From c12d2cda6c043a3f40e23dcbf770ba054ae16f92 Mon Sep 17 00:00:00 2001
From: Michael Pilosov <consistentbayes@gmail.com>
Date: Wed, 22 Apr 2026 17:04:50 -0600
Subject: [PATCH] flow: hash user-supplied generator_kwargs, not the merged
 dict
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

The flow previously merged _DEFAULT_GENERATOR_KWARGS={random_state:0} and
n_samples=num_points into generator_kwargs BEFORE hashing. Prefect only
records the user-supplied form, so the web app's synth_output_paths
disagreed with the flow's output name — a plain swiss_roll run showed
'embedding: n/a' in the runs list despite completing, because the web
looked for the hash that excluded those defaults.

Now we keep the user-supplied generator_kwargs around for hashing +
metadata, and use the merged dict only for the actual generator call.
n_samples is already captured in the stem as 'N<n>', and random_state=0
is a flow constant — neither belongs in the semantic identity.
---
 flows/embedding_flow.py | 15 +++++++++++----
 1 file changed, 11 insertions(+), 4 deletions(-)

diff --git a/flows/embedding_flow.py b/flows/embedding_flow.py
index 9ebe7a2..3d638ec 100644
--- a/flows/embedding_flow.py
+++ b/flows/embedding_flow.py
@@ -296,9 +296,16 @@ def embedding_flow(
     reference_speedup: float = 10.0,
     samples: int = 10_000,
 ):
+    # Preserve the user-supplied generator_kwargs for hashing / metadata —
+    # the merged dict (with random_state defaults + n_samples) goes to the
+    # generator itself but those aren't part of the run's semantic identity
+    # (random_state=0 is a flow constant; n_samples is captured as `N` in
+    # the stem). If the merged dict were hashed, the web app would disagree
+    # with the flow because Prefect only records the user-supplied form.
+    user_generator_kwargs = dict(generator_kwargs or {})
     generator_kwargs = {
         **_DEFAULT_GENERATOR_KWARGS,
-        **(generator_kwargs or {}),
+        **user_generator_kwargs,
         "n_samples": num_points,
     }
     embed_columns = (
@@ -311,7 +318,7 @@ def embedding_flow(
     output_ref: str = (
         f"{output_dir.strip('/')}/{_generator}_Reference_N{num_points}_T{num_timesteps}_J{jitter_scale}_s{seed}.html"
     )
-    _args_tag = _run_args_hash(embed_args, generator_kwargs)
+    _args_tag = _run_args_hash(embed_args, user_generator_kwargs)
     output_embed: str = (
         f"{output_dir.strip('/')}/{_generator}_{embedder.split('.')[-1]}_N{num_points}_T{num_timesteps}_J{jitter_scale}_s{seed}_{_args_tag}.html"
     )
@@ -405,7 +412,7 @@ def embedding_flow(
             "jitter_scale": jitter_scale,
             "seed": seed,
             "generator_path": generator_path,
-            "generator_kwargs": generator_kwargs or {},
+            "generator_kwargs": user_generator_kwargs,
             "embedder": embedder,
             "embed_args": merged_embed_args,
         },
@@ -428,7 +435,7 @@ def embedding_flow(
         frames = parse_plotly_run(emb_path_result)
         # Persist generator_kwargs so the server's label enrichment can
         # regenerate the correct dataset variant (swiss_roll vs hole).
-        frames.setdefault("meta", {})["generator_kwargs"] = generator_kwargs or {}
+        frames.setdefault("meta", {})["generator_kwargs"] = user_generator_kwargs
         Path(output_frames).write_text(
             json.dumps(frames, separators=(",", ":")), encoding="utf-8"
         )