sharepoint-knowledge 0.2.1__tar.gz → 0.2.3__tar.gz

{sharepoint_knowledge-0.2.1 → sharepoint_knowledge-0.2.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sharepoint-knowledge
-Version: 0.2.1
+Version: 0.2.3
 Summary: SharePoint connector for the grounded-Q&A engine — site pages + document libraries, Entra device-code sign-in, local models, cited answers
 License: Apache-2.0
 Project-URL: Homepage, https://github.com/nkunal1992/agents
@@ -11,7 +11,7 @@ Requires-Dist: msal>=1.31
 Requires-Dist: httpx>=0.27
 Requires-Dist: mcp>=1.8
 Provides-Extra: models
-Requires-Dist: confluence-knowledge[models]>=0.5.1; extra == "models"
+Requires-Dist: confluence-knowledge[models]>=0.5.3; extra == "models"
 Provides-Extra: dev
 Requires-Dist: pytest>=8.0; extra == "dev"
 Requires-Dist: ruff>=0.5; extra == "dev"

{sharepoint_knowledge-0.2.1 → sharepoint_knowledge-0.2.3}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sharepoint-knowledge"
-version = "0.2.1"
+version = "0.2.3"
 description = "SharePoint connector for the grounded-Q&A engine — site pages + document libraries, Entra device-code sign-in, local models, cited answers"
 readme = "README.md"
 requires-python = ">=3.11"
@@ -18,7 +18,7 @@ dependencies = [
 
 [project.optional-dependencies]
 models = [
-    "confluence-knowledge[models]>=0.5.1",
+    "confluence-knowledge[models]>=0.5.3",
 ]
 dev = [
     "pytest>=8.0",

{sharepoint_knowledge-0.2.1 → sharepoint_knowledge-0.2.3}/src/sharepoint_knowledge/cli_sp.py

@@ -153,6 +153,9 @@ def main() -> None:
         import subprocess
         cmd = [sys.executable, "-m", "pip", "install", "-U",
                "sharepoint-knowledge[models]"]
+        # a running MCP server holds open files — quit the IDE first
+        print("note: quit the IDE (or stop the MCP server) before updating —"
+              " a running server can't be replaced cleanly", flush=True)
         if os.name == "nt":
             print("on Windows, upgrade via the venv python directly:\n  "
                   + " ".join(f'"{c}"' if " " in c else c for c in cmd))

{sharepoint_knowledge-0.2.1 → sharepoint_knowledge-0.2.3}/src/sharepoint_knowledge/crawl_sp.py

@@ -23,7 +23,7 @@ from confluence_knowledge.settings import SetupError
 MAX_FILE_BYTES = 25 * 1024 * 1024
 MAX_FILES = 20_000
 
-from confluence_knowledge.retrieval import EMBED_MODEL_ID as _MODEL_ID
+from confluence_knowledge.retrieval import MODEL_MARKER as _MODEL_ID
 
 
 def resolve_root(client: ConfluenceClient, url: str) -> tuple[str, str]:
@@ -197,6 +197,21 @@ class SPCrawler:
 
         queue: list[str] = job["queue"]
         errors: list[str] = job["errors"]
+        # deny/gone errors drop the item from the snapshot; transient
+        # failures keep the previously-indexed content
+        keep_failed: set[str] = set()
+        # Mid-crawl model change: already-embedded pages live in the OLD
+        # vector space — finishing would publish mixed dims. Wipe + fail;
+        # start_crawl rebuilds under the new model.
+        stamped = self.index.model_id()
+        if stamped is not None and stamped != _MODEL_ID and queue:
+            self.index.clear_scope()
+            errors = errors + [
+                (f"model changed mid-crawl ({stamped} → {_MODEL_ID}) — "
+                 "index cleared; run start_crawl")]
+            self.index.save_job(job_id, "error", [], errors)
+            return {"job_id": job_id, "state": "error",
+                    "remaining": 0, "errors": errors}
         seen_raw = self.index.meta_get(f"seen:{job_id}")
         seen: set[str] = set(json.loads(seen_raw)) if seen_raw else set()
         levels_raw = self.index.meta_get(f"levels:{job_id}")
@@ -252,6 +267,13 @@ class SPCrawler:
                     continue
             except AccessError as e:
                 errors.append(f"{entry}: {e}")
+                if not any(m in str(e) for m in
+                           ("401", "403", "404", "permission", "not found")):
+                    # seen stores the built page_id — page: / file: prefixed
+                    last = entry.split(":")[-1]
+                    keep_failed.add(
+                        f"page:{last}" if entry.startswith("page:")
+                        else f"file:{last}")
                 continue
             pid = page["page_id"]
             if pid in seen:
@@ -273,7 +295,7 @@ class SPCrawler:
             seen.add(pid)
 
         self._checkpoint(job_id, queue, errors, seen, levels, warns)
-        self.index.publish(seen)
+        self.index.publish(seen | keep_failed)
         self.index.set_model_id(_MODEL_ID)
         self.index.save_job(job_id, "done" if not queue else "running",
                             queue, errors)

{sharepoint_knowledge-0.2.1 → sharepoint_knowledge-0.2.3}/src/sharepoint_knowledge/server_sp.py

@@ -10,7 +10,9 @@ import threading
 from confluence_knowledge.client import AccessError, SecurityException
 from confluence_knowledge.index import Index
 from confluence_knowledge.retrieval import (
-    EMBED_MODEL_ID,
+    MODEL_MARKER as EMBED_MODEL_ID,
+)
+from confluence_knowledge.retrieval import (
     LocalEmbedder,
     LocalReranker,
     LocalSparseEmbedder,
@@ -28,7 +30,7 @@ from .settings_sp import APP_NAME, SPProfile, graph_base, home, index_root, work
 mcp = FastMCP("sharepoint-knowledge")
 
 _scopes: dict[str, dict] = {}
-_evidence_sets: dict[str, list[dict]] = {}
+_evidence_sets: dict[str, dict] = {}  # set_id → {scope, ev}
 _indexes: dict[str, Index] = {}
 _clients: dict[str, object] = {}
 _EMBEDDER = None
@@ -309,7 +311,10 @@ def search(scope_handle: str, question: str) -> dict:
                         "— run start_crawl once to rebuild it, then search again"}
     ev = retrieve(idx, question, _embedder(), _reranker(), sparse=_sparse())
     set_id = f"ev:{len(_evidence_sets)}:{s['key'][:8]}"
-    _evidence_sets[set_id] = ev
+    # bound the session cache — sets are never reused after the turn ends
+    while len(_evidence_sets) >= 32:
+        _evidence_sets.pop(next(iter(_evidence_sets)))
+    _evidence_sets[set_id] = {"scope": s["key"], "ev": ev}
     return {
         "evidence_set": set_id,
         "coverage": _coverage(idx),
@@ -317,7 +322,9 @@ def search(scope_handle: str, question: str) -> dict:
         "evidence": [
             {"chunk_id": e["chunk_id"], "page_title": e["page_title"],
              "page_url": e["page_url"], "heading": e["heading"],
-             "score": e.get("rerank_score"), "text": e["text"][:2000]}
+             "score": e.get("rerank_score"), "text": e["text"][:12000],
+             "low_confidence": e.get("low_confidence", False),
+             "truncated": len(e["text"]) > 12000}
             for e in ev
         ],
     }
@@ -343,12 +350,12 @@ def get_evidence(scope_handle: str, chunk_ids: list[str]) -> dict:
 @mcp.tool()
 def verify_citations(scope_handle: str, evidence_set: str,
                      mappings: list[dict]) -> dict:
-    _scope(scope_handle)
-    ev = _evidence_sets.get(evidence_set)
-    if ev is None or not evidence_set.endswith(_scopes[scope_handle]["key"][:8]):
+    s = _scope(scope_handle)
+    rec = _evidence_sets.get(evidence_set)
+    if rec is None or rec["scope"] != s["key"]:
         raise SecurityException(
             "unknown or cross-scope evidence set — run search again and cite its evidence_set")
-    return check_citations(ev, mappings)
+    return check_citations(rec["ev"], mappings)
 
 
 def main() -> None:

{sharepoint_knowledge-0.2.1 → sharepoint_knowledge-0.2.3}/src/sharepoint_knowledge.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sharepoint-knowledge
-Version: 0.2.1
+Version: 0.2.3
 Summary: SharePoint connector for the grounded-Q&A engine — site pages + document libraries, Entra device-code sign-in, local models, cited answers
 License: Apache-2.0
 Project-URL: Homepage, https://github.com/nkunal1992/agents
@@ -11,7 +11,7 @@ Requires-Dist: msal>=1.31
 Requires-Dist: httpx>=0.27
 Requires-Dist: mcp>=1.8
 Provides-Extra: models
-Requires-Dist: confluence-knowledge[models]>=0.5.1; extra == "models"
+Requires-Dist: confluence-knowledge[models]>=0.5.3; extra == "models"
 Provides-Extra: dev
 Requires-Dist: pytest>=8.0; extra == "dev"
 Requires-Dist: ruff>=0.5; extra == "dev"

{sharepoint_knowledge-0.2.1 → sharepoint_knowledge-0.2.3}/src/sharepoint_knowledge.egg-info/requires.txt

@@ -12,4 +12,4 @@ openpyxl>=3.1
 pypdf>=5.0
 
 [models]
-confluence-knowledge[models]>=0.5.1
+confluence-knowledge[models]>=0.5.3