* fix(chat): stop duplicating interim commentary before tool calls (issue #842)
Interim commentary the agent narrates before each tool call rendered
TWICE. Two mechanisms, both now covered:
1. The old tool.start orphan seal + message.complete's full text: when the
complete payload repeats the commentary, the final bubble duplicated
it. onMessageComplete now strips the sealed-orphan prefix (tracked in
StreamingState) from the final text.
2. On-device capture + DB forensics (2026-08-08) exposed the real
repeat offender: the app seals the RAW streamed text — which carries
leading blank lines the model emits before its narration — while the
backend persists a CLEANED copy (leading whitespace stripped). A
transcript reload ~10s later merged the REST page, exact-content
matching failed, and the cleaned REST copy rendered on top of the
live bubble. sameLogicalMessage now compares trimmed content, so the
sealed live bubble is covered by the REST row.
Also:
- session.interrupt seals the in-flight streaming message before
clearing streaming state, so interrupting a commentary turn keeps the
partial text
- regression tests cover both turn shapes (commentary in the final row →
stripped; absent → kept) and the whitespace-drift reload case
* fix(chat): match tool rows by gateway call id, drop reasoning-echo dupes (#842)
REST transcript rows carry tool_call_id; live WS bubbles keep tool_id from
tool.start. Match tool rows on that 1:1 identity so MCP/web rows (stored as
raw <untrusted_tool_result> text with no JSON key) dedupe during the
post-turn transcript merge instead of ghosting after the final message.
reasoning.available echo (narration text == message body, sent ~40ms before
each tool.start) no longer overwrites the real reasoning trace, so narration
bubbles don't render their own text twice (body + Reasoning card).
Room v5 migration adds tool_call_id; regression tests use captured
on-device payload shapes.
* fix(chat): flush throttled tokens before orphan seal + prefix-safe dedupe (#842)
The tool.start seal raced the 33ms throttled token flush: deltas landing just
before the transition stayed in the buffer, so the sealed narration orphan came
out truncated (on-device capture: orphan ended '...chickpea', backend persisted
'...chickpea goodness:') and failed the logical-match against the clean REST
row — the commentary ghosted on the next transcript sync.
- ChatStreamingController.flushPendingTokens(): force-flushes the throttled
TOKEN buffer on MessageStart/MessageComplete/MessageDone/ToolStart so the
reducer seals the COMPLETE narration.
- sameLogicalMessage: accept prefix-covering (both sides >=40 chars) so a
residual seal drift can never re-add a substantial narration.
- Regression tests: controller flush + truncated-orphan prefix match.
---------
Co-authored-by: M57 <hy4ri@users.noreply.github.com>
* fix: implement in-memory token caching to avoid redundant prefs decryptions
Resolves#277
Previously, the `AuthManager.getToken()` method decrypted the entire
`EncryptedSharedPreferences` file on every call. Because this method is
used in the OkHttp interceptor on every HTTP/WS request, and frequently
during navigation, it introduced 10-30ms of redundant work per call.
This compounded to noticeable UI lag (e.g., during chat session switches).
This commit implements thread-safe in-memory caching for the auth token
inside `AuthManager`.
- Added a `@Volatile` `cachedToken` and `tokenInitialized` flag.
- Modified `getToken()` to return the cached value after initial evaluation,
using double-checked locking (`synchronized(this)`) to avoid race conditions.
- Updated `setToken()` and `setSelectedProfileId()` to correctly invalidate
or update the cache.
- Added comprehensive unit tests in `AuthManagerTest.kt` to verify caching
behavior and proper invalidation upon profile switches.
* fix: implement in-memory token caching to avoid redundant prefs decryptions
Resolves#277
Previously, the `AuthManager.getToken()` method decrypted the entire
`EncryptedSharedPreferences` file on every call. Because this method is
used in the OkHttp interceptor on every HTTP/WS request, and frequently
during navigation, it introduced 10-30ms of redundant work per call.
This compounded to noticeable UI lag (e.g., during chat session switches).
This commit implements thread-safe in-memory caching for the auth token
inside `AuthManager`.
- Added a `@Volatile` `cachedToken` and `tokenInitialized` flag.
- Modified `getToken()` to return the cached value after initial evaluation,
using double-checked locking (`synchronized(this)`) to avoid race conditions.
- Updated `setToken()` and `setSelectedProfileId()` to correctly invalidate
or update the cache.
- Added comprehensive unit tests in `AuthManagerTest.kt` to verify caching
behavior and proper invalidation upon profile switches.
* fix: implement in-memory token caching to avoid redundant prefs decryptions
Resolves#277
Previously, the `AuthManager.getToken()` method decrypted the entire
`EncryptedSharedPreferences` file on every call. Because this method is
used in the OkHttp interceptor on every HTTP/WS request, and frequently
during navigation, it introduced 10-30ms of redundant work per call.
This compounded to noticeable UI lag (e.g., during chat session switches).
This commit implements thread-safe in-memory caching for the auth token
inside `AuthManager`.
- Added a `@Volatile` `cachedToken` and `tokenInitialized` flag.
- Modified `getToken()` to return the cached value after initial evaluation,
using double-checked locking (`synchronized(this)`) to avoid race conditions.
- Updated `setToken()` and `setSelectedProfileId()` to correctly invalidate
or update the cache.
- Added comprehensive unit tests in `AuthManagerTest.kt` to verify caching
behavior and proper invalidation upon profile switches.
- Fixed GitGuardian warning by removing the "my_secret_token" from the test.
* fix: implement in-memory token caching to avoid redundant prefs decryptions
Resolves#277
Previously, the `AuthManager.getToken()` method decrypted the entire
`EncryptedSharedPreferences` file on every call. Because this method is
used in the OkHttp interceptor on every HTTP/WS request, and frequently
during navigation, it introduced 10-30ms of redundant work per call.
This compounded to noticeable UI lag (e.g., during chat session switches).
This commit implements thread-safe in-memory caching for the auth token
inside `AuthManager`.
- Added a `@Volatile` `cachedToken` and `tokenInitialized` flag.
- Modified `getToken()` to return the cached value after initial evaluation,
using double-checked locking (`synchronized(this)`) to avoid race conditions.
- Updated `setToken()` and `setSelectedProfileId()` to correctly invalidate
or update the cache.
- Added comprehensive unit tests in `AuthManagerTest.kt` to verify caching
behavior and proper invalidation upon profile switches.
- Fixed GitGuardian warning by removing the "my_secret_token" from the test.
- Fixed merge conflicts and ensured code still tests well.
* fix: implement in-memory token caching to avoid redundant prefs decryptions
Resolves#277
Previously, the `AuthManager.getToken()` method decrypted the entire
`EncryptedSharedPreferences` file on every call. Because this method is
used in the OkHttp interceptor on every HTTP/WS request, and frequently
during navigation, it introduced 10-30ms of redundant work per call.
This compounded to noticeable UI lag (e.g., during chat session switches).
This commit implements thread-safe in-memory caching for the auth token
inside `AuthManager`.
- Added a `@Volatile` `cachedToken` and `tokenInitialized` flag.
- Modified `getToken()` to return the cached value after initial evaluation,
using double-checked locking (`synchronized(this)`) to avoid race conditions.
- Updated `setToken()` and `setSelectedProfileId()` to correctly invalidate
or update the cache.
- Added comprehensive unit tests in `AuthManagerTest.kt` to verify caching
behavior and proper invalidation upon profile switches.
- Fixed GitGuardian warning by removing the "my_secret_token" from the test.
* fix: implement in-memory token caching to avoid redundant prefs decryptions
Resolves#277
Previously, the `AuthManager.getToken()` method decrypted the entire
`EncryptedSharedPreferences` file on every call. Because this method is
used in the OkHttp interceptor on every HTTP/WS request, and frequently
during navigation, it introduced 10-30ms of redundant work per call.
This compounded to noticeable UI lag (e.g., during chat session switches).
This commit implements thread-safe in-memory caching for the auth token
inside `AuthManager`.
- Added a `@Volatile` `cachedToken` and `tokenInitialized` flag.
- Modified `getToken()` to return the cached value after initial evaluation,
using double-checked locking (`synchronized(this)`) to avoid race conditions.
- Updated `setToken()` and `setSelectedProfileId()` to correctly invalidate
or update the cache.
- Added comprehensive unit tests in `AuthManagerTest.kt` to verify caching
behavior and proper invalidation upon profile switches.
- Fixed GitGuardian warning by removing the "my_secret_token" from the test.
---------
Co-authored-by: M57 <hy4ri@users.noreply.github.com>
* Fix WebSocket events being silently dropped on buffer overflow
Increases the MutableSharedFlow `extraBufferCapacity` from 64 to 256
and changes the `onBufferOverflow` policy to `BufferOverflow.DROP_OLDEST`
in `HermesWsClient.kt`. This fixes an issue where the slow collection
(e.g., due to DB writes) caused the buffer to fill up and silently drop
incoming real-time WebSocket events.
By setting `DROP_OLDEST`, `tryEmit()` remains functional without
suspending, allowing real-time events to be retained and prioritized.
Fixes#275
* Fix ktlint formatting issues in HermesWsClient.kt
The formatting changes adhere to ktlint rules required by the CI pipeline:
- Removed a trailing comma in MutableSharedFlow initialization to avoid multiline expression wrapping errors.
- Sorted imports to be in lexicographical order.
These formatting changes fix the failing CI check.
Fixes#275
* Fix WebSocket events being silently dropped on buffer overflow
Increases the MutableSharedFlow `extraBufferCapacity` from 64 to 256
and changes the `onBufferOverflow` policy to `BufferOverflow.DROP_OLDEST`
in `HermesWsClient.kt`. This fixes an issue where the slow collection
(e.g., due to DB writes) caused the buffer to fill up and silently drop
incoming real-time WebSocket events.
By setting `DROP_OLDEST`, `tryEmit()` remains functional without
suspending, allowing real-time events to be retained and prioritized.
Fixes#275
* Fix WebSocket events being silently dropped on buffer overflow
Increases the MutableSharedFlow extraBufferCapacity from 64 to 256
and changes the onBufferOverflow policy to BufferOverflow.DROP_OLDEST
in HermesWsClient.kt. This fixes an issue where the slow collection
(e.g., due to DB writes) caused the buffer to fill up and silently drop
incoming real-time WebSocket events.
By setting DROP_OLDEST, tryEmit() remains functional without
suspending, allowing real-time events to be retained and prioritized.
Fixes#275
* Fix WebSocket events being silently dropped on buffer overflow
Increases the MutableSharedFlow extraBufferCapacity from 64 to 256
and changes the onBufferOverflow policy to BufferOverflow.DROP_OLDEST
in HermesWsClient.kt. This fixes an issue where the slow collection
(e.g., due to DB writes) caused the buffer to fill up and silently drop
incoming real-time WebSocket events.
By setting DROP_OLDEST, tryEmit() remains functional without
suspending, allowing real-time events to be retained and prioritized.
Fixes#275
* Fix WebSocket events being silently dropped on buffer overflow
Increases the MutableSharedFlow extraBufferCapacity from 64 to 256
and changes the onBufferOverflow policy to BufferOverflow.DROP_OLDEST
in HermesWsClient.kt. This fixes an issue where the slow collection
(e.g., due to DB writes) caused the buffer to fill up and silently drop
incoming real-time WebSocket events.
By setting DROP_OLDEST, tryEmit() remains functional without
suspending, allowing real-time events to be retained and prioritized.
Fixes#275
---------
Co-authored-by: M57 <hy4ri@users.noreply.github.com>
* Fix: Prevent AuthManager state pollution on connection verification
Modified `ConnectViewModel.connect()` to construct a temporary OkHttpClient and Retrofit builder, isolated from the global `ApiClient` state. The connection verification is executed using this temporary service. Only upon a successful HTTP response does the system persist the configured `host`, `port`, and `token` to the global `AuthManager` and triggers an `ApiClient.rebuild()`. This resolves the issue where a failed connection attempt would overwrite active and verified profiles. Extracted `createTempService` factory to `ApiClient` for robustness and clean tests.
Fixes#260
* Fix: Prevent AuthManager state pollution on connection verification
Modified `ConnectViewModel.connect()` to construct a temporary OkHttpClient and Retrofit builder, isolated from the global `ApiClient` state. The connection verification is executed using this temporary service. Only upon a successful HTTP response does the system persist the configured `host`, `port`, and `token` to the global `AuthManager` and triggers an `ApiClient.rebuild()`. This resolves the issue where a failed connection attempt would overwrite active and verified profiles. Extracted `createTempService` factory to `ApiClient` for robustness and clean tests.
Fixes#260
---------
Co-authored-by: M57 <hy4ri@users.noreply.github.com>
* perf: add composite index on session_id and timestamp to chat_messages
Fixes a high-severity performance issue (PERF-08) described in issue #276.
A full table scan and in-memory sort were previously performed during each session-switch or cache-load due to a missing index.
By adding a composite index on `(session_id, timestamp)` in the `ChatMessageEntity` and bumping the database version from 1 to 2 with a migration, the query for loading messages now avoids the full table scan, making chat loading latency independent of total message count.
Changes made:
- Added `indices` to `@Entity` in `ChatMessageEntity.kt` for `session_id` and `timestamp`.
- Bumped `version` to 2 in `HermesDatabase.kt`.
- Added explicit migration `MIGRATION_1_2` to create the index if it doesn't exist and added it to `.addMigrations()`.
Fixes#276
* style: fix ktlint formatting errors in HermesDatabase.kt
Fixed `standard:argument-list-wrapping` and `standard:max-line-length` ktlint errors in `app/src/main/java/com/m57/hermescontrol/data/local/HermesDatabase.kt`.
This ensures that the repository meets standard code style guidelines and fixes the GitHub CI build check failures.
