fix last modification timestamp field calculation in xpack (#5566)

* fix last modification timestamp field calculation

* bring path back

GitOrigin-RevId: 5ffa1bd1a2642865898cc4d96a7f9ff13406ba0b

CHANGELOG.md

@@ -5,6 +5,9 @@ All notable changes to this project will be documented in this file.
 This project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 ## [Unreleased]
 
+### Fixed
+- `pw.io.xpacks.llm.VectorStoreServer` now correctly computes the last modification timestamp of an indexed document.
+
 ## [0.8.0] - 2024-02-01
 
 ### Added

python/pathway/xpacks/llm/vector_store.py

@@ -173,8 +173,8 @@ def embedder(txt):
         )
 
         parsed_docs += parsed_docs.select(
-            modified=pw.this.data["metadata"]["modified_at"],
-            path=pw.this.data["metadata"]["path"],
+            modified=pw.this.data["metadata"]["modified_at"].as_int(),
+            path=pw.this.data["metadata"]["path"].as_str(),
         )
 
         stats = parsed_docs.reduce(
@@ -203,7 +203,7 @@ def statistics_query(
         @pw.udf
         def format_stats(counts, last_modified) -> pw.Json:
             if counts is not None:
-                response = {"file_count": counts, "last_modified": last_modified.value}
+                response = {"file_count": counts, "last_modified": last_modified}
             else:
                 response = {"file_count": 0, "last_modified": None}
             return pw.Json(response)
@@ -336,8 +336,9 @@ def run_server(
         port,
         threaded: bool = False,
         with_cache: bool = True,
-        cache_backend: pw.persistence.Backend
-        | None = pw.persistence.Backend.filesystem("./Cache"),
+        cache_backend: (
+            pw.persistence.Backend | None
+        ) = pw.persistence.Backend.filesystem("./Cache"),
     ):
         """
         Builds the document processing pipeline and runs it.