5caeb299a4
原本嵌在 fam-edge 里的问答模型降级链（NVIDIA 文字模型 -> Gemini 非 flash 文字 模型 -> 本地 Ollama 兜底，含 key 轮换/熔断）跟视频分析业务无关，是通用能力， 抽成独立 ai-gateway 服务（OpenAI 兼容协议），除了 fam-edge 自己，别的项目也能 直接接入。 - qa.py 重写为 HTTP 转发客户端，调 ai-gateway 的 /v1/chat/completions，翻译回 原有 run_qa/run_qa_stream 契约，api_gateway.py 和 fam-core 调用方零改动 - 删除 model_adapters/ollama_adapter.py 及其测试（问答专用，视频分析不需要本地模型） - gemini_adapter.py / nvidia_adapter.py 移除 chat()/chat_stream() 及问答专用超时 （只保留视频分析用的 analyze_video） - app.py 移除 Ollama 预热逻辑（现在由 ai-gateway 自己负责） - config.yaml 移除 3 个问答专用 model 条目，新增 ai_gateway 客户端配置块 Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
111 lines
4.2 KiB
Python
111 lines
4.2 KiB
Python
from fam_edge.model_adapters.gemini_adapter import GeminiAdapter
|
||
|
||
|
||
def _cfg(**overrides):
|
||
base = {
|
||
"provider": "gemini",
|
||
"model_name": "gemini-flash-latest",
|
||
"api_key": "key-primary",
|
||
"circuit_breaker": {"enabled": False},
|
||
}
|
||
base.update(overrides)
|
||
return base
|
||
|
||
|
||
def test_single_key_backward_compat():
|
||
a = GeminiAdapter(_cfg())
|
||
assert a.api_keys == ["key-primary"]
|
||
assert a.api_key == "key-primary"
|
||
|
||
|
||
def test_extra_keys_appended_in_order():
|
||
a = GeminiAdapter(_cfg(extra_api_keys=["key-2", "key-3", "key-4"]))
|
||
assert a.api_keys == ["key-primary", "key-2", "key-3", "key-4"]
|
||
|
||
|
||
def test_extra_keys_dedup_against_primary():
|
||
a = GeminiAdapter(_cfg(extra_api_keys=["key-primary", "key-2"]))
|
||
assert a.api_keys == ["key-primary", "key-2"]
|
||
|
||
|
||
def test_missing_env_var_keys_are_dropped():
|
||
"""extra_api_keys 里未设置的 ${ENV_VAR} 解析为空字符串，不应该混进 api_keys 列表。"""
|
||
a = GeminiAdapter(_cfg(extra_api_keys=["${SOME_UNSET_GEMINI_KEY_VAR}", "key-2"]))
|
||
assert a.api_keys == ["key-primary", "key-2"]
|
||
|
||
|
||
def test_no_keys_at_all():
|
||
a = GeminiAdapter(_cfg(api_key=""))
|
||
assert a.api_keys == []
|
||
assert a.api_key == ""
|
||
|
||
|
||
def test_key_labels_default_to_generic_when_not_configured():
|
||
a = GeminiAdapter(_cfg(extra_api_keys=["key-2", "key-3"]))
|
||
assert a.key_labels == ["key1", "key2", "key3"]
|
||
|
||
|
||
def test_key_labels_use_configured_project_names():
|
||
a = GeminiAdapter(_cfg(
|
||
extra_api_keys=["key-2", "key-3", "key-4"],
|
||
key_labels=["智能摄像头-1", "智能摄像头-2", "智能摄像头-3", "智能摄像头-4"],
|
||
))
|
||
assert a.key_labels == ["智能摄像头-1", "智能摄像头-2", "智能摄像头-3", "智能摄像头-4"]
|
||
|
||
|
||
def test_key_labels_stay_aligned_when_a_key_is_dropped():
|
||
"""如果某个 ${ENV_VAR} 没设置被丢弃，剩下的 key_labels 要跟着剩下的 key 对齐，
|
||
不能因为下标错位把别的项目名安到错的 key 上。"""
|
||
a = GeminiAdapter(_cfg(
|
||
extra_api_keys=["${SOME_UNSET_GEMINI_KEY_VAR}", "key-2"],
|
||
key_labels=["智能摄像头-1", "智能摄像头-2", "智能摄像头-3"],
|
||
))
|
||
assert a.api_keys == ["key-primary", "key-2"]
|
||
assert a.key_labels == ["智能摄像头-1", "智能摄像头-3"]
|
||
|
||
|
||
def test_rotated_keys_cascades_from_random_start(monkeypatch):
|
||
"""起点由 random.randrange 决定；固定住随机数就能验证级联顺序是"从起点绕一圈"。"""
|
||
a = GeminiAdapter(_cfg(extra_api_keys=["key-2", "key-3", "key-4"]))
|
||
monkeypatch.setattr(
|
||
"fam_edge.model_adapters.gemini_adapter.random.randrange", lambda n: 2)
|
||
order = a._rotated_keys()
|
||
assert [k for _, k in order] == ["key-3", "key-4", "key-primary", "key-2"]
|
||
assert [i for i, _ in order] == [2, 3, 0, 1]
|
||
|
||
|
||
def test_rotated_keys_uses_full_key_range(monkeypatch):
|
||
"""random.randrange 的调用范围必须是 len(api_keys)，否则会漏掉某些 key 永远选不到。"""
|
||
a = GeminiAdapter(_cfg(extra_api_keys=["key-2", "key-3", "key-4"]))
|
||
seen_n = []
|
||
|
||
def fake_randrange(n):
|
||
seen_n.append(n)
|
||
return 0
|
||
|
||
monkeypatch.setattr(
|
||
"fam_edge.model_adapters.gemini_adapter.random.randrange", fake_randrange)
|
||
a._rotated_keys()
|
||
assert seen_n == [4]
|
||
|
||
|
||
def test_rotated_keys_starts_roughly_uniform_over_many_calls():
|
||
"""核心诉求: 每次调用应该是真随机（每个 key 命中概率均等 1/n），而不是像旧的
|
||
round-robin 那样顺序推进——用大样本统计每个 key 被选为起点的频率，应该接近
|
||
1/4，且不应该出现某个 key 明显被冷落或独占（对应此前"全部流量压在同一个
|
||
key"的 bug）。"""
|
||
a = GeminiAdapter(_cfg(extra_api_keys=["key-2", "key-3", "key-4"]))
|
||
n_trials = 4000
|
||
counts = {"key-primary": 0, "key-2": 0, "key-3": 0, "key-4": 0}
|
||
for _ in range(n_trials):
|
||
counts[a._rotated_keys()[0][1]] += 1
|
||
for key, c in counts.items():
|
||
share = c / n_trials
|
||
assert 0.20 <= share <= 0.30, f"{key} 起点占比 {share} 明显偏离 1/4"
|
||
|
||
|
||
def test_rotated_keys_single_key_never_errors():
|
||
a = GeminiAdapter(_cfg())
|
||
for _ in range(3):
|
||
assert a._rotated_keys() == [(0, "key-primary")]
|