feat: support reasoning effort and budget

[urmauur]

Describe Your Changes

This pull request introduces enhanced support for "reasoning" capabilities in models, focusing on both backend and frontend integration. The changes include adding new properties to model configurations, updating capability detection logic, and improving the user interface to allow toggling and configuring reasoning effort for models that support it. Additionally, the capability detection logic is generalized and refactored for better extensibility.

Backend enhancements for reasoning capabilities:

• Added reasoning_budget and chat_template_kwargs properties to the LlamacppConfig type, and ensured these are passed as command-line arguments when starting the model in llamacpp_extension. This enables dynamic configuration of reasoning features for supported models. [1] [2] [3]
• Refactored capability detection by replacing the isToolSupported method with a more generic isModelCapabilitySupported, allowing detection of various capabilities (including reasoning and reasoning_effort) for models. This change is reflected across engine, service, and provider layers. [1] [2] [3] [4] [5] [6]

Frontend improvements for reasoning configuration:

• Updated the UI in ChatInput.tsx to allow users to toggle the reasoning budget and select reasoning effort (low/medium/high) for models with reasoning capability. This includes new popover and icon components, and ensures the model is restarted with new settings when changed. [1] [2]
• Changed the icon for reasoning capability from IconAtom to IconBrain for better clarity in both capability lists and chat input controls. [1] [2] [3]

Capability and model configuration logic:

• Enhanced provider and model configuration logic to correctly detect and assign capabilities, including reasoning and reasoning_effort, during provider/model loading. This ensures models are properly annotated and configured for use in the UI and backend. [1] [2] [3] [4]
• Added REASONING to the ModelCapabilities enum for consistent capability handling throughout the codebase.

Fixes Issues

• Closes #
• Closes #

Self Checklist

• Added relevant comments, esp in complex areas
• Updated docs (for bug fixes / features)
• Created issues for follow-up changes or refactoring needed

---

Important

Add reasoning capabilities to models, updating backend configurations and frontend UI for enhanced reasoning support.

• Backend Enhancements:
  • Add reasoning_budget and chat_template_kwargs to LlamacppConfig in index.ts.
  • Pass these as command-line arguments in llamacpp_extension.
  • Replace isToolSupported with isModelCapabilitySupported in AIEngine.ts and index.ts.
• Frontend Improvements:
  • Update ChatInput.tsx to toggle reasoning budget and select reasoning effort.
  • Change reasoning icon from IconAtom to IconBrain in Capabilities.tsx and ChatInput.tsx.
• Model Configuration:
  • Enhance logic to detect and assign capabilities, including reasoning, in providers.ts.
  • Add REASONING to ModelCapabilities enum in models.ts.

^{This description was created by for dce5af4. You can customize this summary. It will automatically update as commits are pushed.}

[github-actions]

Barecheck - Code coverage report

Total: 35.68%

Your code coverage diff: -0.13% ▾

Uncovered files and lines

File	Lines
web-app/src/containers/Capabilities.tsx	21, 23-26, 28-40, 42-50, 52-61, 63, 65-66, 68
web-app/src/containers/ChatInput.tsx	97-99, 122-128, 130-132, 148-149, 190-191, 213-215, 222-223, 226-229, 232-242, 245, 247-249, 255, 257, 259-260, 262-266, 269-270, 273, 275-278, 280-284, 286-298, 300-310, 312-321, 323-326, 329-330, 332-335, 338-339, 342-346, 349-350, 352-355, 358-360, 363-365, 368-371, 373-374, 376-380, 382-384, 388-390, 393-396, 398-399, 401-402, 404-409, 412-418, 420-424, 427, 429-433, 436-443, 445-448, 450-452, 454-465, 467-473, 475-481, 484-487, 489, 525-533, 535-540, 542-544, 546-548, 550-551, 609, 649-663, 667-677, 679-680, 683-692, 694-699, 701-718, 720-730, 732-735, 737-749, 751-760, 762-767, 769, 785-787, 792-796, 811-815, 830-844, 854, 872, 889, 891-894
web-app/src/containers/DropdownModelProvider.tsx	43-50, 53-61, 90-94, 103, 109-111, 113-120, 123-128, 130-138, 147-150, 152-157, 159-160, 163-171, 173-178, 206-215, 218-219, 224-227, 229-232, 237-238, 248, 250-256, 258, 260-262, 265-266, 268-274, 290, 298-305, 307-311, 322-323, 329-332, 335-336, 338, 346-353, 356-361, 363-364, 367-375, 403-405, 418-424, 456-462, 469-471, 476, 478-482, 485-490, 492-502, 504-517, 519-520, 522-523, 528, 533-535, 537, 539-542, 545-561, 563-568, 571, 573, 575-581, 583-593, 595-598, 600-601, 603-607, 609-610, 612, 614
web-app/src/hooks/useModelProvider.ts	33-38, 40-41, 44-54, 56, 58-65, 67-69, 71-108, 110-140, 142-153, 163-165, 167, 169-171, 174-177, 179-180, 182, 184-186, 188-201, 203-206, 208-213, 219, 234-235, 237-245, 248-250, 253-256, 259-266, 268-280, 282-283
web-app/src/services/models.ts	137-139, 201, 203-207, 271-279, 283-285, 287-288, 290, 292, 294, 296-302, 305-325, 327-328, 331-340, 426, 428-429, 432-433, 452-456, 458-459, 471-475, 477-478, 481-482, 485-488, 490-491, 494, 496-516, 519-524, 526-535, 537, 539-549, 551-565, 567-573, 582-583, 586-593
web-app/src/services/providers.ts	30, 55, 74, 77, 100-110, 192-193