datanotes 3.4.0 → 3.4.1

package/dist/docs/guides/read-and-write/import-notes.md

@@ -314,7 +314,7 @@ Answer:
   "op": "create_table",
   "planned_texts": ["datanotes/metastore.yaml"],
   "commits": [{ "table": "main.public.film", "version": 0, "operation": "CREATE TABLE" }],
-  "migration_id": "20261024T063914-create_table-create-table-main-public-film"
+  "migration_id": "20261024T081156-create_table-create-table-main-public-film"
 }
 ```
 
@@ -359,7 +359,7 @@ Answer (70 lines)
       "path": "Films/Charade.md",
       "set": {
         "table": "main.public.film",
-        "id": "01a16ca2-8b15-7b17-9621-47ebd428dcb2",
+        "id": "01a16cf7-6b19-78cd-bae8-ca1a8fd27c66",
         "director": "Stanley Donen",
         "title": "Charade"
       },
@@ -369,7 +369,7 @@ Answer (70 lines)
       "path": "Films/His Girl Friday.md",
       "set": {
         "table": "main.public.film",
-        "id": "01a16ca2-8b15-736e-b2e9-b44812311845",
+        "id": "01a16cf7-6b19-7117-90e4-2b59555b974a",
         "director": "Howard Hawks",
         "title": "His Girl Friday"
       },
@@ -379,7 +379,7 @@ Answer (70 lines)
       "path": "Films/Metropolis.md",
       "set": {
         "table": "main.public.film",
-        "id": "01a16ca2-8b16-7de3-b3c7-f99fea697aef",
+        "id": "01a16cf7-6b19-73f5-a95f-7200d392529b",
         "director": "Fritz Lang",
         "title": "Metropolis"
       },
@@ -389,7 +389,7 @@ Answer (70 lines)
       "path": "Films/Nosferatu.md",
       "set": {
         "table": "main.public.film",
-        "id": "01a16ca2-8b16-7f5e-a304-e715a1815b66",
+        "id": "01a16cf7-6b19-74d9-a8c2-353ae0902e8c",
         "director": "F. W. Murnau",
         "title": "Nosferatu"
       },
@@ -399,7 +399,7 @@ Answer (70 lines)
       "path": "Films/The Gold Rush.md",
       "set": {
         "table": "main.public.film",
-        "id": "01a16ca2-8b16-747f-9e49-aefd3f27fd58",
+        "id": "01a16cf7-6b19-7a60-b0ac-0a9dcdc412f1",
         "director": "Charlie Chaplin",
         "title": "The Gold Rush"
       },
@@ -458,7 +458,7 @@ Answer (73 lines)
       "path": "Films/Charade.md",
       "set": {
         "table": "main.public.film",
-        "id": "01a16ca2-8b18-7e55-9cbc-a983a43ae96c",
+        "id": "01a16cf7-6b1e-7993-9c87-2ed68d76eba0",
         "director": "Stanley Donen",
         "title": "Charade"
       },
@@ -468,7 +468,7 @@ Answer (73 lines)
       "path": "Films/His Girl Friday.md",
       "set": {
         "table": "main.public.film",
-        "id": "01a16ca2-8b18-739c-9058-e32e9f3a61d4",
+        "id": "01a16cf7-6b1e-79db-8bd7-c55a21c2d29f",
         "director": "Howard Hawks",
         "title": "His Girl Friday"
       },
@@ -478,7 +478,7 @@ Answer (73 lines)
       "path": "Films/Metropolis.md",
       "set": {
         "table": "main.public.film",
-        "id": "01a16ca2-8b18-71f0-92e4-4e2492953fa5",
+        "id": "01a16cf7-6b1e-7e74-9fb2-5f094a71a39f",
         "director": "Fritz Lang",
         "title": "Metropolis"
       },
@@ -488,7 +488,7 @@ Answer (73 lines)
       "path": "Films/Nosferatu.md",
       "set": {
         "table": "main.public.film",
-        "id": "01a16ca2-8b18-78be-9435-87eca4fa962e",
+        "id": "01a16cf7-6b1e-79e9-b36d-92f76a21ce59",
         "director": "F. W. Murnau",
         "title": "Nosferatu"
       },
@@ -498,7 +498,7 @@ Answer (73 lines)
       "path": "Films/The Gold Rush.md",
       "set": {
         "table": "main.public.film",
-        "id": "01a16ca2-8b18-76bd-b55f-ba6aadec67b1",
+        "id": "01a16cf7-6b1e-7d6d-8c78-3104c4071c47",
         "director": "Charlie Chaplin",
         "title": "The Gold Rush"
       },
@@ -519,7 +519,7 @@ Answer (73 lines)
     }
   ],
   "commits": [{ "table": "main.public.film", "version": 1, "operation": "ADOPT ROWS" }],
-  "migration_id": "20261024T063914-adopt_rows-adopt-5-notes-into-film"
+  "migration_id": "20261024T081156-adopt_rows-adopt-5-notes-into-film"
 }
 ```
 
@@ -536,7 +536,7 @@ Answer:
 ```markdown
 ---
 table: main.public.film
-id: 01a16ca2-8b18-7e55-9cbc-a983a43ae96c
+id: 01a16cf7-6b1e-7993-9c87-2ed68d76eba0
 year: 1963
 minutes: 113
 status: watched
@@ -623,7 +623,7 @@ Answer:
   "files_created": 1,
   "planned_creates": ["datanotes/main/public/film/Battleship Potemkin.md"],
   "commits": [{ "table": "main.public.film", "version": 2, "operation": "WRITE" }],
-  "migration_id": "20261024T063914-insert_row-insert-into-film"
+  "migration_id": "20261024T081156-insert_row-insert-into-film"
 }
 ```
 
@@ -655,7 +655,7 @@ Answer:
   "op": "add_column",
   "planned_texts": ["datanotes/metastore.yaml"],
   "commits": [{ "table": "main.public.film", "version": 4, "operation": "ADD COLUMNS" }],
-  "migration_id": "20261024T063915-add_column-add-column-added_by-to-film"
+  "migration_id": "20261024T081158-add_column-add-column-added_by-to-film"
 }
 ```
 
@@ -683,7 +683,7 @@ Answer:
   "files_created": 1,
   "planned_creates": ["datanotes/main/public/film/Night of the Living Dead.md"],
   "commits": [{ "table": "main.public.film", "version": 5, "operation": "WRITE" }],
-  "migration_id": "20261024T063915-insert_row-insert-into-film"
+  "migration_id": "20261024T081158-insert_row-insert-into-film"
 }
 ```
 
@@ -692,7 +692,7 @@ Answer:
 ```markdown
 ---
 table: main.public.film
-id: 01a16ca2-915b-736c-a394-c98699327151
+id: 01a16cf7-71f8-7c36-a392-61386c267a0e
 director: George A. Romero
 minutes: 96
 status: to watch
@@ -734,7 +734,7 @@ Answer:
 ```markdown
 ---
 table: main.public.film
-id: 01a16ca2-943a-7248-9a53-5c7f7532b5af
+id: 01a16cf7-74bb-7a3e-ad0e-2dac8433ce30
 year: 1924
 minutes: 45
 status: to watch
@@ -762,7 +762,7 @@ Answer:
   "op": "alter_table",
   "planned_texts": ["datanotes/metastore.yaml"],
   "commits": [{ "table": "main.public.film", "version": 7, "operation": "ALTER TABLE" }],
-  "migration_id": "20261024T063918-alter_table-alter-table-main-public-film"
+  "migration_id": "20261024T081200-alter_table-alter-table-main-public-film"
 }
 ```
 
@@ -825,7 +825,7 @@ This is the table as you inferred, corrected and extended it on this page. Its d
 
 ```yaml
 # Metastore (Datanotes): catalogs, schemas, tables, columns, views.
-id: 01a16ca2-8b02-76f7-8686-40530eaf4e87
+id: 01a16cf7-6b02-7b71-b2d6-cede5690ed1f
 config:
   history: true
 

package/dist/docs/guides/read-and-write/write-rows.md

@@ -31,7 +31,7 @@ Answer:
   "files_created": 1,
   "planned_creates": ["datanotes/main/public/film/Metropolis.md"],
   "commits": [{ "table": "main.public.film", "version": 1, "operation": "WRITE" }],
-  "migration_id": "20261024T063903-insert_row-insert-into-film"
+  "migration_id": "20261024T081147-insert_row-insert-into-film"
 }
 ```
 
@@ -40,7 +40,7 @@ Answer:
 ```markdown
 ---
 table: main.public.film
-id: 01a16ca2-615c-78e9-af82-5d7ee065ac91
+id: 01a16cf7-4642-7759-8b98-011e31e573c9
 title: Metropolis
 year: 1927
 minutes: 153
@@ -103,7 +103,7 @@ Answer:
     }
   ],
   "commits": [{ "table": "main.public.film", "version": 6, "operation": "UPDATE" }],
-  "migration_id": "20261024T063907-update_row-update-film"
+  "migration_id": "20261024T081151-update_row-update-film"
 }
 ```
 
@@ -174,7 +174,7 @@ Answer:
     }
   ],
   "commits": [{ "table": "main.public.film", "version": 8, "operation": "UPDATE" }],
-  "migration_id": "20261024T063907-update_row-update-film-3",
+  "migration_id": "20261024T081151-update_row-update-film-3",
   "ambiguous_mentions": [
     {
       "path": "datanotes/main/public/film/Nosferatu.md",
@@ -193,7 +193,7 @@ Answer:
 ```markdown
 ---
 table: main.public.film
-id: 01a16ca2-718e-7219-9638-3d26346a9a88
+id: 01a16cf7-5635-75b3-8e95-68aa1e5d8b03
 title: Nosferatu
 year: 1922
 status: watched
@@ -246,7 +246,7 @@ Answer:
     }
   ],
   "commits": [{ "table": "main.public.film", "version": 9, "operation": "UPDATE" }],
-  "migration_id": "20261024T063907-update_row-update-film-4"
+  "migration_id": "20261024T081151-update_row-update-film-4"
 }
 ```
 
@@ -332,7 +332,7 @@ Answer:
     }
   ],
   "commits": [{ "table": "main.public.film", "version": 10, "operation": "UPDATE" }],
-  "migration_id": "20261024T063907-update_rows-update-film"
+  "migration_id": "20261024T081151-update_rows-update-film"
 }
 ```
 
@@ -370,7 +370,7 @@ Answer:
   ],
   "planned_creates": ["datanotes/main/public/film/Battleship Potemkin.md"],
   "commits": [{ "table": "main.public.film", "version": 11, "operation": "MERGE" }],
-  "migration_id": "20261024T063907-merge_rows-merge-2-rows-into-film"
+  "migration_id": "20261024T081151-merge_rows-merge-2-rows-into-film"
 }
 ```
 
@@ -431,7 +431,7 @@ Answer:
     }
   ],
   "commits": [{ "table": "main.public.film", "version": 12, "operation": "DELETE" }],
-  "migration_id": "20261024T063907-delete_rows-archive-1-rows-of-film"
+  "migration_id": "20261024T081151-delete_rows-archive-1-rows-of-film"
 }
 ```
 
@@ -568,6 +568,6 @@ Answer (51 lines)
     { "table": "main.public.film", "version": 22, "operation": "MOVE ROWS" },
     { "table": "main.public.watchlist", "version": 1, "operation": "MOVE ROWS" }
   ],
-  "migration_id": "20261024T063907-move_rows-move-4-rows-from-film-to-watch"
+  "migration_id": "20261024T081151-move_rows-move-4-rows-from-film-to-watch"
 }
 ```

package/dist/docs/guides/search/text.md

@@ -87,7 +87,7 @@ Answer (62 lines)
         {
           "start": 81,
           "end": 86,
-          "context": "-9707-5a03c150e424\nname: Cary Grant\n---\n\n"
+          "context": "-8b5b-1f2bd9e696e3\nname: Cary Grant\n---\n\n"
         }
       ]
     },

package/dist/docs/guides/security/tokens.md

@@ -48,6 +48,20 @@ A pattern of `read` or `write` takes one of these forms:
 
 Some routes stay the owner’s, whatever a named token’s permissions. A named token can’t create or delete tokens and webhooks, list tokens, rotate them or read the engine’s status. A token may list webhooks and read its own entry with `tokens/me/`. With `tokens/update/`, it may change its own `description` and `autolink`, never its permissions.
 
+## A write into another table needs write too
+
+A row write can change rows of tables it doesn’t name. The engine checks those rows against `write` too. These writes change other tables:
+
+* **`delete_rows` with `on_referenced: "set_null"`**: it clears the links to the deleted rows in the rows that link to them.
+* **Inverse links**: a link to a row with an [inverse column](https://datanotes.dev/guides/database-objects/relations/) adds the link back on that row.
+* **`align_relations`**: it writes the missing links on the other side.
+
+When one of those rows belongs to a table outside the token’s `write`, the engine refuses the whole write with 403. Nothing is written, and a dry run gives the same answer. The refusal names the table, such as `token legacy may not write personal.health.weight: this write changes rows of it`. When the token may not read that table, the refusal doesn’t name it, so a token never learns a table name that its listings hide. In a batch, the whole batch is refused with 403 and nothing stays written.
+
+To allow the write, add that table to the token’s `write` with `tokens/update/`. The engine also checks its `read`: a token must read every table it writes.
+
+The owner token and a token with `"write": ["*"]` write every table, so these checks never refuse them. Schema changes with `ddl`, and undo and restore with `maintain`, aren’t row writes and keep their own rules.
+
 ## Limit what a token can read
 
 By default, a named token reads every table. Give it `read` patterns, and it reads only the tables they match. Use them for an agent that works on one part of your notes, so it never sees the rest.
@@ -84,7 +98,7 @@ The engine follows these rules:
 * **`ddl` and `maintain` need every table.** A token with either one can’t have a limited `read`. Cloning a table, adding a constraint, undo and restore read rows of any table. The engine refuses the combination with 400.
 * **A view needs its tables too.** To read a view, a token must read the view’s name and every table the view reads. That includes nested views, joins, unions and the links its filters follow. `list_tables` lists a view only when the engine can tell its tables from the view’s query alone. It hides a view over a union, even when every table of the union is readable.
 * **Free notes stay hidden.** Only a token that reads every table reads a note outside tables, or any other file of the vault. A token with a limited `read` never reads free notes.
-* **A link stays visible as text.** A readable row may link to a row the token can’t read. The link text, such as `[[Pasta]]`, stays in the answer. `expand` doesn’t expand it, so the linked note’s properties stay hidden. `file_backlink_count` stays a plain count.
+* **A link stays visible as text.** A readable row may link to a row the token can’t read. The link text, such as `[[Pasta]]`, stays in the answer. `expand` doesn’t expand it, so the linked note’s properties stay hidden.
 
 ### What the engine filters and what it refuses
 
@@ -107,6 +121,7 @@ A set you did not list is filtered to what the token may read, with no error:
 * **Wildcards**: a pattern in `tables`, or every table when you leave `table` out
 * **Search and the graph**: `search`, `search_notes`, `semantic_search`, and the neighbours and paths of the graph
 * **History and changes**: `list_changes`, `list_migrations`, `describe_history` and the change feed’s stream
+* **Counts**: `file_backlink_count` counts only the links from readable rows, in `columns`, `where` and `order_by`. The files, passages and pending notes of `describe_semantic_index` count only readable notes.
 
 The engine filters before it applies `limit`, so a search still returns up to `limit` results. A token sees its own tool calls on the change feed, never another token’s. `list_webhooks` shows only the webhooks whose tables the token may all read.
 
@@ -146,9 +161,7 @@ Read patterns limit tables, not columns or rows. These gaps remain:
 
 * **No write-only tokens.** A token always reads what it writes.
 * **No column or row limits.** A token reads every column and every row of a readable table.
-* **Writes into other tables.** A write can change rows of other tables: `on_referenced: "set_null"`, inverse links and `align_relations`. The engine doesn’t check those changes against `write` yet. The answer still hides the rows the token can’t read.
-* **Counts.** `describe_semantic_index` and `file_backlink_count` count every note, readable or not. They may tell how many notes exist.
-* **Tool calls.** Any token may record a tool call with `tool-calls/record/`, in the name of any agent.
+* **Tool calls.** A token reports what its agents say about themselves, such as the harness and the tool. The engine doesn’t check it. A token can’t report in another token’s session, as [A session belongs to one token](https://datanotes.dev/guides/security/who-did-what/#a-session-belongs-to-one-token) explains.
 * **Guessed names.** As [Names are hidden, not secret](#names-are-hidden-not-secret) explains, a refusal confirms that a table exists.
 
 ## Turn automatic links off for a sync program

package/dist/docs/guides/security/who-did-what.md

@@ -33,6 +33,14 @@ Tool calls appear on the change feed as events with `kind: "tool_call"`, sent on
 
 A large result costs less to report when the program cuts it first. The engine answers `takes_summary: true` when it accepts a cut call. Then send the value cut at 1,000 characters, and the result’s `summary` next to it. The event is the same as for the whole call. `datanotes mcp` does this after its first report.
 
+### A session belongs to one token
+
+The first token that reports a session owns it. Both routes count as a report. After that, only that token and the owner can report the session. Another named token gets a 403 error, and the engine stores nothing. So no token can rename another agent’s model or add tool calls to its session.
+
+Give each agent session its own id, such as a new UUID. The engine compares the first 16 characters of the id, so make them unique. When one program serves agents with several tokens, give each token its own sessions.
+
+The engine keeps the owners of the last 2,000 sessions in memory, and forgets them when it restarts.
+
 ## Read who did what
 
 Three operations read the record:

package/dist/docs/reference/cli.md

@@ -3,7 +3,7 @@
 > The datanotes command and its options.
 
 ```text
-datanotes 3.4.0: Markdown notes as database tables, for you and your agents.
+datanotes 3.4.1: Markdown notes as database tables, for you and your agents.
 
 
 Usage:

package/dist/docs/release-notes.md

@@ -4,6 +4,21 @@
 
 This page lists each version of Datanotes, newest first. Datanotes follows [semantic versioning](https://semver.org): a major version can break what you rely on, and its notes list those changes first. The package is on [npm](https://www.npmjs.com/package/datanotes), and each version has a [GitHub release](https://github.com/davidefornelli/datanotes/releases).
 
+## 3.4.1
+
+Datanotes 3.4.1 closes three gaps in token permissions. It only adds to 3.4: no operation, argument or setting goes away. One fix can refuse a write that 3.4.0 allowed, so check your tokens before you update.
+
+### Fixes
+
+* **Writes into other tables need `write`**: a row write can change rows of tables it doesn’t name, through `on_referenced: "set_null"`, inverse links or `align_relations`. A named token now needs `write` on those tables too. Otherwise the engine refuses the whole write with 403, and nothing is written. The refusal names the table, unless the token may not read it. [A write into another table needs write too](https://datanotes.dev/guides/security/tokens/#a-write-into-another-table-needs-write-too) explains it.
+* **A session belongs to one token**: the first token that sends a session id to `sessions/identify/` or `tool-calls/record/` owns it. Another named token gets 403 for that session, and nothing is stored. The owner may use any session. [A session belongs to one token](https://datanotes.dev/guides/security/who-did-what/#a-session-belongs-to-one-token) explains it.
+* **Counts follow `read`**: for a token with limited `read`, `file_backlink_count` counts only links from notes it may read. `describe_semantic_index` counts only those notes in `files`, `passages` and `pending`.
+
+### Before you update
+
+* **Token `write`**: an integration that relies on those side effects gets 403 after the update. Add the other table to the token’s `write`, and to its `read` when `read` is limited.
+* **Session ids**: give each token its own session ids. `datanotes mcp` already does: it makes a random id for each session.
+
 ## 3.4.0
 
 Datanotes 3.4.0 brings three additions:

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "datanotes",
-  "version": "3.4.0",
+  "version": "3.4.1",
   "mcpName": "io.github.davidefornelli/datanotes",
   "description": "The engine of your second brain: Markdown notes as database tables, for you and your agents.",
   "keywords": [