feat(tui): split settings into focused panels, real SIP panel, voice to Session, wiki sync

Settings now opens as separate LLM & Provider / Email (SMTP+IMAP) / General panels, each with its own Save+validation. Provider panel: active-provider dropdown is the single way to change the active provider (persists across restarts), active one marked (attivo), removed the Set-default button, Edit opens the selected provider, API key masked. SIP is now a full editable panel (was read-only). Voice moved from Settings to Session. TTS engine panel gained reset-to-default. Docs updated to the new structure and mirrored to the GitHub wiki via sync-wiki.yml; Help/Docs points at the wiki.

Author: Andrea-Bruno

.github/workflows/release.yml

@@ -520,34 +520,38 @@ jobs:
             artifacts/**/*.tar.gz
             artifacts/**/*.msi
           body: |
-            ### Start-up options: an upgrade note
+            ### Settings, rebuilt around separate panels
 
-            Options given on the command line were **silently ignored** when they followed a
-            valueless switch — `agent --headless --LLM:Anonymize true` made the switch swallow
-            the next key, so `--Urls`, `--Sip:ListenPort`, `--Tts:Engine`, `--LLM:Anonymize`,
-            `--SkipIndexingOnStartup` and any other `--Key value` after `--headless`, `--tui`,
-            `--no-update`, `--no-gui` or `--enable-log` did nothing (this also affected the
-            systemd unit shipped by the Linux installer). Fixed: every documented override now
-            applies wherever it appears. The same settings passed as environment variables or
-            written in `appsettings.json` worked before and are unaffected.
+            The single tabbed "Main settings" window is gone. Settings now opens as focused
+            panels — **LLM & Provider**, **Email (SMTP + IMAP)** and **General** — each with
+            its own **Save** that closes only when its values validate, so a bad value in one
+            area never blocks the others.
 
-            ### Chat sessions recover instead of failing
+            ### Provider management is clearer
 
-            A chat that has been idle for over an hour (or a server restart) no longer returns
-            "session not found": the conversation continues on a new session, and the previous
-            turns are restored when the client resends its transcript — the terminal UI always
-            does, so you just keep talking. API clients get the new `session_id` (plus
-            `session_resumed: true` when the earlier turns came back) instead of a 404.
+            The **active-provider dropdown** is now the single way to change which provider is
+            active: pick one, press Save, and the choice persists across restarts. The active
+            provider is marked **(attivo)** in the list, the dropdown and the list always agree,
+            and the old "Set default" button is gone. **Edit** opens the provider selected in
+            the list, and the API-key field shows the stored key masked.
 
-            ### Privacy
+            ### SIP telephony is now a real settings panel
 
-            With anonymisation enabled, the transcript a client resends is now anonymised like
-            a normal prompt instead of entering the conversation in clear text.
+            The SIP page used to be read-only. It is now a full editable panel: enable/disable,
+            listen port, registrar, credentials, answer mode and PIN, plus call / hang-up /
+            reload actions.
 
-            ### Documentation
+            ### Voice moved to the Session menu
 
-            New plain-language guide **How AgentBridge works** (what an AI agent is, its tools,
-            the documents area, attachments) under `docs/guide/`, shipped with this archive.
+            Voice starts a one-shot dictation, so it is no longer a "setting": it now lives
+            under **Session**, alongside the model switch and session features. The TTS engine
+            panel gained a **reset to default**.
+
+            ### Documentation on the wiki
+
+            The user guides now live on the GitHub wiki, kept in sync automatically from the
+            repository. **Help → Documentation** (and `/docs`) opens
+            https://github.com/Graphene-Lab/AgentBridge/wiki.
           generate_release_notes: true
 
   # Microsoft Store submission: once the release exists with the MSI attached, point the

.github/workflows/sync-wiki.yml

@@ -0,0 +1,70 @@
+name: Sync docs to wiki
+
+# Mirrors the user-facing docs/ tree into the GitHub wiki (Graphene-Lab/AgentBridge.wiki).
+# Runs on every push that touches docs/ and on manual dispatch. The wiki is a separate git
+# repo; this job clones it, regenerates the flat pages from docs/ with tools/wiki/sync-wiki.js,
+# and pushes. Keeping the wiki in the repo means it stays in sync automatically whenever a
+# guide changes — see the docs rule in AGENTS.md.
+on:
+  push:
+    branches: [master]
+    paths:
+      - 'docs/**'
+      - 'tools/wiki/**'
+  workflow_dispatch:
+
+permissions:
+  contents: write
+
+concurrency:
+  group: sync-wiki
+  cancel-in-progress: false
+
+jobs:
+  sync:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Checkout main repo (docs + sync script)
+        uses: actions/checkout@v4
+        with:
+          path: main
+
+      - name: Setup Node
+        uses: actions/setup-node@v4
+        with:
+          node-version: '20'
+
+      - name: Clone wiki (or init if empty)
+        shell: bash
+        run: |
+          WIKI_URL="https://x-access-token:${GITHUB_TOKEN}@github.com/${GITHUB_REPOSITORY}.wiki.git"
+          if git clone "$WIKI_URL" wiki 2>/dev/null; then
+            echo "wiki cloned"
+          else
+            echo "wiki not found — initialising"
+            mkdir -p wiki
+            ( cd wiki && git init -q && git remote add origin "$WIKI_URL" )
+          fi
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+
+      - name: Generate wiki pages from docs
+        shell: bash
+        run: node main/tools/wiki/sync-wiki.js main/docs wiki
+
+      - name: Commit and push
+        shell: bash
+        working-directory: wiki
+        run: |
+          git config user.name "github-actions[bot]"
+          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+          git add -A
+          if git diff --cached --quiet; then
+            echo "no wiki changes"
+            exit 0
+          fi
+          git commit -q -m "docs: sync wiki from docs/ (${{ github.sha }})"
+          # The wiki default branch may be main or master; push whatever HEAD is on.
+          BRANCH=$(git rev-parse --abbrev-ref HEAD)
+          [ "$BRANCH" = "HEAD" ] && BRANCH=main
+          git push -u origin "HEAD:$BRANCH"

AGENTS.md

@@ -39,6 +39,21 @@ Rules:
 - Before writing any guide, ask: *who reads this — the person who installs the app, or the
   developer who maintains the code?* User → `docs/`. Developer → `docs-dev/`.
 
+### The GitHub wiki mirrors `docs/` (keep it in sync)
+
+The public wiki at **https://github.com/Graphene-Lab/AgentBridge/wiki** is generated from
+`docs/` by `.github/workflows/sync-wiki.yml` (runs on every push that touches `docs/` or
+`tools/wiki/`, and on manual dispatch). The workflow clones the wiki repo, regenerates the
+flat pages with `tools/wiki/sync-wiki.js` (each `docs/**/*.md` becomes a wiki page named by
+its file basename; `INDEX.md` → `Home`; relative `.md` links are rewritten to the flat page
+name), and pushes.
+
+**Rule: the wiki is never edited by hand — edit `docs/` and let the workflow sync it.** Any
+change to a user guide in `docs/` must be pushed so the wiki regenerates; do not make
+divergent edits directly in the wiki repo (they will be overwritten on the next sync). When
+you add or rename a `docs/` page, the sync picks it up automatically; just make sure the
+links between guides use the file paths so the rewriter can map them to the wiki page names.
+
 ## Release gate: IsPrerelease flag
 
 `AgentBridge.csproj` carries `<IsPrerelease>` (default `true`). It decides whether a GitHub

AgentBridge.csproj

@@ -16,13 +16,13 @@
          off triggers release.yml automatically (tag v1.yy.MM.dd created by the workflow).
          Date-based auto version, same scheme as UISupportBlazor/UISupportGeneric
          (1.yy.MM.dd, e.g. 1.26.08.09). -->
-    <IsPrerelease>true</IsPrerelease>
+    <IsPrerelease>false</IsPrerelease>
     <!-- NuGet wait marker (release.ps1 / release.yml): true = the release build must wait for
          today's core dependency packages on nuget.org (a core repo changed today and its
          package is still propagating); false = no core repo changed since its last tag, so the
          wait is skipped. Default true (conservative: a manual gate-off push without
          release.ps1 still waits). -->
-    <NuGetWait>true</NuGetWait>
+    <NuGetWait>false</NuGetWait>
     <!-- NuGetWaitPackages: the comma-separated package ids the release build waits for when
          <NuGetWait> is true. release.ps1 narrows it to the repos that actually changed today so
          the wait exits as soon as those propagate instead of burning the full 30 minutes on
@@ -37,7 +37,7 @@
          pins it to the LOCAL push date in the gate-off commit, so CI (DateTime.Now = UTC)
          can never derive the previous UTC day across the local/UTC-midnight boundary and
          reuse an already-released version (release.yml would skip it). -->
-    <ReleaseDate Condition="'$(ReleaseDate)' == ''">$([System.DateTime]::Now.ToString("yy.MM.dd"))</ReleaseDate>
+    <ReleaseDate Condition="'$(ReleaseDate)' == ''">26.09.15</ReleaseDate>
     <Version Condition="'$(IsPrerelease)' == 'true'">1.$(ReleaseDate)-prerelease</Version>
     <Version Condition="'$(IsPrerelease)' != 'true'">1.$(ReleaseDate)</Version>
   </PropertyGroup>