Skip to content

Speakr is a personal, self-hosted web application designed for transcribing audio recordings

Python

docs: cover this round's features - agentic Inquire with timestamped click-to-play citations and privacy toggles, recorder ASR hints, username watch folders, Help Center link, SSO UserInfo claim resolution

Murtaza Nasir42aff45b1440
diff --git a/README.md b/README.md
index 345f7d4a..0b5958b3 100644
--- a/README.md
+++ b/README.md
@@ -37,13 +37,13 @@ Speakr turns a recording into organized, searchable, shareable knowledge. Here i
### Capture
- **Flexible input** - record from your microphone, your computer's system or browser-tab audio, or both mixed together; or drag and drop existing files. A per-OS setup guide and a virtual-device picker surface Pulse / PipeWire monitors, BlackHole, VB-Cable, Voicemeeter, and Stereo Mix as inputs.
- **Long sessions** - in-app recordings stream to the server during capture, so sessions can run for hours and survive a page reload.
-- **Hands-off intake** - a watched "black hole" folder auto-imports and processes any audio dropped into it.
+- **Hands-off intake** - a watched "black hole" folder auto-imports and processes any audio dropped into it, with optional per-user subfolders named by user id or username.
### Transcribe
- **Bring your own engine** - self-hosted WhisperX (recommended; it is what enables the speaker features below), OpenAI, Mistral / Voxtral, AssemblyAI, or any custom ASR webservice. The right connector is auto-detected from your configuration.
- **Speaker diarization** - automatic who-said-what labeling (WhisperX, or OpenAI's diarizing models).
- **Voice profiles** - recognize the same person across different recordings via voice embeddings (requires the WhisperX ASR backend).
-- **Custom vocabulary and hotwords** (most effective with the WhisperX backend) - bias the transcriber toward names, jargon, and acronyms it would otherwise mishear; configurable globally or per tag / folder.
+- **Custom vocabulary and hotwords** (most effective with the WhisperX backend) - bias the transcriber toward names, jargon, and acronyms it would otherwise mishear; configurable globally, per tag / folder, or right in the upload and in-app recording dialogs, with reusable saved templates.
- **Synced playback** - click any line to jump to that moment, follow-along highlighting during playback, and a chat-style bubble view.
- **Language support** - automatic language detection plus a quick-pick of 11 common languages.
@@ -51,7 +51,7 @@ Speakr turns a recording into organized, searchable, shareable knowledge. Here i
- **Summaries** - generated automatically, with prompts you can fully customize per recording, tag, or folder (including reusable prompt variables).
- **Event extraction** - surface action items and calendar-worthy events from a transcript.
- **Per-recording chat** - ask questions about a single recording in a floating, dockable panel.
-- **Inquire Mode** - semantic search and natural-language chat across your entire library at once.
+- **Inquire Mode** - semantic search and natural-language chat across your entire library at once. An optional **agentic mode** (opt-in beta) upgrades this to an AI researcher that iteratively searches, lists, and reads your recordings until it can answer - showing every step live, and citing the exact moment with numbered source links that open the recording and play from that point. You decide what it may read: transcripts always, summaries and private notes only if you allow them.
### Organize
- **Folders and bulk operations** to keep a large library tidy.
@@ -69,7 +69,7 @@ Speakr turns a recording into organized, searchable, shareable knowledge. Here i
- **Signed webhooks** - HMAC-signed, SSRF-guarded, retrying outbound notifications on recording lifecycle events.
- **Usage budgets** for LLM tokens and transcription minutes, per user.
-Speakr is also an installable Progressive Web App - mobile-first, offline-capable, with a phone share-target - and ships light/dark themes, an incognito mode, and a UI translated into seven languages.
+Speakr is also an installable Progressive Web App - mobile-first, offline-capable, with a phone share-target - and ships light/dark themes, an incognito mode, a configurable Help Center link for your own documentation, and a UI translated into seven languages.
## Real-World Use Cases
diff --git a/docs/admin-guide/sso-setup.md b/docs/admin-guide/sso-setup.md
index 0adf33cc..c4c373e5 100644
--- a/docs/admin-guide/sso-setup.md
+++ b/docs/admin-guide/sso-setup.md
@@ -40,6 +40,8 @@ Restart Speakr after updating environment variables.
- `email` (recommended): used for matching and domain allowlist.
- `preferred_username` or `name`: used for username/full name if provided.
+The claims may live in the ID token or at the UserInfo endpoint; both work. When the ID token carries only the bare claims (`sub`, `iss`, `aud`), as Keycloak, Azure AD, and LemonLDAP::NG do by default, Speakr fetches the UserInfo endpoint automatically and merges its claims, verifying the endpoint's `sub` matches the ID token's. No mapper configuration is needed on the IdP side to embed claims in the ID token.
+
## Keycloak quick start
1. In Keycloak, create a new client (e.g., `speakr`) with:
diff --git a/docs/admin-guide/system-settings.md b/docs/admin-guide/system-settings.md
index a470eac5..2cbe0af2 100644
--- a/docs/admin-guide/system-settings.md
+++ b/docs/admin-guide/system-settings.md
@@ -126,6 +126,12 @@ Processing backlogs might indicate your timeout is too high. If the system waits
Beyond the UI-configurable settings above, several environment variables in your `.env` file control fundamental system behaviors. These require instance restart to take effect.
+### Interface Customization
+
+**HELP_CENTER_URL**: Adds a Help Center link to the user menu (on both the main app and the account/admin pages), opening in a new tab. Point it at your own FAQ, tutorial, or intranet page so users find your organization's guidance instead of the technical documentation. Unset by default (no link shown).
+
+**HELP_CENTER_LABEL**: Optional custom label for the Help Center link. When unset, a localized "Help Center" label is used.
+
### Collaboration & Sharing
**ENABLE_INTERNAL_SHARING**: Controls user-to-user sharing capabilities. Set to `true` to enable internal sharing features, allowing users to share recordings with specific colleagues. Required for group functionality. Default: `false`.
diff --git a/docs/features.md b/docs/features.md
index e4f902f7..09ba1d1c 100644
--- a/docs/features.md
+++ b/docs/features.md
@@ -69,6 +69,8 @@ Transform static transcripts into dynamic conversations with the integrated chat
Search across all your recordings using natural language questions instead of keywords through [Inquire Mode](user-guide/inquire-mode.md). The [vector store](admin-guide/vector-store.md) must be configured for this feature to work. The semantic search understands meaning and context, finding relevant content even when exact words don't match. Ask questions like "When did we discuss the budget increase?" and get results from any recording that covered that topic, regardless of the specific terminology used.
+An optional [agentic mode](user-guide/inquire-mode.md#agentic-inquire-beta) (opt-in beta, `ENABLE_INQUIRE_AGENT=true`) replaces the single retrieval pass with an AI agent that searches, lists, and reads your recordings iteratively until it can answer. The interface shows each step live with a stop button, keeps a collapsible activity record on every answer, and cites sources as numbered markers with timestamps; clicking a citation opens the recording and starts playback at the cited moment. Per-user privacy settings control what the agent may read: transcripts are always available, while summaries and private notes are individual opt-ins enforced by removing the corresponding capability from the agent entirely.
+
## Organization and Management
### Folders
@@ -216,7 +218,7 @@ Store recording audio files in S3-compatible object storage (AWS S3, MinIO, or a
### Black Hole Processing
-Set up a watched directory where dropped audio files are automatically processed. Configure this in [system settings](admin-guide/system-settings.md) for automated workflows. Perfect for automation workflows or batch processing, the black hole directory monitors for new files and queues them for transcription without manual intervention.
+Set up a watched directory where dropped audio files are automatically processed. Configure this in [system settings](admin-guide/system-settings.md) for automated workflows. Perfect for automation workflows or batch processing, the black hole directory monitors for new files and queues them for transcription without manual intervention. In `user_directories` mode, each user gets their own subfolder, which may be named after the user id (`user123` or `123`) or the username itself (matching the naming of the per-user export folders), a clean fit for multi-tenant setups where other systems drop files per user.
### Automated Export
@@ -229,7 +231,7 @@ Improve transcription accuracy for domain-specific content with two complementar
- **Hotwords** - Provide comma-separated terms (brand names, acronyms, technical jargon) that the transcription engine should prioritize. This biases the model toward correctly recognizing words it might otherwise misspell.
- **Initial Prompt** - Give the transcription engine context about the recording content, helping it make better overall word choices when audio is ambiguous.
-Set these at any level - [per-user defaults](user-guide/settings.md#transcription-hints), [per-tag](user-guide/settings.md#transcription-hints-hotwords--initial-prompt), [per-folder](user-guide/folders.md), or per-upload in the Advanced ASR Options. Values cascade with upload form taking highest priority, then tags, folders, and finally user defaults. See the [FAQ](faq.md#certain-specialized-words-dont-get-transcribed-properly-how-do-i-fix-this) for practical usage tips.
+Set these at any level - [per-user defaults](user-guide/settings.md#transcription-hints), [per-tag](user-guide/settings.md#transcription-hints-hotwords--initial-prompt), [per-folder](user-guide/folders.md), or per-recording in the Advanced ASR Options of both the upload dialog and the in-app recorder (including a saved-template picker in each). Values cascade with the upload or recording form taking highest priority, then tags, folders, and finally user defaults. See the [FAQ](faq.md#certain-specialized-words-dont-get-transcribed-properly-how-do-i-fix-this) for practical usage tips.
### Custom ASR Configuration
diff --git a/docs/getting-started.md b/docs/getting-started.md
index aeea2b1b..35e1dab2 100644
--- a/docs/getting-started.md
+++ b/docs/getting-started.md
@@ -192,6 +192,8 @@ ENABLE_INQUIRE_MODE=true
Then restart the container with `docker compose restart` for the change to take effect.
+Once Inquire Mode works, you can additionally try the [agentic mode beta](user-guide/inquire-mode.md#agentic-inquire-beta) (`ENABLE_INQUIRE_AGENT=true`), which answers by iteratively searching and reading your recordings, with live activity display and timestamped, click-to-play citations.
+
### Enable User Registration
By default, only the admin can create new users. Learn more about [user management](admin-guide/user-management.md) in the admin guide. To allow self-registration, set:
diff --git a/docs/getting-started/installation.md b/docs/getting-started/installation.md
index 41ca4e0f..db6d00d8 100644
--- a/docs/getting-started/installation.md
+++ b/docs/getting-started/installation.md
@@ -391,7 +391,7 @@ AUTO_PROCESS_WATCH_DIR=/data/auto-process
AUTO_PROCESS_CHECK_INTERVAL=30
```
-When enabled, Speakr checks the watch directory every 30 seconds for new audio files. Any files found are automatically moved to the uploads directory and processed using your configured transcription settings. The `admin_only` mode assigns all processed files to the admin user, but you can also configure it for multi-user scenarios with separate directories for each user.
+When enabled, Speakr checks the watch directory every 30 seconds for new audio files. Any files found are automatically moved to the uploads directory and processed using your configured transcription settings. The `admin_only` mode assigns all processed files to the admin user, but you can also configure it for multi-user scenarios with separate directories for each user (`AUTO_PROCESS_MODE=user_directories`). Per-user subfolders may be named after the user id (`user123` or `123`) or after the username itself, mirroring the naming of the per-user export folders.
To use this feature, you'll need to mount an additional volume in your Docker Compose configuration, which we'll cover in the next steps.
diff --git a/docs/user-guide/recording.md b/docs/user-guide/recording.md
index 773f0661..db20a62e 100644
--- a/docs/user-guide/recording.md
+++ b/docs/user-guide/recording.md
@@ -244,6 +244,8 @@ The **Initial Prompt** field provides context to the transcription engine about
Both fields follow a precedence hierarchy: values you set in the upload form override tag defaults, which override folder defaults, which override your personal defaults in [account settings](settings.md#custom-prompts-tab). If you regularly transcribe similar content, set your defaults at the tag or user level so you don't have to enter them every time.
+The same fields, along with the saved-template picker, are also available in the in-app recorder: after stopping a recording, expand **Upload Settings**, then **Advanced ASR Options**. Values set there travel with the recording through every upload path, including crash recovery and incognito processing.
+
#### Transcription Model
When your administrator has configured `TRANSCRIPTION_MODELS_AVAILABLE` (or curated a model list from the admin dashboard), the upload form gains a **Transcription model** dropdown that lets you pick the specific model used for this recording. The default option uses whichever model the resolved tag, folder, or admin default specifies. The dropdown is hidden if only one option would be available, since there is no real choice to offer.
diff --git a/docs/user-guide/settings.md b/docs/user-guide/settings.md
index 67f5f70e..4cd0cd51 100644
--- a/docs/user-guide/settings.md
+++ b/docs/user-guide/settings.md
@@ -83,6 +83,15 @@ These defaults apply to all your recordings unless overridden by tag defaults, f
!!! tip "When to use hotwords vs initial prompt"
Use **hotwords** for specific terms the model tends to misspell - proper nouns, brand names, technical acronyms. Use **initial prompt** for broader context that helps the model understand the domain and make better overall word choices.
+### Inquire Privacy
+
+The **Inquire Privacy** section controls what the Inquire AI may read when answering your questions. Transcripts are always available, since they are the minimum Inquire needs to function; everything else is your choice:
+
+- **Allow access to summaries** (on by default) - lets Inquire read your recordings' AI summaries, which is a fast way for the agentic mode to survey a long recording before deciding whether to read its transcript.
+- **Allow access to your notes** (off by default) - your private notes stay unavailable unless you enable this. Only your own notes are ever used; notes on recordings shared with you are never exposed, regardless of this setting.
+
+These settings are enforced structurally: a disabled content type is removed from the AI's capabilities entirely rather than merely discouraged. See [Inquire Mode](inquire-mode.md#agentic-inquire-beta) for how the agentic mode uses these permissions.
+
### Timestamp Availability for Summary and Chat
At the top of the Prompt Options you can control whether the AI receives per-line timestamps with the transcript. There are two independent toggles, one for **summarization** and one for **chat**, so you can enable the feature only where you need it: