Agent Zero AI framework
Keep numpy scalars out of memory metadata and API JSON
Jehuc26ed8064e87
diff --git a/helpers/api.py b/helpers/api.pyindex 6047f3eb9..653f0804e 100644--- a/helpers/api.py+++ b/helpers/api.py@@ -83,7 +83,7 @@ class ApiHandler:if isinstance(output, Response):return outputelse:- response_json = json.dumps(output)+ response_json = json.dumps(output, default=float)return Response(response=response_json, status=200, mimetype="application/json")diff --git a/plugins/_memory/helpers/memory.py b/plugins/_memory/helpers/memory.pyindex 364bfe76c..e1d8339cc 100644--- a/plugins/_memory/helpers/memory.py+++ b/plugins/_memory/helpers/memory.py@@ -594,7 +594,7 @@ class Memory:res = max(0, min(1, res)) # float precision can cause values like 1.0000000596046448- return res+ return float(res) # native float, not numpy scalar (JSON serializable)@staticmethoddef format_docs_plain(docs: list[Document]) -> list[str]:diff --git a/plugins/_memory/helpers/memory_consolidation.py b/plugins/_memory/helpers/memory_consolidation.pyindex 227fb11e8..c2eb243b9 100644--- a/plugins/_memory/helpers/memory_consolidation.py+++ b/plugins/_memory/helpers/memory_consolidation.py@@ -344,7 +344,7 @@ class MemoryConsolidator:filter=f"area == '{area}'")for doc, score in semantic_results:- doc.metadata['_consolidation_similarity'] = score+ doc.metadata['_consolidation_similarity'] = float(score) if score is not None else 0.0all_similar.append(doc)# Step 3: Keyword-based searches with real scores@@ -358,7 +358,7 @@ class MemoryConsolidator:filter=f"area == '{area}'")for doc, score in keyword_results:- doc.metadata['_consolidation_similarity'] = score+ doc.metadata['_consolidation_similarity'] = float(score) if score is not None else 0.0all_similar.append(doc)# Step 4: Deduplicate by document ID, keep highest score per memory ID