)
from free_claude_code.providers.openai_chat.tool_calls import (
OpenAIToolCallAssembler,
)
@pytest.fixture(params=["messages", "responses"])
def output(request) -> ChatStreamOutput:
if request.param == "messages":
return AnthropicChatStreamOutput(
message_id="msg_test", model="test-model", input_tokens=1
)
prepared = build_responses_chat_request(
OpenAIResponsesRequest(model="test-model", input="hello"),
reasoning_replay=ReasoningReplayMode.DISABLED,
)
return ResponsesChatStreamOutput(prepared.tool_adapter, input_tokens=1)
def _argument_deltas(frames: list[str]) -> list[str]:
parts = []
for event in parse_sse_text("".join(frames)):
if event.event == "response.function_call_arguments.delta":
parts.append(event.data["delta"])
elif event.data.get("delta", {}).get("type") == "input_json_delta":
parts.append(event.data["delta"]["partial_json"])
return parts
@pytest.mark.parametrize("name", ["Task", "ordinary_tool"])
@pytest.mark.parametrize(
"arguments",
[
'{"run_in_background":true,"prompt":"inspect"}',
"{}",
'{"run_in_background":null}',
"[]",
"not json",
'{"broken":',
"",
],
)
def test_tool_arguments_stream_without_name_specific_rewrites(output, name, arguments):
assembler = OpenAIToolCallAssembler(reserved_tool_ids=())
frames = output.start_events()
split = len(arguments) // 2
for part, tool_name in [(arguments[:split], name), (arguments[split:], None)]:
emitted = list(
assembler.process_tool_call(
{
"index": 0,
"id": "call_task",
"function": {"name": tool_name, "arguments": part},
},
output,
)
)
if isinstance(output, AnthropicChatStreamOutput):
assert _argument_deltas(emitted) == ([part] if part else [])
frames.extend(emitted)
frames.extend(
output.finish_success(
stop_reason="tool_use",
usage=ChatStreamUsage(input_tokens=1, output_tokens=1),
)
)
assert output.tool_states[0].content == arguments
assert output.tool_states[0].tool_id == "call_task"
events = parse_sse_text("".join(frames))
assert events[-1].event == (
"response.completed"
if isinstance(output, ResponsesChatStreamOutput)
else "message_stop"
)
assert "".join(_argument_deltas(frames)) == arguments
def test_task_argument_aliases_are_restored_recursively(output):
assembler = OpenAIToolCallAssembler(reserved_tool_ids=())
buffers = {}
frames = []
for name, part in [
("Task", '{"run_in_background":true,"wire_prompt":"inspect",'),
(None, '"nested":[{"wire_prompt":"child"}]}'),
]:
frames.extend(
assembler.process_tool_call(
{
"index": 0,
"id": "call_task",
"function": {"name": name, "arguments": part},
},
output,
tool_argument_aliases={"Task": {"wire_prompt": "prompt"}},
tool_argument_alias_buffers=buffers,
)
)
frames.extend(
assembler.flush_tool_argument_alias_buffers(
output, {"Task": {"wire_prompt": "prompt"}}, buffers
)
)
frames.extend(output.close_all_blocks())
assert json.loads("".join(_argument_deltas(frames))) == {
"run_in_background": True,
"prompt": "inspect",
"nested": [{"prompt": "child"}],
}
assert not buffers