v0.2.0: point SDK at api.latence.ai gateway

Breaking: default base_url changed from localhost to https://api.latence.ai.
Update endpoint paths to /v1/grounding, /v1/redact, /v1/compress.
RunPod-direct mode still works when LATENCE_TRACE_DEPLOYMENT=runpod.

Author: ddickmann

CHANGELOG.md

@@ -0,0 +1,21 @@
+# Changelog
+
+## 0.2.0 (2026-05-08)
+
+**Breaking**: default `base_url` changed from `http://localhost:8090` to
+`https://api.latence.ai`. All traffic now routes through the authenticated
+gateway which handles API key verification, rate limiting, balance gating,
+and usage logging.
+
+- **Endpoint paths** aligned with the public gateway surface:
+  - `/groundedness` -> `/v1/grounding`
+  - `/v1/compliance/redact` -> `/v1/redact`
+  - `/v1/compression` -> `/v1/compress`
+- RunPod-direct mode (`LATENCE_TRACE_DEPLOYMENT=runpod`) still works for
+  local or self-hosted deployments; the legacy action map is preserved.
+- Code scoring (`grounding.code()`) and sessions remain `NotImplementedError`.
+  Use `grounding.rag()` for all groundedness scoring.
+
+## 0.1.6
+
+Initial public release.

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "latence"
-version = "0.1.6"
+version = "0.2.0"
 description = "Thin Python SDK for Latence TRACE."
 readme = "README.md"
 requires-python = ">=3.10"

src/latence/_transport.py

@@ -25,16 +25,18 @@
     _Envelope,
 )
 
-DEFAULT_USER_AGENT = "latence/0.1.6"
+DEFAULT_USER_AGENT = "latence/0.2.0"
 DEFAULT_TIMEOUT_SECONDS = 30.0
 DEFAULT_MAX_RETRIES = 4
 RETRYABLE_STATUS = frozenset({429, 500, 502, 503, 504})
 RUNPOD_PRODUCT_ACTIONS = {
+    "/v1/grounding": "score",
+    "/v1/redact": "redact",
+    "/v1/compress": "compress",
+    # Legacy paths — kept for RunPod-direct users during migration
     "/groundedness": "score",
     "/v1/compliance/redact": "redact",
     "/v1/compression": "compress",
-    "/v1/memory/update": "memory.update",
-    "/groundedness/rollup": "rollup",
 }
 
 
@@ -212,7 +214,7 @@ def _envelope_from_body(body: Any) -> _Envelope | None:
 
 
 def coerce_base_url(base_url: str | None) -> str:
-    base = base_url or os.environ.get("LATENCE_TRACE_URL", "http://localhost:8090")
+    base = base_url or os.environ.get("LATENCE_TRACE_URL", "https://api.latence.ai")
     return base.rstrip("/")
 
 

src/latence/async_client.py

@@ -105,7 +105,7 @@ def __init__(self, owner: AsyncLatence) -> None:
     async def text(self, text: str, **options: Any) -> CompressionResponse:
         return await self._owner._request(
             "POST",
-            "/v1/compression",
+            "/v1/compress",
             json={"text": text, **options},
             expected_model=CompressionResponse,
         )
@@ -117,7 +117,7 @@ async def messages(
     ) -> CompressionResponse:
         return await self._owner._request(
             "POST",
-            "/v1/compression",
+            "/v1/compress",
             json={"action": "compress_messages", "messages": list(messages), **options},
             expected_model=CompressionResponse,
         )
@@ -307,7 +307,7 @@ async def score_groundedness(
         )
         return await self._request(
             "POST",
-            "/groundedness",
+            "/v1/grounding",
             json=payload,
             expected_model=GroundednessResponse,
         )
@@ -358,7 +358,7 @@ async def redact_compliance(
             body.update(dict(extra))
         return await self._request(
             "POST",
-            "/v1/compliance/redact",
+            "/v1/redact",
             json=body,
             expected_model=ComplianceRedactionResponse,
         )

src/latence/client.py

@@ -248,7 +248,7 @@ def __init__(self, owner: Latence) -> None:
     def text(self, text: str, **options: Any) -> CompressionResponse:
         return self._owner._request(
             "POST",
-            "/v1/compression",
+            "/v1/compress",
             json={"text": text, **options},
             expected_model=CompressionResponse,
         )
@@ -260,7 +260,7 @@ def messages(
     ) -> CompressionResponse:
         return self._owner._request(
             "POST",
-            "/v1/compression",
+            "/v1/compress",
             json={"action": "compress_messages", "messages": list(messages), **options},
             expected_model=CompressionResponse,
         )
@@ -472,7 +472,7 @@ def score_groundedness(
         )
         return self._request(
             "POST",
-            "/groundedness",
+            "/v1/grounding",
             json=payload,
             expected_model=GroundednessResponse,
         )
@@ -525,7 +525,7 @@ def redact_compliance(
             body.update(dict(extra))
         return self._request(
             "POST",
-            "/v1/compliance/redact",
+            "/v1/redact",
             json=body,
             expected_model=ComplianceRedactionResponse,
         )