The system prompt for the medical assistant LLM has been updated to
improve clarity and explicitly state the desired formatting for
corrections and uncertainties. This includes:
- Consolidating similar correction examples.
- Specifying that only "==text==" annotations are allowed for
corrections and uncertainties.
- Explicitly disallowing other annotation formats like "(unsicher)" or
"[TODO]".
Avoid redundant LLM calls by waiting until all recordings in a batch
have been processed before regenerating the oneliner. This change
introduces a helper function `has_pending_recordings` to check for any
`.m4a` files without a corresponding transcript, and the
`update_oneliner` logic now only proceeds if no pending recordings are
found.
This commit integrates the Gazetteer into the transcription pipeline to
normalize terminology.
The Gazetteer, loaded at startup, is now passed to the transcription
worker. Before persisting the transcription text, it is processed
through the Gazetteer's `replace` method, ensuring standardized
terminology. This normalization is also applied to the generated
"oneliner" text.
Additionally, the `regenerate_missing_oneliners` function in the
recovery module has been updated to accept and utilize the Gazetteer,
ensuring that regenerated oneliners also benefit from terminology
normalization.
A new integration test, `transcribe_worker_normalizes_whisper_output`,
has been added to verify that the Whisper output is correctly normalized
by the Gazetteer before being saved to the transcript file.
The gazetteer has been refactored to act as a post-AI normalization
filter. Previously, it was used to annotate LLM input with potential
corrections from a vocabulary. This approach was ineffective because
LLMs often override such hints.
The new approach applies the gazetteer *after* the LLM has generated its
output. This allows for deterministic correction of known terminology,
including fixing LLM output drift (e.g., anglicized drug names).
Key changes:
- `annotate` function renamed to `replace`.
- The output format changes from `Canonical [?original]` to simply the
`Canonical` form.
- The gazetteer now operates on the final LLM output before persistence,
ensuring consistency.
- Tests have been updated to reflect this new behavior, focusing on the
final output rather than the LLM request payload.
The gazetteer now annotates text with `Canonical [?original]`. This
prioritizes the corrected term for the LLM while keeping the original
transcription as a fallback.
This change aligns the gazetteer's annotation strategy with the LLM's
prompt, which expects corrections to be marked for review. Previously,
the format was `original [?Canonical]`, which could lead to the LLM
using the potentially incorrect original term.
The build scripts for generating `anatomy.txt` and related files have
been removed as they are no longer used or maintained. The Kölner
Phonetik module was also removed as it was not being utilized.
The `filter-hunspell-de.sh` script now uses a `MIN_LEN` variable that is
synchronized with the `MIN_TOKEN_LEN` constant in
`src/gazetteer/mod.rs`.
This ensures that short tokens, which are prone to false positives due
to
collisions with common German words, are filtered out by both the
gazetteer
loading and the Hunspell dictionary processing.
Additionally, the script's dependency instructions have been updated to
be
more comprehensive, listing package manager commands for Arch/CachyOS,
Debian/Ubuntu, and Fedora. The `awk` command has been introduced to
filter
out tokens shorter than `MIN_LEN` before sorting and deduplicating.
The `Gazetteer::insert` method has been updated to skip entries shorter
than `MIN_TOKEN_LEN` to prevent noisy phonetic collisions. The
`best_candidate`
helper function has been extracted to improve readability and structure.
The test cases have been updated to reflect the change in
`MIN_TOKEN_LEN`
from 4 to 5 and to use more relevant examples for the updated
functionality,
such as "Cerebrum" and "Zerebrum". The `anatomy.txt` file has been
updated
to reflect that anatomical terms are currently disabled. The
`medications.txt` file has been significantly expanded with a curated
list
of German medication brand and active ingredient names.
Adds three new vocabulary files to the server's vocabulary directory:
`anatomy.txt`, `medications.txt`, and `substances.txt`. A README.md is
also added to explain their purpose, format, and population strategy.
The `server/vocab/raw/` directory is added to `.gitignore`.
This commit introduces functionality to annotate the user prompt with
potential proper name corrections from a gazetteer. The LLM will then
use these annotations to improve the accuracy of transcriptions,
especially for medical terms and names.
The `SYSTEM_PROMPT` has also been updated to inform the LLM about these
new annotations and how to handle them.
Pass a `Gazetteer` to the analyze worker to enable proper-name
correction. The gazetteer is loaded from disk at application startup. If
the directory is missing or unreadable, the worker will start without
proper-name correction capabilities.
This commit introduces the `VOCAB_DIR` configuration option, which
specifies the directory for gazetteer files. It also adds the `strsim`
crate, which is used for calculating string similarity and is essential
for the gazetteer's functionality.
The gazetteer module has been significantly expanded to include the
`Gazetteer` struct, its loading mechanism from `.txt` files, and an
annotation function. This functionality allows for pre-LLM correction of
proper names and technical vocabulary by identifying potential
misspellings based on phonetic similarity and edit distance.
Implements the Kölner Phonetik algorithm for German word sound-encoding.
This phonetic algorithm is used to find gazetteer entries that sound
similar to a potentially misspelled token, enabling correction
suggestions.
The implementation includes:
- `encode`: The main function that orchestrates the encoding process.
- `prepare`: Prepares the input string by expanding umlauts,
uppercasing, and filtering non-alphabetic characters.
- `raw_encode`: Performs the character-by-character phonetic encoding
with context-aware rules.
- `postprocess`: Cleans up the raw output by collapsing duplicate digits
and removing leading zeros.
Includes unit tests to verify the correctness of the encoding for
various German words and edge cases, ensuring phonetic similarity and
handling of special character combinations.
This commit introduces a new module `analyze::render` to handle the
rendering of Markdown content produced by the LLM into safe HTML.
The LLM output now includes a custom `==text==` highlighting convention
to indicate sections that require doctor review due to potential
transcription errors or incomplete information. This highlighting is
converted into `<mark>` tags in the final HTML.
To ensure security, all LLM-generated Markdown is first HTML-escaped.
This prevents any malicious HTML or script injection from being executed
in the browser. Only the custom `<mark>` tags are preserved as
functional HTML elements.
The process is as follows:
1. The raw Markdown from the LLM is processed.
2. All HTML special characters (`<`, `>`, `&`, `"`, `'`) are escaped.
3. The `==text==` highlights are replaced with `<mark>text</mark>`.
4. The resulting string is parsed as Markdown by `pulldown-cmark`.
5. The parsed Markdown is converted to HTML, which is then safe to
inject into the Askama template using the `|safe` filter.
The `Cargo.toml` and `Cargo.lock` files have been updated to include the
`pulldown-cmark` dependency. The `document.html` template has been
modified to use a `div` with the class `doc-content` instead of a `pre`
tag, allowing the rendered HTML to be displayed correctly. The
`handle_document_view` function now calls the new `md_to_html` rendering
function.
The system prompt for the transcription Ollama LLM has been updated to
improve robustness and clarity, especially for handling empty or
non-medical transcripts.
The prompt has been refactored to:
- Use English for instructions to improve robustness with smaller LLMs.
- Clearly define the behavior for empty or non-medical transcripts,
requiring an empty string response.
- Reduce the maximum character limit from 120 to 60 to better fit
smartwatch displays.
- Ensure the output is German, adhering to the original language
requirement for the output.
- Add an explicit instruction to correct transcription errors in the
input dictation for the consolidation LLM.
This commit introduces a recovery mechanism to regenerate `oneliner.txt`
files for cases where transcription was successful but the oneliner
generation failed due to a crash between writing the transcript and the
oneliner.
The `transcribe::recovery::regenerate_missing_oneliners` function is
added and called at startup after `scan_and_enqueue`. This function
identifies cases with non-empty transcripts but missing oneliners and
triggers their regeneration.
The `transcribe::worker::update_oneliner` function is refactored to
always regenerate the oneliner from all available transcripts in a case,
ensuring that later recordings can correct earlier ones. This replaces
the previous `ensure_oneliner` logic which only generated the oneliner
if it didn't exist.
New helper functions `cases_needing_oneliner` and
`all_transcripts_joined` are introduced to support these changes.
Comprehensive unit tests are included for the new recovery and file
processing logic.
The LLM client now conditionally adds the `Authorization` header only
when an API key is provided. The `llm_configured` check is updated to
reflect that an API key is not strictly required for Ollama-style
endpoints. A new integration test verifies the functionality against an
Ollama-compatible endpoint without an API key.
Introduce a new "reset" action for bulk operations and a dedicated
endpoint for individual case resets. These features allow administrators
to revert a case to its raw audio state by deleting derived artifacts
like transcripts, analysis input, and documents. The functionality also
includes renaming `.m4a.failed` files back to `.m4a` to re-trigger
transcription.
This commit also:
- Adds a `reset_case_artefacts` helper function to `case_actions.rs`.
- Implements the `handle_reset_case` endpoint.
- Adds the "Reset" button to the UI for administrators.
- Includes unit and integration tests to verify the new functionality
and access control.
The system prompt has been updated to include an explicit instruction
for the LLM to return an empty string if no suitable keyword is found.
This improves robustness by handling cases where the input transcript
might not contain easily extractable medical terms.
Additionally, the `normalize` function's logic for finding the first
non-empty line has been slightly refined to use a more idiomatic Rust
approach by chaining `.lines().find(...)` and `.unwrap_or("")`. This
change improves code clarity and maintainability without altering the
functional behavior of finding the first non-empty line.
The versioning of analysis input and document files
(`analysis_input_v{N}.json`, `document_v{N}.md`) has been removed. All
analysis inputs will now use `analysis_input.json` and generated
documents will use `document.md`.
This simplifies file management, as there's no longer a need to track
and manage multiple versions of these files within a case directory. The
analysis worker will now overwrite the existing `document.md` if it
exists, ensuring that the latest analysis result is always present. The
`version` field has also been removed from `AnalyzeJob` and
`AnalysisInput`.
The case listing on the "My Cases" page is now grouped by local date,
with labels for "Heute", "Gestern", or the ISO date. This improves
readability and organization.
Additionally, the handling of silent recordings has been refined.
Previously, the absence of usable recordings would result in an error.
Now, the analysis worker gracefully handles this by writing a stub
document and skipping the LLM call. This prevents unnecessary errors and
provides a clearer status for silent cases.
The `dictate.sh` script has been updated to simplify the case directory
lookup logic. Instead of iterating through `open` and `done`
subdirectories, it now directly checks for the case ID and ensures it's
not marked as deleted. This simplifies the script and improves
efficiency.
The `server/Cargo.toml` and `server/Cargo.lock` have been updated to
include the `num_threads` dependency and enable additional features for
the `time` crate, which are necessary for proper local time zone
handling.
The `WorkerBusy` type was previously a single `Arc<AtomicBool>` shared
between the analyze and transcribe workers. This commit refactors this
to:
- Introduce `AnalyzeBusy` and `TranscribeBusy` newtype wrappers around
`WorkerBusy` to distinguish between the two flags. This allows
`axum::extract::State` to target them individually.
- Move the `BusyGuard` RAII guard into `src/lib.rs` and make it generic
to work with any `WorkerBusy` instance.
- Update the `main.rs`, `worker.rs`, and `routes/user_web.rs` files to
use the new types and guards, ensuring correct state management for
both pipelines.
- Enhance the transcription recovery scan
(`transcribe::recovery::scan_and_enqueue`) to iterate over user
directories and call `enqueue_pending_for_user` for each, making it
more robust and aligned with the per-user self-healing mechanism.
- Add a check for `transcribe_busy` in the `case_detail.html` template
to correctly display "Transkription läuft…" only when the transcribe
worker is actually active.
The analyze recovery scan has been refactored to iterate through user
directories and enqueue pending analysis jobs more efficiently. The
`WorkerBusy` status has been introduced as an `Arc<AtomicBool>` to track
whether the analyze worker is currently processing a job.
This `WorkerBusy` flag is used in the `analyze::worker::run` function
and managed by a `BusyGuard` RAII struct, ensuring the flag is correctly
set and unset even in case of panics.
The `handle_my_cases` and `handle_case_detail` handlers in `user_web.rs`
now utilize the `WorkerBusy` flag to accurately display the "analyzing"
status and to trigger a self-heal of orphaned analysis inputs when the
worker is idle.
Additionally, the `WorkerBusy` type is now exported from
`server/src/lib.rs` to be accessible by other modules. The test cases
have been updated to include the `WorkerBusy` parameter when spawning
the analyze worker.
Introduce a "Back to Cases" button in the document view and update the
case detail page to conditionally display the "Analyze" button. Also,
add a copy-to-clipboard functionality for the document content.
Introduces a new `/web/cases/bulk` endpoint to handle multiple case
actions simultaneously.
This change adds the following:
- A new `bulk` module for handling bulk operations.
- The `handle_bulk_action` function in `bulk.rs` to dispatch to specific
actions.
- `bulk_analyze` and `bulk_delete` functions to process the actions.
- Updates `Cargo.toml` and `Cargo.lock` to include necessary
dependencies: `form_urlencoded`, `serde_core`, `serde_html_form`, and
`serde_path_to_error`.
- Modified `axum-extra` features to include `form`.
- Adds checks for deleted cases in `collect_pending` for both analysis
and transcription recovery.
- Renames and modifies `handle_close_case` to `handle_analyze_case` in
`case_actions.rs` to better reflect its functionality.
- Adds a new `handle_delete_case` and `handle_undo_delete` to
`case_actions.rs`.
- Updates `my_cases.html` and `case_detail.html` to support bulk
actions, including checkboxes, a bulk action bar, and an "Undo last
delete" feature.
- Modifies `handle_audio` and `scan_cases` in `web.rs` to respect delete
markers.
- Updates `analyze_test.rs` with new request builders for analyze and
delete actions.
Introduce DeleteMarker struct and associated functions for managing
soft-delete markers within case directories. This enables tracking
deleted cases and supports undo functionality.
feat: Add delete marker support to paths
Introduces a `.deleted` file to mark cases as soft-deleted. The marker
contains a `batch` UUID for grouping deletions and a `deleted_at`
timestamp for ordering.
New functions `is_deleted`, `read_delete_marker`, and
`write_delete_marker` are added to manage this marker. The `locate_case`
function is updated to also consider soft-deleted cases as not found.
The distinction between `open/` and `done/` directories for cases has
been removed.
All cases for a user now reside directly under the user's directory
(e.g., `<data_path>/<slug>/<case_id>/`).
This simplifies path management and eliminates redundant directory
traversals.
Key changes include:
- Removed `open/` and `done/` subdirectories in path resolution.
- Introduced a `paths::case_dir` function as a single source of truth
for case directory layout.
- Updated various modules (`recovery`, `auth`, `routes`, `transcribe`,
`tests`) to use the new path structure.
- Adjusted templates to reflect the simplified case status
representation.
Allow overriding the LLM system prompt via the `LLM_SYSTEM_PROMPT`
environment variable. This provides flexibility for customizing LLM
behavior without code changes. A default prompt is used if the
environment variable is unset or empty.
Introduce a new endpoint for reanalyzing cases. This feature allows
administrators
to rebuild analysis input files and enqueue new analysis jobs. The
system will
then generate a new document version, which will become the displayed
document.
Extract the logic for building `AnalysisInput` into a new function
`build_analysis_input`. This function encapsulates the process of
collecting recordings, reading transcripts, and determining the last
recording modification time.
This change improves code organization and reusability by centralizing
the input building logic.
Updates the `LLM_URL` in the project plan and `.env.example` to the
correct value.
Also, adds a warning in `main.rs` to inform the user if the LLM is not
configured, which effectively disables the 'Fall abschließen' feature.
This commit introduces a comprehensive suite of integration tests for
the case closing functionality.
These tests cover various scenarios including:
- Successful case closure with proper file writing and redirection.
- Handling of invalid inputs like non-UUIDs or cases without recordings.
- Authentication and authorization checks, ensuring only the case owner
can close it.
- Error handling for attempting to close a case twice or when the LLM
service is unavailable.
- Verification of the analysis worker's interaction with the mocked LLM.
- Testing the recovery mechanism for re-enqueuing pending analyses.
- Validating the document view displays the latest version.
Introduce a `DocumentTemplate` struct for rendering the document view,
replacing inline HTML.
Introduce a `CaseFlags` struct to encapsulate the logic for determining
various UI states of a case (e.g., `has_document`, `analyzing`,
`can_close`, `llm_missing`).
Update `handle_case_detail` to use `compute_flags` for populating the
`CaseDetailTemplate`.
Update `scan_user_cases` to compute `analyzing` and `has_document` flags
for the `UserCaseView`.
Add new template logic in `case_detail.html` to display actions based on
`CaseFlags`.
Add new template logic in `my_cases.html` to display `analyzing` and
`done-doc` labels for cases.
Create a new `document.html` template.
Add a helper function `any_document_exists` to check for the presence of
a document file.
Adds API endpoints for closing a case, which triggers an analysis job,
and for viewing the generated document.
This also includes:
- New `AppError` variants: `Conflict` and `ServiceUnavailable`.
- Helper functions for file handling, date parsing, and document
searching.
- Unit tests for helper functions.
Introduces a new channel for the analyze pipeline, including its sender
in the `AppState`.
Also updates the test router to create both transcribe and analyze
channels.
This commit introduces the `analyze` module, which orchestrates
communication with Large Language Models (LLMs) for text summarization
and analysis.
Key components include:
- `llm.rs`: Contains the `LlmError` enum and the `chat_once` function
for interacting with OpenAI-compatible LLM APIs.
- `prompt.rs`: Defines the system prompt and logic for rendering user
content from analysis inputs.
- `recovery.rs`: Implements logic to scan for and re-enqueue pending
analysis jobs upon server startup.
- `worker.rs`: The core worker that consumes analysis jobs, processes
them with the LLM, and writes the output.
- `mod.rs`: Defines `AnalyzeJob` and associated channel types for
inter-component communication.
This module enables the server to process dictated recordings, summarize
them using an LLM, and store the results.
Introduced a `test_default` method to `Config` to provide sane defaults
for integration tests. This refactors several test files to use this new
method, reducing boilerplate and improving maintainability.
Added `LLM_TIMEOUT_SECONDS` to the environment variables and `Config`
struct.
This commit introduces several improvements to the `hash-password`
binary:
- **New command-line interface:** Supports printing hashes directly,
hashing via prompts, or patching `web_password` in `users.toml` in
place.
- **Atomic file patching:** Utilizes `toml_edit` to preserve comments,
ordering, and formatting, and writes changes atomically via a
temporary file and rename operation.
- **Robust argument parsing:** Handles various command-line scenarios
and provides informative usage messages.
- **Error handling:** Improves error reporting for failed operations.
- **Added unit tests:** Includes tests for patching behavior and error
conditions.
This commit introduces the foundation for web-based authentication and
user interface. It includes:
- **Session Management**: Securely handling user sessions using
cryptographically generated tokens, cookies with appropriate security
flags, and server-side storage.
- **Login/Logout Functionality**: Endpoints for users to log in with
their credentials and log out, clearing their session.
- **User Interface**: Basic templates for the login page, a list of
cases, and a detailed view of a single case, allowing users to
navigate and view their data.
- **Authorization**: An `AuthenticatedWebUser` extractor to ensure only
logged-in users can access protected web routes.
- **Configuration**: Added a `cookie_secure` option to the configuration
to control the `Secure` flag on session cookies, useful for local
development.
- **Dependencies**: Added necessary dependencies for password hashing,
time handling, and TOML file manipulation.
- **Helper Binary**: Included a `hash-password` binary to simplify
generating bcrypt hashes for user passwords.
Add a new `failed` field to `RecordingView` and corresponding logic in
the worker.
When FFmpeg remuxing or Whisper transcription fails, the worker now
renames the
original `.m4a` file to `.m4a.failed`. This prevents the recovery scan
from
reprocessing these files and signals to the UI that transcription
failed.
The UI displays a message indicating failure and suggests renaming the
file
back to `.m4a` to retry. A new integration test
`worker_renames_audio_to_failed_on_whisper_error`
is included to verify this behavior.
This commit introduces the capability to configure user-specific Whisper
settings, including language, hotwords, and initial prompts. These
settings are stored in `users.toml` and are passed to the Whisper
service for more tailored transcription results.
New unit tests have been added to verify that the `transcribe` function
correctly forwards these optional settings to the Whisper client and
omits them when they are not provided.
Additionally, a new directory `tests/fixtures/dictations` has been
created to store M4A audio files, their expected transcripts, and
associated hotword files. This serves as a regression corpus for the
Whisper pipeline. A README file explains the structure and conventions
for adding new fixtures. Several new fixtures have been added, covering
various medical domains and transcription scenarios.
Brummel