- Limit max attachments to 5 files to prevent context overflow
- Fix URL update timing: use SvelteKit's replaceState in onComplete
callback instead of history.replaceState before streaming
- Load attachment content from IndexedDB in conversation history
so follow-up messages have access to file content
- Show error messages in chat when Ollama fails instead of stuck
"Processing..." indicator
- Force file analysis when >3 files attached to reduce context usage
Clear 'Processing...' text only when first token arrives, not before
the LLM request. This keeps the indicator visible during prompt
resolution, RAG retrieval, and LLM initialization.
- Add "Processing X files..." indicator in chat while handling attachments
- Indicator transitions to "Analyzing X files..." for large files needing LLM summarization
- Reuse streaming message for seamless transition to LLM response
- Add FileAnalyzer service for large file summarization with 10s timeout
- Skip analysis for borderline files (within 20% of 8K threshold)
- Read up to 50KB from original file for analysis (not just truncated content)
- Remove base64 blobs from JSON before analysis to reduce prompt size
- Add AttachmentDisplay component for showing file badges on messages
- Persist attachments to IndexedDB with message references
- Add chat state methods: setStreamContent, removeMessage
- Clean up debug logging
- Add postinstall script to copy worker to static/
- Update Dockerfile to copy worker during build
- Update file-processor to try local worker first, fallback to CDN
- Bump version to 0.4.11
Adds two related features for enhanced model customization:
**Model-Specific System Prompts:**
- Assign prompts to models via Settings > Model Prompts
- Capability-based default prompts (vision, tools, thinking, code)
- Auto-select appropriate prompt when switching models in chat
- Per-model prompt mappings stored in IndexedDB
**Custom Ollama Model Creation:**
- Create custom models with embedded system prompts via Models page
- Edit system prompts of existing custom models
- Streaming progress during model creation
- Visual "Custom" badge for models with embedded prompts
- Backend handler for Ollama /api/create endpoint
New files:
- ModelEditorDialog.svelte: Create/edit dialog for custom models
- model-creation.svelte.ts: State management for model operations
- model-prompt-mappings.svelte.ts: Model-to-prompt mapping store
- model-info-service.ts: Fetches and caches model info from Ollama
- modelfile-parser.ts: Parses system prompts from Modelfiles
Custom tools were displayed as enabled in the UI but never sent to
Ollama because getEnabledToolDefinitions() only queried the builtin
tool registry. Now iterates customTools and includes enabled ones.
Fixes #4
- Fix TypeError when check updates returns null updates array
- Display verified capabilities from Ollama runtime in Local Models tab
- Fetch all model capabilities on page mount
- Add data-dev to gitignore
- Add capability verification for installed models using /api/show
- SyncModels now updates real capabilities when fetchDetails=true
- Model browser shows verified/unverified badges for capabilities
- Add info notice that capabilities are sourced from ollama.com
- Fix incorrect capability data (e.g., deepseek-r1 "tools" badge)
Capabilities from ollama.com website may be inaccurate. Once a model
is installed, Vessel fetches actual capabilities from Ollama runtime
and displays a "verified" badge in the model details panel.
- Add size filter (≤3B, 4-13B, 14-70B, >70B) based on model tags
- Add model family filter dropdown with dynamic family list
- Display last updated date on model cards (scraped from ollama.com)
- Add /api/v1/models/remote/families endpoint
- Convert relative time strings ("2 weeks ago") to timestamps during sync
- Add customMaxTokens override to ContextManager
- maxTokens is now derived from custom override or model default
- ChatWindow syncs settings.num_ctx to context manager
- Progress bar now shows custom context length when enabled
- Fetch actual model defaults from Ollama's /api/show endpoint
- Parse YAML-like parameters field (e.g., "temperature 0.7")
- Cache model defaults to avoid repeated API calls
- Initialize sliders with model's actual values when enabling custom params
- Show asterisk indicator when parameter differs from model default
- Reset button now restores to model defaults, not hardcoded values
Tool usage was not showing after page reload because the toolCalls
field was not being included when saving assistant messages to the
database. Now toolCalls are properly persisted and restored.
- Add CodeMirror editor with syntax highlighting for JavaScript and Python
- Add 8 starter templates (4 JS, 4 Python) for common tool patterns
- Add inline documentation panel with language-specific guidance
- Add tool testing UI to run tools with sample inputs before saving
- Add Python tool execution via backend API with 30s timeout
- Add POST /api/v1/tools/execute endpoint for backend tool execution
- Update Dockerfile to include Python 3 for tool execution
- Bump version to 0.4.0
Backend:
- Add /api/v1/version endpoint returning current and latest version
- Fetch latest release from GitHub API with 1-hour cache
- Semver comparison to detect available updates
- Configurable via GITHUB_REPO env var (default: vikingowl/vessel)
Frontend:
- Add VersionState store with 12-hour periodic checking
- Check on app load and periodically for new versions
- Persist dismissed versions in localStorage
- Add UpdateBanner component with teal styling
- Slide-in animation from top, dismissible
The notification appears below TopNav when a new version is available
and remembers dismissals per-version across sessions.
- Remove curl --max-filesize to avoid hard failures on large pages
- Add Truncated/OriginalSize fields to FetchResult for all fetch methods
- Return truncation info in proxy response (truncated, originalSize, returnedSize)
- Add timeout parameter to fetch_url tool (default 30s, max 120s)
- Increase default maxLength from 5KB to 50KB, allow up to 2MB
- Include _hint in response guiding LLM to retry with larger maxLength
Instead of failing when content exceeds limits, the tool now returns
truncated content with guidance for the LLM to request more if needed.
Auto-scroll now stops when the top of the assistant's response
reaches the top of the viewport, allowing users to read from
the beginning while more content streams below.
The SvelteKit hooks proxy was only forwarding pathname without
the query string (event.url.search), causing filters and pagination
to be silently dropped in production builds.
Fixes "COMPOSE_CMD: unbound variable" error when running
--uninstall or --update flags. Previously the variable was only
set during check_prerequisites(), which runs after flag handling.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Ollama must listen on 0.0.0.0 for Docker containers to connect.
Added instructions for both systemd and manual configuration.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Simplify setup by requiring local Ollama installation:
- docker-compose.yml now connects to host Ollama via host.docker.internal
- Remove ollama service from compose (no longer included)
- install.sh now requires Ollama to be installed
- Update README with clear prerequisites
- Add Docker Ollama support to roadmap for future
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove depends_on from docker-compose.yml (services handle reconnection)
- This allows the override to disable ollama without errors
- Fix prompt display when running via curl | bash (print to stderr)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When disabling the ollama service, we also need to clear
depends_on references to avoid "undefined service" errors.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
SvelteKit production build needs server-side request handling
since Vite's dev proxy isn't available. The hooks.server.ts file
proxies:
- /api/v1/* -> backend service
- /api/* -> Ollama service
- /health -> backend health check
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When running via `curl | bash`, stdin is not a terminal.
Read from /dev/tty to allow interactive prompts.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add install.sh for one-line installation (Linux/macOS)
- Detects local Ollama and lets user choose system vs Docker
- Generates docker-compose.override.yml for system Ollama mode
- Supports --update and --uninstall flags
- Fix backend not reading OLLAMA_URL from environment variable
- Add getEnvOrDefault() helper for PORT, DB_PATH, OLLAMA_URL
- Update Dockerfile to use env vars instead of hardcoded flags
- Update README with new Quick Start instructions
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add GPL-3.0 license with copyright notice (2026 VikingOwl)
- Update license badge from MIT to GPL-3.0
- Add copyright to license section
- Remove completed roadmap items (system prompts, search, shortcuts, export)
- Add Hugging Face integration to roadmap
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
