npm.io
0.0.1 • Published 1 week ago

@dpopsuev/alef-adapter-nodesh

Licence
MIT
Version
0.0.1
Deps
2
Size
28 kB
Vulns
0
Weekly
179
DeprecatedThis package is deprecated
SummaryDependencyVersions

Alef Agent

Discord

This fork is maintained BDFL-style; outside contributions are not accepted. Source is open to read and to fork (MIT). See CONTRIBUTING.md.

Alef Agent Harness Monorepo

This repository contains the Alef CLI app, runtime, blueprint, and supporting packages.

Attribution

Alef Agent is a fork of Pi (the upstream Pi coding agent / terminal harness). Pi was created by Mario Zechner (@badlogic). The upstream open-source tree is earendil-works/pi-mono.

This fork keeps Mario's design and implementation as its foundation; it adds Alef branding (@dpopsuev/alef-* packages, alef CLI, pkg.alef extensions) and fork-owned defaults (optional version checks and install pings only when you set ALEF_LATEST_VERSION_URL / ALEF_REPORT_INSTALL_URL). Use the upstream repository for the original project line; use dpopsuev/alef for Alef packaging and fork-specific issues.

Share your OSS coding agent sessions

If you use Pi, Alef, or other coding agents for open source work, please share your sessions.

Public OSS session data helps improve coding agents with real-world tasks, tool use, failures, and fixes instead of toy benchmarks.

For the full explanation, see this post on X.

To publish sessions, use badlogic/pi-share-hf. Read its README.md for setup instructions. All you need is a Hugging Face account, the Hugging Face CLI, and pi-share-hf.

You can also watch this video, where I show how I publish my pi-mono sessions.

I regularly publish my own pi-mono work sessions here:

All Packages

Package Description
@dpopsuev/alef-kernel Microkernel — buses, adapter framework, contributions
@dpopsuev/alef-runtime Agent, AgentController, delegation, tool shell
@dpopsuev/alef-llm Unified multi-provider LLM API (Anthropic, OpenAI, Google, etc.)
@dpopsuev/alef-session Session store, turn assembly, context compaction
@dpopsuev/alef-reasoner LLM turn loop, tool dispatch, budget signals
@dpopsuev/alef-tui Terminal UI library with differential rendering
@dpopsuev/alef-web-ui Web components for AI chat interfaces
@dpopsuev/alef-coding-agent Coding agent blueprint
packages/agent Headless agent server + CLI

For Slack/chat automation and workflows see earendil-works/pi-chat.

Contributing

See CONTRIBUTING.md for governance (BDFL; read/fork only for outsiders) and AGENTS.md for project-specific rules (for both humans and agents).

XDG Directory Structure

Alef follows the XDG Base Directory Specification:

$XDG_CONFIG_HOME/alef/          # User configuration (~/.config/alef)
  ├── theme.yaml                # Custom TUI theme
  ├── config.yaml               # User preferences
  └── skills/                   # User skill library
      └── debug-alef/SKILL.md   # Debug diagnostics skill

$XDG_DATA_HOME/alef/            # User data (~/.local/share/alef)
  ├── sessions/<cwd-hash>/      # Session JSONL logs
  └── prototypes/               # User-written adapter prototypes

$XDG_STATE_HOME/alef/           # Logs & state (~/.local/state/alef)
  ├── debug.log                 # Pino debug trace (rotates at 10MB)
  ├── alef.db                   # SQLite (sessions, events, daemon registry)
  └── last-session.json         # Most recent session metadata

$XDG_CACHE_HOME/alef/           # Cache (~/.cache/alef)
  ├── lsp/                      # TypeScript LSP cache
  └── embeddings/               # Vector embedding cache

<cwd>/.alef/                    # Project-local configuration
  ├── directives/               # Project-specific system prompts
  └── skills/                   # Project-specific skills
Setup XDG Directories
make xdg-setup    # Create all XDG directories and migrate legacy ~/.alef
make xdg-info     # Show current XDG paths and status

Development

pnpm install         # Install all dependencies
npm run check:fast   # Lint, format, and type check (pre-commit)
npm run check        # Full check including unit tests (CI)
./alef-test.sh       # Run alef from sources

Debug Mode

Run Alef with comprehensive debugging:

make debug           # Run with debug logging, auto-load debug-alef skill
make debug-watch     # Watch debug.log in real-time (requires jq)
make debug-errors    # Show only errors from debug.log
make debug-tools     # Show tool execution timing
make debug-clean     # Clear debug.log and session files

Debug logs are written to $XDG_STATE_HOME/alef/debug.log (typically ~/.local/state/alef/debug.log).

Watch logs live:

tail -f ~/.local/state/alef/debug.log | jq .

License

MIT

Timeout Configuration

Alef's default LLM timeout is 90 seconds (increased from 60s to accommodate Claude Sonnet 4-5 thinking mode). Override per session:

# Increase timeout to 3 minutes for complex analysis
ALEF_LLM_TIMEOUT_MS=180000 alef

# Debug mode uses 2 minutes by default
make debug

Keywords

aleforgannodeshrepljavascript