* implement native embedder job queue
* persist embedding progress across renders
* add development worker timeouts
* change to static method
* native reranker
* remove useless return
* lint
* simplify
* make embedding worker timeout value configurable by admin
* add event emission for missing data
* lint
* remove onProgress callback argument
* make rerank to rerankDirect
* persists progress state across app reloads
* remove chunk level progress reporting
* remove unuse dvariable
* make NATIVE_RERANKING_WORKER_TIMEOUT user configurable
* remove dead code
* scope embedding progress per-user and clear stale state on SSE reconnect
* lint
* revert vector databases and embedding engines to call their original methods
* simplify rerank
* simplify progress fetching by removing updateProgressFromApi
* remove duplicate jsdoc
* replace sessionStorage persistence with server-side history replay for embedding progress
* fix old comment
* fix: ignore premature SSE all_complete when embedding hasn't started yet
The SSE connection opens before the embedding API call fires, so the
server sees no buffered history and immediately sends all_complete.
Firefox dispatches this eagerly enough that it closes the EventSource
before real progress events arrive, causing the progress UI to clear
and fall back to the loading spinner. Chrome's EventSource timing
masks the race.
Track slugs where startEmbedding was called but no real progress event
has arrived yet via awaitingProgressRef. Ignore the first all_complete
for those slugs and keep the connection open for the real events.
* reduce duplication with progress emissions
* remove dead code
* refactor: streamline embedding progress handling
Removed unnecessary tracking of slugs for premature all_complete events in the EmbeddingProgressProvider. Updated the server-side logic to avoid sending all_complete when no embedding is in progress, allowing the connection to remain open for real events. Adjusted the embedding initiation flow to ensure the server processes the job before the SSE connection opens, improving the reliability of progress updates.
* fix stale comment
* remove unused function
* fix event emissions for document creation failure
* refactor: move Reranking Worker Idle Timeout input to LanceDBOptions component
Extracted the Reranking Worker Idle Timeout input from GeneralEmbeddingPreference and integrated it into the LanceDBOptions component. This change enhances modularity and maintains a cleaner structure for the settings interface.
* lint
* remove unused hadHistory vars
* refactor workspace directory by hoisting component and converting into functions
* moved EmbeddingProgressProvider to wrap Document Manager Modal
* refactor embed progress SSE connection to use fetchEventSource instead of native EventSource API.
* refactor message handlng into a function and reduce duplication
* refactor: utilize writeResponseChunk for event emissions in document embedding progress SSE
* refactor: explicit in-proc embedding and rerank methods that are called by workers instead of process.send checks
* Abstract EmbeddingProgressBus and Worker Queue into modules
* remove error and toast messages on embed process result
* use safeJsonParse
* add chunk-level progress events with per-document progress bar in UI
* remove unused parameter
* rename all worker timeout references to use ttl | remove ttl updating from UI
* refactor: pass embedding context through job payload instead of global state
* lint
* add graceful shutdown for workers
* apply figma styles
* refactor embedding worker to use bree
* use existing WorkerQueue class as the management layer for jobs
* lint
* revert all reranking worker changes back to master state
Removes the reranking worker queue, rerankViaWorker/rerankInProcess
renames, and NATIVE_RERANKING_WORKER_TTL config so this branch
only contains the embedding worker job queue feature.
* remove breeManaged flag — WorkerQueue always spawns via Bree
* fix prompt embedding bug
* have embedTextInput call embedChunksInProcess
* add message field to `process.send()`
* remove nullish check and error throw
* remove bespoke graceful shutdown logix
* add spawnWorker method and asbtract redudant flows into helper methods
* remove unneeded comment
* remove recomputation of TTL value
* frontend cleanup and refactor
* wip on backend refactor
* backend overhaul
* small lint
* second pass
* add logging, update endpoint
* simple refactor
* add reporting to all embedder providers
* fix styles
---------
Co-authored-by: Timothy Carambat <rambat1010@gmail.com>
* fix(lemonade): throw on embedding failures instead of returning empty vectors
* use class logger
---------
Co-authored-by: Timothy Carambat <rambat1010@gmail.com>
* add API key param to Lemonade LLM Provider and Embedding Provider
* add LEMONADE_LLM_API_KEY to .env.example
* add api key to aibitat provider
* fix api key from being sent to frontend
* fix tooltip id
* add null fallback for `apiKey`
* remove console log
* add missing api keys
---------
Co-authored-by: Timothy Carambat <rambat1010@gmail.com>
* add ask to run prompt for tools
* border-none on buttons
* translations
* linting
* i18n (#5263)
* extend approve/deny requests to telegram
* break up handler
* Add User-Agent header for Anthropic API calls
Passes User-Agent: AnythingLLM/{version} to the Anthropic SDK
so Anthropic can identify traffic from AnythingLLM.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* remove test, simplify header default
* unset change to spread
---------
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Timothy Carambat <rambat1010@gmail.com>
* Beta Intelligent Tooling
todo: Agent Skill banner warning when tool # is high or % of content window?
* forgot files
* add UI controls and maxToolCallStack setting
* update docs link
* ISS i18n (#5237)
i18n
* Add automatic chat mode with native tool calling support
Introduces a new automatic chat mode (now the default) that automatically invokes tools when the provider supports native tool calling. Conditionally shows/hides the @agent command based on whether native tooling is available.
- Add supportsNativeToolCalling() to AI providers (OpenAI, Anthropic, Azure always support; others opt-in via ENV)
- Update all locale translations with new mode descriptions
- Enhance translator to preserve Trans component tags
- Remove deprecated ability tags UI
* rebase translations
* WIP on image attachments. Supports initial image attachment + subsequent attachments
* persist images
* Image attachments and updates for providers
* desktop pre-change
* always show command on failure
* add back gemini streaming detection
* move provider native tooling flag to Provider func
* whoops - forgot to delete
* strip "@agent" from prompts to prevent weird replies
* translations for automatic-mode (#5145)
* translations for automatic-mode
* rebase
* translations
* lint
* fix dead translations
* change default for now to chat mode just for rollout
* remove pfp for workspace
* passthrough workspace for showAgentCommand detection and rendering
* Agent API automatic mode support
* ephemeral attachments passthrough
* support reading of pinned documents in agent context
* feat(agents): Add Perplexity Search API as web search provider
Adds Perplexity as a search provider for the agent web-browsing plugin,
using the Perplexity Search API (POST /search) which returns raw ranked
web results — distinct from the existing Perplexity LLM integration.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
* chore: replace docs.perplexity.ai with console.perplexity.ai
* chore: replace docs.perplexity.ai with console.perplexity.ai
---------
Co-authored-by: kesku <kesku@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
Co-authored-by: Timothy Carambat <rambat1010@gmail.com>
Validate all ZIP entries before extraction in importCommunityItemFromUrl()
to prevent path traversal attacks (CWE-22). Malicious ZIP entries with
paths like "../../" could write files outside the intended plugin folder.
Requires admin privileges and explicit opt-in to unverified hub downloads.
GHSA-rh66-4w74-cf4m
Previously, suspended users could continue using browser extension
endpoints if they had created an API key before suspension. The normal
JWT session path blocked suspended users, but the browser extension
middleware did not.
Changes:
- Add suspension and user existence checks to validBrowserExtensionApiKey
- Delete browser extension API keys when a user is deleted
- Add deleteAllForUser method to BrowserExtensionApiKey model
GHSA-7754-8jcc-2rg3
Replace string concatenation with parameterized queries in all database
connectors to prevent SQL injection through LLM-generated table names.
Changes:
- PostgreSQL: Use $1, $2 placeholders with pg client parameterization
- MySQL: Use ? placeholders with mysql2 execute() prepared statements
- MSSQL: Use @p0 placeholders with request.input() parameterization
- Update handlers to support parameterized query objects
- Add formatQueryForDisplay() for logging parameterized queries
Security: Mitigates potential SQL injection when LLM passes unsanitized
user input as table_name parameter to getTableSchemaSql/getTablesSql.
GHSA-jwjx-mw2p-5wc7
