docs(site): use current msgvault favicon

docs/api-server.md

@@ -10,7 +10,7 @@ description: Daemon-served analytical Web UI and REST API for your msgvault arch
 `msgvault serve` starts an HTTP server that exposes your archive through the
 first-party Web UI at `/` and a REST API under `/api`. It optionally runs a
 background sync scheduler to keep accounts up to date on a cron-based schedule.
-The complete UI is embedded in the release binary; see [Web UI](/web-ui/) for
+The complete UI is embedded in the release binary; see [Web UI](/docs/web-ui/) for
 browser login, secure remote deployment, search states, and keyboard controls.
 
 The API is registered through Huma and exposes a generated OpenAPI document at `/openapi.json`. You can also run `msgvault openapi` to print the same checked-in contract without starting a daemon or opening the archive database. The OpenAPI `info.version` is the API schema version used for client/server compatibility; the current schema is 2.4.0. The running daemon binary version is exposed separately in the generated document metadata. The API queries the same archive database and attachment store as the CLI, Web UI, and TUI. SQLite is the default archive database; PostgreSQL is supported when `[data].database_url` is a PostgreSQL DSN. Keyword search and ordinary archive reads stay local to that database. If vector search is enabled, semantic and hybrid search also call the embedding endpoint configured in `[vector.embeddings]`. The server is designed for interactive archive use, local integrations, dashboards, and automation scripts.
@@ -160,7 +160,7 @@ the building generation while a rebuild is in flight, otherwise the active
 generation. During a rebuild the old active generation keeps serving vector
 and hybrid search, but active-generation top-ups are frozen until the
 building generation activates. See
-[Vector Search](/usage/vector-search/) for the end-to-end workflow.
+[Vector Search](/docs/usage/vector-search/) for the end-to-end workflow.
 
 ---
 
@@ -1046,8 +1046,8 @@ that signal (BM25 missed it or the ANN pool did not include it).
 nothing to fuse). `subject_boosted` is true when the subject-line
 boost was applied.
 
-See [Searching](/usage/searching/) for the full query syntax
-reference and [Vector Search](/usage/vector-search/) for vector /
+See [Searching](/docs/usage/searching/) for the full query syntax
+reference and [Vector Search](/docs/usage/vector-search/) for vector /
 hybrid setup.
 
 ---
@@ -1634,7 +1634,7 @@ The same HTTP server backs configured remote CLI access and the local background
     messages. Teams and Discord importers detect and checkpoint their own
     first-run history backfills.
 
-`msgvault serve` also runs scheduled SyncTech SMS Backup & Restore Drive sources configured under `[[synctech_sms.sources]]`; see [Configuration](/configuration/#synctech-sms-sources).
+`msgvault serve` also runs scheduled SyncTech SMS Backup & Restore Drive sources configured under `[[synctech_sms.sources]]`; see [Configuration](/docs/configuration/#synctech-sms-sources).
 
 ## Security Model
 
@@ -1682,7 +1682,7 @@ requires a usable Parquet cache and keeps analytics unavailable until it is
 ready; a build or open failure is fatal rather than a silent SQL fallback.
 `auto_build_cache = false` leaves cache rebuilds to explicit
 `msgvault build-cache` runs. These settings replace the TUI/MCP analytics flags
-deprecated in 0.17.0; see [Configuration: analytics](/configuration/#analytics).
+deprecated in 0.17.0; see [Configuration: analytics](/docs/configuration/#analytics).
 
 `min_rebuild_interval` limits only automatic post-sync rebuilds. Explicit
 builds, startup maintenance, query-required builds, and unusable-cache recovery
@@ -1700,4 +1700,4 @@ repeated archive-scale work on frequently synced archives. Changes under
 | `schedule` | — | Cron expression for sync schedule |
 | `enabled` | `true` | Whether scheduled sync is active |
 
-See the [Configuration](/configuration/) page for the full config file reference.
+See the [Configuration](/docs/configuration/) page for the full config file reference.

docs/architecture/overview.md

@@ -14,7 +14,7 @@ embedding endpoint configured in `[vector.embeddings]` to build/query semantic
 vectors, then stores those vectors in `vectors.db` on SQLite or pgvector tables
 on PostgreSQL.
 
-<img src="/assets/static/how-it-works.svg" alt="msgvault architecture: Gmail API syncs to SQLite, then offline Parquet analytics, FTS5 search, TUI, and MCP Server" style="width: 100%; max-width: 960px; margin: 1.5rem auto; display: block;" />
+<img src="/docs/assets/static/how-it-works.svg" alt="msgvault architecture: Gmail API syncs to SQLite, then offline Parquet analytics, FTS5 search, TUI, and MCP Server" style="width: 100%; max-width: 960px; margin: 1.5rem auto; display: block;" />
 
 ## Package Structure
 

docs/architecture/postgresql.md

@@ -147,7 +147,7 @@ workspaces; and Relationships' ranking, timeline, and identity link/unlink
 cache refresh) have no PostgreSQL equivalent: those endpoints detect the
 missing DuckDB/Parquet cache and return a named unavailable-cache state
 rather than falling back to live SQL. If you see that state on a PostgreSQL
-backend, it is expected — the [cache troubleshooting guidance](/web-ui/#cache-states)
+backend, it is expected — the [cache troubleshooting guidance](/docs/web-ui/#cache-states)
 applies to SQLite archives only.
 
 ## Current Scope
@@ -192,5 +192,5 @@ that generation's embedding rows so old vectors do not consume search candidate
 budget for the active generation. Frequent full rebuilds can create dead tuples,
 so monitor autovacuum on the embedding tables and run maintenance when needed.
 
-See [Search Ranking Across Backends](/architecture/search-ranking/) for
+See [Search Ranking Across Backends](/docs/architecture/search-ranking/) for
 ranking differences between SQLite, PostgreSQL, sqlite-vec, and pgvector.

docs/architecture/storage.md

@@ -128,7 +128,7 @@ string when checking already-imported source items.
 
 SQLite uses an FTS5 virtual table named `messages_fts`. PostgreSQL uses a `search_fts` `tsvector` column on `messages` with a GIN index.
 
-Both power `msgvault search`, but the rankers differ. See [Search Ranking Across Backends](/architecture/search-ranking/).
+Both power `msgvault search`, but the rankers differ. See [Search Ranking Across Backends](/docs/architecture/search-ranking/).
 
 ### Relationships
 
@@ -146,7 +146,7 @@ PostgreSQL uses native types such as `BIGINT GENERATED ALWAYS AS IDENTITY`, `TIM
 
 For semantic search, pgvector stores index generations, pending embedding work, and embedding vectors in the same PostgreSQL database. There is no separate `vectors.db` on PostgreSQL.
 
-There is currently no SQLite to PostgreSQL migration command. Use PostgreSQL for a new archive or re-sync/import into an empty PostgreSQL database. See [PostgreSQL Backend](/architecture/postgresql/) for setup and operational notes.
+There is currently no SQLite to PostgreSQL migration command. Use PostgreSQL for a new archive or re-sync/import into an empty PostgreSQL database. See [PostgreSQL Backend](/docs/architecture/postgresql/) for setup and operational notes.
 
 ## Parquet (Analytics Cache)
 
@@ -258,7 +258,7 @@ Use `pack-attachments` to migrate the eligible loose backlog immediately,
 `unpack-attachments` to restore cataloged packed objects to loose files before
 downgrading. The last command is local-only and requires the daemon to be
 stopped because it removes production pack files. See the [CLI
-reference](/cli-reference/#pack-attachments) and [Backup](/usage/backup/) guide
+reference](/docs/cli-reference/#pack-attachments) and [Backup](/docs/usage/backup/) guide
 for maintenance and restore behavior.
 
 Set `[data].loose_attachments = true` when file-oriented backup or storage

docs/changelog.md

@@ -444,7 +444,7 @@ All notable changes to msgvault, grouped by release.
   the single archive writer: concurrent operations queue with a visible
   `Waiting:` message, read-only commands run immediately, and scheduled
   syncs yield to interactive commands. See the
-  [Daemon Migration Guide](/guides/daemon-migration/).
+  [Daemon Migration Guide](/docs/guides/daemon-migration/).
 - Daemon lifecycle management via `msgvault serve start|status|stop|restart`,
   with automatic restart of older local daemons on binary upgrade
   (`[server].daemon_auto_restart`).
@@ -634,7 +634,7 @@ All notable changes to msgvault, grouped by release.
 
 **New features**
 
-- **Vector search (semantic and hybrid).** msgvault can now embed your archive using a configured OpenAI-compatible embedding endpoint (Ollama, llama.cpp `server`, LM Studio, etc.) and search it by meaning, not just keywords. `msgvault search --mode vector` runs pure semantic search; `--mode hybrid` fuses BM25 and vector similarity via Reciprocal Rank Fusion. Exposed through local CLI search (`msgvault search`), the HTTP API (`GET /api/v1/search?mode=vector|hybrid`), and the MCP server (`search_messages` mode argument plus a new `find_similar_messages` tool). See [Vector Search](/usage/vector-search/).
+- **Vector search (semantic and hybrid).** msgvault can now embed your archive using a configured OpenAI-compatible embedding endpoint (Ollama, llama.cpp `server`, LM Studio, etc.) and search it by meaning, not just keywords. `msgvault search --mode vector` runs pure semantic search; `--mode hybrid` fuses BM25 and vector similarity via Reciprocal Rank Fusion. Exposed through local CLI search (`msgvault search`), the HTTP API (`GET /api/v1/search?mode=vector|hybrid`), and the MCP server (`search_messages` mode argument plus a new `find_similar_messages` tool). See [Vector Search](/docs/usage/vector-search/).
 - `msgvault build-embeddings` command to generate and maintain the local vector index. Incremental by default; `--full-rebuild` creates a new generation and atomically activates it once coverage reaches zero. Same-model rebuilds keep answering against the previous active generation while the new one is built, with active-generation top-ups frozen until activation; model or dimension changes return `index_stale` until activation.
 - Background embedding via the daemon scheduler. A new `[vector.embed.schedule]` config block drives the embed worker on cron and/or after every successful scheduled sync, so `msgvault serve` can keep the vector index current without manual intervention.
 - `/api/v1/stats` gains a `vector_search` sub-object reporting the active generation, any in-flight rebuild, and the actionable missing embedding count for the generation the worker will target next.
@@ -643,7 +643,7 @@ All notable changes to msgvault, grouped by release.
 **Improvements**
 
 - `search` command gains `--mode fts|vector|hybrid` and `--explain` flags. `--explain` includes per-signal scores (RRF, BM25, vector) in table and JSON output for ranking inspection.
-- Configuration gains a full `[vector]` block with sub-tables for the embedding endpoint, message preprocessing, hybrid ranking, and the embed scheduler. See [Configuration: vector](/configuration/#vector).
+- Configuration gains a full `[vector]` block with sub-tables for the embedding endpoint, message preprocessing, hybrid ranking, and the embed scheduler. See [Configuration: vector](/docs/configuration/#vector).
 - `remove-account` deletes attachment files from disk when they were unique to the removed account. Files shared across multiple accounts are preserved automatically, and an in-progress sync on any account skips file deletion to avoid racing new attachment writes.
 
 **Bug fixes**
@@ -672,7 +672,7 @@ All notable changes to msgvault, grouped by release.
 
 **New features**
 
-- Structured file logging with per-run correlation IDs. Every CLI invocation gets a unique `run_id` on every log line, making it easy to trace a single run across shared log files. New `msgvault logs` command for viewing and tailing logs. File logging is opt-in; see [Configuration: Log](/configuration/#log) for setup.
+- Structured file logging with per-run correlation IDs. Every CLI invocation gets a unique `run_id` on every log line, making it easy to trace a single run across shared log files. New `msgvault logs` command for viewing and tailing logs. File logging is opt-in; see [Configuration: Log](/docs/configuration/#log) for setup.
 
 **Improvements**
 
@@ -725,9 +725,9 @@ All notable changes to msgvault, grouped by release.
 
 **New features**
 
-- SQL query interface via `msgvault query`. Run arbitrary SQL against DuckDB over Parquet with `--format json|csv|table`. See [SQL Queries](/usage/querying/).
+- SQL query interface via `msgvault query`. Run arbitrary SQL against DuckDB over Parquet with `--format json|csv|table`. See [SQL Queries](/docs/usage/querying/).
 - Microsoft 365 OAuth2 support via `msgvault add-o365` for Outlook.com and organizational accounts. Auto-detects personal vs. org IMAP hosts.
-- Text message import: `import-whatsapp`, `import-imessage`, and `import-gvoice` for WhatsApp, iMessage, and Google Voice. See [Text Messages](/usage/text-messages/).
+- Text message import: `import-whatsapp`, `import-imessage`, and `import-gvoice` for WhatsApp, iMessage, and Google Voice. See [Text Messages](/docs/usage/text-messages/).
 - TUI text mode: press `m` to toggle between Email and Texts for browsing imported text conversations.
 - `--after` and `--before` date filters for `sync-full` with IMAP accounts.
 - CC and BCC recipients exposed in the message API responses.