Skip to content

~95% on SimpleQA (e.g. Qwen3.6-27B on a 3090). Supports all local and cloud LLMs (llama.cpp, Ollama, Google, ...). 10+ search engines - arXiv, PubMed, your private documents. Everything Local & Encrypted.

chore: clear changelog fragments for 1.10.6 (#5946)

github-actions[bot]156048ec64e0
diff --git a/changelog.d/+benchmark-run-cross-user-isolation.security.md b/changelog.d/+benchmark-run-cross-user-isolation.security.md
deleted file mode 100644
index c7c85e741b..0000000000
--- a/changelog.d/+benchmark-run-cross-user-isolation.security.md
+++ /dev/null
@@ -1,11 +0,0 @@
-Benchmark runs are now isolated per user in the in-process run registry. That
-registry was keyed by the benchmark run id — a **per-user** autoincrement
-integer, so two different users can each own a run with the same id (both have
-run 1). Because the id was trusted on its own, one signed-in user could cancel
-another user's in-flight benchmark (a denial of service), trigger a
-read/persist of another user's in-memory results, or observe their run state;
-and two users running a benchmark with the same id at the same time would
-**overwrite each other's entry**, crossing their results and — via the run's
-stored password — their credentials. The registry is now keyed by
-`(username, run id)`, so every operation only ever reaches the caller's own run
-and colliding per-user ids can no longer collide.
diff --git a/changelog.d/+citation-warning-polish.misc.md b/changelog.d/+citation-warning-polish.misc.md
deleted file mode 100644
index 2761480f58..0000000000
--- a/changelog.d/+citation-warning-polish.misc.md
+++ /dev/null
@@ -1 +0,0 @@
-The missing-inline-citations warnings added for #4969 are now more precise: the iteration-limit failure sentinel no longer triggers the skip warning (an agent failure is not a citation gap), lenticular-bracket citations (`【N】`) count as markers, the zero-marker warning is suppressed when the citation handler itself failed (its own warning already fires, including new ones for non-dict handler results and for dicts missing both the `content` and `response` keys — each naming the returned type or missing keys and the query), and the log wording states facts without referencing internal issue numbers.
diff --git a/changelog.d/+dedup-contract-replace-on-collision.misc.md b/changelog.d/+dedup-contract-replace-on-collision.misc.md
deleted file mode 100644
index bee9ada781..0000000000
--- a/changelog.d/+dedup-contract-replace-on-collision.misc.md
+++ /dev/null
@@ -1 +0,0 @@
-Clarified the `VectorStore.apply()` contract: `dedup` de-duplicates ids within one call and never means "skip an add whose id is already live" — an already-live id must be replaced, because SQLite recycles AUTOINCREMENT ids on ROLLBACK and keeping the stale vector makes search rehydrate the wrong document's text.
diff --git a/changelog.d/+drop-maintainer-path-punchlist.misc.md b/changelog.d/+drop-maintainer-path-punchlist.misc.md
deleted file mode 100644
index b61125ecea..0000000000
--- a/changelog.d/+drop-maintainer-path-punchlist.misc.md
+++ /dev/null
@@ -1 +0,0 @@
-Removed a maintainer's local filesystem path from the test-review completion report.
diff --git a/changelog.d/+followup-parent-ownership.security.md b/changelog.d/+followup-parent-ownership.security.md
deleted file mode 100644
index 40ea70e825..0000000000
--- a/changelog.d/+followup-parent-ownership.security.md
+++ /dev/null
@@ -1,7 +0,0 @@
-Starting a follow-up research now verifies you own the parent research.
-`POST /api/followup/start` prepared and spawned a follow-up without checking
-that `parent_research_id` belonged to the caller (unlike `/api/followup/prepare`,
-which returns 404). Research ids are per-user, so this could not read another
-user's research data (it lives in a separate encrypted database), but it let a
-user start a research thread referencing a parent that isn't theirs. `start`
-now applies the same 404 ownership check as `prepare`.
diff --git a/changelog.d/+nat64-embedded-ip-policy.security.md b/changelog.d/+nat64-embedded-ip-policy.security.md
deleted file mode 100644
index 3cf9cb541c..0000000000
--- a/changelog.d/+nat64-embedded-ip-policy.security.md
+++ /dev/null
@@ -1 +0,0 @@
-The `LDR_SECURITY_ALLOW_NAT64` operator opt-in (for IPv6-only deployments that reach IPv4 via DNS64/NAT64) no longer widens the private-IP policy, and now decodes NAT64-wrapped addresses at the correct RFC 6052 §2.2 byte positions. A NAT64-wrapped destination is evaluated exactly as a direct connection to its embedded IPv4 would be: a wrapped RFC1918 or loopback address is blocked unless the matching private-IP / localhost allowance is also enabled, the six always-blocked cloud-metadata literals stay blocked unconditionally, and link-local wraps are blocked wherever a direct connection to the embedded address would be — including in the RFC 8215 `/48` local-use form, where the embedded IPv4 lives at different byte positions than the trailing 32 bits (previously a crafted `/48` address could hide a metadata target behind a public-looking suffix). Note that the general fetch path calls `validate_url`, which does not set `block_link_local`, so a NAT64-wrapped link-local address outside those six literals (for example a wrap of Scaleway's `169.254.42.42`) is reachable once `allow_private_ips` is enabled — exactly as a direct connection to it already is. That is a pre-existing limitation of the private-IP opt-in, not something this change widens; the point of this fix is that enabling NAT64 grants nothing beyond it. NAT64-wrapped public addresses — the opt-in's intended use — remain reachable. Closes a gap where enabling NAT64 could reach internal or metadata IPv4 targets through the general fetch path.
diff --git a/changelog.d/+news-feed-non-string-source-url.bugfix.md b/changelog.d/+news-feed-non-string-source-url.bugfix.md
deleted file mode 100644
index 06a9fbec50..0000000000
--- a/changelog.d/+news-feed-non-string-source-url.bugfix.md
+++ /dev/null
@@ -1 +0,0 @@
-The news feed no longer fails with a generic "Database error" when a stored source row's URL is not a string; such rows are now skipped instead of aborting the whole feed.
diff --git a/changelog.d/+redact-notification-service-url.security.md b/changelog.d/+redact-notification-service-url.security.md
deleted file mode 100644
index ac60b19acf..0000000000
--- a/changelog.d/+redact-notification-service-url.security.md
+++ /dev/null
@@ -1,34 +0,0 @@
-The notification service URL is now redacted on the settings API read paths.
-`notifications.service_url` (apprise-style) embeds credentials — e.g.
-`mailto://user:pass@host`, `discord://webhook_id/token`, Slack/ntfy tokens —
-and was returned in plaintext by the settings API (single, bulk, and full GET,
-and the save-all echo), unlike API keys which are masked. It is now treated as
-a secret and returned as `[REDACTED]` on read (an unconfigured/empty URL stays
-readable; whitespace-only counts as unconfigured), so the credential no longer
-reaches the browser at all and cannot be picked up by anything that records a
-settings response — devtools, a saved HAR, or a logging proxy. Clearing the URL
-from the UI now stores the empty value instead of silently retaining the old
-URL, and Test Notification uses the stored URL when the field contains the
-redaction sentinel. The app already masked it in notification call-site logs;
-the settings debug logger now redacts it too.
-
-`notifications.service_url` is the first sensitive setting that is not a
-password input, so it is the first one whose control renders its value into an
-editable field. The settings UI now renders any redacted control blank with a
-"Saved — type to replace, or press Enter while empty to clear" placeholder,
-the way password inputs already do, and the write routes reject a submitted
-value that merely *contains* `[REDACTED]` with a 400 instead of storing it.
-Without both, a stale tab produced `[REDACTED],discord://webhook/tok` — not an
-exact sentinel match, so the existing no-op did not fire, and the saved list
-made `NotificationManager` refuse every URL in it, silently stopping all
-notifications. The reverse order, `discord://webhook/tok,[REDACTED]`, passed
-URL validation and appended the sentinel to the webhook token. An exact
-sentinel keeps its existing idempotent no-op behaviour on update; on the
-create/recreate path, where there is no stored value to preserve, it is
-rejected too.
-
-Because Test Notification can now fall back to the stored URL, it accepts an
-empty body and becomes a zero-argument send trigger, so that path gets its own
-rate-limit bucket; testing a URL supplied in the request stays unlimited. The
-endpoint also treats a whitespace-only URL as unconfigured, matching the
-notification manager, instead of handing Apprise literal whitespace.
diff --git a/changelog.d/+release-gate-cpython-cve-batch.misc.md b/changelog.d/+release-gate-cpython-cve-batch.misc.md
deleted file mode 100644
index b6e4021a2c..0000000000
--- a/changelog.d/+release-gate-cpython-cve-batch.misc.md
+++ /dev/null
@@ -1 +0,0 @@
-Unblock the release gate: pin the rebuilt `python:3.14.7-slim` base-image digest (2026-08-25 rebuild) and add Grype suppressions for the 2026-08-18 CPython security batch (CVE-2026-19672 tarfile, CVE-2026-15806 urllib HTTPPasswordMgr, CVE-2026-17084 stringprep). The fixes ship first in Python 3.14.8, which has no release or Docker image yet, and none of the three stdlib APIs is reachable in `src/`.
diff --git a/changelog.d/+relevance-filter-sanitize-debug.security.md b/changelog.d/+relevance-filter-sanitize-debug.security.md
deleted file mode 100644
index 034a2b0272..0000000000
--- a/changelog.d/+relevance-filter-sanitize-debug.security.md
+++ /dev/null
@@ -1 +0,0 @@
-Sanitize credential-shaped content in relevance-filter parser diagnostics before persisting or streaming the response preview.
diff --git a/changelog.d/+research-terminate-cross-user.security.md b/changelog.d/+research-terminate-cross-user.security.md
deleted file mode 100644
index 7ac9c26c06..0000000000
--- a/changelog.d/+research-terminate-cross-user.security.md
+++ /dev/null
@@ -1,6 +0,0 @@
-Terminating a research run now verifies you own it. `cancel_research` (reached
-from `POST /research/api/terminate/<id>`) drove the process-global termination
-registry — keyed by research id alone and shared across users — before any
-ownership check, so a signed-in user who knew another user's research id could
-terminate that user's in-progress research. Ownership is now confirmed against
-the caller's own database before any global termination state is touched.
diff --git a/changelog.d/+verified-mkdir.misc.md b/changelog.d/+verified-mkdir.misc.md
deleted file mode 100644
index d45c423589..0000000000
--- a/changelog.d/+verified-mkdir.misc.md
+++ /dev/null
@@ -1 +0,0 @@
-All production directory creation now routes through `security.create_directory` (audited chokepoint with opt-in path containment), enforced by a pre-commit/CI check that forbids raw `mkdir`/`os.makedirs`.
diff --git a/changelog.d/5381.breaking.md b/changelog.d/5381.breaking.md
deleted file mode 100644
index a33d9fb6fa..0000000000
--- a/changelog.d/5381.breaking.md
+++ /dev/null
@@ -1 +0,0 @@
-**`SearchResultsCollector.add_results()` now returns `tuple[int, list[dict]]` instead of `int`.** Callers that did `start = collector.add_results(results)` must migrate to `start, indexed = collector.add_results(results)` and pass `indexed` to `_format_results()` so deduped citation indices (`[N]`) stay in sync with the collector. The old `int`-only return is no longer available; this is a breaking change for external programmatic users of `SearchResultsCollector`. Internal call sites (`web_search`, `specialized_search`, `fetch_content`) have been updated.
diff --git a/changelog.d/5381.bugfix.md b/changelog.d/5381.bugfix.md
deleted file mode 100644
index 8a1ec2882e..0000000000
--- a/changelog.d/5381.bugfix.md
+++ /dev/null
@@ -1,3 +0,0 @@
-Fixed search result collection in `LangGraphAgentStrategy` to deduplicate URLs so repeated search hits across queries reuse their existing citation index instead of assigning duplicate numbers. Also fixes RAG chunk URI fragments for library search hits to use correct metadata keys (`source_id`/`document_id`) and adds chunk-targeted HTML anchors (`id="chunk-{{ chunk.index }}"`) so citations scroll to the cited chunk.
-
-Further hardened the chunk-anchor contract: chunk-URL fragments are now validated centrally in `src/local_deep_research/utilities/chunk_anchor.py` and reused by `CollectionSearchEngine`, `LibraryRAGSearchEngine`, and `SearchResultsCollector.add_results`. A malformed `chunk_index`/`chunk_id` (UUID, boolean, negative, etc.) cannot leak into a citation; a missing/rejected `doc_id` no longer appends `#chunk-...` to an unrelated route; and any producer-supplied unvalidated `#chunk-...` fragment is stripped rather than trusted. RAG producers also sanitise `doc_id`/`source_id` (including the primary `r.source_id` path) before URL construction. `SearchResultsCollector` now allocates citation indices past the highest seeded index so sparse seeds are not overwritten; the `_format_results` fallback warning routes the URL through `redact_url_for_log` so credentials never appear raw in logs; and the constructor skips legacy non-dict seeded entries.
diff --git a/changelog.d/5485.security.md b/changelog.d/5485.security.md
deleted file mode 100644
index b45168b399..0000000000
--- a/changelog.d/5485.security.md
+++ /dev/null
@@ -1 +0,0 @@
-`/library/document/...` URL resolution now requires the path segment to be a canonical document id (32/64-hex or UUID), rejecting guessable filenames at the regex instead of relying solely on a downstream database miss.
diff --git a/changelog.d/5495.bugfix.md b/changelog.d/5495.bugfix.md
deleted file mode 100644
index bddcdc3681..0000000000
--- a/changelog.d/5495.bugfix.md
+++ /dev/null
@@ -1,10 +0,0 @@
-Collection uploads now correctly report each file's outcome when a single
-request batch contains two files with identical bytes: dropped intra-batch
-duplicates are listed under a new "Skipped — duplicate of another file
-in this upload" bucket using the uploaded filename, instead of being
-mis-reported under the first occurrence's filename as "Skipped — already
-in collection". In addition, uploads matching existing library documents
-now report the newly uploaded filename rather than the stored document's
-canonical filename. Multi-file uploads also isolate failures per file
-via savepoints instead of rolling back the whole request, ensuring that
-a subsequent failing file no longer discards earlier successful uploads.
diff --git a/changelog.d/5517.bugfix.md b/changelog.d/5517.bugfix.md
deleted file mode 100644
index a249873d82..0000000000
--- a/changelog.d/5517.bugfix.md
+++ /dev/null
@@ -1 +0,0 @@
-Skip SearXNG rate-limit locks that are in use when the URL tracker evicts at capacity, so a held lock is not replaced by a fresh one that ignores the configured delay.
diff --git a/changelog.d/5520.security.md b/changelog.d/5520.security.md
deleted file mode 100644
index 6418bbafac..0000000000
--- a/changelog.d/5520.security.md
+++ /dev/null
@@ -1,9 +0,0 @@
-Benchmark progress is no longer delivered to the wrong user's browser. Socket.IO
-subscriptions were held in a process-global map keyed by the research id alone,
-but `BenchmarkRun.id` is a **per-user** autoincrement integer — so two users who
-each own a benchmark run with the same id (both have run 1) were filed under the
-same subscriber set, and each user's browser received the other's
-`research_progress_<id>` events: live queries, partial results, rate-limit
-warnings. Subscriptions are now keyed by `(username, run id)` for numeric
-benchmark ids, matching the same fix applied to the benchmark run registry.
-Ordinary research is unaffected — those ids are UUIDs and cannot collide.
diff --git a/changelog.d/5531.bugfix.md b/changelog.d/5531.bugfix.md
deleted file mode 100644
index bf8d31682c..0000000000
--- a/changelog.d/5531.bugfix.md
+++ /dev/null
@@ -1 +0,0 @@
-Enforce the settings lock (`app.lock_settings`) on the delete, import and reset-to-defaults endpoints, which previously ignored it: deleting or bulk-importing settings while locked now returns 403, and reset-to-defaults can no longer clear the lock as a side effect. Startup paths that load newly shipped defaults are unaffected.
diff --git a/changelog.d/5576.bugfix.md b/changelog.d/5576.bugfix.md
deleted file mode 100644
index 7a2f603abe..0000000000
--- a/changelog.d/5576.bugfix.md
+++ /dev/null
@@ -1 +0,0 @@
-Telegram notifications now use Apprise's canonical `tgram://` scheme; the unregistered `telegram://` alias is rejected. Apprise diagnostic log output that could contain notification credentials is suppressed. Multi-target URLs such as `tgram://...?to=id1,id2` are no longer split into partial notifications, and `telegram://` URLs now get a dedicated migration error pointing at the `tgram://` form.
diff --git a/changelog.d/5589.bugfix.md b/changelog.d/5589.bugfix.md
deleted file mode 100644
index 3dd2dad4c2..0000000000
--- a/changelog.d/5589.bugfix.md
+++ /dev/null
@@ -1 +0,0 @@
-`import_settings()` no longer accepts values that are invalid under the current defaults schema. Both the DB-backed `SettingsManager` and `InMemorySettingsManager` validate every file-supplied value for a known setting key against that key's current-defaults constraints (select options, numeric min/max, checkbox type) and skip the entry — with a warning — when it does not match, so importing a pre-upgrade export can no longer resurrect stale field shapes (e.g. a free-text `default_params.time_period` value from before the field gained its options list). Keys not in the current defaults keep their previous behavior, dynamically populated dropdowns (`llm.provider`, `llm.model`, `search.tool`) stay exempt, values equal to the current default always pass, and values retained from existing storage (the version-bump reconciliation path and environment-locked settings) are deliberately not re-validated. (#5589)
diff --git a/changelog.d/5623.bugfix.md b/changelog.d/5623.bugfix.md
deleted file mode 100644
index f138d39b79..0000000000
--- a/changelog.d/5623.bugfix.md
+++ /dev/null
@@ -1,5 +0,0 @@
-Token metrics now persist the total the provider reported instead of
-recomputing it as prompt + completion, so the metrics dashboard and the
-in-session counter no longer disagree about the same response. A provider
-that reports usage without a total is now counted as prompt + completion
-rather than as zero tokens.
diff --git a/changelog.d/5647.misc.md b/changelog.d/5647.misc.md
deleted file mode 100644
index a293bee311..0000000000
--- a/changelog.d/5647.misc.md
+++ /dev/null
@@ -1 +0,0 @@
-Docs now state concretely that the private engine-URL allowlist requires v1.10.5 or newer, replacing the pre-release "newer than v1.10.4" phrasing, and add the same caveat to the two `docker run` recipes (FAQ, installation guide) that recommended the variable with no version note.
diff --git a/changelog.d/5649.misc.md b/changelog.d/5649.misc.md
deleted file mode 100644
index 25b628edeb..0000000000
--- a/changelog.d/5649.misc.md
+++ /dev/null
@@ -1 +0,0 @@
-Docs and `.env.template` now state clearly that LDR itself never loads a `.env` file — it reads only the environment of the server process — while pointing at the file-based mechanisms that *do* work (`docker run --env-file`, Compose `env_file:`, systemd `EnvironmentFile=`), with a new per-deployment persistence table in the environment-configuration guide and a note on the `crawl4ai` dependency's own `load_dotenv()` call.
diff --git a/changelog.d/5660.bugfix.md b/changelog.d/5660.bugfix.md
deleted file mode 100644
index 87da7e2855..0000000000
--- a/changelog.d/5660.bugfix.md
+++ /dev/null
@@ -1 +0,0 @@
-Clarified the help text and tooltips on the embedding settings page so they match which chunk-size, chunk-overlap, and text-separator settings each text splitter type actually uses. Previously, the descriptions implied every splitter honored every setting — for example, the chunk-size field was described as "characters per chunk" even though the `token` and `sentence` splitters interpret it as tokens, and the `semantic` splitter silently forces chunk overlap to 0.
diff --git a/changelog.d/5661.misc.md b/changelog.d/5661.misc.md
deleted file mode 100644
index df79240d82..0000000000
--- a/changelog.d/5661.misc.md
+++ /dev/null
@@ -1 +0,0 @@
-The remediation text shown when a research run fails on a missing API key no longer implies a `.env` file is an option — LDR itself reads only the server process's environment. To keep values in a file, hand the file to whatever starts LDR (`docker run --env-file`, Compose `env_file:`, systemd `EnvironmentFile=`).
diff --git a/changelog.d/5669.bugfix.md b/changelog.d/5669.bugfix.md
deleted file mode 100644
index 156c7c1789..0000000000
--- a/changelog.d/5669.bugfix.md
+++ /dev/null
@@ -1 +0,0 @@
-The SearXNG rate limiter now keeps each URL's lock and its last-request time in one tracked entry, so a URL evicted while one of its requests is still in flight can no longer leave its timestamp behind. Tracked state is now bounded by `MAX_TRACKED_URLS` as the module documented, rather than only the locks being bounded.
diff --git a/changelog.d/5685.bugfix.md b/changelog.d/5685.bugfix.md
deleted file mode 100644
index 1bcf8d14e9..0000000000
--- a/changelog.d/5685.bugfix.md
+++ /dev/null
@@ -1,29 +0,0 @@
-Fixed three downstream effects of the chunk-anchor citation URLs introduced in #5381.
-
-`fetch_content` can read library documents again: the RAG engines now cite `/library/document/<id>/chunks#chunk-<n>`, but the fetch tool's library resolver only matched an optional `/pdf` suffix and never stripped the fragment, so every such citation fell through to the egress gate and was rejected as `unsupported_scheme`. The resolver now accepts the `chunks` route and drops the fragment before matching.
-
-`SearchResultsCollector.sources` no longer loses deduplicated URLs after `reset()`. `reset()` clears `_sources` per report subsection while the dedup maps deliberately persist, so a subsection that re-cited an earlier section's hits reported no sources at all — an empty `sources` list from the MCP server, and `sources_count: 0` in news cards.
-
-Library citations collapse to one bibliography entry per document again. `canonical_url_key` stripped fragments only from absolute URLs, so each cited chunk of a document became a separate `## Sources` line. Library routes now key per document — `/library/document/<id>`, its `/pdf` view and its `/chunks#chunk-<n>` view are one source — which also collapses the two entries a document used to get when some of its chunks carried `chunk_index` metadata and others fell back to `/pdf`. The rendered `URL:` line keeps a `#chunk-<n>` anchor, preferring an anchored view over the anchor-less `/pdf` one regardless of which was registered first, so collapsing the list does not cost the scroll-to-chunk behaviour the anchor exists for. Note the entry carries one URL for the whole group: it anchors at the first cited chunk, so a document cited across several chunks scrolls to the first of them rather than to each citation's own chunk. Per-citation anchors would need a different Sources layout and are not attempted here. Query strings are not part of a library document's identity — `/library/document/<id>?v=2` keys with `/library/document/<id>` — which no current producer emits.
-
-Citations of one library document via the absolute `https://library.document/<id>` alias — the form the agent sometimes emits and the fetch resolver accepts — now share a bibliography entry with that document's other citations instead of rendering as a separate, anchor-less external-looking link. Only the exact `<id>[/pdf|/chunks]` shape normalizes: look-alike hosts, deeper paths, queries and traversal are rejected, so a crafted alias cannot merge into a real document's entry and replace its link.
-
-Library/RAG citations no longer point at the wrong document when a chunk's denormalised metadata disagrees with its `DocumentChunk` foreign key. The authoritative document id is now carried at the top level of each search result and preferred over metadata wherever the citation URL is built or rebuilt.
-
-`format_links_to_markdown` no longer crashes the whole bibliography when a citation index contains a non-ASCII digit character (`"¹".isdigit()` is true while `int("¹")` raises).
-
-Bibliography exports (BibTeX/Quarto and LaTeX) keep library documents that are cited from several chunks. Collapsing those citations onto one Sources line produced a `[1, 2, 3]` index group that the exporters' single-index patterns skipped, silently dropping the source from the exported bibliography.
-
-`sources_count` now agrees with the rendered bibliography. The bibliography groups sources with `canonical_url_key`, but `SearchResultsCollector` still keyed its dedup map and its `sources` list on the raw link, so the three views of one library document (`/library/document/<id>`, `/<id>/pdf`, `/<id>/chunks#chunk-<n>`, plus the absolute `https://library.document/<id>` alias) produced three citation indices and a `sources_count` of 3 for a report whose `## Sources` block renders one line — the count the MCP server returns and the news impact score reads. The collector now dedups on the same canonical key while recording the display URL, anchor included. This also means a search hit and a later `fetch_content` of the same document share one citation index no matter which spelling of the URL the agent typed back.
-
-A subsection that only re-fetches previously cited URLs renders its `## Sources` block again. `find_or_add_result` recorded the source on all three branches but appended to `_results` only when it allocated a new citation, so after `reset()` such a subsection reported N sources, rendered none, and tripped the #4969 "citation pass skipped" warning.
-
-Library-route parsing is now one code path for the relative route, its `/lib/document/<id>` abbreviation and the absolute alias, so all three enforce the same validation. Previously only the alias validated its document id: `/library/document/7/../../../admin/wipe#chunk-1` keyed as document 7, inherited that document's title and citation numbers, and replaced its link. The `/lib/` spelling and percent-encoded ids (`a%2Db` vs `a-b`) now key with the document they denote instead of fanning it across extra entries, and the displayed route is always the registered `/library/document/<id>` one. The bare `[<uuid>]` alias the agent sometimes pastes into `fetch_content` is re-cited as the resolved document's canonical route instead of verbatim, closing the third spelling that used to get its own (dead-linking) entry. A route carrying a control character is now refused outright rather than having its document-id segment truncated: truncating discarded the payload but keyed the crafted URL as the real `/library/document/<id>` it was prefixed with, merging it into that document's bibliography entry and lending it the document's title and citation number. The payload is still neutralised where it is rendered.
-
-An external URL that merely contains `/library/document/` — `https://evil.example/library/document/7/chunks` — is no longer relabelled as a local library document. The eligibility test is anchored at the start of the URL instead of being a substring match.
-
-Chunk indices and document ids are validated against ASCII character classes with a magnitude bound, so non-ASCII digits (`٧`, `𝟝`), whitespace-padded values, absurd magnitudes (`10**30`, `1e30`) and fullwidth document ids can no longer reach a URL fragment or path. The bibliography's "prefer an anchored URL" rule now looks for `#chunk-` specifically, so an empty or unrelated fragment can no longer lock out the real chunk anchor.
-
-`sources` in the strategy result keeps first-seen order. It was passed through `list(set(...))`, which deduped nothing (the collector already guarantees uniqueness) and randomized the ordering differently on every process.
-
-Note: RIS export does not yet handle merged index lines, and mis-attributes rather than merely skipping them — a record can carry one source's title with another source's URL. It needs a different fix from the other exporters and is tracked in #5687.
diff --git a/changelog.d/5894.bugfix.md b/changelog.d/5894.bugfix.md
deleted file mode 100644
index 7b2662411a..0000000000
--- a/changelog.d/5894.bugfix.md
+++ /dev/null
@@ -1,3 +0,0 @@
-Fixed `SearchResultsCollector` (LangGraph agent strategy) discarding a source's later excerpts. Citations were deduplicated on the canonical URL alone, so a URL reached by several queries kept only the first occurrence's snippet — the agent read three excerpts and the bibliography, `research_resources` and the API `sources` payload retained one. The dedup key is now the pair `(canonical URL, snippet)`: a repeat with the *same* snippet still collapses onto its existing `[N]` (the #5381 behaviour), while a repeat with a genuinely different snippet becomes its own entry with its own `[N]`. Each entry owns exactly one citation index, so `CitationFormatter`'s index → URL map stays 1:1. Snippets are compared after normalising whitespace, case and trailing ellipsis/punctuation, over a `blake2b` digest of the *whole* passage — never a prefix, which cannot tell "one passage truncated twice" from "two passages that open alike". There is deliberately no cap on how many excerpts one source may contribute: folding a distinct excerpt onto an existing citation would show the model text under a number that resolves to different text, and the count is already bounded by the number of queries run.
-
-`all_links_of_system` therefore holds one entry per distinct excerpt rather than one per source, and every "N sources" number now counts distinct canonical URLs via the new `search_utilities.count_distinct_sources` — the agent step heartbeat, `_finalize`'s accumulated-sources milestone and its two citation-skip warnings, and the detailed report's `initial_sources` metadata. `initial_sources` now counts distinct sources rather than raw entries for **every** strategy, not only the LangGraph agent: `report_generator` reads `all_links_of_system` whatever produced it, and `source_based`, `focused_iteration` and `topic_organization` have no cross-call dedup at all, so a run that revisits a URL across `analyze_topic()` calls will report a lower — and now correct — number than before (e.g. 4 raw entries reported as 3 sources). `format_links_to_markdown` still renders one `## Sources` line per source, now listing every citation number that source owns (`[1, 4] Title`), and it groups on the same field the collector keys on (`link` first, via the new `source_url_field`) so a result carrying divergent `url` and `link` cannot split one source across two lines.
diff --git a/docs/release_notes/1.10.6.md b/docs/release_notes/1.10.6.md
new file mode 100644
index 0000000000..8d87cf5a38
--- /dev/null
+++ b/docs/release_notes/1.10.6.md
@@ -0,0 +1,148 @@
+
+### 💥 Breaking Changes
+
+- **`SearchResultsCollector.add_results()` now returns `tuple[int, list[dict]]` instead of `int`.** Callers that did `start = collector.add_results(results)` must migrate to `start, indexed = collector.add_results(results)` and pass `indexed` to `_format_results()` so deduped citation indices (`[N]`) stay in sync with the collector. The old `int`-only return is no longer available; this is a breaking change for external programmatic users of `SearchResultsCollector`. Internal call sites (`web_search`, `specialized_search`, `fetch_content`) have been updated. ([#5381](https://github.com/LearningCircuit/local-deep-research/pull/5381))
+
+### 🔒 Security
+
+- `/library/document/...` URL resolution now requires the path segment to be a canonical document id (32/64-hex or UUID), rejecting guessable filenames at the regex instead of relying solely on a downstream database miss. ([#5485](https://github.com/LearningCircuit/local-deep-research/pull/5485))
+- Benchmark progress is no longer delivered to the wrong user's browser. Socket.IO
+ subscriptions were held in a process-global map keyed by the research id alone,
+ but `BenchmarkRun.id` is a **per-user** autoincrement integer — so two users who
+ each own a benchmark run with the same id (both have run 1) were filed under the
+ same subscriber set, and each user's browser received the other's
+ `research_progress_<id>` events: live queries, partial results, rate-limit
+ warnings. Subscriptions are now keyed by `(username, run id)` for numeric
+ benchmark ids, matching the same fix applied to the benchmark run registry.
+ Ordinary research is unaffected — those ids are UUIDs and cannot collide. ([#5520](https://github.com/LearningCircuit/local-deep-research/pull/5520))
+- Benchmark runs are now isolated per user in the in-process run registry. That
+ registry was keyed by the benchmark run id — a **per-user** autoincrement
+ integer, so two different users can each own a run with the same id (both have
+ run 1). Because the id was trusted on its own, one signed-in user could cancel
+ another user's in-flight benchmark (a denial of service), trigger a
+ read/persist of another user's in-memory results, or observe their run state;
+ and two users running a benchmark with the same id at the same time would
+ **overwrite each other's entry**, crossing their results and — via the run's
+ stored password — their credentials. The registry is now keyed by
+ `(username, run id)`, so every operation only ever reaches the caller's own run
+ and colliding per-user ids can no longer collide.
+- Sanitize credential-shaped content in relevance-filter parser diagnostics before persisting or streaming the response preview.
+- Starting a follow-up research now verifies you own the parent research.
+ `POST /api/followup/start` prepared and spawned a follow-up without checking
+ that `parent_research_id` belonged to the caller (unlike `/api/followup/prepare`,
+ which returns 404). Research ids are per-user, so this could not read another
+ user's research data (it lives in a separate encrypted database), but it let a
+ user start a research thread referencing a parent that isn't theirs. `start`
+ now applies the same 404 ownership check as `prepare`.
+- Terminating a research run now verifies you own it. `cancel_research` (reached
+ from `POST /research/api/terminate/<id>`) drove the process-global termination
+ registry — keyed by research id alone and shared across users — before any
+ ownership check, so a signed-in user who knew another user's research id could
+ terminate that user's in-progress research. Ownership is now confirmed against
+ the caller's own database before any global termination state is touched.
+- The `LDR_SECURITY_ALLOW_NAT64` operator opt-in (for IPv6-only deployments that reach IPv4 via DNS64/NAT64) no longer widens the private-IP policy, and now decodes NAT64-wrapped addresses at the correct RFC 6052 §2.2 byte positions. A NAT64-wrapped destination is evaluated exactly as a direct connection to its embedded IPv4 would be: a wrapped RFC1918 or loopback address is blocked unless the matching private-IP / localhost allowance is also enabled, the six always-blocked cloud-metadata literals stay blocked unconditionally, and link-local wraps are blocked wherever a direct connection to the embedded address would be — including in the RFC 8215 `/48` local-use form, where the embedded IPv4 lives at different byte positions than the trailing 32 bits (previously a crafted `/48` address could hide a metadata target behind a public-looking suffix). Note that the general fetch path calls `validate_url`, which does not set `block_link_local`, so a NAT64-wrapped link-local address outside those six literals (for example a wrap of Scaleway's `169.254.42.42`) is reachable once `allow_private_ips` is enabled — exactly as a direct connection to it already is. That is a pre-existing limitation of the private-IP opt-in, not something this change widens; the point of this fix is that enabling NAT64 grants nothing beyond it. NAT64-wrapped public addresses — the opt-in's intended use — remain reachable. Closes a gap where enabling NAT64 could reach internal or metadata IPv4 targets through the general fetch path.
+- The notification service URL is now redacted on the settings API read paths.
+ `notifications.service_url` (apprise-style) embeds credentials — e.g.
+ `mailto://user:pass@host`, `discord://webhook_id/token`, Slack/ntfy tokens —
+ and was returned in plaintext by the settings API (single, bulk, and full GET,
+ and the save-all echo), unlike API keys which are masked. It is now treated as
+ a secret and returned as `[REDACTED]` on read (an unconfigured/empty URL stays
+ readable; whitespace-only counts as unconfigured), so the credential no longer
+ reaches the browser at all and cannot be picked up by anything that records a
+ settings response — devtools, a saved HAR, or a logging proxy. Clearing the URL
+ from the UI now stores the empty value instead of silently retaining the old
+ URL, and Test Notification uses the stored URL when the field contains the
+ redaction sentinel. The app already masked it in notification call-site logs;
+ the settings debug logger now redacts it too.
+
+ `notifications.service_url` is the first sensitive setting that is not a
+ password input, so it is the first one whose control renders its value into an
+ editable field. The settings UI now renders any redacted control blank with a
+ "Saved — type to replace, or press Enter while empty to clear" placeholder,
+ the way password inputs already do, and the write routes reject a submitted
+ value that merely *contains* `[REDACTED]` with a 400 instead of storing it.
+ Without both, a stale tab produced `[REDACTED],discord://webhook/tok` — not an
+ exact sentinel match, so the existing no-op did not fire, and the saved list
+ made `NotificationManager` refuse every URL in it, silently stopping all
+ notifications. The reverse order, `discord://webhook/tok,[REDACTED]`, passed
+ URL validation and appended the sentinel to the webhook token. An exact
+ sentinel keeps its existing idempotent no-op behaviour on update; on the
+ create/recreate path, where there is no stored value to preserve, it is
+ rejected too.
+
+ Because Test Notification can now fall back to the stored URL, it accepts an
+ empty body and becomes a zero-argument send trigger, so that path gets its own
+ rate-limit bucket; testing a URL supplied in the request stays unlimited. The
+ endpoint also treats a whitespace-only URL as unconfigured, matching the
+ notification manager, instead of handing Apprise literal whitespace.
+
+### 🐛 Bug Fixes
+
+- Fixed search result collection in `LangGraphAgentStrategy` to deduplicate URLs so repeated search hits across queries reuse their existing citation index instead of assigning duplicate numbers. Also fixes RAG chunk URI fragments for library search hits to use correct metadata keys (`source_id`/`document_id`) and adds chunk-targeted HTML anchors (`id="chunk-{{ chunk.index }}"`) so citations scroll to the cited chunk.
+
+ Further hardened the chunk-anchor contract: chunk-URL fragments are now validated centrally in `src/local_deep_research/utilities/chunk_anchor.py` and reused by `CollectionSearchEngine`, `LibraryRAGSearchEngine`, and `SearchResultsCollector.add_results`. A malformed `chunk_index`/`chunk_id` (UUID, boolean, negative, etc.) cannot leak into a citation; a missing/rejected `doc_id` no longer appends `#chunk-...` to an unrelated route; and any producer-supplied unvalidated `#chunk-...` fragment is stripped rather than trusted. RAG producers also sanitise `doc_id`/`source_id` (including the primary `r.source_id` path) before URL construction. `SearchResultsCollector` now allocates citation indices past the highest seeded index so sparse seeds are not overwritten; the `_format_results` fallback warning routes the URL through `redact_url_for_log` so credentials never appear raw in logs; and the constructor skips legacy non-dict seeded entries. ([#5381](https://github.com/LearningCircuit/local-deep-research/pull/5381))
+- Collection uploads now correctly report each file's outcome when a single
+ request batch contains two files with identical bytes: dropped intra-batch
+ duplicates are listed under a new "Skipped — duplicate of another file
+ in this upload" bucket using the uploaded filename, instead of being
+ mis-reported under the first occurrence's filename as "Skipped — already
+ in collection". In addition, uploads matching existing library documents
+ now report the newly uploaded filename rather than the stored document's
+ canonical filename. Multi-file uploads also isolate failures per file
+ via savepoints instead of rolling back the whole request, ensuring that
+ a subsequent failing file no longer discards earlier successful uploads. ([#5495](https://github.com/LearningCircuit/local-deep-research/pull/5495))
+- Skip SearXNG rate-limit locks that are in use when the URL tracker evicts at capacity, so a held lock is not replaced by a fresh one that ignores the configured delay. ([#5517](https://github.com/LearningCircuit/local-deep-research/pull/5517))
+- Enforce the settings lock (`app.lock_settings`) on the delete, import and reset-to-defaults endpoints, which previously ignored it: deleting or bulk-importing settings while locked now returns 403, and reset-to-defaults can no longer clear the lock as a side effect. Startup paths that load newly shipped defaults are unaffected. ([#5531](https://github.com/LearningCircuit/local-deep-research/pull/5531))
+- Telegram notifications now use Apprise's canonical `tgram://` scheme; the unregistered `telegram://` alias is rejected. Apprise diagnostic log output that could contain notification credentials is suppressed. Multi-target URLs such as `tgram://...?to=id1,id2` are no longer split into partial notifications, and `telegram://` URLs now get a dedicated migration error pointing at the `tgram://` form. ([#5576](https://github.com/LearningCircuit/local-deep-research/pull/5576))
+- `import_settings()` no longer accepts values that are invalid under the current defaults schema. Both the DB-backed `SettingsManager` and `InMemorySettingsManager` validate every file-supplied value for a known setting key against that key's current-defaults constraints (select options, numeric min/max, checkbox type) and skip the entry — with a warning — when it does not match, so importing a pre-upgrade export can no longer resurrect stale field shapes (e.g. a free-text `default_params.time_period` value from before the field gained its options list). Keys not in the current defaults keep their previous behavior, dynamically populated dropdowns (`llm.provider`, `llm.model`, `search.tool`) stay exempt, values equal to the current default always pass, and values retained from existing storage (the version-bump reconciliation path and environment-locked settings) are deliberately not re-validated. (#5589) ([#5589](https://github.com/LearningCircuit/local-deep-research/pull/5589))
+- Token metrics now persist the total the provider reported instead of
+ recomputing it as prompt + completion, so the metrics dashboard and the
+ in-session counter no longer disagree about the same response. A provider
+ that reports usage without a total is now counted as prompt + completion
+ rather than as zero tokens. ([#5623](https://github.com/LearningCircuit/local-deep-research/pull/5623))
+- Clarified the help text and tooltips on the embedding settings page so they match which chunk-size, chunk-overlap, and text-separator settings each text splitter type actually uses. Previously, the descriptions implied every splitter honored every setting — for example, the chunk-size field was described as "characters per chunk" even though the `token` and `sentence` splitters interpret it as tokens, and the `semantic` splitter silently forces chunk overlap to 0. ([#5660](https://github.com/LearningCircuit/local-deep-research/pull/5660))
+- The SearXNG rate limiter now keeps each URL's lock and its last-request time in one tracked entry, so a URL evicted while one of its requests is still in flight can no longer leave its timestamp behind. Tracked state is now bounded by `MAX_TRACKED_URLS` as the module documented, rather than only the locks being bounded. ([#5669](https://github.com/LearningCircuit/local-deep-research/pull/5669))
+- Fixed three downstream effects of the chunk-anchor citation URLs introduced in #5381.
+
+ `fetch_content` can read library documents again: the RAG engines now cite `/library/document/<id>/chunks#chunk-<n>`, but the fetch tool's library resolver only matched an optional `/pdf` suffix and never stripped the fragment, so every such citation fell through to the egress gate and was rejected as `unsupported_scheme`. The resolver now accepts the `chunks` route and drops the fragment before matching.
+
+ `SearchResultsCollector.sources` no longer loses deduplicated URLs after `reset()`. `reset()` clears `_sources` per report subsection while the dedup maps deliberately persist, so a subsection that re-cited an earlier section's hits reported no sources at all — an empty `sources` list from the MCP server, and `sources_count: 0` in news cards.
+
+ Library citations collapse to one bibliography entry per document again. `canonical_url_key` stripped fragments only from absolute URLs, so each cited chunk of a document became a separate `## Sources` line. Library routes now key per document — `/library/document/<id>`, its `/pdf` view and its `/chunks#chunk-<n>` view are one source — which also collapses the two entries a document used to get when some of its chunks carried `chunk_index` metadata and others fell back to `/pdf`. The rendered `URL:` line keeps a `#chunk-<n>` anchor, preferring an anchored view over the anchor-less `/pdf` one regardless of which was registered first, so collapsing the list does not cost the scroll-to-chunk behaviour the anchor exists for. Note the entry carries one URL for the whole group: it anchors at the first cited chunk, so a document cited across several chunks scrolls to the first of them rather than to each citation's own chunk. Per-citation anchors would need a different Sources layout and are not attempted here. Query strings are not part of a library document's identity — `/library/document/<id>?v=2` keys with `/library/document/<id>` — which no current producer emits.
+
+ Citations of one library document via the absolute `https://library.document/<id>` alias — the form the agent sometimes emits and the fetch resolver accepts — now share a bibliography entry with that document's other citations instead of rendering as a separate, anchor-less external-looking link. Only the exact `<id>[/pdf|/chunks]` shape normalizes: look-alike hosts, deeper paths, queries and traversal are rejected, so a crafted alias cannot merge into a real document's entry and replace its link.
+
+ Library/RAG citations no longer point at the wrong document when a chunk's denormalised metadata disagrees with its `DocumentChunk` foreign key. The authoritative document id is now carried at the top level of each search result and preferred over metadata wherever the citation URL is built or rebuilt.
+
+ `format_links_to_markdown` no longer crashes the whole bibliography when a citation index contains a non-ASCII digit character (`"¹".isdigit()` is true while `int("¹")` raises).
+
+ Bibliography exports (BibTeX/Quarto and LaTeX) keep library documents that are cited from several chunks. Collapsing those citations onto one Sources line produced a `[1, 2, 3]` index group that the exporters' single-index patterns skipped, silently dropping the source from the exported bibliography.
+
+ `sources_count` now agrees with the rendered bibliography. The bibliography groups sources with `canonical_url_key`, but `SearchResultsCollector` still keyed its dedup map and its `sources` list on the raw link, so the three views of one library document (`/library/document/<id>`, `/<id>/pdf`, `/<id>/chunks#chunk-<n>`, plus the absolute `https://library.document/<id>` alias) produced three citation indices and a `sources_count` of 3 for a report whose `## Sources` block renders one line — the count the MCP server returns and the news impact score reads. The collector now dedups on the same canonical key while recording the display URL, anchor included. This also means a search hit and a later `fetch_content` of the same document share one citation index no matter which spelling of the URL the agent typed back.
+
+ A subsection that only re-fetches previously cited URLs renders its `## Sources` block again. `find_or_add_result` recorded the source on all three branches but appended to `_results` only when it allocated a new citation, so after `reset()` such a subsection reported N sources, rendered none, and tripped the #4969 "citation pass skipped" warning.
+
+ Library-route parsing is now one code path for the relative route, its `/lib/document/<id>` abbreviation and the absolute alias, so all three enforce the same validation. Previously only the alias validated its document id: `/library/document/7/../../../admin/wipe#chunk-1` keyed as document 7, inherited that document's title and citation numbers, and replaced its link. The `/lib/` spelling and percent-encoded ids (`a%2Db` vs `a-b`) now key with the document they denote instead of fanning it across extra entries, and the displayed route is always the registered `/library/document/<id>` one. The bare `[<uuid>]` alias the agent sometimes pastes into `fetch_content` is re-cited as the resolved document's canonical route instead of verbatim, closing the third spelling that used to get its own (dead-linking) entry. A route carrying a control character is now refused outright rather than having its document-id segment truncated: truncating discarded the payload but keyed the crafted URL as the real `/library/document/<id>` it was prefixed with, merging it into that document's bibliography entry and lending it the document's title and citation number. The payload is still neutralised where it is rendered.
+
+ An external URL that merely contains `/library/document/` — `https://evil.example/library/document/7/chunks` — is no longer relabelled as a local library document. The eligibility test is anchored at the start of the URL instead of being a substring match.
+
+ Chunk indices and document ids are validated against ASCII character classes with a magnitude bound, so non-ASCII digits (`٧`, `𝟝`), whitespace-padded values, absurd magnitudes (`10**30`, `1e30`) and fullwidth document ids can no longer reach a URL fragment or path. The bibliography's "prefer an anchored URL" rule now looks for `#chunk-` specifically, so an empty or unrelated fragment can no longer lock out the real chunk anchor.
+
+ `sources` in the strategy result keeps first-seen order. It was passed through `list(set(...))`, which deduped nothing (the collector already guarantees uniqueness) and randomized the ordering differently on every process.
+
+ Note: RIS export does not yet handle merged index lines, and mis-attributes rather than merely skipping them — a record can carry one source's title with another source's URL. It needs a different fix from the other exporters and is tracked in #5687. ([#5685](https://github.com/LearningCircuit/local-deep-research/pull/5685))
+- Fixed `SearchResultsCollector` (LangGraph agent strategy) discarding a source's later excerpts. Citations were deduplicated on the canonical URL alone, so a URL reached by several queries kept only the first occurrence's snippet — the agent read three excerpts and the bibliography, `research_resources` and the API `sources` payload retained one. The dedup key is now the pair `(canonical URL, snippet)`: a repeat with the *same* snippet still collapses onto its existing `[N]` (the #5381 behaviour), while a repeat with a genuinely different snippet becomes its own entry with its own `[N]`. Each entry owns exactly one citation index, so `CitationFormatter`'s index → URL map stays 1:1. Snippets are compared after normalising whitespace, case and trailing ellipsis/punctuation, over a `blake2b` digest of the *whole* passage — never a prefix, which cannot tell "one passage truncated twice" from "two passages that open alike". There is deliberately no cap on how many excerpts one source may contribute: folding a distinct excerpt onto an existing citation would show the model text under a number that resolves to different text, and the count is already bounded by the number of queries run.
+
+ `all_links_of_system` therefore holds one entry per distinct excerpt rather than one per source, and every "N sources" number now counts distinct canonical URLs via the new `search_utilities.count_distinct_sources` — the agent step heartbeat, `_finalize`'s accumulated-sources milestone and its two citation-skip warnings, and the detailed report's `initial_sources` metadata. `initial_sources` now counts distinct sources rather than raw entries for **every** strategy, not only the LangGraph agent: `report_generator` reads `all_links_of_system` whatever produced it, and `source_based`, `focused_iteration` and `topic_organization` have no cross-call dedup at all, so a run that revisits a URL across `analyze_topic()` calls will report a lower — and now correct — number than before (e.g. 4 raw entries reported as 3 sources). `format_links_to_markdown` still renders one `## Sources` line per source, now listing every citation number that source owns (`[1, 4] Title`), and it groups on the same field the collector keys on (`link` first, via the new `source_url_field`) so a result carrying divergent `url` and `link` cannot split one source across two lines. ([#5894](https://github.com/LearningCircuit/local-deep-research/pull/5894))
+- The news feed no longer fails with a generic "Database error" when a stored source row's URL is not a string; such rows are now skipped instead of aborting the whole feed.
+
+### 📝 Other Changes
+
+- Docs now state concretely that the private engine-URL allowlist requires v1.10.5 or newer, replacing the pre-release "newer than v1.10.4" phrasing, and add the same caveat to the two `docker run` recipes (FAQ, installation guide) that recommended the variable with no version note. ([#5647](https://github.com/LearningCircuit/local-deep-research/pull/5647))
+- Docs and `.env.template` now state clearly that LDR itself never loads a `.env` file — it reads only the environment of the server process — while pointing at the file-based mechanisms that *do* work (`docker run --env-file`, Compose `env_file:`, systemd `EnvironmentFile=`), with a new per-deployment persistence table in the environment-configuration guide and a note on the `crawl4ai` dependency's own `load_dotenv()` call. ([#5649](https://github.com/LearningCircuit/local-deep-research/pull/5649))
+- The remediation text shown when a research run fails on a missing API key no longer implies a `.env` file is an option — LDR itself reads only the server process's environment. To keep values in a file, hand the file to whatever starts LDR (`docker run --env-file`, Compose `env_file:`, systemd `EnvironmentFile=`). ([#5661](https://github.com/LearningCircuit/local-deep-research/pull/5661))
+- All production directory creation now routes through `security.create_directory` (audited chokepoint with opt-in path containment), enforced by a pre-commit/CI check that forbids raw `mkdir`/`os.makedirs`.
+- Clarified the `VectorStore.apply()` contract: `dedup` de-duplicates ids within one call and never means "skip an add whose id is already live" — an already-live id must be replaced, because SQLite recycles AUTOINCREMENT ids on ROLLBACK and keeping the stale vector makes search rehydrate the wrong document's text.
+- Removed a maintainer's local filesystem path from the test-review completion report.
+- The missing-inline-citations warnings added for #4969 are now more precise: the iteration-limit failure sentinel no longer triggers the skip warning (an agent failure is not a citation gap), lenticular-bracket citations (`【N】`) count as markers, the zero-marker warning is suppressed when the citation handler itself failed (its own warning already fires, including new ones for non-dict handler results and for dicts missing both the `content` and `response` keys — each naming the returned type or missing keys and the query), and the log wording states facts without referencing internal issue numbers.
+- Unblock the release gate: pin the rebuilt `python:3.14.7-slim` base-image digest (2026-08-25 rebuild) and add Grype suppressions for the 2026-08-18 CPython security batch (CVE-2026-19672 tarfile, CVE-2026-15806 urllib HTTPPasswordMgr, CVE-2026-17084 stringprep). The fixes ship first in Python 3.14.8, which has no release or Docker image yet, and none of the three stdlib APIs is reachable in `src/`.