Conversation
- Updated ScoredMemoryEntry structure to include distinct fields for Dense, Keyword, RRF, Composite, and Rerank scores, along with corresponding Has* flags. - Modified formatScoredExperience function to display relevance details based on Dense and Rerank scores. - Adjusted the RerankEvolvingMemory function to set Rerank scores and flags appropriately. - Implemented minimum rerank score configuration in EvolvingMemory and EvolvingMemoryConfig. - Enhanced keyword search functionality to drop stopwords and short tokens, ensuring only relevant terms are processed. - Added tests for new scoring and reranking logic, ensuring proper functionality and edge case handling. - Updated configuration validation to include checks for minimum rerank score. - Introduced new tests for keyword query term processing and retrieval accuracy.
… parameters - Implemented OpenRouter and llama.cpp as new LLM providers in the factory. - Added default request parameters for each provider, including OpenRouter's Anthropic-backed settings. - Updated onboarding and specialist creation flows to apply provider-specific defaults. - Enhanced frontend components to accommodate new providers and their default parameters. - Added comprehensive tests for provider defaults, onboarding, and model resolution. - Refactored provider backend logic to streamline provider handling and configuration.
- Updated SetupView.vue to dynamically render LLM provider options from a constants file. - Introduced ConfigGroupsForm.vue for managing configuration groups with save functionality. - Created ConfigValueField.vue to handle various input types for configuration values. - Added llmProviders.ts to centralize LLM provider options and backend mapping. - Implemented tests for ConfigValueField and router functionality. - Enhanced SettingsView tests to ensure proper rendering and functionality of settings sections. - Added theme defaults test to verify default theme settings.
…out and search functionality
- Introduced a new implementation plan for descriptive friendly tool names in Manifold, updating the `friendlyToolTitles` map and adding regression tests for runtime mapping. - Created a new Vue component `FileTreeIcon` to visually represent file types in the UI, supporting various file extensions and directory states.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…e, editor components, project_id) Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
…pter implementation
- Introduced `lexminify` package for deterministic character/word-level minification of provider-visible messages. - Added `protect.go` for scanning and protecting spans of text (code, URLs, UUIDs, etc.). - Created `tables.go` for managing filler phrases, stopwords, and protected words. - Documented architecture decisions and flow of the lexminify process in the wiki. - Established testing guidelines and maintenance logs for ongoing development. - Configured default settings for lexminify to enhance message compression while preserving essential content.
…ming capabilities - Added new assets for Supertonic TTS including ONNX models and voice styles. - Implemented the Supertonic engine for text-to-speech synthesis using ONNX Runtime. - Created a streaming mechanism to handle speech synthesis in real-time. - Introduced settings management for TTS configurations, including custom voice styles. - Developed a store using Pinia to manage TTS state and settings within the application. - Added utility functions for asset fetching and caching to optimize performance. - Implemented tests for asset paths and settings normalization to ensure reliability.
Python onnxruntime sidecar chosen over in-process Go: born-ml/born spike confirmed 8 unimplemented ops (Pad in all 4 graphs). Stable /tts SSE contract across Manifold Go route and Minibot ManifoldTts engine. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Introduced a new service for rendering Supertonic text-to-speech audio. - Implemented API endpoints for audio synthesis and health checks. - Added support for PCM16 streaming and WAV container responses. - Included environment configuration for model directory, voice, and language settings. - Implemented sentence-level streaming for efficient audio generation. - Added tests for API functionality, audio processing, and text chunking. - Created necessary project files including README, .gitignore, and pyproject.toml.
…ts and engine files - Deleted assets.ts and engine.ts, consolidating constants and types into constants.ts. - Introduced serverEngine.ts to handle TTS requests via a server endpoint instead of in-browser processing. - Updated imports across the application to reflect the new structure and removed references to onnxruntime-web. - Adjusted tests to accommodate the new server-based synthesis approach and ensure functionality remains intact.
Finalizes the in-process ("born") TTS engine to consume our published fork
instead of a local vendored copy, and documents the /tts work in the repo wiki.
- go.mod/go.sum: require github.com/intelligencedev/born (pinned to the
feat/supertonic-tts fork commit); drop the ./services/born-supertonic-fork/born
replace and remove the ~3.8 MB vendored subset (409 files).
- handlers_tts_born.go: import .../intelligencedev/born/supertonic.
- services/born-supertonic-fork/STATUS.md: fork is now the source of truth.
- wiki: add components/tts.md and flows/tts-path.md; update index, evidence,
and maintenance-log to cover the three-engine /tts route, the pure-Go born
engine, and the Chat swap off in-browser WebGPU.
Verified: TestTTSBornEngineLive passes against the fork module; go build ./...
green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds stt.engine=moonshine (+ stt.modelDir): the /stt route transcribes
multipart WAV in-process via the Born fork's moonshine package (greedy decode,
KV cache, SentencePiece detok) instead of proxying to OpenAI — voice input is
now fully local. Includes a WAV PCM16 decoder + linear resampler to 16 kHz and
the same {"text"} response shape as the proxy path. Bumps the fork dep
(Range/Neg/InstanceNorm/Trilu ops, Reshape 0-dim fix, moonshine package).
Verified live: TestSTTMoonshineLive transcribes supertonic-generated speech —
a full pure-Go TTS→STT roundtrip.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
…fy runtime management
Approved spike design: port qwen_image (Krea 2 Turbo 12B MMDiT) to the born fork via component-by-component parity gates, transliterating stable-diffusion.cpp, targeting an end-to-end image on 64GB Apple Silicon. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Added OnStreamRollback callback to Engine for handling character rollbacks. - Enhanced engineStreamObserver to trigger rollback on rejected attempts. - Updated RunStreamInference to reset streamed output on retries. - Implemented tests for streaming rollbacks in engine harness and TTS. - Modified chat execution and UI components to support delta rollback events. - Added rollback handling in SupertonicStreamer to manage audio buffer adjustments.
…ment - Removed unused CockpitToolRow type and related tool message handling from chatActivity composable. - Simplified the useChatActivity composable by eliminating toolMessages and associated logic. - Introduced new responseParts management functions to handle streamed text and tool calls more effectively. - Updated chat message structure to include responseParts for better organization of streamed content. - Enhanced ChatTranscript component to render response parts in order, including tool calls with sticky activity. - Added tests for response parts functionality to ensure correct rendering and behavior.
- Introduced a new theme "Desert Night" with corresponding tokens and styles. - Updated theme cycling logic to include the new theme. - Implemented the ApproveCommand component for handling command approvals in chat. - Enhanced ChatView to display pending approvals using the new ApproveCommand component. - Added accessibility improvements, including aria-labels for better screen reader support. - Updated tests to cover new theme and component functionalities.
# Conflicts: # web/agentd-ui/package.json # web/agentd-ui/pnpm-lock.yaml
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds the Desert Night theme and command-approval UI, refreshes related chat navigation and styling, resolves outstanding Dependabot and
pnpm auditadvisories, and updates stale UI tests for the current application flows.Changes
Security and dependencies
golang.org/x/netgoogle.golang.org/grpcVerification
go test ./...pnpm audit: zero known vulnerabilitiesgit diff --checkpassedNotes
The UI test run may print a non-failing jsdom warning about browser navigation from an anchor interaction. All tests still pass.