Skip to main content

corespine

Spine 家族的薄共享核(见 ADR 0001)。 只装 domain-neutral 的底层原语——既不属于 RAG 也不属于 agent 的稳定地基,被 ragspine / spineagent 兄弟包各自依赖,含任何它们的领域概念。

刻意地薄。按证据(rule of three)增长,不预先造框架。详见 CLAUDE.md 宪章。

🤖 给 AI / LLM: 用本库前先读 llms.txt(精简索引)与 docs/llms/(完整 API / recipes / 陷阱);pip install 后这些文档随包位于 site-packages/corespine/_llms/

缝的元模式

每条缝都长一个样,核心 import 零 SDK、离线可跑:

Protocol + 离线确定性默认 + Registry / make_* 工厂 + 参数化 conformance

里面有什么

模块 原语
seam/registry.py Registry:name→factory 解析(大小写/留白/连字符不敏感)+ entry-point 自动发现(corespine.<seam> group)+ 未知 spec 列清可用名 + lazy_extra_import(缺 extra 给"pip install …"友好提示)
observability/trace.py TraceSink 协议 + InProcessPrivacyTraceSink:只记 code/计数/耗时,拒绝任何携带正文(answer/value/text/content…)的载荷
llm/provider.py LLMProvider 协议 + 离线确定性 MockProvider(零网络、零 key、可复现)
config/env.py load_from_env:把 PREFIX_* 环境变量读进一个 frozen dataclass(范式同 ragspine from_env)
queue/task_queue.py TaskQueue 协议 + FakeQueue:同步内联执行 + 记录,离线/测试用
conformance/harness.py ConformanceSuite × InvariantPack:把"实现 × 不变量"绑成笛卡尔积逐格执行(机制,具体不变量由各 app 自己绑)

本地开发(始终从包根)

uv venv .venv
VIRTUAL_ENV="$(pwd)/.venv" uv pip install -e ".[dev]"
.venv/bin/python -m pytest -q
.venv/bin/python -c "import corespine"

30 秒上手

from corespine import Registry, MockProvider, InProcessPrivacyTraceSink, FakeQueue

# 缝:一个 spec 选实现(大小写/留白不敏感;找不到列清可用名;还能 entry-point 自动发现)
reg: Registry = Registry("llm")
reg.register("mock", lambda **kw: MockProvider(**kw))
provider = reg.make("  MOCK ")
# OpenAI chat-completions 规范:messages 进,OpenAI 形状的 ChatCompletion 出(确定性可复现)
out = provider.chat([{"role": "user", "content": "hello"}])
print(out.choices[0].message.content)

# 隐私 trace:只记元数据;塞正文会被直接拒绝(raise TraceError)
sink = InProcessPrivacyTraceSink()
sink.emit("retrieve", count=3, took_ms=12)

# 任务队列:同步内联执行
q = FakeQueue()
jid = q.enqueue(lambda p: {"doubled": p["n"] * 2}, {"n": 21})
print(q.get(jid).result)                 # {'doubled': 42}

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

corespine-0.1.1.tar.gz (57.9 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

corespine-0.1.1-py3-none-any.whl (41.5 kB view details)

Uploaded Python 3

File details

Details for the file corespine-0.1.1.tar.gz.

File metadata

  • Download URL: corespine-0.1.1.tar.gz
  • Upload date:
  • Size: 57.9 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.2

File hashes

Hashes for corespine-0.1.1.tar.gz
Algorithm Hash digest
SHA256 8f3c3cf54bbc43282cc23bce228bdbe5db9a9e5d65a434d922bfc68e5df14685
MD5 c4cf05a58388ea7683f932b4d38a1d81
BLAKE2b-256 cf95516e456906ac250a0eb4e0e427747ed1e038256a7acd6b0b4e30e5350f5d

See more details on using hashes here.

File details

Details for the file corespine-0.1.1-py3-none-any.whl.

File metadata

  • Download URL: corespine-0.1.1-py3-none-any.whl
  • Upload date:
  • Size: 41.5 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.13.2

File hashes

Hashes for corespine-0.1.1-py3-none-any.whl
Algorithm Hash digest
SHA256 896c35286d25b8dddf52c42334d45409d7224910bde3a71bd0aab0abecb1f5fb
MD5 8ab20dbc6e619f146d14a7b5e7a1bc2b
BLAKE2b-256 fdc02ecc6164f6c1b8d07cf2cae9664c773407316923a2533188136a4586846b

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page