hermes-agent/tests/hermes_cli/test_openai_picker_curated.py
Teknium 6b81590c55
test: prune low-value tests suite-wide (wave 1) — 46,820 → 28,106 test functions
Systematic prune per AGENTS.md test policy, one pass over every major
test tree (gateway, hermes_cli, tools, agent, run_agent, plugins, cli,
cron, tui_gateway, honcho/openviking, root-level):

- DELETE: source-reading tests (read_text/getsource on prod files),
  change-detector tests (exact catalog counts, model-name snapshots,
  config version literals), mock-echo tests (assert a mock returns what
  it was told), assertion-free/trivial tests, near-duplicate
  parametrizations (boundaries + one representative kept), async/sync
  twin duplicates, cosmetic within-file variations.
- KEEP (mandatory): security/redaction/approval guards, message-role
  alternation invariants, prompt-caching/deterministic-call-id
  invariants, issue-number regression tests (deduped), E2E tests.
- 6 test files deleted outright (script-style/no-assert or fully
  redundant); conftest.py, fakes/, fixtures/ untouched.
- tests/acp/conftest.py added: autouse fixture stubs the live
  models.dev/GitHub/Copilot/Anthropic inventory fetches that ACP server
  tests performed on every session create — test_server.py 147s → 3.4s,
  and the tests are now genuinely hermetic.
- Sleep-based slowness shrunk where safe (codex_ttfb_watchdog,
  compression_concurrent_fork, etc.); no wall-clock assertion tightened.

Verification: full hermetic suite via scripts/run_tests.sh —
2439 files, 31,130 tests passed, 0 failed, 0 flaky retries, 315s wall
(baseline: 583s wall, 13,564s subprocess CPU).
2026-07-29 13:10:23 -07:00

72 lines
3 KiB
Python

"""Regression tests for two OpenAI/OpenRouter model-picker bugs.
Bug 1 — OpenAI picker dumped the raw ``/v1/models`` catalog
``provider_model_ids("openai")`` hit ``api.openai.com/v1/models`` and
returned the full 120+ entry catalog (embeddings, whisper, tts, dall-e,
moderation, gpt-3.5, …). The ``hermes model`` CLI shows only the curated
agentic list. The picker now intersects the live default-endpoint catalog
with the curated list (preserving curated order) so both surfaces match.
Custom OpenAI-compatible endpoints (proxies, gateways) keep the live list
verbatim so discovery still works.
Bug 2 — OpenRouter appeared authenticated whenever OPENAI_API_KEY was set
OpenRouter's HermesOverlay carried ``extra_env_vars=("OPENAI_API_KEY",)``.
``list_authenticated_providers`` reads ``extra_env_vars`` to decide whether
a provider has credentials, so any OpenAI user saw a phantom OpenRouter
row. The overlay entry is removed; runtime credential resolution still
falls back to OPENAI_API_KEY for explicitly-selected OpenRouter (handled
in runtime_provider.py, independent of the overlay).
"""
import os
from unittest.mock import patch
import pytest
from hermes_cli import models as M
from hermes_cli.providers import HERMES_OVERLAYS
# --- Bug 2: overlay no longer lists OPENAI_API_KEY --------------------------
def test_openrouter_overlay_does_not_list_openai_api_key():
overlay = HERMES_OVERLAYS["openrouter"]
assert "OPENAI_API_KEY" not in overlay.extra_env_vars
# --- Bug 1: default OpenAI endpoint filters to curated agentic models -------
def test_default_openai_endpoint_filters_to_curated(monkeypatch):
"""The 126-model /v1/models dump is intersected with the curated list."""
monkeypatch.setenv("OPENAI_API_KEY", "sk-fake")
monkeypatch.delenv("OPENAI_BASE_URL", raising=False)
curated = M._PROVIDER_MODELS["openai-api"]
# Live catalog: every curated model PLUS a pile of non-agentic junk.
live = list(curated) + [
"text-embedding-3-large", "whisper-1", "tts-1", "dall-e-3",
"gpt-3.5-turbo", "davinci-002", "omni-moderation-latest",
]
with patch.object(M, "fetch_api_models", return_value=live):
result = M.provider_model_ids("openai-api", force_refresh=True)
# Only curated models survive, in curated order, no junk.
assert result == list(curated)
for m in result:
assert m in curated
def test_default_openai_endpoint_intersects_account_access(monkeypatch):
"""Curated models the account can't access are dropped (intersection)."""
monkeypatch.setenv("OPENAI_API_KEY", "sk-fake")
monkeypatch.delenv("OPENAI_BASE_URL", raising=False)
curated = M._PROVIDER_MODELS["openai-api"]
# Account only serves the first two curated models.
live = list(curated[:2]) + ["text-embedding-3-large", "whisper-1"]
with patch.object(M, "fetch_api_models", return_value=live):
result = M.provider_model_ids("openai-api", force_refresh=True)
assert result == list(curated[:2])