-
7cf01b9Thanks @coopbri! - Record AI token usage to Aether for billing after each LLM completion -
a94576aThanks @coopbri! - Add@filemention expansion to inline file contents as a fenced code block in the prompt -
7cf01b9Thanks @coopbri! - Addomni auth loginandomni auth logoutcommands for cloud Synapse authentication -
a94576aThanks @coopbri! - Add /compact slash command to summarize old messages and free context window space -
a94576aThanks @coopbri! - Add /cost slash command to display per-session token usage and estimated cost -
7cf01b9Thanks @coopbri! - Add knowledge pack support with Manifold resolver, local caching, and tag-based chunk selection for persona system prompts -
11cd883Thanks @coopbri! - Add MCP STDIO reconnect-and-retry on transport failure, and animated braille spinner for tool calls in the TUI (inline block + status bar) -
9ece8deThanks @coopbri! - Addmulti_searchtool for parallel multi-source web search and TUI display for tool call activity -
65d45beThanks @coopbri! - Add Synapse client integration, agent-core provider registry, and Beacon browser tool -
a94576aThanks @coopbri! - Add /undo slash command to revert last agent file changes using shadow git snapshots
1ddde64Thanks @coopbri! - Fix hardcoded Sonnet pricing with model-aware cost lookup, add auth token extraction fallback chain, and add missing tests
-
7923c3cThanks @coopbri! - Add Kimi 2.5 (Moonshot AI) provider support- Add
kimiprovider with OpenAI-compatible API atapi.moonshot.cn - Add models:
kimi-k2.5,moonshot-v1-128k,moonshot-v1-32k - Add prefix detection for
kimi-*andmoonshot-*model names - Set
MOONSHOT_API_KEYenvironment variable to use
- Add
-
0783ebdThanks @coopbri! - Add unified LLM provider, session sharing, skill system, and LSP integration- Add unified LLM provider via
llmcrate supporting Anthropic, OpenAI, Google Gemini, Groq, and Mistral - Add session sharing with URL tokens, secrets, and optional TTL expiration
- Add CLI commands:
omni session shareandomni session unshare - Add skill system for loading reusable instructions from SKILL.md files
- Discover skills from
.omni/skill/,.opencode/skill/, and.claude/skills/directories - Add
skilltool for agents to load specialized instructions on demand - Add LSP integration with support for 13 language servers
- Add
lsptool for code intelligence (hover, definition, references, symbols) - Add agent loop detection to prevent infinite tool call loops
- Add secret masking for API keys, tokens, and credentials in tool output
- Add MCP server integration to discover and execute external tools
- Add plugin auto-discovery and integration into agent tool registry
- Add unified LLM provider via
-
80a07ebThanks @coopbri! - Add GitHub integration tools- Add
github_prtool for creating pull requests - Add
github_issuetool for creating, viewing, listing, and closing issues - Add
github_pr_reviewtool for viewing PRs, diffs, checks, and adding comments - All tools use the
ghCLI and respect plan mode restrictions
- Add
-
46bedb3Thanks @coopbri! - Add persistent memory system for cross-session context- Add
memory_addtool for storing facts about user preferences, project patterns, and corrections - Add
memory_searchtool for finding relevant memories by query or category - Add
memory_deletetool for removing outdated memories - Memories persist across sessions and can be injected into system prompts
- Support for pinned memories that are always included in context
- Add
-
08b5595Thanks @coopbri! - Add sandboxed execution tool- Add
sandbox_exectool for running untrusted code safely - Uses Docker containers with resource limits when available
- Falls back to timeout + restricted env when Docker not present
- Supports Python, Node, Ruby, and shell execution
- Configurable timeout, network access, and workdir mounting
- Add
-
761b54aThanks @coopbri! - Add session continuity flags for resuming conversations- Add
--continue/-cflag to resume most recent session - Add
--session/-sflag to resume specific session by ID or slug - Works with both
tuiandagentcommands - Session list now shows slugs for easier CLI use
- Fail fast with clear error for invalid session IDs
- Add
-
6e4f1acThanks @coopbri! - Add new agent tools and LLM providers- Add
apply_patchtool for applying unified diffs to files - Add
multi_edittool for editing multiple files in a single operation - Add support for Groq, Google Gemini, Mistral, LMStudio, OpenRouter, and Together AI providers
- Add
-
ccaf64fThanks @coopbri! - Improve terminal compatibility and provider configuration UX- Fix key event handling for Termux and other terminals that don't report KeyEventKind correctly
- Show provider-agnostic messages when no provider is configured
- Update placeholder text to guide users toward configuration
-
5820f20Thanks @coopbri! - Improve TUI experience- Fix message cutoff by removing bottom padding and correcting height calculations
- Show activity status (e.g., "Using Bash...") instead of just "Thinking..."
- Add vertical and horizontal padding to user messages for better visual separation
- Lighten user message background color for a "previous message" look
- Implement smooth line-by-line scrolling for messages
- Allow typing while agent is responding (can prepare next message)
- Add rich diff rendering for all tool output (muted green/red for diff-like content)
- Fix paste behavior: pasted text now inserts without auto-submitting
- Add Ctrl+Left/Right for word-by-word cursor movement
-
04e91f8Thanks @coopbri! - Add model autocomplete dropdown with automatic provider switching- Type
/modelto see available models with provider info - Arrow keys navigate, Tab/Enter to select
- Automatically switches provider when model requires it (e.g., gpt-4o switches to OpenAI)
- Type
-
7b14e8fThanks @coopbri! - Add session management to TUI- Switch between sessions with Ctrl+S and Enter
- Create new sessions with 'n' key
- Delete sessions with 'd' key
- Session list blocked while streaming
-
770af7aThanks @coopbri! - Fix autoscroll, text wrapping, and model identity- Fix autoscroll during streaming responses
- Fix text wrapping in tool output (was overflowing on long lines)
- Fix model self-identification when switching models mid-conversation
-
f87c14dThanks @coopbri! - Make clipboard support optional viaclipboardfeature flag for Termux/Android compatibility