mirror of
https://github.com/NousResearch/hermes-agent.git
synced 2026-07-31 19:16:29 +00:00
The `text_to_speech` tool schema accepted only `text` and `output_path`, so style direction (tone, emotion, pacing, whispering) could never reach the OpenAI backend — even though `gpt-4o-mini-tts` (Hermes's OpenAI provider default) treats `instructions` as its primary voice-design control. This plumbs an optional `instructions` argument through the tool schema, the handler lambda, and `text_to_speech_tool()` into `_generate_openai_tts`, where it is forwarded to `client.audio.speech.create()` only when truthy. Empty/None values still omit the key entirely, preserving behavior on `tts-1`/`tts-1-hd` and strict OpenAI-compatible servers. The same passthrough unblocks self-hosted OpenAI-compatible voice-design servers (Qwen3-TTS-VoiceDesign on oMLX, etc.) that are already wired in via `tts.openai.base_url` — the established convention per #9004 and the TTS config docs — without inventing a new provider backend. Tests: `tests/tools/test_tts_instructions.py` covers backend passthrough, tool-level threading, schema declaration, and the empty-string/absent omission cases. `tests/tools/test_tts_max_text_length.py` fake_openai signature widened to accept the new kwarg. Refs NousResearch/hermes-agent#14196
124 lines
5.2 KiB
Python
124 lines
5.2 KiB
Python
"""Tests for the OpenAI TTS `instructions` field passthrough.
|
|
|
|
Covers #14196: forwarding the OpenAI-spec `instructions` parameter through the
|
|
`text_to_speech` tool so the agent can control tone/emotion/pacing on
|
|
gpt-4o-mini-tts and OpenAI-compatible voice-design servers.
|
|
"""
|
|
|
|
import json
|
|
from unittest.mock import MagicMock, patch
|
|
|
|
import pytest
|
|
|
|
|
|
@pytest.fixture(autouse=True)
|
|
def clean_env(monkeypatch):
|
|
for key in ("OPENAI_API_KEY", "HERMES_SESSION_PLATFORM"):
|
|
monkeypatch.delenv(key, raising=False)
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Backend-level passthrough (_generate_openai_tts)
|
|
# ---------------------------------------------------------------------------
|
|
|
|
class TestOpenaiBackendInstructions:
|
|
def _run(self, tmp_path, monkeypatch, *, tts_config=None, instructions=None):
|
|
monkeypatch.setenv("OPENAI_API_KEY", "test-key")
|
|
mock_client = MagicMock()
|
|
mock_client.audio.speech.create.return_value = MagicMock()
|
|
mock_cls = MagicMock(return_value=mock_client)
|
|
|
|
with patch("tools.tts_tool._import_openai_client", return_value=mock_cls), \
|
|
patch("tools.tts_tool._resolve_openai_audio_client_config",
|
|
return_value=("test-key", None)):
|
|
from tools.tts_tool import _generate_openai_tts
|
|
kwargs = {}
|
|
if instructions is not None:
|
|
kwargs["instructions"] = instructions
|
|
_generate_openai_tts(
|
|
"Hello", str(tmp_path / "out.mp3"), tts_config or {}, **kwargs
|
|
)
|
|
return mock_client.audio.speech.create
|
|
|
|
def test_instructions_forwarded_when_provided(self, tmp_path, monkeypatch):
|
|
"""Tool arg `instructions` is passed to audio.speech.create as-is."""
|
|
create = self._run(tmp_path, monkeypatch, instructions="Speak cheerfully.")
|
|
assert create.call_args[1]["instructions"] == "Speak cheerfully."
|
|
|
|
def test_instructions_absent_by_default(self, tmp_path, monkeypatch):
|
|
"""No instructions arg -> key not present in create kwargs.
|
|
|
|
Preserves behavior on `tts-1`/`tts-1-hd` and strict servers that
|
|
reject unknown kwargs.
|
|
"""
|
|
create = self._run(tmp_path, monkeypatch)
|
|
assert "instructions" not in create.call_args[1]
|
|
|
|
def test_empty_string_instructions_omitted(self, tmp_path, monkeypatch):
|
|
"""Empty string is treated as absent (not forwarded)."""
|
|
create = self._run(tmp_path, monkeypatch, instructions="")
|
|
assert "instructions" not in create.call_args[1]
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Tool-level plumbing (text_to_speech_tool -> _generate_openai_tts)
|
|
# ---------------------------------------------------------------------------
|
|
|
|
class TestToolLevelInstructions:
|
|
def _invoke_tool(self, tmp_path, monkeypatch, *, instructions=None):
|
|
monkeypatch.setenv("OPENAI_API_KEY", "test-key")
|
|
|
|
mock_client = MagicMock()
|
|
|
|
def fake_stream(path):
|
|
# Mimic OpenAI SDK's stream_to_file by writing a tiny payload.
|
|
with open(path, "wb") as f:
|
|
f.write(b"ID3\x03\x00\x00\x00\x00\x00\x00")
|
|
|
|
response = MagicMock()
|
|
response.stream_to_file.side_effect = fake_stream
|
|
mock_client.audio.speech.create.return_value = response
|
|
|
|
mock_cls = MagicMock(return_value=mock_client)
|
|
|
|
with patch("tools.tts_tool._import_openai_client", return_value=mock_cls), \
|
|
patch("tools.tts_tool._resolve_openai_audio_client_config",
|
|
return_value=("test-key", None)), \
|
|
patch("tools.tts_tool._load_tts_config",
|
|
return_value={"provider": "openai"}):
|
|
from tools.tts_tool import text_to_speech_tool
|
|
kwargs = {"output_path": str(tmp_path / "out.mp3")}
|
|
if instructions is not None:
|
|
kwargs["instructions"] = instructions
|
|
result = text_to_speech_tool("Hello world", **kwargs)
|
|
return mock_client.audio.speech.create, json.loads(result)
|
|
|
|
def test_tool_threads_instructions_to_openai_create(
|
|
self, tmp_path, monkeypatch
|
|
):
|
|
create, result = self._invoke_tool(
|
|
tmp_path, monkeypatch, instructions="Whisper conspiratorially."
|
|
)
|
|
assert result.get("success") is True
|
|
assert create.call_args[1]["instructions"] == "Whisper conspiratorially."
|
|
|
|
def test_tool_omits_instructions_when_not_supplied(
|
|
self, tmp_path, monkeypatch
|
|
):
|
|
create, result = self._invoke_tool(tmp_path, monkeypatch)
|
|
assert result.get("success") is True
|
|
assert "instructions" not in create.call_args[1]
|
|
|
|
|
|
# ---------------------------------------------------------------------------
|
|
# Schema
|
|
# ---------------------------------------------------------------------------
|
|
|
|
class TestSchema:
|
|
def test_schema_exposes_instructions_parameter(self):
|
|
from tools.tts_tool import TTS_SCHEMA
|
|
props = TTS_SCHEMA["parameters"]["properties"]
|
|
assert "instructions" in props
|
|
assert props["instructions"]["type"] == "string"
|
|
# Must stay optional — current behavior must be preserved.
|
|
assert "instructions" not in TTS_SCHEMA["parameters"].get("required", [])
|