main

Changelog

0.10.0 - 2026-01-27

Changed

  • Breaking: Plugin API - plugins now receive agent instead of toolbox
    • Old: Elelem::Plugins.register(:name) { |toolbox| toolbox.add(...) }
    • New: Elelem::Plugins.register(:name) { |agent| agent.toolbox.add(...) }
    • Plugins can now access agent.terminal, agent.commands, agent.conversation
  • Extracted Conversation class from Agent for better separation of concerns
  • Extracted Commands class for slash command handling
  • Refactored LLM fetch interface to emit separate events for thinking/content/tool_calls
  • Simplified system prompt with inline ERB template
  • Renamed confirm plugin to zz_confirm to ensure it loads last
  • MCP logs now write to ~/.elelem/mcp.log instead of working directory
  • Tool schema now frozen to prevent mutation
  • Uses Open3.capture2 instead of backticks for thread safety
  • Improved ANSI escape sequence stripping in /shell transcripts

Added

  • Async MCP loading for faster startup - tools load in background thread
  • HTTP MCP servers with SSE support and session management
  • OAuth authentication for MCP servers with PKCE, automatic token refresh
  • Global hooks - toolbox.before/toolbox.after without tool name applies to all tools
  • /context improvements: /context <n> to view entry, /context json for full dump
  • ast-grep (sg) support for building repo maps - faster and more accurate than ctags
  • New tools: glob, grep, list, git, task, /tools command
  • Permissions system (lib/elelem/permissions.rb) for tool access control
  • OpenAI reasoning mode - enables Reasoning: high for o-series models
  • Test coverage for OAuth, token storage, HTTP MCP, SSE parsing, global hooks

0.9.2 - 2026-01-22

Fixed

  • Suppress pathname gem constant redefinition warnings on Ruby 4.0

0.9.1 - 2026-01-22

Fixed

  • Use next instead of return to exit block in execute plugin after-hook
  • Use break instead of return to exit inject block in verify plugin

0.9.0 - 2026-01-21

Changed

  • Breaking: Requires Ruby >= 4.0.0 (was 3.4.0)
  • Breaking: Removed net-llm dependency - LLM clients now inline in lib/elelem/net/
    • Elelem::Net::Claude (Anthropic and Vertex AI)
    • Elelem::Net::OpenAI
    • Elelem::Net::Ollama
  • Breaking: Simplified LLM client fetch contract
    • Yields {content:, thinking:} deltas
    • Returns tool_calls array directly
  • Breaking: Tool schema uses OpenAI format ({type: "function", function: {...}})
  • Breaking: Tool definitions use description: key (was desc:)
  • Breaking: Removed modes and permissions system entirely
  • Breaking: Removed slash commands (/mode, /env, /provider, /model)
    • Remaining: /clear, /context, /init, /reload, /shell, /exit, /help
  • Breaking: Removed many dependencies
    • Removed: thor, cli-ui, erb, cgi, set, timeout, logger, net-llm, json-schema
    • Added: json_schemer, optparse, tempfile, stringio, uri
  • Consolidated multiple exe files into single exe/elelem entry point
  • Tools are now defined via plugins instead of hardcoded in Toolbox
  • System prompt includes hints for rg, fd, sg (ast-grep), sed, patch
  • System prompt regenerated on each fetch (includes dynamic repo map)
  • Default tool set: read, write, edit, execute, eval, verify, task
  • System prompt encourages using eval to create tools for repetitive tasks

Added

  • Plugin system with support for custom tool definitions
    • Load plugins from lib/elelem/plugins/, ~/.elelem/plugins/, and .elelem/plugins/
    • Elelem::Plugins.register(name) { |toolbox| ... } API
    • Built-in plugins: read, write, edit, execute, eval, verify, confirm, mcp
  • MCP (Model Context Protocol) server support via .mcp.json configuration
  • AGENTS.md file support - searches up directory tree for project instructions
  • /init command to generate an AGENTS.md file for the current project
  • /shell command to drop into a shell session and capture the transcript to context
  • /reload command to hot-reload source code without restarting the process
  • task tool for delegating subtasks to focused sub-agents
  • edit tool for replacing first occurrence of text in a file
  • eval tool for executing Ruby code and dynamically registering new tools
  • verify tool for syntax checking and running project tests
  • Pre/post tool hooks (toolbox.before/toolbox.after) for extensibility
  • Confirmation prompt before executing shell commands (when TTY)
  • Context compaction for long conversations (summarizes old messages)
  • Repo map via ctags included in system prompt
  • Markdown rendering with glow for LLM responses
  • CLI improvements: optparse-based interface with -p/-m flags
    • elelem chat - Interactive REPL (default)
    • elelem ask <prompt> - One-shot query (reads stdin if piped)
    • elelem files - Output files as XML
  • JSON Schema validation for tool call arguments (via json_schemer)
  • Tool aliases support (e.g., bash, sh, exec → execute)
  • Dependencies documentation in README with installation links

Removed

  • lib/elelem/application.rb - CLI now in exe/elelem
  • lib/elelem/conversation.rb - simplified into Agent
  • lib/elelem/git_context.rb - inlined into Agent
  • lib/elelem/system_prompt.erb - now generated in Agent
  • web_fetch, web_search, fetch, search_engine tools
  • patch tool (use edit or execute with sed/patch)
  • grep, list tools (use execute with rg, fd)
  • Modes and permissions system
  • Events module
  • GitHub Actions CI workflow

Fixed

  • Handle missing args in Claude provider
  • Tool alias resolution (use canonical tool name, not alias)
  • Unknown tool error now suggests using execute and lists available tools
  • Duplicate write operations in edit flow

0.8.0 - 2026-01-14

Added

  • fetch tool for HTTP GET requests (returns status and body)
  • search_engine tool for DuckDuckGo Instant Answer API searches
  • Tool aliases: get/web → fetch, ddg/duckduckgo → search_engine
  • net-hippie and cgi dependencies for HTTP requests

0.7.0 - 2026-01-14

Changed

  • Renamed internal mode concept to permissions for clarity (read/write/execute are permissions, plan/build/verify are modes)
  • Refactored Toolbox#run_tool to accept permissions: parameter

Added

  • ASCII spinner animation while waiting for LLM responses
  • Terminal#waiting method with automatic cleanup on next output
  • Decision-making principles in system prompt (prefer reversible actions, ask when uncertain)
  • Mode enforcement tests

Fixed

  • Security: Mode restrictions now enforced at execution time, not just schema time
    • Previously, LLMs could call tools outside their mode by guessing tool names
    • Now run_tool validates the tool is allowed for the current permission set

0.6.0 - 2026-01-12

Changed

  • Renamed bash tool to exec
  • Tuned system prompt
  • Changed thinking prompt to ellipsis
  • Removed username from system prompt
  • Use pessimistic constraint on net-llm dependency
  • Extracted Terminal class for IO abstraction (enables E2E testing)

Added

  • /env slash command to capture environment variables for provider connections
  • /shell slash command
  • /provider and /model slash commands
  • Tab completion for commands
  • Help output for /mode and /env commands

Fixed

  • Prevent infinite looping errors
  • Provide function schema when tool is called with invalid arguments
  • Tab completion for pass entries without requiring show subcommand
  • Password store symlink support in tab completion

0.5.0 - 2026-01-07

Changed

  • Requires net-llm >= 0.5.0 with unified fetch interface
  • Updated gem description to reflect multi-provider support

Added

  • Multi-provider support: Ollama, Anthropic, OpenAI, and VertexAI
  • --provider CLI option to select LLM provider (default: ollama)
  • --model CLI option to override default model
  • Tool aliases (bash also accepts exec, shell, command, terminal, run)
  • Thinking text output for models that support extended thinking

0.4.2 - 2025-12-01

Changed

  • Renamed exec tool to bash for clarity
  • Improved system prompt with iterative refinements
  • Added environment context variables to system prompt

0.4.1 - 2025-11-26

Changed

  • Refactor tool-call formatting to a more compact JSON payload for better LLM parsing
  • Updated CI and documentation to use GitHub instead of previous hosting
  • Runtime validation of command-line parameters against a JSON schema

Added

  • elelem files subcommand: generates Claude-compatible XML file listings
  • Rake task files:prompt to output a ready-to-copy list of files for prompts

Fixed

  • Minor documentation and CI workflow adjustments

0.4.0 - 2025-11-10

Changed

  • Architecture Improvements: Significant refactoring for better separation of concerns
    • Extracted Tool class to separate file (lib/elelem/tool.rb)
    • Extracted Toolbox class to separate file (lib/elelem/toolbox.rb)
    • Extracted Shell class for command execution
    • Improved tool registration through #add_tool method
    • Tool constants moved to Toolbox for better organization
    • Agent class simplified by delegating to Tool instances

Added

  • Eval Tool: Meta-programming tool that allows the LLM to dynamically create and register new tools at runtime
    • Eval tool has access to the toolbox for enhanced capabilities
  • Comprehensive test coverage with RSpec
    • Agent specs
    • Conversation specs
    • Toolbox specs

Fixed

  • /context command now correctly accounts for the current mode

0.3.0 - 2025-11-05

Changed

  • Breaking: Removed llm-ollama and llm-openai standalone executables (use main elelem chat command)
  • Breaking: Simplified architecture - consolidated all logic into Agent class
    • Removed Configuration class
    • Removed Toolbox system
    • Removed MCP client infrastructure
    • Removed Tool and Tools classes
    • Removed TUI abstraction layer (direct puts/Reline usage)
    • Removed API wrapper class
    • Removed state machine
  • Improved execute tool description to guide LLM toward direct command execution
  • Extracted tool definitions from long inline strings to readable private methods
  • Updated README with clear philosophy and usage examples
  • Reduced total codebase from 417 to 395 lines (-5%)

Added

  • Mode System: Control agent capabilities with workflow modes
    • /mode plan - Read-only mode (grep, list, read)
    • /mode build - Read + Write mode (grep, list, read, patch, write)
    • /mode verify - Read + Execute mode (grep, list, read, execute)
    • /mode auto - All tools enabled
    • Each mode adapts system prompt to guide appropriate behavior
  • Improved output formatting
    • Suppressed verbose thinking/reasoning output
    • Clean tool call display (e.g., date instead of full JSON hash)
    • Mode switch confirmation messages
    • Clear command feedback
  • Design philosophy documentation in README
  • Mode system documentation

Removed

  • exe/llm-ollama (359 lines)
  • exe/llm-openai (340 lines)
  • lib/elelem/configuration.rb
  • lib/elelem/toolbox.rb and toolbox/* files
  • lib/elelem/mcp_client.rb
  • lib/elelem/tool.rb and lib/elelem/tools.rb
  • lib/elelem/tui.rb
  • lib/elelem/api.rb
  • lib/elelem/states/* (state machine infrastructure)
  • Removed ~750 lines of unused/redundant code

Fixed

  • Working directory handling for execute tool (handles empty string cwd)
  • REPL EOF handling (graceful exit when input stream ends)
  • Tool call formatting now shows clean, readable commands

0.2.1 - 2025-10-15

Fixed

  • Added missing exe/llm-ollama and exe/llm-openai files to gemspec
  • These executables were added in 0.2.0 but not included in the packaged gem

0.2.0 - 2025-10-15

Changed

  • Breaking: Migrated from custom Net::HTTP implementation to net-llm gem
  • API client now uses Net::Llm::Ollama for better reliability and maintainability
  • Removed direct dependencies on net-http and uri (now transitive through net-llm)
  • Maps Ollama’s thinking field to internal reasoning field
  • Maps Ollama’s done_reason to internal finish_reason
  • Improved system prompt for better agent behavior
  • Enhanced error handling and logging

Added

  • New llm-ollama executable - minimal coding agent with streaming support for Ollama
  • New llm-openai executable - minimal coding agent for OpenAI/compatible APIs
  • Memory feature for persistent context storage and retrieval
  • Web fetch tool for retrieving and analyzing web content
  • Streaming responses with real-time token display
  • Visual “thinking” progress indicators with dots during reasoning phase

Fixed

  • Response processing for Ollama’s native message format
  • Tool argument parsing to handle both string and object formats
  • Safe navigation operator usage to prevent nil errors

0.1.2 - 2025-08-14

Fixed

  • Fixed critical bug where bash tool had nested parameters schema causing tool calls to fail with “no implicit conversion of nil into String” error

0.1.1 - 2025-08-12

Fixed

  • Fixed infinite loop bug after tool execution - loop now continues until assistant provides final response
  • Fixed conversation history accumulating streaming chunks as separate entries - now properly combines same-role consecutive messages
  • Improved state machine logging with better debug output

0.1.0 - 2025-08-08

  • Initial release