Add resilient Azure offline transcription backlog

docs/meeting-assistant-configuration.md

@@ -129,7 +129,7 @@ The default profile is always named `default`. Non-default profile hotkeys are r
 
 During recording, Meeting Assistant captures microphone and system loopback separately, buffers both streams to align samples, cleans the microphone stream through a local adaptive echo canceller using loopback as the far-end reference, then mixes the cleaned microphone and system streams into the normal 16 kHz mono PCM chunks. If one source is quiet beyond the alignment timeout, the available source is mixed with synthetic silence so microphone-only speech and system-only playback keep flowing to transcription.
 
-`Recording:MicrophoneMixGain` and `Recording:SystemAudioMixGain` are applied during the final mix and default to `1`. `Recording:TemporaryRecordingsFolder` controls where the temporary mixed WAV is written while the run is active. Temporary WAV files are deleted after the run completes, and stale temporary recordings from interrupted runs are deleted when the application starts.
+`Recording:MicrophoneMixGain` and `Recording:SystemAudioMixGain` are applied during the final mix and default to `1`. `Recording:TemporaryRecordingsFolder` controls where the temporary mixed WAV is written while the run is active. Temporary WAV files are deleted after the run completes, and stale temporary recordings from interrupted runs are deleted when the application starts. If an Azure Speech meeting cannot drain transcription before `Recording:StopProcessingTimeout`, Meeting Assistant keeps the WAV and writes a durable backlog item under `TemporaryRecordingsFolder\offline-transcription-backlog`. The background backlog worker retries those queued meetings, replays each WAV through a fresh speech pipeline, rewrites the original transcript, completes meeting metadata and summary generation, then removes the backlog item and WAV.
 
 `Recording:MaxMetadataAttendeeImportCount` limits how many attendees Outlook metadata enrichment imports into meeting-note frontmatter. The default is `30`; when an appointment has more attendees than that, Meeting Assistant still imports title, agenda, and scheduled end time, but leaves attendees empty because large invites are usually presentation-style meetings.
 
@@ -234,10 +234,12 @@ Pyannote diarization settings are shared by local Whisper finalization and speak
 
 ## Azure Speech
 
-`azure-speech` streams Meeting Assistant's captured PCM chunks to Azure AI Speech using the Speech SDK `ConversationTranscriber`; it does not use an Azure-owned microphone capture or MAS/AEC input path.
+`azure-speech` streams Meeting Assistant's captured PCM chunks to Azure AI Speech using the Speech SDK `ConversationTranscriber`; it does not use MSAL, an Azure-owned microphone capture, or a separate MAS/AEC input path.
 
 Configure `AzureSpeech:Endpoint`, `Region`, `Language`, and either `Key` or `KeyEnv`; this repository uses `AZURE_SPEECH_KEY` by default and does not store the key in appsettings. `DiarizeIntermediateResults` enables diarization for intermediate and final conversation transcription results.
 
+If Azure Speech reports a transient connection failure, Meeting Assistant starts a new SDK session instead of completing the transcript with an error. The transcript note shows a single status line such as `<Reconnecting... 1/5>`, updates that line across retry attempts, and replaces it with the next real transcript line after Azure emits speech again. After the configured immediate attempts are exhausted, the same line changes to an Azure disconnected marker while retries continue. Reconnect starts a fresh Azure conversation session, so live speaker mappings are reset before processing the first resumed speech line.
+
 | Setting | Purpose |
 | --- | --- |
 | `Endpoint` | Optional Azure Speech endpoint. When blank, `Region` plus subscription key is used. |
@@ -248,6 +250,8 @@ Configure `AzureSpeech:Endpoint`, `Region`, `Language`, and either `Key` or `Key
 | `Key` | Optional inline Azure Speech key. Prefer `KeyEnv`. |
 | `KeyEnv` | Environment variable name that contains the Azure Speech key. |
 | `RecognitionStopTimeout` | Timeout while stopping the ConversationTranscriber after closing the audio stream. |
+| `ReconnectAttemptsBeforeDisconnectedMarker` | Immediate reconnect-marker attempt count before the transcript status changes to the longer disconnected marker. |
+| `ReconnectDelay` | Delay between Azure Speech SDK session reconnect attempts. |
 | `DiarizeIntermediateResults` | Requests diarized intermediate results from Azure. |
 | `PostProcessingOption` | Accepted by config but currently skipped for the live `ConversationTranscriber` path; the app logs a warning if set. |
 | `PhraseListWeight` | Weight applied to dictation-word phrase-list hints. |

docs/meeting-workflow-engine.md

@@ -68,7 +68,7 @@ The engine runs when `MeetingRecordingCoordinator` emits a meeting workflow even
 - `created`: after the meeting note and assistant context artifact are created.
 - `state_transition`: after the assistant context state moves forward.
 - `speaker_identified`: when live or final speaker identification reports a display name.
-- `transcript_line`: before a formatted transcript line is appended or used in a transcript rewrite.
+- `transcript_line`: after a formatted live transcript line is durably appended, before any changed line is rewritten in place, and before lines are used in full transcript rewrites.
 
 For every event, the engine:
 
@@ -80,9 +80,9 @@ For every event, the engine:
 6. Saves the meeting note once if any meeting-note step changed it.
 7. Updates assistant-context frontmatter links/title from the saved meeting note after note changes.
 
-For `transcript_line` events, the engine returns the possibly updated transcript line to the caller. Live transcript appends, live speaker relabel rewrites, final diarization rewrites, and final speaker-identity rewrites all pass their formatted lines through this event before writing markdown.
+For `transcript_line` events, the engine returns the possibly updated transcript line to the caller. Live transcript appends first write the original formatted line and keep a reference to that exact body line, then run transcript-line rules out of band. Later transcript segments can continue to be appended while the workflow runs. If rules return a changed line, Meeting Assistant rewrites the referenced line in the transcript file. If a transcript-line rule fails during live recording, Meeting Assistant logs the failure and keeps the original line so later transcript segments can continue to be written.
 
-Rules are best-effort automation. The settings/logs write tool rejects invalid YAML, unknown steps, unsupported set-property fields, trigger or condition entries without a supported key, and step value Razor templates that fail against the workflow template model before writing the rules file. Invalid NCalc expressions can still fail at workflow runtime and should be covered by tests before the engine is broadened.
+Rules are best-effort automation for live transcript persistence. The settings/logs write tool rejects invalid YAML, unknown steps, unsupported set-property fields, trigger or condition entries without a supported key, and step value Razor templates that fail against the workflow template model before writing the rules file. Invalid NCalc expressions can still fail at workflow runtime. Workflow execution logs every triggered rule with its rule name and event type, logs completion with whether the note or transcript line changed, and logs rule failures with the rule name and event type.
 
 ## YAML Shape
 
@@ -227,6 +227,10 @@ valid email address token. Literal email addresses such as `Support@contoso.com`
 unchanged; if the same value also contains a Razor expression, the email `@` is escaped
 before rendering so the address still appears normally in the rendered result.
 
+Rendered values are plain text for markdown artifacts, not HTML. UTF-8 characters such as
+`ß`, `ö`, and `ä` are preserved after Razor rendering instead of being persisted as HTML
+entities.
+
 ## Steps
 
 ### `add_attendee`
@@ -251,7 +255,7 @@ steps:
 
 ### `set_property`
 
-Sets a supported property. Supported meeting properties are `title` and `meeting.title`. During `transcript_line` events, `transcript.line` is also supported and replaces the single formatted line that will be written.
+Sets a supported property. Supported meeting properties are `title` and `meeting.title`. During `transcript_line` events, `transcript.line` is also supported. For live transcription, changing `transcript.line` rewrites the referenced formatted line after the workflow task completes.
 
 ```yaml
 steps: