devora-python 0.1.0__py3-none-any.whl → 0.1.1__py3-none-any.whl

devora_python-0.1.1.dist-info/METADATA

@@ -0,0 +1,167 @@
+Metadata-Version: 2.5
+Name: devora-python
+Version: 0.1.1
+Summary: Devora Python backend SDK for secure impersonation
+Project-URL: Documentation, https://docs.devora.sh
+Project-URL: Repository, https://github.com/getdevora/devora-sdks
+Project-URL: Issues, https://github.com/getdevora/devora-sdks/issues
+Author: Devora
+License-Expression: MIT
+License-File: LICENSE
+Keywords: backend,devora,impersonation,sdk,security
+Classifier: Development Status :: 5 - Production/Stable
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Programming Language :: Python :: 3.14
+Classifier: Typing :: Typed
+Requires-Python: <4.0,>=3.10
+Description-Content-Type: text/markdown
+
+# devora-python
+
+Recording, masking and capture policy are configured in the Devora dashboard and
+authorized server-side for each session. This backend SDK takes no capture
+settings, and `devora_sdk()` ignores keyword arguments it does not recognize,
+so check option names carefully. See
+[capture settings](https://github.com/getdevora/devora-sdks/blob/main/SETTINGS.md).
+
+Python backend core SDK for Devora customer integrations.
+
+## Requirements
+
+- Python `>=3.10,<4.0`
+- In production, a replay store shared by every worker (the example below uses
+  Redis 6.2 or newer, for `SET ... PXAT`)
+
+## Install
+
+```bash
+pip install devora-python redis
+```
+
+## Quick Start
+
+```python
+import os
+from dataclasses import asdict
+
+import redis
+from devora_sdk import DEVORA_ENDPOINTS, devora_sdk
+
+# Replay protection shared by every worker and instance (required in production).
+redis_client = redis.Redis.from_url(os.environ["REDIS_URL"])
+
+
+class RedisReplayStore:
+    def consume(self, namespace: str, request_id: str, expires_at: int) -> bool:
+        # Atomic insert-if-absent kept until expires_at; an exception makes the SDK fail closed (503).
+        key = f"devora:replay:{namespace}:{request_id}"
+        return bool(redis_client.set(key, b"1", nx=True, pxat=expires_at))
+
+
+sdk = devora_sdk(
+    api_key=os.environ["DEVORA_API_KEY"],  # pk_server_live_...
+    secret_key=os.environ["DEVORA_SECRET_KEY"],  # sk_server_live_...
+    org_id=os.environ["DEVORA_ORG_ID"],
+    replay_store=RedisReplayStore(),
+)
+
+
+@sdk.register(DEVORA_ENDPOINTS.USER_SEARCH)
+def search_users(req):
+    return {"users": search_customer_users(req.query.get("term", ""))}
+
+
+@sdk.register(DEVORA_ENDPOINTS.IMPERSONATE)
+def impersonate(req):
+    ctx = req.devora_context
+    if not ctx:
+        raise ValueError("Missing impersonation context")
+    # Store the whole verified context in the token: the scope guard reads every
+    # field of it back. Expire the token no later than ctx.expires_at.
+    devora = {**asdict(ctx), "is_impersonation": True}
+    token = create_customer_token(ctx.target_user["id"], devora)
+    return {"token": token}
+```
+
+Register every handler, then mount the SDK with the
+[Django](https://pypi.org/project/devora-django/) or
+[FastAPI](https://pypi.org/project/devora-fastapi/) adapter. For any other
+framework, pass the request exactly as received to `process_request` (sync) or
+`await async_process_request` (async):
+
+```python
+from devora_sdk import AdapterRequest, process_request
+from devora_sdk.utils import get_error_status_code
+
+
+def handle_devora(method: str, path: str, query: str, body: bytes, headers) -> tuple[int, dict]:
+    # path: still percent-encoded and relative to your Devora mount, e.g. "/user/search"
+    # query: the raw query string without "?"; body: the raw bytes (cap the read yourself)
+    # headers: a mapping, or (name, value) pairs so duplicate headers stay visible
+    result = process_request(sdk, sdk.get_routes(), AdapterRequest(method, path, query, body, headers))
+    status = 200 if result["success"] else get_error_status_code(result.get("errorCode"))
+    return status, result  # send as JSON with Cache-Control: private, no-store
+```
+
+`async_process_request` runs signature verification, the replay store and
+synchronous handlers in a worker thread, so it never blocks the event loop.
+`process_request` cannot run `async def` handlers.
+
+### Production replay store
+
+Every signed request carries a single-use id. In production the SDK needs a
+shared, atomic `replay_store` so a captured request cannot be replayed against
+another worker or instance. The environment comes from `environment=`, else
+`DEVORA_ENV`, else `NODE_ENV`. Anything other than `development` or `test`
+(including unset) is production, and production refuses to start without a
+`replay_store`. Any store whose `consume` is an atomic insert-if-absent works;
+see the
+[replay-store contract](https://github.com/getdevora/devora-sdks/blob/main/SIGNING.md#replay-store).
+
+**Local development only:** to run without Redis, omit `replay_store` and set
+`DEVORA_ENV=development` (or pass `environment="development"`). The SDK then
+keeps request ids in memory, which protects a single process only. Never use
+this in production.
+
+`consume` must be a regular method returning `True` or `False`; the SDK calls
+it synchronously, including from `async_process_request` (in a worker thread).
+Use a synchronous client such as `redis.Redis`: an `async def consume` fails
+closed with a 503. Keep it a single short round trip, and do not let the store
+evict these keys before they expire. A unique-key database insert with an
+expiry column works too.
+
+The guard's session-liveness checker caches at most 1,024 results and permits at
+most 64 distinct concurrent lookups per checker. Requests for the same session
+share a lookup. Live/ended results use the configured TTL (five seconds by
+default); unavailable results use at most one second. Cache hits never extend a
+verdict's lifetime. Capacity exhaustion follows the configured unavailable
+policy, which denies access by default.
+
+### Cold-start latency
+
+The impersonation guard's scope policy is fetched lazily on first use, so the
+first request handled by a freshly started worker process pays for that fetch
+synchronously (and any request racing it gets a `503
+IMPERSONATION_POLICY_UNAVAILABLE` rather than waiting). If your deployment
+runs multiple worker processes (gunicorn, uWSGI, etc.), pass
+`prefetch_scope_config=True` to start that fetch in a background thread as soon
+as `devora_sdk(...)` is constructed, so the cache is usually warm by the first
+request:
+
+```python
+sdk = devora_sdk(
+    api_key=os.environ["DEVORA_API_KEY"],
+    secret_key=os.environ["DEVORA_SECRET_KEY"],
+    org_id=os.environ["DEVORA_ORG_ID"],
+    replay_store=RedisReplayStore(),
+    prefetch_scope_config=True,
+)
+```
+
+Options, error codes and the guard are documented in the
+[Python reference](https://docs.devora.sh/reference/python).

{devora_python-0.1.0.dist-info → devora_python-0.1.1.dist-info}/RECORD

@@ -1,7 +1,7 @@
 devora_sdk/__init__.py,sha256=X3_377Yyj4fesPUdFmVRG_4MLQCMqaVyjl1uQkXLlUg,2214
 devora_sdk/api_url_gen.py,sha256=wfYa1I23mbENK9IoDQrw0ZT_6ivkszRI9Jhs48yTcYI,116
 devora_sdk/browser_session.py,sha256=Get6voUbs4dg9ZvqJtBgiMr4gOqtHZh-Gi9xHX6E-tY,2007
-devora_sdk/constants.py,sha256=rj4y8s79eNl1tcWeZljSO16FtKVE3HJFAXW30vLvHvk,1395
+devora_sdk/constants.py,sha256=HBoHYI5dqPL34Eg5pjGCnSw-e8mzjwJ65dLpyprcCDw,1395
 devora_sdk/guard.py,sha256=idlLUrt-HKstvuLJvwy4oCiW0BQ90n_WD3XiYMS6lCU,18284
 devora_sdk/handler.py,sha256=Tk98Z7c26gORnlE1CQoKrI9CnwGXuW56Is5oV0cI4K4,10295
 devora_sdk/hmac.py,sha256=a0OyYEij39j_rWXsgIyKsbgydl0dpVgiIs-BYZR198k,1889
@@ -14,7 +14,7 @@ devora_sdk/security.py,sha256=gfKwz_rF9tGXFejNcwOPBqlczK_g1AZrKZe1SMxdKCA,1437
 devora_sdk/signing.py,sha256=TzB_DEDX6hkN5cZeJVXXp7Q6Z8OKFXLD8GP9OvopZuA,8906
 devora_sdk/transport.py,sha256=plTQ0J4B_-LGUs6waOqBk8wL4P06KuS_LilyEYr9sW4,5528
 devora_sdk/utils.py,sha256=ZC1kmHMB6RVocv1uqpG9srRsQciOO0SzZEKyswY5NLk,5527
-devora_python-0.1.0.dist-info/METADATA,sha256=nr33H63Ulv-mhYHk_iGNsS0ZMm7RAEGA9zQxgucakr4,4547
-devora_python-0.1.0.dist-info/WHEEL,sha256=W3fkpkm7-wf9vBI5Z-7s0eWkeM-spu78I8Neb98DeEg,87
-devora_python-0.1.0.dist-info/licenses/LICENSE,sha256=ZrqlZJezuSZhSOx4R8S5JOwwQ-sLmssj2XW7-3coRjA,1063
-devora_python-0.1.0.dist-info/RECORD,,
+devora_python-0.1.1.dist-info/METADATA,sha256=fvnBQv-yW_0I3UWUcVcFumBw-VHH3o71g6Fv6j37Zos,6879
+devora_python-0.1.1.dist-info/WHEEL,sha256=W3fkpkm7-wf9vBI5Z-7s0eWkeM-spu78I8Neb98DeEg,87
+devora_python-0.1.1.dist-info/licenses/LICENSE,sha256=ZrqlZJezuSZhSOx4R8S5JOwwQ-sLmssj2XW7-3coRjA,1063
+devora_python-0.1.1.dist-info/RECORD,,

devora_sdk/constants.py

@@ -34,7 +34,7 @@ class SECURITY_HEADERS:
 	SESSION_ID = "x-devora-session-id"
 
 
-SDK_VERSION = "0.1.0"
+SDK_VERSION = "0.1.1"
 DEFAULT_API_URL = DEVORA_API_ORIGIN
 DEFAULT_TIMESTAMP_TOLERANCE_SECONDS = 300
 DEFAULT_MAX_BODY_SIZE_BYTES = 1024 * 1024

devora_python-0.1.0.dist-info/METADATA

@@ -1,132 +0,0 @@
-Metadata-Version: 2.5
-Name: devora-python
-Version: 0.1.0
-Summary: Devora Python backend SDK for secure impersonation
-Project-URL: Documentation, https://docs.devora.sh
-Project-URL: Repository, https://github.com/getdevora/devora-sdks
-Project-URL: Issues, https://github.com/getdevora/devora-sdks/issues
-Author: Devora
-License-Expression: MIT
-License-File: LICENSE
-Keywords: backend,devora,impersonation,sdk,security
-Classifier: Development Status :: 5 - Production/Stable
-Classifier: Intended Audience :: Developers
-Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3.11
-Classifier: Programming Language :: Python :: 3.12
-Classifier: Programming Language :: Python :: 3.13
-Classifier: Programming Language :: Python :: 3.14
-Classifier: Typing :: Typed
-Requires-Python: <4.0,>=3.10
-Description-Content-Type: text/markdown
-
-# devora-python
-
-Recording, masking and activity preferences are configured in Devora Settings.
-SDK initialization overrides are ignored. New sessions retain the server's policy
-snapshot across exchange and resume. Developer privacy labels take effect only
-when selected in Settings; sensitive-field protection remains mandatory.
-See [migration details](https://github.com/getdevora/devora-sdks/blob/main/SETTINGS.md).
-
-Python backend core SDK for Devora customer integrations.
-
-## Requirements
-
-- Python `>=3.10,<4.0`
-
-## Install
-
-```bash
-pip install devora-python
-```
-
-## Quick Start
-
-```python
-from devora_sdk import DEVORA_ENDPOINTS, devora_sdk
-
-sdk = devora_sdk(
-    api_key="pk_server_live_...",
-    secret_key="sk_server_live_...",
-    org_id="org_...",
-)
-
-@sdk.register(DEVORA_ENDPOINTS.USER_SEARCH)
-def search_users(req):
-    return {"users": search_customer_users(req.query.get("term", ""))}
-
-@sdk.register(DEVORA_ENDPOINTS.IMPERSONATE)
-def impersonate(req):
-    context = req.devora_context
-    token = create_customer_token(context.target_user["id"], context)
-    return {"token": token}
-```
-
-Use `process_request()` for sync frameworks, `async_process_request()` for
-async frameworks, or use the Django and FastAPI adapter packages.
-
-### Production replay store
-
-Every signed request carries a single-use id. In production the SDK needs a
-shared, atomic `replay_store` so a captured request cannot be replayed against
-another worker or instance; the in-memory store is for
-`environment="development"` or `"test"` only, and any other environment refuses
-to start without one.
-
-```python
-import os
-
-import redis
-from devora_sdk import devora_sdk
-
-client = redis.Redis.from_url(os.environ["REDIS_URL"])
-
-
-class RedisReplayStore:
-    def consume(self, namespace: str, request_id: str, expires_at: int) -> bool:
-        # Atomic insert-if-absent that lives until expires_at (Unix ms).
-        # Exceptions propagate: the SDK then fails closed with a 503.
-        return bool(
-            client.set(f"devora:replay:{namespace}:{request_id}", b"1", nx=True, pxat=expires_at)
-        )
-
-
-sdk = devora_sdk(
-    api_key=os.environ["DEVORA_API_KEY"],
-    secret_key=os.environ["DEVORA_SECRET_KEY"],
-    org_id=os.environ["DEVORA_ORG_ID"],
-    replay_store=RedisReplayStore(),
-)
-```
-
-`consume` is called synchronously, including from `async_process_request`;
-keep it a single short round trip. Do not let the store evict these keys
-before they expire. A unique-key database insert with an expiry column works
-too.
-
-The guard's session-liveness checker caches at most 1,024 results and permits at
-most 64 distinct concurrent lookups per checker. Requests for the same session
-share a lookup. Live/ended results use the configured TTL (five seconds by
-default); unavailable results use at most one second. Cache hits never extend a
-verdict's lifetime. Capacity exhaustion follows the configured unavailable
-policy, which denies access by default.
-
-### Cold-start latency
-
-The impersonation guard's scope policy is fetched lazily on first use, so the
-first request handled by a freshly started worker process pays for that fetch
-synchronously (and any request racing it gets a `503
-IMPERSONATION_POLICY_UNAVAILABLE` rather than waiting). If your deployment
-runs multiple worker processes (gunicorn, uWSGI, etc.), pass
-`prefetch_scope_config=True` to warm the cache in the background as soon as
-`devora_sdk(...)` is constructed, before the process starts serving traffic:
-
-```python
-sdk = devora_sdk(
-    api_key="pk_server_live_...",
-    secret_key="sk_server_live_...",
-    org_id="org_...",
-    prefetch_scope_config=True,
-)
-```