cci-sdk 0.1.0__py3-none-any.whl

cci_sdk-0.1.0.dist-info/METADATA

@@ -0,0 +1,430 @@
+Metadata-Version: 2.5
+Name: cci-sdk
+Version: 0.1.0
+Summary: Python SDK and standalone statistics engine for Conformal Logit Inference (CLI) — conformal prediction, Venn-Abers calibration, and e-values for LLM decisions.
+Project-URL: Homepage, https://cci.gitdate.ink
+Project-URL: Documentation, https://cci-docs.gitdate.ink
+Project-URL: Repository, https://github.com/rahvis/cci-sdk
+Author-email: CCI <rahul.vishwakarma@workonward.org>
+License: Apache-2.0
+License-File: LICENSE
+Keywords: calibration,conformal-prediction,e-values,llm,uncertainty-quantification,venn-abers
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Programming Language :: Python :: 3
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.9
+Requires-Dist: httpx>=0.27
+Requires-Dist: numpy>=1.24
+Provides-Extra: agent-framework
+Requires-Dist: agent-framework-core>=1.19; extra == 'agent-framework'
+Provides-Extra: all
+Requires-Dist: agent-framework-core>=1.19; extra == 'all'
+Requires-Dist: anthropic>=1.0; extra == 'all'
+Requires-Dist: google-adk>=2.0; extra == 'all'
+Requires-Dist: langchain-core>=1.0; extra == 'all'
+Requires-Dist: langchain>=1.0; extra == 'all'
+Requires-Dist: langgraph>=1.0; extra == 'all'
+Requires-Dist: openai>=1.40; extra == 'all'
+Provides-Extra: anthropic
+Requires-Dist: anthropic>=1.0; extra == 'anthropic'
+Provides-Extra: dev
+Requires-Dist: mypy>=1.10; extra == 'dev'
+Requires-Dist: pytest-asyncio>=0.23; extra == 'dev'
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Requires-Dist: ruff>=0.5; extra == 'dev'
+Provides-Extra: google-adk
+Requires-Dist: google-adk>=2.0; extra == 'google-adk'
+Provides-Extra: langchain
+Requires-Dist: langchain-core>=1.0; extra == 'langchain'
+Requires-Dist: langchain>=1.0; extra == 'langchain'
+Requires-Dist: langgraph>=1.0; extra == 'langchain'
+Provides-Extra: langgraph
+Requires-Dist: langchain-core>=1.0; extra == 'langgraph'
+Requires-Dist: langgraph>=1.0; extra == 'langgraph'
+Provides-Extra: openai
+Requires-Dist: openai>=1.40; extra == 'openai'
+Description-Content-Type: text/markdown
+
+# cci-sdk — Python SDK for Conformal Logit Inference
+
+Conformal Logit Inference (CLI) turns the raw output of any LLM into
+finite-sample statistical guarantees. This package contains:
+
+- **`cli_sdk`** — sync and async clients for the hosted API, the seven
+  guarantee-bearing query primitives, typed answers, calibration profiles,
+  and drift monitors.
+- **`cli_sdk.stats`** — the standalone statistics engine (conformal
+  prediction, Venn-Abers calibration, e-values). It needs only NumPy and
+  makes no network calls, so it runs inside air-gapped environments on
+  precomputed score arrays.
+
+- **`cli_sdk.local`** and **`cli_sdk.evidence`** — local mode: calibrate and
+  evaluate the same queries in-process on your own model (OpenAI, Azure
+  OpenAI, Anthropic, Gemini, or open weights such as Gemma on vLLM or
+  SGLang), with calibration profiles stored as reviewable JSON files.
+- **`cli_sdk.integrations`** — calibrated guardrails for agent tool calls in
+  LangChain, LangGraph, Google ADK and Microsoft Agent Framework: allow,
+  escalate to a human, or block, each with a stated guarantee.
+
+Every guarantee is marginal or group-conditional over data exchangeable with
+a stated calibration set. None is a promise about one decision in isolation.
+
+## Install
+
+```bash
+pip install cci-sdk          # or: uv add cli-sdk
+```
+
+Optional extras for model providers and agent frameworks (Python 3.10+ for
+the frameworks):
+
+```bash
+pip install "cci-sdk[openai]"            # OpenAI, Azure OpenAI, vLLM, SGLang evidence backends
+pip install "cci-sdk[anthropic]"         # Claude evidence backend
+pip install "cci-sdk[langchain,openai]"  # or [langgraph], [google-adk], [agent-framework], [all]
+```
+
+For development in this repository:
+
+```bash
+cd sdk/python
+pip install -e ".[dev]"
+pytest
+```
+
+## Authenticate
+
+```bash
+export CLI_API_KEY=sk_live_...
+# optional, for a self-hosted deployment:
+export CLI_BASE_URL=https://cli.internal.example.com/v1
+```
+
+A key is not required when `CLI_BASE_URL` points at `localhost`.
+
+## Accessing the API
+
+### Over HTTP
+
+Every call is `POST https://cci.gitdate.ink/api/v1/evaluate` with a bearer token:
+
+```bash
+curl https://cci.gitdate.ink/api/v1/evaluate \
+  -H "Authorization: Bearer $CLI_API_KEY" \
+  -H "Content-Type: application/json" \
+  -d '{
+    "context": {"ticket": "My payouts have been failing for 3 days."},
+    "backend": {"provider": "openai", "model": "gpt-4.1-2025-04-14"},
+    "queries": {
+      "department": {
+        "type": "set",
+        "instructions": "Which team should handle this ticket?",
+        "options": {"billing": null, "technical": null, "sales": null},
+        "calibration_profile": "support-routing-v3",
+        "alpha": 0.10
+      }
+    }
+  }'
+```
+
+| Endpoint | Purpose |
+|---|---|
+| `POST /v1/evaluate` | Evaluate a context against one or more queries |
+| `POST /v1/calibration-profiles` | Create a calibration profile |
+| `GET /v1/calibration-profiles/{name}` | Size, coverage interval, audit history |
+| `POST /v1/calibration-profiles/{name}/examples` | Add labelled examples |
+| `POST /v1/calibration-profiles/{name}/label-with-judge` | Judge-labelled pool plus a human sample, combined validly |
+| `POST /v1/calibration-profiles/{name}/audit` | Audit against fresh labelled examples |
+| `POST /v1/calibration-profiles/{name}/monitors` | Create an anytime-valid drift monitor |
+| `GET /v1/calibration-profiles/{name}/monitors/{id}/alerts` | Poll monitor alerts |
+| `GET /v1/backends` | Configured backends and their detected access level |
+
+### With the SDK
+
+```python
+from cli_sdk import CLIClient, Set, Gate
+
+with CLIClient() as client:
+    result = client.evaluate(
+        context={"ticket": "My payouts have been failing for 3 days."},
+        backend={"provider": "openai", "model": "gpt-4.1-2025-04-14"},
+        queries={
+            "department": Set(
+                instructions="Which team should handle this ticket?",
+                options={"billing": None, "technical": None, "sales": None},
+                calibration_profile="support-routing-v3",
+                alpha=0.10,
+            ),
+            "route": Gate(
+                instructions="Auto-route this ticket without human review?",
+                calibration_profile="support-routing-v3",
+                guarantee="fdr",
+                target=0.05,
+            ),
+        },
+    )
+
+department = result.answers["department"]
+print(department.set, department.guarantee.describe())
+
+if result.answers["route"].approved:
+    assign_queue(department.top)
+else:
+    send_to_human_triage(department.set)
+```
+
+The async client has the same interface, with `await` on each call (`monitors.poll` returns a list rather than a generator):
+
+```python
+from cli_sdk import AsyncCLIClient
+
+async with AsyncCLIClient() as client:
+    result = await client.evaluate(context=..., backend=..., queries=...)
+```
+
+## The seven primitives
+
+| Primitive | Returns | Guarantee |
+|---|---|---|
+| `Belief` | Venn-Abers interval `[p0, p1]` for one statement | calibrated probability bracket |
+| `Set` | a set of options | coverage >= 1 - alpha |
+| `Interval` | a conformalized ordinal/continuous interval | coverage >= 1 - alpha |
+| `Gate` | `auto_approve` / `escalate` / `abstain` | risk <= alpha, risk with probability 1 - delta, or batch FDR <= q |
+| `Claim` | long-form output filtered to supported claims | P(all retained claims true) >= 1 - alpha |
+| `Judge` | a verdict with an escalation cascade | human agreement >= 1 - alpha on accepted verdicts |
+| `Route` | which backend in a cascade served the request | cost or accuracy bound |
+
+Every answer carries `answer.guarantee`, a `Guarantee` with `type`,
+`method`, `alpha`/`delta`/`target`, `calibration_profile`, `calibration_n`,
+`coverage_ci`, and `last_audited`. `guarantee.is_heuristic` is `True` when an
+answer carries no formal guarantee, for example because its profile is still
+below the minimum size. Pass `strict_guarantees=True` to the client to raise
+`InsufficientCalibrationError` instead of receiving heuristic answers.
+
+## Typed responses
+
+```python
+from cli_sdk import EvaluateResponse, SetAnswer
+
+class RoutingResponse(EvaluateResponse):
+    department: SetAnswer
+
+result = client.evaluate(..., response_model=RoutingResponse)
+result.department  # a SetAnswer; a missing or mistyped answer raises at parse time
+```
+
+## Calibration profiles
+
+```python
+client.calibration_profiles.create(
+    name="support-routing-v3",
+    backend={"provider": "openai", "model": "gpt-4.1-2025-04-14"},
+    method="APS",
+    alpha=0.10,
+    group_by="account_tier",       # optional Mondrian (per-group) calibration
+)
+client.calibration_profiles.add_examples(
+    "support-routing-v3",
+    examples=[{"context": {"ticket": "I was charged twice."}, "label": "billing"}],
+)
+profile = client.calibration_profiles.get("support-routing-v3")
+profile.n, profile.minimum_n, profile.recommended_n, profile.realized_coverage_ci
+```
+
+| alpha | hard minimum n | recommended n |
+|---|---|---|
+| 0.20 | 4 | ~300 |
+| 0.10 | 9 | ~1,000 |
+| 0.05 | 19 | ~1,500 |
+| 0.01 | 99 | ~2,500 |
+
+Label-efficient calibration labels a pool with a judge and a random human
+subset, combined so the result is valid regardless of judge quality:
+
+```python
+client.calibration_profiles.label_with_judge(
+    "support-routing-v3",
+    judge={"provider": "openai", "model": "gpt-4.1-2025-04-14"},
+    unlabelled_examples=pool,
+    human_labelled_sample_size=300,
+)
+```
+
+## Drift monitoring
+
+```python
+client.calibration_profiles.monitors.create(
+    "support-routing-v3", type="coverage", target=0.90,
+    false_alarm_rate=0.05, labelled_sample_rate=0.02,
+)
+for alert in client.calibration_profiles.monitors.poll("support-routing-v3"):
+    pause_auto_actioning()
+```
+
+Monitors are e-processes: the false-alarm rate holds no matter how often or
+for how long you check them. `LocalMonitor` runs the same test in-process:
+
+```python
+from cli_sdk import LocalMonitor
+
+monitor = LocalMonitor(type="coverage", target=0.90, false_alarm_rate=0.05)
+for label, prediction_set in stream:
+    if alert := monitor.update(label in prediction_set):
+        page_oncall(alert)
+```
+
+## Backends
+
+```python
+from cli_sdk import OpenAIBackend, AnthropicBackend, VLLMBackend
+
+OpenAIBackend(model="gpt-4.1-2025-04-14")                  # L1 on non-reasoning configs
+AnthropicBackend(model="claude-sonnet-5", sample_count=20) # L0: sampling only
+VLLMBackend(model="Qwen/Qwen3-8B", base_url="http://vllm:8000")  # up to L4
+```
+
+Plain dicts (`{"provider": "openai", "model": ...}`) work anywhere a backend
+is accepted.
+
+### Bring your own model
+
+A `CustomBackend` runs your model locally and sends only the evidence
+(probabilities, samples, scores) to CLI. Implement the methods your access
+level supports:
+
+```python
+from cli_sdk import CLIClient, CustomBackend
+
+class MyEngine(CustomBackend):
+    access_level = "L1"
+
+    def score_options(self, context, instructions, options):
+        return my_model.option_probabilities(context, instructions, list(options))
+
+    def sample(self, context, instructions, n):
+        return [my_model.generate(context, instructions) for _ in range(n)]
+
+client = CLIClient(backend=MyEngine())
+```
+
+## Local mode
+
+`LocalCLIClient` runs the same queries fully in-process: it scores labelled
+examples with a model you bring, stores the calibration profile as a JSON
+file, and answers with the same typed answers and guarantee cards as the
+hosted client. Nothing leaves your process except calls to your own model
+provider.
+
+```python
+from cli_sdk import Gate, LocalCLIClient
+from cli_sdk.evidence import OpenAIEvidenceBackend, vllm_backend
+
+evidence = OpenAIEvidenceBackend("gpt-4.1-mini", api_key="YOUR_OPENAI_API_KEY")
+# or open weights:  evidence = vllm_backend("google/gemma-4-12B-it", base_url="http://localhost:8000/v1")
+
+client = LocalCLIClient(evidence, store=".cli_profiles")
+refund = Gate(instructions="Is issuing this refund correct under the policy?",
+              calibration_profile="refund-approvals-v1", guarantee="risk", target=0.05)
+client.calibrate(refund, labelled_examples)   # [{"context": {...}, "label": True}, ...]
+answer = client.evaluate(context=case, queries={"refund": refund}).answers["refund"]
+print(answer.decision, answer.guarantee.describe())
+```
+
+Profiles are fingerprinted with the backend configuration and the query's
+prompt: change either and answers fall back to their safe heuristic form
+(escalate, every option, `[0, 1]`) until you recalibrate. See the
+[local mode guide](https://cci-docs.gitdate.ink/agents/local-mode).
+
+## Agent frameworks
+
+`ToolGuard` decides each proposed tool call from a calibrated answer and
+maps it onto the framework's native human-in-the-loop mechanism:
+
+```python
+from cli_sdk.integrations import GuardRule, ToolGuard
+from cli_sdk.integrations.langchain import cli_middleware
+
+guard = ToolGuard(client, [GuardRule(tool="issue_refund", query=refund, context=refund_context)])
+agent = create_agent(model, tools=[lookup_order, issue_refund],
+                     middleware=cli_middleware(guard), checkpointer=InMemorySaver())
+```
+
+| Framework | Module | Hook | Human review |
+|---|---|---|---|
+| LangChain 1.x | `cli_sdk.integrations.langchain` | `wrap_tool_call` middleware | `HumanInTheLoopMiddleware` |
+| LangGraph | `cli_sdk.integrations.langgraph` | guard node before `ToolNode` | `interrupt()` / `Command(resume=...)` |
+| Google ADK 2.x | `cli_sdk.integrations.google_adk` | `before_tool_callback` or plugin | tool confirmation |
+| Microsoft Agent Framework | `cli_sdk.integrations.agent_framework` | `FunctionMiddleware` | function approval requests |
+
+Eight runnable examples in risk-sensitive domains (refunds, patient triage,
+insurance claims, discharge summaries, AML holds, trial screening, credit
+tiers, drug safety) are in `examples/agents/`. Each runs offline with
+`--provider mock` and with your own keys for OpenAI, Azure, Anthropic,
+Gemini, or Gemma on vLLM or SGLang. See the
+[agent frameworks guide](https://cci-docs.gitdate.ink/agents).
+
+## Offline statistics engine
+
+```python
+from cli_sdk.stats.conformal import aps, crc
+from cli_sdk.stats.venn_abers import ivap
+from cli_sdk.stats.evalues import ebh, ppi, CoverageMonitor
+
+q_hat = aps.calibrate(cal_probs, cal_labels, alpha=0.10)
+sets = aps.predict(test_probs, q_hat)
+
+p0, p1 = ivap.calibrate_and_predict(cal_scores, cal_labels, test_scores)
+
+selected = ebh.select(e_values, q=0.10)
+```
+
+| Module | Methods |
+|---|---|
+| `stats.conformal` | LAC, APS, RAPS, CQR, CRC, RCPS, Learn-then-Test, Mondrian |
+| `stats.venn_abers` | IVAP, CVAP, interval merging and width |
+| `stats.evalues` | anytime-valid risk and coverage monitors, e-BH, prediction-powered inference |
+
+## Command-line tool
+
+```bash
+cli calibration show support-routing-v3
+cli calibration audit support-routing-v3 --examples fresh.jsonl --fail-below 0.88
+cli calibration audit-local --covered outcomes.txt --target 0.90
+```
+
+`audit` exits with status 1 on failure, so it can block a deploy in CI.
+
+## Errors
+
+| Exception | When |
+|---|---|
+| `AuthenticationError` | 401: missing or invalid API key |
+| `ValidationError` | 422: the request failed validation |
+| `InsufficientCalibrationError` | a profile is too small for the guarantee (raised only with `strict_guarantees=True`) |
+| `RateLimitError` | 429 after retries are exhausted |
+| `BackendError` | 502 / 529: the model backend errored or reported a lower access level than required |
+| `ConfigurationError` | client-side misconfiguration, raised before any request |
+
+`429`, `502` and `529` are retried with full-jitter exponential backoff,
+honoring `retry-after`. Configure with `CLIClient(retry=RetryConfig(max_attempts=5, base_delay=0.5))`.
+
+## Package layout
+
+```
+src/cli_sdk/
+  client/         sync + async clients, retry policy
+  queries/        Belief, Set, Interval, Gate, Claim, Judge, Route
+  answers/        typed answers, the Guarantee card, EvaluateResponse
+  calibration/    profiles, examples, audits, label-efficient calibration
+  monitoring/     hosted monitors, alerts, LocalMonitor
+  backends/       one adapter per provider, plus CustomBackend
+  stats/          offline engine: conformal/, venn_abers/, evalues/
+  local/          LocalCLIClient: in-process calibration and evaluation
+  evidence/       evidence backends: OpenAI, Azure, Anthropic, Gemini, vLLM, SGLang, mock
+  integrations/   ToolGuard and adapters for LangChain, LangGraph, ADK, Agent Framework
+  cli_tool.py     the `cli` command
+examples/         runnable end-to-end examples (agents/: agent-framework examples)
+tests/            engine and client tests
+```

cci_sdk-0.1.0.dist-info/RECORD

@@ -0,0 +1,92 @@
+cli_sdk/__init__.py,sha256=NKRmlmMkj-qMqknHeJFnc9t8EdiDvn6zoKY6MdNlS4k,2988
+cli_sdk/cli_tool.py,sha256=uE36rN-Y4wHu8rffXlNvX2MGav6I6fgPuEIX9ROUM6M,5849
+cli_sdk/constants.py,sha256=86Of269FcxCAddaeotp-oBJXZ3bg6hnQpsXvYYFY1Ok,1048
+cli_sdk/exceptions.py,sha256=kwGHhJ3sbt3E2_tVBC7DTLRP5i_x1p3nLXGXsJOqD3w,2002
+cli_sdk/py.typed,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+cli_sdk/answers/__init__.py,sha256=ZJ637QK1TEClnoSrcJgUOiMyLGImxvOlSvkDLgZrxO0,910
+cli_sdk/answers/base.py,sha256=QtQ4SRehQY596dSn6SVWBCS4OmjA48jwQxbemZnACKM,634
+cli_sdk/answers/belief_answer.py,sha256=M0RcRB_ZMDwHR9-B_phzcJ4NP10ZNDngR-bOuS1PVgc,1603
+cli_sdk/answers/claim_answer.py,sha256=DKR4GY6dFxFBRCpRUgyyECxzDfQFQYg9rye4ySvosTc,1727
+cli_sdk/answers/gate_answer.py,sha256=YduF85j3jPyfTn3Zx7SjrsNslG9euf5WSbRMV83j4wc,963
+cli_sdk/answers/guarantee.py,sha256=AM5l1Y0IaI1rgZ8KRrIAbsO2wGDHWhqGSlo3F30qtC0,6648
+cli_sdk/answers/interval_answer.py,sha256=8sGmnLtj2W0hm4hM7wbavZQUv9wi87F7RXOn-TlL7pA,1088
+cli_sdk/answers/judge_answer.py,sha256=3muSNv4aQnhJmROkeapR_qKlCSBzWPNykEI9UJiXgS8,894
+cli_sdk/answers/response.py,sha256=h4P9vzrEI2DL5rLQ67LkQKosfvDQGI6cH1L7cOie_vo,4821
+cli_sdk/answers/route_answer.py,sha256=JaFQ01oWIVXnsCTCpspsQzRp-1w-GnPosjjtqUvlmw8,928
+cli_sdk/answers/set_answer.py,sha256=EAWSuYE-4F34DxUR8H6zU6q6UBOeIWxFTtGGWp19F1I,1633
+cli_sdk/backends/__init__.py,sha256=tm7kl9d7mR-YTFf4Tc90FjqdVu9JUbYce7Hjfo6LW74,1007
+cli_sdk/backends/anthropic.py,sha256=wta0lKxqwhHwbS0IKrfMde4Rs2U7Tbn2_43_QwmbtkM,734
+cli_sdk/backends/azure_openai.py,sha256=xskNICPOUT1ztNi0GgPIKP7ncGHzP6LUsgHymWZVLM8,699
+cli_sdk/backends/base.py,sha256=nTJL2mDGotsHv4SuoyAs4bzodSecOMgDTsNpqNXmPtk,8541
+cli_sdk/backends/bedrock.py,sha256=mmu8VgcVf7FFPtF3tZqpwgJmyMCaGP7ioCDn2nttG1k,544
+cli_sdk/backends/custom.py,sha256=eTqxQ8NQHTgGBQYhlKA7UHyIOt5HBowzNt8bt1un2lY,179
+cli_sdk/backends/gemini.py,sha256=qtIb7vq9QsANd8iF3ETaDpTjAec1NRYCnEdqgXWYbZk,483
+cli_sdk/backends/openai.py,sha256=XPnO40JEMCyt1SmsRVS1VKZFJ9FKZykycornPdFp2Ts,760
+cli_sdk/backends/openrouter.py,sha256=HjrZO9Lg826Kc-0WlXvFgraA7LYIrmZ4xE9CG6V6izY,735
+cli_sdk/backends/sglang.py,sha256=GhJ7hKuZ4kum4kHC5WprXtoeM5cPqTFhFR97zE640bU,1023
+cli_sdk/backends/vllm.py,sha256=Iw9ZP5DBllySMQevu80ejcUglaVG3UQFukuOHDX6TeU,1061
+cli_sdk/calibration/__init__.py,sha256=-xgLiKg4Qr4O9NHT_3_6pPYW266qSO_47efVINnKzJU,832
+cli_sdk/calibration/audit.py,sha256=xdEMnueB4Iwv9kP0dRezDoTu7nMkskYCSJDVOxtan8o,3266
+cli_sdk/calibration/examples.py,sha256=kOHgf8nQxUJbXEEzzMLmdN0R69UXMOoG-TQp2MbMvvc,2066
+cli_sdk/calibration/label_efficient.py,sha256=06B9XRgXvjhM06mfH8ZqxS-Sl68Mcj4y10fO9hFmd0w,3370
+cli_sdk/calibration/profile.py,sha256=du5QWsHkFLgE8N4yccgvgOGVxKJsDZ7GkSvKT--sJDU,9305
+cli_sdk/client/__init__.py,sha256=3F-eSatsWf9RN93nOHXbMxhzDNkgkFvpG1dBFmkbKD4,267
+cli_sdk/client/async_client.py,sha256=FVxsXDIDlNbqFCIk28aIF-antj5cGtvYeCsx8c42YEM,4470
+cli_sdk/client/base.py,sha256=f4z5gSkzqcmcK8CZCx5obOeiDjD1q76YCxvs_Et0F8Q,8352
+cli_sdk/client/retries.py,sha256=BxgH0-MmUgqV9_7LiSyYKm14W-tiLBdxsNPj3uM3ERE,2824
+cli_sdk/client/sync_client.py,sha256=Jx0z3LCeYoSKuhazsK2wMF1s61i8lLu-Cw-I_5MD_00,5445
+cli_sdk/evidence/__init__.py,sha256=pLbfnSp0CZDnr-kJIR5bzQ6sfq3NDCjIafY0irSMFxk,2115
+cli_sdk/evidence/_prompts.py,sha256=SiZJQv4yAXg65UgaXXekwdOlgVvANlXHyD1wzHuPKIk,6758
+cli_sdk/evidence/anthropic.py,sha256=3Y1WblGJmGbaCOTyRKbKxgBs_OXgN8f12RmTQCJv6c4,6221
+cli_sdk/evidence/base.py,sha256=EIleGWAllfLB-tjNELQ944EXIaYy9TVFjtvi4zntB3w,5506
+cli_sdk/evidence/langchain.py,sha256=2MfVqDNqSVaBzP4NDnmox8diSNJkweO_LShovmxpsS0,5061
+cli_sdk/evidence/mock.py,sha256=EsEivEQFXVsV-jDIEIlRgbNuU540j7p4k3ZFh60HdOQ,5977
+cli_sdk/evidence/openai_compatible.py,sha256=2-mHBZv8DL5ileSiuOkOCeQUDi4zXptxkP1LRPETCJQ,11790
+cli_sdk/integrations/__init__.py,sha256=P5--yExihe4npc0OWukB27qdwz7NSg__eGhhgHOxNnw,1410
+cli_sdk/integrations/_guard.py,sha256=FsyPIZTZ-tQTYeJFF5lMV4YBNIaRetXU6BrlLQwPtZc,16767
+cli_sdk/integrations/agent_framework.py,sha256=tZnSqJDnEU5hWSWju6q2fiFd6IWkSiDxoeXeAnNV5sU,6358
+cli_sdk/integrations/google_adk.py,sha256=_ty3U_hCDRvXiYuKFRC7TuFGoK3car5ns7A8_Tsikl8,6891
+cli_sdk/integrations/langchain.py,sha256=cXvpbeA-sfrr0xLr368p03hPpuo2VGHRgN7liHOaoEo,11684
+cli_sdk/integrations/langgraph.py,sha256=iTdKMyMJw7rYJ9nCQaLP2dFeELy3EhuhRunD2TXPsEk,9889
+cli_sdk/local/__init__.py,sha256=_UPZ3--X3SjKW8wGuYOlNJHMKs06ntb1B5xNwjyARUc,696
+cli_sdk/local/client.py,sha256=r7IX48IKsmdahQlxr1c7o60UQmVe6zwsTVLDEpmrRSc,28139
+cli_sdk/local/engine.py,sha256=bPgPQrBeOqvWBPpCTQn4D_pliXUlcUaYHSX76sLhsFM,29644
+cli_sdk/local/scoring.py,sha256=8XWAP-sFNsZMqCpc8CuP2ldOYIzaro0bGiLzHE6Z44I,6698
+cli_sdk/local/store.py,sha256=2TjZSdHzqjnw1swA90YdVaOan2TTOUfdxNTXtC3Lvo8,5304
+cli_sdk/monitoring/__init__.py,sha256=jqc85muXbGlV9FtX0xJMW9HWze-u0ailhCwKQ8cVuak,297
+cli_sdk/monitoring/alerts.py,sha256=__FSsV4S7r2SNQcvnZWAtkTECHwYfnRaPlhHem54gHQ,1637
+cli_sdk/monitoring/monitor.py,sha256=EB-Z8IKw_hqTRg9g58rDf4F-3ZckRYu5vMnrgqwYbFA,7047
+cli_sdk/queries/__init__.py,sha256=CXOXWGcL346ax4ReaWaX-kvx_vEEolgff4pw5P2AiDQ,529
+cli_sdk/queries/base.py,sha256=lr9lUme_cRmLRBOTm36WevcPuTrYEaVZf9CpKU58C9c,2180
+cli_sdk/queries/belief.py,sha256=DFy3dstKONZ-9ocGIRZaqNKhWwdkEzVoVtNtAsn5lVY,806
+cli_sdk/queries/claim.py,sha256=MeZfUWCecndI9bqf70CJF8UqM96reK9SxdL297ErZrg,893
+cli_sdk/queries/gate.py,sha256=QZhZimscShYbxr3szgTRhbqqUD-I7H4E0yCU1uhwrRE,1391
+cli_sdk/queries/interval.py,sha256=LADAdWOGryQIMiy5EZJEhQdrnMS4E840uWxT4V0ba4I,1280
+cli_sdk/queries/judge.py,sha256=s1rd9goNKu49uiziOF_LRU9ktkmJAD-jCVinChjrpTk,922
+cli_sdk/queries/route.py,sha256=qk9bFk5nSoSMrEbR-rkCHhfnK3geMbt4MllfJsSKXQo,1423
+cli_sdk/queries/set.py,sha256=fooAqL8QxcIdlua_pksllZhK36tHeDLuY-8RejB-juk,1336
+cli_sdk/stats/__init__.py,sha256=-bxfJECa684Btsid-YEWXVA_r8aHf38zXHZauHYAQ4s,1090
+cli_sdk/stats/conformal/__init__.py,sha256=yEy_7Gdnne9NsLPistAMNkLFticZq2bgtGbUuTy_mzU,767
+cli_sdk/stats/conformal/_quantile.py,sha256=p2lyXus7J_z-MQg-K0HDS5l5Xy6v6SjeNjgqkZSQuFw,4929
+cli_sdk/stats/conformal/aps.py,sha256=HggLTQZ98TJQsNmvWo8KhTlUD9iZLDKzAZIRV8d7qh8,3769
+cli_sdk/stats/conformal/cqr.py,sha256=Jbxml3-32kj3BK-31QA8q2Xip9lWoIcIiLmQwAkedDg,1564
+cli_sdk/stats/conformal/crc.py,sha256=bEw3pp_hBvUvhX2-0M3tJOZAio58__UWzt8J5uA8npE,2710
+cli_sdk/stats/conformal/lac.py,sha256=1LLReheW0ec729xzIhhC1TVouts44Ms6ciOXB9rp6G4,1341
+cli_sdk/stats/conformal/ltt.py,sha256=Wk9I7KVxH_elrKrAFAodegXHjmLCAfebFzonf_37-DQ,2906
+cli_sdk/stats/conformal/mondrian.py,sha256=CZDzsJrJZjctMBDeh34rPgBiM1YSGbZ_Do6-UFOhf0I,2738
+cli_sdk/stats/conformal/raps.py,sha256=IahBUpDtIuh5R4DJf9A6w23-TSBsEBuxYXMNXhH-T3I,3250
+cli_sdk/stats/conformal/rcps.py,sha256=2V9BZAWBCqY7SkOlePwln1rtrCfpUFBOZm6Zx6iPJ_I,3743
+cli_sdk/stats/evalues/__init__.py,sha256=Y0gNFZt124NenfuRrm1aRFsTez6JEh63_diN0NSkbmo,747
+cli_sdk/stats/evalues/anytime.py,sha256=iUcwAcBOMIR6yTp7Gg5rzb4Rg5ka-UnePSJAIbnXIJk,5147
+cli_sdk/stats/evalues/ebh.py,sha256=VCStp3_1iSvU8IKnd7n1VO9SGk0krGz871eITsX-7i0,1935
+cli_sdk/stats/evalues/eprocess.py,sha256=KhoPbKGxvqq5B58THSJPF93p3WBdTS5sdisOuAasjx0,2231
+cli_sdk/stats/evalues/ppi.py,sha256=bBZLC1B6Ns3Vm0AiiUVDocnrEYtGg34ZwfCHsaRqvjg,3222
+cli_sdk/stats/venn_abers/__init__.py,sha256=Ho1UEkOyNR94lF5vOIEM6b1p3oTYo1ukvViZfgAsk0A,463
+cli_sdk/stats/venn_abers/_isotonic.py,sha256=ux3r0_CcjsuRj3hvCie8ZWgHxLd9R_CEueiQ9IhHqls,1473
+cli_sdk/stats/venn_abers/cvap.py,sha256=f3fORbBC0ShnUyYNw747kcHpT4lLvR2fK9-BQ3jQ9TE,2356
+cli_sdk/stats/venn_abers/ivap.py,sha256=Yua1H2OHiDUKw6SMHhEeDmx4sIsGzk40_wkvwZCGyNU,4650
+cci_sdk-0.1.0.dist-info/METADATA,sha256=t5TKvtlBmZxuGHrrLurlIBXbyJrZL4Gs9Nh0dTcC6zs,16746
+cci_sdk-0.1.0.dist-info/WHEEL,sha256=W3fkpkm7-wf9vBI5Z-7s0eWkeM-spu78I8Neb98DeEg,87
+cci_sdk-0.1.0.dist-info/entry_points.txt,sha256=3Habu-N3nF77zkm0IqAWenL3Fuvc300ucebniR74FgY,46
+cci_sdk-0.1.0.dist-info/licenses/LICENSE,sha256=z8d0m5b2O9McPEK1xHG_dWgUBT6EfBDz6wA0F7xSPTA,11358
+cci_sdk-0.1.0.dist-info/RECORD,,

cci_sdk-0.1.0.dist-info/WHEEL

@@ -0,0 +1,4 @@
+Wheel-Version: 1.0
+Generator: hatchling 1.32.4
+Root-Is-Purelib: true
+Tag: py3-none-any

cci_sdk-0.1.0.dist-info/entry_points.txt

@@ -0,0 +1,2 @@
+[console_scripts]
+cli = cli_sdk.cli_tool:main