mirror of
https://github.com/LibreChat-AI/librechat.ai.git
synced 2026-03-27 10:48:32 +07:00
888a9f1e8d
- Introduced `context` capability for agents endpoint, enabling file uploads in chat. - Enhanced `fileConfig` with OCR, text parsing, STT separation, and `fileTokenLimit` parameter. - Added `transactions` object structure for token usage tracking. - Included `jinaApiUrl` in `webSearch` configuration for custom Jina instances. - Updated documentation for `fileConfig` to reflect new fields and processing methods.
89 lines
6.8 KiB
Plaintext
89 lines
6.8 KiB
Plaintext
- Added `azure_mistral_ocr` strategy option for OCR configuration
|
|
- Enables use of Mistral OCR models deployed on Azure AI Foundry
|
|
- Supports Azure-specific deployment configurations with custom endpoints
|
|
- Requires Azure deployment name to be specified in `mistralModel` field
|
|
- Compatible with Mistral OCR 2503 model available on Azure AI Foundry
|
|
- See [OCR Configuration](/docs/configuration/librechat_yaml/object_structure/ocr) for details
|
|
|
|
- Added `vertexai_mistral_ocr` strategy option for OCR configuration
|
|
- Enables use of Mistral OCR models deployed on Google Cloud Vertex AI
|
|
- Supports multiple methods for providing Google Cloud service account credentials:
|
|
- File path to service account JSON
|
|
- URL to fetch service account JSON
|
|
- Base64 encoded service account JSON
|
|
- Raw JSON string
|
|
- Automatically extracts project ID from service account credentials
|
|
- Uses JWT authentication for secure access to Vertex AI endpoints
|
|
- See [OCR Configuration](/docs/configuration/librechat_yaml/object_structure/ocr) for details
|
|
|
|
- Enhanced MCP (Model Context Protocol) server management with comprehensive new features:
|
|
- Added `serverInstructions` to MCP Servers object structure - controls whether server instructions are included in agent context when using MCP Server tools
|
|
- Added user placeholder variables support:
|
|
- `{{LIBRECHAT_USER_ID}}` - replaced with the current user's ID, enabling multi-user support
|
|
- `{{LIBRECHAT_USER_*}}` - dynamic user field placeholders (replace `*` with the UPPERCASE version of any allowed field)
|
|
- Supported in `headers` and `url` fields
|
|
- Added `customUserVars` for per-user credentials:
|
|
- Allows users to securely provide their own API keys and credentials for MCP servers
|
|
- Configurable through UI via settings icon in tool selection or MCP Settings panel
|
|
- Ensures authentication details remain private in multi-user environments
|
|
- Added `mcpServers` to the `interface` configuration:
|
|
- Centralizes UI-related settings for Model Context Protocol servers
|
|
- Includes `placeholder` option to customize text shown in server selection dropdown when no MCP is selected
|
|
- Enhanced server management with connection status tracking and OAuth support:
|
|
- Added dynamic status icons showing server state (connected, disconnected, OAuth required, error, initializing)
|
|
- Implemented server (re)initialization with OAuth flow support through dropdown and settings panel
|
|
- Added set/unset state tracking for `customUserVars`
|
|
- Unified server listing across MCPSelect and MCPPanel with consistent status display
|
|
- See [MCP Servers Object Structure](/docs/configuration/librechat_yaml/object_structure/mcp_servers) for complete details
|
|
|
|
- Added user placeholder variables support to Custom Endpoint Headers:
|
|
- Users can now use the same placeholder variables (`{{LIBRECHAT_USER_ID}}`, `{{LIBRECHAT_USER_EMAIL}}`, etc.) in custom endpoint headers
|
|
- See [Custom Endpoint Object Structure - Headers](/docs/configuration/librechat_yaml/object_structure/custom_endpoint#headers) for details
|
|
|
|
- Added `temporaryChatRetention` to interface configuration:
|
|
- Configures the retention period for temporary chats in hours (min: 1, max: 8760)
|
|
- Default retention is 720 hours (30 days) if not specified
|
|
- Can be set via environment variable `TEMP_CHAT_RETENTION_HOURS` or in `librechat.yaml`
|
|
- Applies to newly created temporary chats only; existing chats retain their original expiration
|
|
- See [Interface Object Structure - temporaryChatRetention](/docs/configuration/librechat_yaml/object_structure/interface#temporarychatretention) for details
|
|
|
|
- Added `clientImageResize` to `fileConfig` configuration:
|
|
- Enables automatic client-side image resizing before upload to prevent failures and optimize performance
|
|
- Configurable maximum dimensions (`maxWidth`/`maxHeight`) with aspect ratio preservation
|
|
- Adjustable JPEG/WebP compression quality (0.1-1.0) for file size optimization
|
|
- Supports output format selection between JPEG and WebP
|
|
- Helps reduce bandwidth usage and improve upload speeds
|
|
- See [File Config Object Structure - clientImageResize](/docs/configuration/librechat_yaml/object_structure/file_config#clientimageresize) for details
|
|
|
|
- Enhanced `webSearch` configuration with comprehensive Firecrawl scraper options:
|
|
- Added detailed configuration options for Firecrawl scraper including `formats`, `includeTags`, `excludeTags`, `headers`, `waitFor`, `timeout`, `maxAge`, `mobile`, `skipTlsVerification`, `parsePDF`, `removeBase64Images`, `blockAds`, `storeInCache`, `zeroDataRetention`, `onlyMainContent`, `location`, and `changeTrackingOptions`
|
|
- See [Web Search Configuration](/docs/configuration/librechat_yaml/object_structure/web_search) for details
|
|
|
|
- Added title generation configuration with advanced options for all endpoints:
|
|
- Added `titleMethod` to control title generation strategy:
|
|
- `"completion"` (new default) - Uses standard completion API for broader LLM compatibility
|
|
- `"structured"` or `"functions"` (legacy) - Uses structured output/function calling (not all providers support this)
|
|
- Added `titlePrompt` for customizing the main title generation prompt:
|
|
- Must include `{convo}` placeholder for conversation content
|
|
- Allows full control over how the AI generates titles
|
|
- Added `titlePromptTemplate` for customizing conversation formatting:
|
|
- Must include `{input}` and `{output}` placeholders
|
|
- Default: `"User: {input}\nAI: {output}"`
|
|
- Controls how the conversation is formatted when inserted into `titlePrompt`
|
|
- Added `titleEndpoint` to use alternative endpoints for title generation:
|
|
- Supported values: `openAI`, `azureOpenAI`, `google`, `anthropic`, `bedrock`
|
|
- For custom endpoints, use the exact custom endpoint name as defined in configuration
|
|
- Allows using a different, potentially cheaper model/endpoint for titles
|
|
- All title configuration options are now available as [Shared Endpoint Settings](/docs/configuration/librechat_yaml/object_structure/shared_endpoint_settings)
|
|
- Can be configured globally using the `all:` endpoint definition
|
|
- See individual endpoint documentation for specific examples
|
|
|
|
- Added `fileSearch` to interface configuration:
|
|
- Enables or disables the file search button in the chat interface
|
|
- Similar to the `webSearch` option but controls file search (RAG API usage via tool)
|
|
- Does not disable the [Agents File Search Capability](/docs/features/agents#file-search)
|
|
- Default value is `true` if not specified
|
|
- See [Interface Object Structure - fileSearch](/docs/configuration/librechat_yaml/object_structure/interface#filesearch) for details
|
|
|
|
- Improved [Model Specs documentation](/docs/configuration/librechat_yaml/object_structure/model_specs) with parameter support updates:
|
|
- Added support for `disableStreaming`, `thinking`, `thinkingBudget`, `web_search`, and other parameters |