marfrit f22d21d754 repl: :memory summarize — LLM candidate extraction (Phase 4 commit #4)
Phase 4 commit #4 per docs/PHASE4.md §6.

:memory summarize:
  1. Source-of-truth: session log file via history.load(session_path),
     NOT ctx:to_messages() (R-C2). Skips turns tagged meta="summarize"
     so prior summarize exchanges don't self-amplify across multiple
     calls within the same session.
  2. Pick summarizer model from cfg.memory.summarizer_model (default
     active model).
  3. Build a transcript string ("role: content" per turn, 800 chars max
     per turn) and feed it as a single user turn alongside a system
     instruction asking for "(fact|pref|context): <content>" lines.
  4. broker.chat with max_tokens=1024 + timeout_ms=90000 (the deep
     model can take a while; we don't want a 15s probe-cap here).
  5. Log the response as an assistant turn with meta="summarize" so the
     next :memory summarize call filters it out.
  6. Parse response lines tolerating markdown bullets and bold markup:
     ^%s*[-*]?%s*[*_]*(fact|pref|context)[*_]*:%s*(.+)$
  7. Per-candidate prompt: y / N / edit.
       y    → memory:add(kind, content)
       edit → readline prompt for replacement text
       any other → drop
  8. status: "summarize: added N / M candidates".

Live-tested against hossenfelder/fast:
  Pipeline correct end-to-end. Model emitted one candidate; user
  confirmation prompt fired; item persisted; :memory list showed it.
  Candidate quality from the 1.5B model is poor — typical
  small-model behavior; deep/cloud models would do better but this
  isn't an aish bug.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-13 07:53:36 +00:00

aish

aish — AI-augmented conversational shell.

A single REPL that interleaves shell command execution and language-model conversation, backed by a llama.cpp HTTP broker. Implementation is LuaJIT 2.x with FFI bindings to libcurl, GNU readline, and libc — no C extensions, no build step, one source tree.

Why

Three flows that currently live in three windows fold into one:

  1. "Run this command and show me the output" — fast feedback loop, no copy-paste between terminal and chat.
  2. "Explain or write code based on the output we just looked at" — exec output is automatically injected into the model's context.
  3. "Plan and execute a multi-step task with confirmation gates" — landing in Phase 3 as Chuck Norris autonomous mode.

aish is not a wrapper around bash. It's a first-class interactive environment where the shell is one of several execution channels.

Status

Component State
Repository skeleton in this commit
Phase 0 manifest docs/PHASE0.md — locked
Phase 0 implementation 🔜 next session
Phase 1+ 📋 enumerated in PHASE0.md §11

Every module file currently raises not implemented (Phase 0 pending) when called. luajit main.lua fails loudly at the first un-implemented function, never silently.

Quick orientation

Read this If you want to know
docs/PHASE0.md §12 What aish is and what Phase 0 ships
docs/PHASE0.md §3 Technology decisions (LuaJIT, FFI, readline, libcurl, llama.cpp)
docs/PHASE0.md §4 Directory layout — these file names are stable across all phases
docs/PHASE0.md §5 How input is dispatched (meta / shell / AI)
docs/PHASE0.md §6 Broker contract: /v1/chat/completions, CMD: extraction
docs/PHASE0.md §10 Config schema and resolution order
docs/PHASE0.md §11 Phase sequence (what lands when)
docs/PHASE0.md §13 Open questions, tracked per phase
CLAUDE.md Project conventions for AI-assisted contributors

Directory layout

aish/
├── main.lua              # entry point
├── repl.lua              # readline loop, dispatch, prompt
├── broker.lua            # llama.cpp HTTP client
├── router.lua            # input classifier (meta/shell/AI)
├── executor.lua          # command exec + CMD: extraction
├── context.lua           # in-memory turn history
├── history.lua           # disk persistence (Phase 1+)
├── safety.lua            # destructive-op gate (Phase 3+)
├── renderer.lua          # output formatting
├── config.lua            # default model registry + preferences
├── ffi/
│   ├── curl.lua          # libcurl easy interface
│   ├── readline.lua      # GNU readline
│   ├── pty.lua           # forkpty (Phase 1+)
│   └── libc.lua          # chdir, errno, strerror
└── docs/
    └── PHASE0.md         # locked substrate

Build / runtime dependencies

System packages (Debian / ALARM / Arch names):

  • luajit (>= 2.0)
  • libcurl4 / libcurl-openssl-3 runtime
  • libreadline8 runtime
  • libc6 runtime (always present)

No compilation, no luarocks, no make. Just luajit main.lua.

Running

Once Phase 0 ships:

luajit main.lua                          # uses ~/.config/aish/config.lua
luajit main.lua --config ./config.lua    # explicit config path
AISH_CONFIG=/path/to/config.lua luajit main.lua

Config resolution order is documented in docs/PHASE0.md §10.

Configuration

config.lua is a Lua file returning a single table. The committed config.lua in this repo is both the canonical example and the development-fallback config (lowest precedence). Copy it to ~/.config/aish/config.lua and edit endpoints to your local llama.cpp servers, or point AISH_CONFIG at your own.

The default endpoints assume mfritsche's home network:

  • fastdirac.fritz.box:8081 (Qwen2.5-Coder-7B q4 8k ctx)
  • deepdirac.fritz.box:8080 (Qwen2.5-Coder-7B q4 32k ctx)
  • cloudhossenfelder.fritz.box:8082 (forwards to OpenRouter)

Replace these with your own llama.cpp endpoints if you're not on that LAN.

License

Not yet selected. Default-private until decided.

Project conventions

See CLAUDE.md for contribution conventions, commit style, and the phase-loop discipline this project follows.

S
Description
AI-augmented conversational shell — LuaJIT REPL with llama.cpp broker, shell executor, and routed AI inference.
Readme MIT 2.2 MiB
Languages
Lua 99.8%
Shell 0.2%