soulpolicy — Python SDK
LLM 行为可信度 primitive。一行集成,按调用计费。
安装
pip install soulpolicy
30 秒上手
from soulpolicy import Client
sp = Client(api_key="sk_live_xxx") # 或设 SOULPOLICY_API_KEY env
# 1) 用 5+ 条「健康输出」建一个 baseline
bl = sp.baselines.create(samples=[
"Hello, thanks for contacting us. How can I help today?",
"Hi there, I am happy to assist. Could you share more details?",
"Apologies for the trouble. Let me look into that for you.",
"Sure, I can check the status. One moment please.",
"Thank you for your patience. I will follow up by email.",
])
print(bl.id, bl.fingerprint)
# 2) 对每次 LLM 输出做漂移判定
r = sp.drift_checks.check(text="Sure, let me help.", baseline=bl.id)
if r.severity == "critical":
handle_drift(r)
print(r.score, r.severity, r.evidence["nearest_baseline_samples"])
# 3) 任何时刻可复现:用 reproducibility 包重算
replay = sp.drift_checks.replay(
text="Sure, let me help.",
baseline=bl.id,
reproducibility=r.reproducibility,
)
assert replay.matched
# 4) 客户回标 — 帮助我们改进 precision/recall
sp.drift_checks.feedback(r.id, label="true_positive")
错误处理
from soulpolicy import Client, AuthenticationError, IdempotencyError
try:
sp.baselines.create(samples=["x"]) # 少于 3 条
except InvalidRequestError as e:
print(e.code, e.message, e.param)
except AuthenticationError:
print("api key invalid")
except IdempotencyError:
print("idempotency key reused with different payload")
Idempotency
POST 请求默认自动带 Idempotency-Key(sdk-auto-<random>)。手动指定:
sp.drift_checks.create(text=..., baseline=..., idempotency_key="my-key-1")
自托管
sp = Client(api_key="sk_local_xxx", base_url="https://soulpolicy.internal.example.com")
资源对照
| Method | Endpoint |
|---|---|
sp.baselines.create(...) |
POST /api/v2/baselines |
sp.baselines.retrieve(id) |
GET /api/v2/baselines/{id} |
sp.baselines.report(id) |
GET /api/v2/baselines/{id}/report |
sp.drift_checks.check(...) |
POST /api/v2/drift_checks |
sp.drift_checks.replay(...) |
POST /api/v2/drift_checks/replay |
sp.drift_checks.feedback(id, ...) |
POST /api/v2/drift_checks/{id}/feedback |
Metadata
Release files for soulpolicy 0.2.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| soulpolicy-0.2.0.tar.gz | 19.0 kB | Without provenance | Details for soulpolicy-0.2.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| soulpolicy-0.2.0-py3-none-any.whl | Python 3 | none | any | Details for soulpolicy-0.2.0-py3-none-any.whl |
Total release size: 36.6 kB
Release files / soulpolicy-0.2.0.tar.gz
| Download URL | soulpolicy-0.2.0.tar.gz |
|---|---|
| Size | 19.0 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
8d363f8cc80bf66a9af43ed32decd0216ae7374b068c43e7ab59027b0dd0a63a
|
|
BLAKE2b-256 checksum How to use checksums |
1b2d51efa84f56124f7aa5cd9dbf86b57fc8cfb0456705ce2a1ddd61757f040b
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.11.14
|
Release files / soulpolicy-0.2.0-py3-none-any.whl
| Download URL | soulpolicy-0.2.0-py3-none-any.whl |
|---|---|
| Size | 17.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
905a57fd95fa1377a3da78f3ff0cd722c0bfc1d4b8f1ee0367b868b4efa728b3
|
|
BLAKE2b-256 checksum How to use checksums |
560165918386771d707dcf090344858c079320fb388140eac6109b761a51b146
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.11.14
|
Security for soulpolicy 0.2.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.