chore(weave): Legacy Refactor pt1

.github/workflows/test.yaml

@@ -186,7 +186,7 @@ jobs:
         # github actions does something funky with the std file descriptors, they end up
         # being closed. tqdm (for example) raises an exception when the descriptor it
         # wants to write to is closed.
-        run: nohup ./weave_server_test.sh < /dev/null &> /tmp/weave/log/stdout.log &
+        run: nohup ./scripts/weave_server_test.sh < /dev/null &> /tmp/weave/log/stdout.log &
         shell: bash
       - name: Cypress run
         # Use the following to run just a single test

DEVELOPMENT.md

@@ -72,7 +72,7 @@ yarn dev
 Weave backend server. This does not currently auto-reload, so you need to restart it if you change stuff.
 
 ```
-sh weave_server.sh
+sh scripts/weave_server.sh
 ```
 
 Now you should be able to go to localhost:3000 to see the weave home page.
@@ -92,7 +92,6 @@ You can tell everything is working if the weave UI renders and you see your serv
 
 Some ops require environment variables to be set, like OPENAI_API_KEY. You need to set these for the server environment and your Jupyter notebook.
 
-
 ## Unit tests
 
 Some of the unit tests try to run a wandb server container, and will produce 403s if that container is out of date. The container is only accessible to wandb developers currently.

integration_test/README.md

@@ -18,4 +18,4 @@ yarn && yarn run cypress open
 
 Start the weave server with 
 
-bash weave_server_test.sh
+bash scripts/weave_server_test.sh

scripts/build_test_container.sh

@@ -0,0 +1,5 @@
+docker buildx build \
+    --platform linux/amd64 \
+    -t us-east4-docker.pkg.dev/weave-support-367421/weave-images/weave-test:latest \
+    -f Dockerfile.test \
+    .

dd_weave_server.sh → scripts/dd_weave_server.sh

@@ -1,13 +1,16 @@
 #!/bin/sh
 
-DD_SERVICE="weave-python" DD_ENV="dev-$(whoami)" DD_LOGS_INJECTION=true \
-	DD_TRACE_PROPAGATION_STYLE_EXTRACT=b3,datadog \
-	DD_TRACE_PROPAGATION_STYLE_INJECT=b3,datadog \
-	WEAVE_DISABLE_ANALYTICS=true \	
-	WEAVE_SERVER_ENABLE_LOGGING=true \
-	FLASK_ENV=development \
-	FLASK_APP=weave.weave_server \
-	ddtrace-run flask run --port 9994
+DD_SERVICE="weave-python"
+DD_ENV="dev-$(whoami)"
+DD_LOGS_INJECTION=true
+DD_TRACE_PROPAGATION_STYLE_EXTRACT=b3,datadog
+DD_TRACE_PROPAGATION_STYLE_INJECT=b3,datadog
+WEAVE_DISABLE_ANALYTICS=true
+WEAVE_SERVER_ENABLE_LOGGING=true
+FLASK_ENV=development
+FLASK_APP=weave.weave_server
+
+ddtrace-run flask run --port 9994
 
 # This runs with the datadog profiler on, but it is expensive!
 # I noticed a 2x increase in query time for some queries

scripts/dd_weave_server_replay.sh

@@ -0,0 +1,10 @@
+#!/bin/sh
+
+WEAVE_SERVER_DEBUG=true
+DD_SERVICE="weave-python"
+DD_ENV="dev-$(whoami)"
+DD_LOGS_INJECTION=true
+WEAVE_SERVER_ENABLE_LOGGING=true
+FLASK_APP=weave.weave_server
+
+ddtrace-run flask run --port 9994

scripts/weave_server.sh

@@ -0,0 +1,12 @@
+#!/bin/sh
+
+# We could set WEAVE_BACKEND_HOST=http://localhost:9994/__weave but the frontend dev server
+# automatically forwards so it shouldn't be necessary.
+source ./config/auth_modes.sh
+
+WEAVE_DISABLE_ANALYTICS=true
+WEAVE_SERVER_ENABLE_LOGGING=true
+FLASK_DEBUG=1
+FLASK_APP=weave.weave_server
+
+flask run --port 9994

scripts/weave_server_replay.sh

@@ -0,0 +1,7 @@
+#!/bin/sh
+
+export WEAVE_SERVER_DEBUG=true
+export WEAVE_SERVER_ENABLE_LOGGING=true
+export FLASK_APP=weave.weave_server
+
+flask run --port 9994

scripts/weave_server_test.sh

@@ -0,0 +1,10 @@
+#!/bin/sh
+
+export WEAVE_CI=true
+export WEAVE_DISABLE_ANALYTICS=true
+export WEAVE_SERVER_DEBUG=true
+export WEAVE_SERVER_ENABLE_LOGGING=true
+export WEAVE_WANDB_GQL_NUM_TIMEOUT_RETRIES=1
+export FLASK_APP=weave.weave_server
+
+flask run --port 9994 "$@"

supervisord.conf

@@ -27,7 +27,7 @@ stopasgroup = true
 killasgroup = true
 
 [program:weave-python-server]
-command = bash ./weave_server_test.sh
+command = bash ./scripts/weave_server_test.sh
 directory = .
 autostart = true
 autorestart = false

weave-js/src/core/README.md

@@ -424,5 +424,5 @@ See [generateDocs.ts](./generateDocs.ts) for implementation.
 
 # Testing Against Weave Python
 
-1. Start a Weave Server using `./weave_server.sh` from the `weave/weave` directory
+1. Start a Weave Server using `./scripts/weave_server.sh` from the `weave/weave` directory
 2. Run `yarn test:python-backend` from this directory

weave/errors.py

@@ -1,10 +1,6 @@
 from typing import Iterable, Optional
 
 
-class WeaveUnmergableArtifactsError(Exception):
-    pass
-
-
 class WeaveFingerprintErrorMixin:
     fingerprint: Optional[Iterable] = None
 
@@ -30,11 +26,11 @@ class WeaveInternalError(WeaveBaseError):
     pass
 
 
-class WeaveConfigurationError(WeaveBaseError):
+class WeaveSerializeError(WeaveBaseError):
     pass
 
 
-class WeaveSerializeError(WeaveBaseError):
+class WeaveConfigurationError(WeaveBaseError):
     pass
 
 

weave/legacy/errors.py

@@ -0,0 +1,32 @@
+from typing import Iterable, Optional
+
+
+class WeaveUnmergableArtifactsError(Exception):
+    pass
+
+
+class WeaveFingerprintErrorMixin:
+    fingerprint: Optional[Iterable] = None
+
+
+class WeaveBaseError(Exception, WeaveFingerprintErrorMixin):
+    pass
+
+
+class WeaveBaseWarning(Warning):
+    pass
+
+# Only use this if you actually want to return an Http 400
+# to the client. This should only happen in cases where the
+# client is wrong.
+class WeaveBadRequest(WeaveBaseError):
+    pass
+
+
+class WeaveInternalError(WeaveBaseError):
+    """Internal Weave Error (a programming error)"""
+
+    pass
+
+class WeaveSerializeError(WeaveBaseError):
+    pass

weave/legacy/ops_primitives/weave_api.py

@@ -7,7 +7,6 @@
     ref_base,
     registry_mem,
     storage,
-    trace_legacy,
     weave_internal,
 )
 from weave import weave_types as types
@@ -20,6 +19,7 @@
     graph,
     object_context,
     runs,
+    trace_legacy,
     uris,
 )
 from weave.legacy.graph import Node

weave/legacy/clear_cache.py → weave/legacy/scripts/clear_cache.py

@@ -3,7 +3,6 @@
 
 from weave.legacy import cache
 
-# TODO: This should be split out into a scripts dir
 # Script to run to delete expired caches
 if __name__ == "__main__":
     print("Starting clear cache job", flush=True)

weave/trace_legacy.py → weave/legacy/trace_legacy.py

@@ -3,8 +3,9 @@
 
 from weave.legacy import artifact_local, graph, runs
 
-from . import errors, ref_base
-from . import weave_types as types
+from .. import ref_base
+from . import errors
+from .. import weave_types as types
 
 
 def get_obj_creator(ref: ref_base.Ref) -> typing.Optional[runs.Run]:

weave/tests/legacy/test_trace.py

@@ -3,7 +3,8 @@
 from weave.legacy import graph
 
 from ... import api as weave
-from ... import storage, trace_legacy
+from ... import storage
+from ...legacy import trace_legacy
 from ...weave_internal import make_const_node