waveflowdb-client 1.0.2__tar.gz → 1.0.4__tar.gz

{waveflowdb_client-1.0.2/waveflowdb_client.egg-info → waveflowdb_client-1.0.4}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: waveflowdb_client
-Version: 1.0.2
+Version: 1.0.4
 Summary: VectorLake SDK — Deterministic backend engine powering agent workflows
 Author-email: "agentanalytics.ai" <nitin@agentanalytics.ai>
 License: MIT License

{waveflowdb_client-1.0.2 → waveflowdb_client-1.0.4}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "waveflowdb_client"                   # pip install name
-version = "1.0.2"
+version = "1.0.4"
 description = "VectorLake SDK — Deterministic backend engine powering agent workflows"
 readme = "readme.md"
 requires-python = ">=3.8"

{waveflowdb_client-1.0.2 → waveflowdb_client-1.0.4}/waveflowdb_client/client.py

@@ -87,6 +87,24 @@ def _divider(label: str = "", width: int = 60) -> str:
     return f"\n{'─' * width}\n  {label}\n{'─' * width}" if label else f"{'─' * width}"
 
 
+def _source_ext(chunk_name: str, base_path: str) -> str:
+    """
+    Recover the original source file extension from a chunk filename.
+
+    Parses the chunk stem via VersionedChunkName, then looks for the source
+    file in base_path to get its real extension. Falls back to "txt" if the
+    source file is not found (e.g. direct-upload mode).
+    """
+    parsed = VersionedChunkName.parse(chunk_name)
+    if parsed:
+        candidates = list(Path(base_path).glob(f"{parsed.stem}.*"))
+        # Exclude other chunk files in the same dir
+        candidates = [c for c in candidates if not VersionedChunkName.is_versioned(c.name)]
+        if candidates:
+            return candidates[0].suffix.lstrip(".").lower()
+    return Path(chunk_name).suffix.lstrip(".").lower() or "txt"
+
+
 class VectorLakeClient:
     """
     Main entry point for the VectorLake SDK.
@@ -128,7 +146,6 @@ class VectorLakeClient:
         headers    = self._headers()
         request_kb = len(json.dumps(payload).encode()) / 1024
         logger.debug("Batch %d — payload %.1f KB", batch_num, request_kb)
-        print("making request")
         last_error: Optional[APIError] = None
 
         for attempt in range(self.config.max_retries):
@@ -230,7 +247,7 @@ class VectorLakeClient:
         user_id: str,
         vector_lake_description: str,
         start_from_batch: int = 1,
-        end_batch: Optional[int] = None,                      # ← NEW
+        end_batch: Optional[int] = None,
         intelligent_segmentation: bool = True,
         session_id: Optional[str] = None,
         files: Optional[List[str]] = None,
@@ -289,7 +306,7 @@ class VectorLakeClient:
 
         active  = [
             (i, b) for i, b in enumerate(batches)
-            if start_from_batch <= (i + 1) <= resolved_end   # ← respects both bounds
+            if start_from_batch <= (i + 1) <= resolved_end
         ]
         initial = start_from_batch - 1
 
@@ -305,19 +322,22 @@ class VectorLakeClient:
                 t0        = time.time()
 
                 try:
-                    # ── per-batch START marker ─────────────────────────────
                     logger.info(
                         "[BATCH START] op=%s batch=%d/%d files=%s",
                         operation, batch_num, display_total, batch,
                     )
-                    # ──────────────────────────────────────────────────────
 
                     file_contents = []
+                    files_meta    = []
                     for fname in batch:
                         chunk_path = os.path.join(chunks_dir, fname)
                         if not os.path.exists(chunk_path):
                             chunk_path = os.path.join(base_path, fname)
                         file_contents.append(self._file_processor.read_content(chunk_path))
+                        files_meta.append({
+                            "filename":  fname,
+                            "extension": _source_ext(fname, base_path),
+                        })
 
                     payload = {
                         "session_id":               session_id,
@@ -325,9 +345,9 @@ class VectorLakeClient:
                         "vector_lake_description":  vector_lake_description,
                         "files_name":               batch,
                         "files_data":               file_contents,
+                        "files_meta":               files_meta,
                         "intelligent_segmentation": intelligent_segmentation,
                     }
-
                     request_start_ts = time.time()
                     result           = self._make_request(endpoint, payload, operation, batch_num)
                     elapsed_ms       = round((time.time() - request_start_ts) * 1000, 1)
@@ -345,12 +365,10 @@ class VectorLakeClient:
                             batch_num, result.get("message", result.get("error")),
                         )
 
-                    # ── per-batch END marker ───────────────────────────────
                     logger.info(
                         "[BATCH END] op=%s batch=%d/%d success=%s elapsed_ms=%.1f files=%s",
                         operation, batch_num, display_total, succeeded, elapsed_ms, batch,
                     )
-                    # ──────────────────────────────────────────────────────
 
                     batch_outputs.append({
                         "batch_number":    batch_num,
@@ -440,15 +458,13 @@ class VectorLakeClient:
         chunks_dir: str,
     ) -> SyncPlan:
         """
-        Classify each source file by comparing its MD5 against the MD5 of
-        its existing chunk(s) in chunks_dir.
+        Classify each source file by checking whether chunks already exist on disk.
 
         Rules
         ─────
-        No chunk exists on disk        → NEW       → add_docs
-        Chunk exists, MD5 matches      → UNCHANGED → skip
-        Chunk exists, MD5 differs      → CHANGED   → refresh_docs
-        Source file missing from disk  → UNKNOWN   → add_docs (safe default)
+        No chunk exists on disk   → NEW     → add_docs
+        Chunk(s) exist on disk    → CHANGED → refresh_docs
+        Source file missing       → UNKNOWN → add_docs (safe default)
         """
         to_add, to_refresh, to_skip = [], [], []
         classifications = []
@@ -460,50 +476,35 @@ class VectorLakeClient:
                 bar.set_postfix_str(fname[:40])
 
                 if not src.exists():
-                    c = FileClassification(
+                    classifications.append(FileClassification(
                         filename=fname,
                         status=FileStatus.UNKNOWN,
                         endpoint="add_docs",
                         reason="Source file not on disk — defaulting to add_docs",
-                    )
+                    ))
                     to_add.append(fname)
-                    classifications.append(c)
                     continue
 
-                stem            = src.stem
-                existing_chunks = sorted(Path(chunks_dir).glob(f"{stem}_part*.txt"))
+                existing_chunks = sorted(
+                    Path(chunks_dir).glob(f"{src.stem}_part*.txt")
+                )
 
                 if not existing_chunks:
-                    c = FileClassification(
+                    classifications.append(FileClassification(
                         filename=fname,
                         status=FileStatus.NEW,
                         endpoint="add_docs",
                         reason="No chunks on disk — first upload",
-                    )
+                    ))
                     to_add.append(fname)
                 else:
-                    src_md5       = _md5(str(src))
-                    chunk_content = b"".join(Path(cp).read_bytes() for cp in existing_chunks)
-                    chunk_md5     = hashlib.md5(chunk_content).hexdigest()
-
-                    if src_md5 == chunk_md5:
-                        c = FileClassification(
-                            filename=fname,
-                            status=FileStatus.UNCHANGED,
-                            endpoint="skip",
-                            reason="Content matches existing chunks — skipping",
-                        )
-                        to_skip.append(fname)
-                    else:
-                        c = FileClassification(
-                            filename=fname,
-                            status=FileStatus.CHANGED,
-                            endpoint="refresh_docs",
-                            reason="Content differs from existing chunks — refresh required",
-                        )
-                        to_refresh.append(fname)
-
-                classifications.append(c)
+                    classifications.append(FileClassification(
+                        filename=fname,
+                        status=FileStatus.CHANGED,
+                        endpoint="refresh_docs",
+                        reason=f"Chunks exist on disk ({len(existing_chunks)}) — routing to refresh",
+                    ))
+                    to_refresh.append(fname)
 
         print(f"\n   NEW={len(to_add)}  CHANGED={len(to_refresh)}  UNCHANGED={len(to_skip)}")
         print(_divider())
@@ -522,7 +523,7 @@ class VectorLakeClient:
         user_id: str,
         vector_lake_description: str,
         start_from_batch: int = 1,
-        end_batch: Optional[int] = None,                      
+        end_batch: Optional[int] = None,
         intelligent_segmentation: bool = True,
         session_id: Optional[str] = None,
         files: Optional[List[str]] = None,
@@ -556,7 +557,7 @@ class VectorLakeClient:
             grand_total = self._compute_grand_total(self.config.vector_lake_path)
             return self._process_files_in_batches(
                 "add_docs", user_id, vector_lake_description,
-                start_from_batch, end_batch,                  # ← NEW
+                start_from_batch, end_batch,
                 intelligent_segmentation, session_id, files,
                 grand_total_batches=grand_total,
             )
@@ -570,7 +571,7 @@ class VectorLakeClient:
         user_id: str,
         vector_lake_description: str,
         start_from_batch: int = 1,
-        end_batch: Optional[int] = None,                      # ← NEW
+        end_batch: Optional[int] = None,
         intelligent_segmentation: bool = True,
         session_id: Optional[str] = None,
         files: Optional[List[str]] = None,
@@ -604,7 +605,7 @@ class VectorLakeClient:
             grand_total = self._compute_grand_total(self.config.vector_lake_path)
             return self._process_files_in_batches(
                 "refresh_docs", user_id, vector_lake_description,
-                start_from_batch, end_batch,                  # ← NEW
+                start_from_batch, end_batch,
                 intelligent_segmentation, session_id, files,
                 grand_total_batches=grand_total,
             )
@@ -618,7 +619,7 @@ class VectorLakeClient:
         user_id: str,
         vector_lake_description: str,
         start_from_batch: int = 1,
-        end_batch: Optional[int] = None,                      # ← NEW
+        end_batch: Optional[int] = None,
         intelligent_segmentation: bool = True,
         session_id: Optional[str] = None,
         files: Optional[List[str]] = None,
@@ -629,9 +630,8 @@ class VectorLakeClient:
 
         Classification (based on chunks already on disk)
         ────────────────────────────────────────────────
-        No chunk on disk     → NEW       → add_docs
-        Chunk exists, same   → UNCHANGED → skipped (no server call)
-        Chunk exists, differs→ CHANGED   → refresh_docs
+        No chunk on disk     → NEW     → add_docs
+        Chunk(s) exist       → CHANGED → refresh_docs
 
         dry_run=True returns the classification plan without uploading anything.
         """
@@ -684,7 +684,7 @@ class VectorLakeClient:
             if plan.to_add:
                 add_result = self._process_files_in_batches(
                     "add_docs", user_id, vector_lake_description,
-                    start_from_batch, end_batch,              # ← NEW
+                    start_from_batch, end_batch,
                     intelligent_segmentation, session_id,
                     files=plan.to_add,
                     grand_total_batches=grand_total,
@@ -693,7 +693,7 @@ class VectorLakeClient:
             if plan.to_refresh:
                 refresh_result = self._process_files_in_batches(
                     "refresh_docs", user_id, vector_lake_description,
-                    start_from_batch, end_batch,              # ← NEW
+                    start_from_batch, end_batch,
                     intelligent_segmentation, session_id,
                     files=plan.to_refresh,
                     grand_total_batches=grand_total,
@@ -755,12 +755,19 @@ class VectorLakeClient:
                 "files_name and files_data must have the same length"
             ).to_response()
 
+        basenames  = [os.path.basename(n) for n in files_name]
+        files_meta = [
+            {"filename": b, "extension": Path(b).suffix.lstrip(".").lower()}
+            for b in basenames
+        ]
+
         payload = {
             "session_id":               session_id,
             "user_id":                  user_id,
             "vector_lake_description":  vector_lake_description,
-            "files_name":               [os.path.basename(n) for n in files_name],
+            "files_name":               basenames,
             "files_data":               files_data,
+            "files_meta":               files_meta,
             "intelligent_segmentation": intelligent_segmentation,
         }
         return self._make_request(

{waveflowdb_client-1.0.2 → waveflowdb_client-1.0.4}/waveflowdb_client/models.py

@@ -68,9 +68,9 @@ class FileStatus(Enum):
     Classified intent for a single source file before upload.
 
     NEW       → never uploaded; use add_docs
-    CHANGED   → uploaded before but content differs; use refresh_docs
-    UNCHANGED → content identical to last upload; skip
-    UNKNOWN   → no chunks on disk to compare against; caller decides
+    CHANGED   → uploaded before, chunks exist on disk; use refresh_docs
+    UNCHANGED → reserved, not currently used
+    UNKNOWN   → source file missing from disk; caller decides
     """
     NEW       = "new"
     CHANGED   = "changed"

{waveflowdb_client-1.0.2 → waveflowdb_client-1.0.4}/waveflowdb_client/utils.py

@@ -260,8 +260,18 @@ class FileProcessor:
             raise UnsupportedFileTypeError(filename, self.allowed_extensions)
 
     def read_content(self, filepath: str) -> str:
-        """Read and return full text content of any supported file."""
+        """
+        Read and return full text content of any supported file.
+
+        Chunks written by BatchManager are always plain .txt files regardless
+        of the original source format. Running a format-specific extractor
+        (e.g. _extract_pdf) on an already-extracted .txt chunk would either
+        fail or produce garbage, so versioned chunks are always read as plain
+        text directly.
+        """
         self.assert_supported(filepath)
+        if VersionedChunkName.is_versioned(filepath):
+            return _read_plain(filepath)
         return extract_text(filepath)
 
 
@@ -274,7 +284,7 @@ class BatchManager:
     Converts a list of source files into upload-ready batches.
 
     Chunk naming follows the SDK convention:
-        {stem}__v0001_part{part:04d}{ext}
+        {stem}_part{part}.txt
     Chunks already on disk are reused — safe to re-run after a crash.
     """
 

{waveflowdb_client-1.0.2 → waveflowdb_client-1.0.4/waveflowdb_client.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: waveflowdb_client
-Version: 1.0.2
+Version: 1.0.4
 Summary: VectorLake SDK — Deterministic backend engine powering agent workflows
 Author-email: "agentanalytics.ai" <nitin@agentanalytics.ai>
 License: MIT License