quark-ai 0.1.1__tar.gz

quark_ai-0.1.1/.env.example

@@ -0,0 +1,3 @@
+ANTHROPIC_API_KEY=
+OPENAI_API_KEY=
+QUARK_PROVIDER=anthropic

quark_ai-0.1.1/.github/workflows/publish.yml

@@ -0,0 +1,36 @@
+name: Publish to PyPI
+
+on:
+  release:
+    types: [published]
+  push:
+    tags:
+      - "v*"
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - run: pip install build
+      - run: python -m build
+      - uses: actions/upload-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+  publish:
+    needs: build
+    runs-on: ubuntu-latest
+    environment: pypi
+    permissions:
+      id-token: write  # required for PyPI Trusted Publishing (OIDC), no stored secrets
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+      - uses: pypa/gh-action-pypi-publish@release/v1

quark_ai-0.1.1/.github/workflows/tests.yml

@@ -0,0 +1,19 @@
+name: Tests
+
+on:
+  push:
+  pull_request:
+
+jobs:
+  test:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.10", "3.11", "3.12"]
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+      - run: pip install -e ".[dev]"
+      - run: pytest -q

quark_ai-0.1.1/.gitignore

@@ -0,0 +1,218 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[codz]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#   Usually these files are written by a python script from a template
+#   before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py.cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+# Pipfile.lock
+
+# UV
+#   Similar to Pipfile.lock, it is generally recommended to include uv.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+# uv.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+# poetry.lock
+# poetry.toml
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#   pdm recommends including project-wide configuration in pdm.toml, but excluding .pdm-python.
+#   https://pdm-project.org/en/latest/usage/project/#working-with-version-control
+# pdm.lock
+# pdm.toml
+.pdm-python
+.pdm-build/
+
+# pixi
+#   Similar to Pipfile.lock, it is generally recommended to include pixi.lock in version control.
+# pixi.lock
+#   Pixi creates a virtual environment in the .pixi directory, just like venv module creates one
+#   in the .venv directory. It is recommended not to include this directory in version control.
+.pixi
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# Redis
+*.rdb
+*.aof
+*.pid
+
+# RabbitMQ
+mnesia/
+rabbitmq/
+rabbitmq-data/
+
+# ActiveMQ
+activemq-data/
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.envrc
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#   JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#   be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#   and can be added to the global gitignore or merged into this file.  For a more nuclear
+#   option (not recommended) you can uncomment the following to ignore the entire idea folder.
+# .idea/
+
+# Abstra
+#   Abstra is an AI-powered process automation framework.
+#   Ignore directories containing user credentials, local state, and settings.
+#   Learn more at https://abstra.io/docs
+.abstra/
+
+# Visual Studio Code
+#   Visual Studio Code specific template is maintained in a separate VisualStudioCode.gitignore 
+#   that can be found at https://github.com/github/gitignore/blob/main/Global/VisualStudioCode.gitignore
+#   and can be added to the global gitignore or merged into this file. However, if you prefer, 
+#   you could uncomment the following to ignore the entire vscode folder
+# .vscode/
+# Temporary file for partial code execution
+tempCodeRunnerFile.py
+
+# Ruff stuff:
+.ruff_cache/
+
+# PyPI configuration file
+.pypirc
+
+# Marimo
+marimo/_static/
+marimo/_lsp/
+__marimo__/
+
+# Streamlit
+.streamlit/secrets.toml

quark_ai-0.1.1/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Moyter
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

quark_ai-0.1.1/PKG-INFO

@@ -0,0 +1,187 @@
+Metadata-Version: 2.5
+Name: quark-ai
+Version: 0.1.1
+Summary: Lightning-fast micro-agent. A stripped-down alternative to heavy AI agents, executing tasks with minimal prompt overhead.
+Project-URL: Homepage, https://github.com/rmoya81/Quark
+Project-URL: Repository, https://github.com/rmoya81/Quark
+Project-URL: Issues, https://github.com/rmoya81/Quark/issues
+Author: rmoya81
+License-Expression: MIT
+License-File: LICENSE
+Classifier: Development Status :: 3 - Alpha
+Classifier: Environment :: Console
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Software Development :: Libraries
+Requires-Python: >=3.10
+Requires-Dist: httpx>=0.27
+Provides-Extra: dev
+Requires-Dist: pytest>=8.0; extra == 'dev'
+Description-Content-Type: text/markdown
+
+# Quark
+
+Lightning-fast micro-agent. A stripped-down alternative to heavy AI agents, executing tasks with minimal prompt overhead.
+
+Quark is a minimal CLI agent loop: send a message, let the model call tools (read/edit/write files, list directories, optionally run shell commands), and get an answer back — with conversation history persisted to disk between runs. It talks to any LLM provider through a small abstraction layer (Anthropic and OpenAI included) instead of depending on a heavy vendor SDK.
+
+Two things it optimizes for on purpose:
+- **Lightweight**: no framework, no heavy SDKs (just `httpx`), a short default system prompt, and an `edit_file` tool so the model can make targeted changes instead of re-sending whole files.
+- **Productive by default, secure by default**: file tools are confined to a workspace directory (symlink escapes included), `run_shell` is off unless you opt in, and if a `QUARK.md`/`AGENTS.md` file exists it's folded into the system prompt automatically so the agent already knows about your project.
+
+## Install
+
+Requires Python 3.10+. Works anywhere Python does, including WSL2 — nothing platform-specific.
+
+```bash
+./install.sh
+```
+
+Creates `.venv`, installs Quark editable with dev deps, and copies `.env.example` to `.env` if you don't already have one. Safe to re-run (reuses the venv, never overwrites an existing `.env`).
+
+Or by hand:
+
+```bash
+python3 -m venv .venv
+source .venv/bin/activate
+pip install -e ".[dev]"
+```
+
+## Configure
+
+Copy `.env.example` to `.env` (done automatically by `install.sh`) and fill in the key(s) for the provider(s) you want to use:
+
+```bash
+cp .env.example .env
+```
+
+```
+ANTHROPIC_API_KEY=sk-ant-...
+OPENAI_API_KEY=sk-...
+QUARK_PROVIDER=anthropic
+```
+
+`.env` is loaded automatically from the current working directory. You can also export the variables directly, or pass `--api-key` on the command line.
+
+## Usage
+
+```bash
+# One-shot message, prints the reply and exits
+quark chat "list the files in this directory"
+
+# Interactive REPL (Ctrl+D or 'exit' to quit)
+quark chat
+
+# Choose a provider/model
+quark --provider openai --model gpt-4o chat "what's in README.md?"
+
+# Named sessions keep separate histories on disk
+quark --session project-a chat "remember that we're using Postgres"
+quark --session project-a chat "what database are we using?"
+
+# Confine file tools to a specific directory (default: cwd)
+quark --workspace ./my-project chat "list the files here"
+
+# Opt in to the run_shell tool (off by default)
+quark --allow-shell chat "run the test suite"
+
+# Read-only mode: only read_file + list_dir, smallest tool-schema footprint, no escape hatch
+quark --minimal-tools chat "what does this project do?"
+
+# Progressive mode: starts like --minimal-tools, but the model can unlock
+# write_file/edit_file/delete_file itself the moment it actually needs them
+quark --progressive-tools chat "fix the typo in README.md"
+
+# Manage sessions
+quark sessions list
+quark sessions clear project-a
+```
+
+Sessions are stored as JSON under `~/.quark/sessions/<name>.json`.
+
+### Project context
+
+If a `QUARK.md` or `AGENTS.md` file exists at the workspace root, its contents are appended to the system prompt automatically — a cheap way to give the agent standing context (stack, conventions, gotchas) without repeating it every session.
+
+### Progressive tool disclosure
+
+`--progressive-tools` (also `QUARK_PROGRESSIVE_TOOLS=1`) starts a turn with just `read_file`, `list_dir`, and one cheap extra tool, `request_more_tools` (no parameters). Its description tells the model to call it if it needs to write, edit, or delete files. If the model calls it, Quark registers `write_file`/`edit_file`/`delete_file` on the spot, drops `request_more_tools` (no longer needed), and the *same* `agent.step()` call retries with the expanded set — so the model still gets everything it needs, in the same turn, it just asks first.
+
+This is deliberately not a heuristic: matching intent from text (keywords, embeddings, a classifier) always trades accuracy for cost and can silently withhold a tool the model actually needed on some phrasing you didn't anticipate. Letting the model itself request more tools can't misfire that way — the only cost is one extra round trip, and only on turns that actually mutate something. `run_shell` stays independently gated behind `--allow-shell`; progressive mode never unlocks it.
+
+`--minimal-tools` is the stricter sibling: same starting set, but with no `request_more_tools` escape hatch at all, for when you want a hard read-only guarantee. Passing both flags together, `--minimal-tools` wins.
+
+### Token overhead
+
+On the first turn (empty session), before the model generates anything, Quark sends the system prompt plus the JSON schema of every registered tool. Rough token counts (`tiktoken` `cl100k_base`, an approximation — Claude's real tokenizer isn't public — but the right order of magnitude either way):
+
+| Mode | Tools registered | Tool-schema tokens | Total (system + tools + a short message) |
+|---|---|---|---|
+| default | `read_file`, `list_dir`, `write_file`, `edit_file`, `delete_file` | ~380 | ~415 |
+| `--minimal-tools` | `read_file`, `list_dir` | ~120 | ~155 |
+| `--progressive-tools`, before unlocking | `read_file`, `list_dir`, `request_more_tools` | ~165 | ~200 |
+| `--progressive-tools`, after unlocking | same as default | ~380 | ~415 |
+| default + `--allow-shell` | + `run_shell` | ~380 + ~65 | ~480 |
+
+`--minimal-tools` is cheapest but can never write/edit/delete. `--progressive-tools` costs a little more than minimal on read-only turns (~200 vs ~155) but is capability-equivalent to the default the moment it's actually needed — it only pays the full ~415 on turns that mutate something. Every subsequent turn in a session also carries the accumulated conversation history on top of these numbers.
+
+## Architecture
+
+```
+quark/
+  message.py       Internal Message/ToolCall representation
+  session.py        Persistent, named conversation history
+  agent.py           The chat + tool-calling loop
+  providers/          Provider abstraction (unified request/response shape)
+    base.py
+    anthropic.py       Anthropic Messages API
+    openai.py           OpenAI Chat Completions API
+  tools/               Built-in tools the agent can call
+    base.py             Tool + ToolRegistry
+    workspace.py         Path confinement (resolve_within)
+    fs.py                read_file / write_file / edit_file / delete_file / list_dir
+    shell.py             run_shell (opt-in)
+    meta.py              request_more_tools (--progressive-tools)
+  cli.py               argparse-based entry point
+```
+
+Adding a provider means implementing `Provider.complete()` in `quark/providers/` and registering it in `quark/providers/__init__.py`. Adding a tool means subclassing `Tool` in `quark/tools/` and adding it to `default_registry()`.
+
+## Security note
+
+- `read_file`, `write_file`, `edit_file`, `delete_file`, and `list_dir` are confined to `--workspace` (default: the current directory); paths that escape it, including via symlinks, are rejected.
+- `run_shell` executes arbitrary shell commands and is **off by default** — enable it with `--allow-shell` or `QUARK_ALLOW_SHELL=1` only when you trust the prompts/provider you're running against.
+- `write_file` and `delete_file` still mutate the filesystem inside the workspace without asking for confirmation. Only point Quark at trusted providers/prompts, and run it in an environment you're comfortable with an autonomous agent touching.
+
+## Roadmap / not yet implemented
+
+- Streaming responses (print tokens as they arrive instead of waiting for the full reply).
+- Automatic context compaction for long-running sessions that approach the model's context window.
+
+## Development
+
+```bash
+pip install -e ".[dev]"
+pytest
+```
+
+Tests mock the HTTP layer (`httpx.post`), so the suite runs offline with no API keys required.
+
+## Publishing to PyPI
+
+The project builds as `quark-ai` (the CLI command stays `quark` either way — the package name on PyPI doesn't have to match the command it installs). `.github/workflows/publish.yml` publishes to PyPI via [Trusted Publishing](https://docs.pypi.org/trusted-publishers/) (OIDC) whenever a GitHub Release is published — no API token stored in the repo.
+
+One-time setup (done from your own PyPI account, not from CI):
+
+1. Create a PyPI account at [pypi.org](https://pypi.org) if you don't have one, and enable 2FA (PyPI requires it).
+2. Go to [pypi.org/manage/account/publishing](https://pypi.org/manage/account/publishing/) and add a **pending publisher**:
+   PyPI project name `quark-ai`, owner `rmoya81`, repository `Quark`, workflow `publish.yml`, environment `pypi`.
+   This reserves the name immediately — the project is created the moment the first publish succeeds, and only this repo's workflow can publish to it.
+3. In the GitHub repo, create an environment named `pypi` (Settings → Environments), matching the one in the workflow and in the pending publisher config.
+4. Cut a GitHub Release (tag `v0.1.0` or similar) — the workflow builds the sdist/wheel and publishes them automatically.
+
+To claim the name immediately instead of waiting on that setup (e.g. to block squatting right now), a one-off manual publish works too: `pip install build twine && python -m build && twine upload dist/*` with a PyPI API token — `quark-ai`/`quark-cli`/`quarkagent`/`quark-micro-agent`/`quark-llm-agent` are all currently unregistered as of this writing; `quark` and `quark-agent` are already taken by unrelated projects.

quark_ai-0.1.1/README.md

@@ -0,0 +1,162 @@
+# Quark
+
+Lightning-fast micro-agent. A stripped-down alternative to heavy AI agents, executing tasks with minimal prompt overhead.
+
+Quark is a minimal CLI agent loop: send a message, let the model call tools (read/edit/write files, list directories, optionally run shell commands), and get an answer back — with conversation history persisted to disk between runs. It talks to any LLM provider through a small abstraction layer (Anthropic and OpenAI included) instead of depending on a heavy vendor SDK.
+
+Two things it optimizes for on purpose:
+- **Lightweight**: no framework, no heavy SDKs (just `httpx`), a short default system prompt, and an `edit_file` tool so the model can make targeted changes instead of re-sending whole files.
+- **Productive by default, secure by default**: file tools are confined to a workspace directory (symlink escapes included), `run_shell` is off unless you opt in, and if a `QUARK.md`/`AGENTS.md` file exists it's folded into the system prompt automatically so the agent already knows about your project.
+
+## Install
+
+Requires Python 3.10+. Works anywhere Python does, including WSL2 — nothing platform-specific.
+
+```bash
+./install.sh
+```
+
+Creates `.venv`, installs Quark editable with dev deps, and copies `.env.example` to `.env` if you don't already have one. Safe to re-run (reuses the venv, never overwrites an existing `.env`).
+
+Or by hand:
+
+```bash
+python3 -m venv .venv
+source .venv/bin/activate
+pip install -e ".[dev]"
+```
+
+## Configure
+
+Copy `.env.example` to `.env` (done automatically by `install.sh`) and fill in the key(s) for the provider(s) you want to use:
+
+```bash
+cp .env.example .env
+```
+
+```
+ANTHROPIC_API_KEY=sk-ant-...
+OPENAI_API_KEY=sk-...
+QUARK_PROVIDER=anthropic
+```
+
+`.env` is loaded automatically from the current working directory. You can also export the variables directly, or pass `--api-key` on the command line.
+
+## Usage
+
+```bash
+# One-shot message, prints the reply and exits
+quark chat "list the files in this directory"
+
+# Interactive REPL (Ctrl+D or 'exit' to quit)
+quark chat
+
+# Choose a provider/model
+quark --provider openai --model gpt-4o chat "what's in README.md?"
+
+# Named sessions keep separate histories on disk
+quark --session project-a chat "remember that we're using Postgres"
+quark --session project-a chat "what database are we using?"
+
+# Confine file tools to a specific directory (default: cwd)
+quark --workspace ./my-project chat "list the files here"
+
+# Opt in to the run_shell tool (off by default)
+quark --allow-shell chat "run the test suite"
+
+# Read-only mode: only read_file + list_dir, smallest tool-schema footprint, no escape hatch
+quark --minimal-tools chat "what does this project do?"
+
+# Progressive mode: starts like --minimal-tools, but the model can unlock
+# write_file/edit_file/delete_file itself the moment it actually needs them
+quark --progressive-tools chat "fix the typo in README.md"
+
+# Manage sessions
+quark sessions list
+quark sessions clear project-a
+```
+
+Sessions are stored as JSON under `~/.quark/sessions/<name>.json`.
+
+### Project context
+
+If a `QUARK.md` or `AGENTS.md` file exists at the workspace root, its contents are appended to the system prompt automatically — a cheap way to give the agent standing context (stack, conventions, gotchas) without repeating it every session.
+
+### Progressive tool disclosure
+
+`--progressive-tools` (also `QUARK_PROGRESSIVE_TOOLS=1`) starts a turn with just `read_file`, `list_dir`, and one cheap extra tool, `request_more_tools` (no parameters). Its description tells the model to call it if it needs to write, edit, or delete files. If the model calls it, Quark registers `write_file`/`edit_file`/`delete_file` on the spot, drops `request_more_tools` (no longer needed), and the *same* `agent.step()` call retries with the expanded set — so the model still gets everything it needs, in the same turn, it just asks first.
+
+This is deliberately not a heuristic: matching intent from text (keywords, embeddings, a classifier) always trades accuracy for cost and can silently withhold a tool the model actually needed on some phrasing you didn't anticipate. Letting the model itself request more tools can't misfire that way — the only cost is one extra round trip, and only on turns that actually mutate something. `run_shell` stays independently gated behind `--allow-shell`; progressive mode never unlocks it.
+
+`--minimal-tools` is the stricter sibling: same starting set, but with no `request_more_tools` escape hatch at all, for when you want a hard read-only guarantee. Passing both flags together, `--minimal-tools` wins.
+
+### Token overhead
+
+On the first turn (empty session), before the model generates anything, Quark sends the system prompt plus the JSON schema of every registered tool. Rough token counts (`tiktoken` `cl100k_base`, an approximation — Claude's real tokenizer isn't public — but the right order of magnitude either way):
+
+| Mode | Tools registered | Tool-schema tokens | Total (system + tools + a short message) |
+|---|---|---|---|
+| default | `read_file`, `list_dir`, `write_file`, `edit_file`, `delete_file` | ~380 | ~415 |
+| `--minimal-tools` | `read_file`, `list_dir` | ~120 | ~155 |
+| `--progressive-tools`, before unlocking | `read_file`, `list_dir`, `request_more_tools` | ~165 | ~200 |
+| `--progressive-tools`, after unlocking | same as default | ~380 | ~415 |
+| default + `--allow-shell` | + `run_shell` | ~380 + ~65 | ~480 |
+
+`--minimal-tools` is cheapest but can never write/edit/delete. `--progressive-tools` costs a little more than minimal on read-only turns (~200 vs ~155) but is capability-equivalent to the default the moment it's actually needed — it only pays the full ~415 on turns that mutate something. Every subsequent turn in a session also carries the accumulated conversation history on top of these numbers.
+
+## Architecture
+
+```
+quark/
+  message.py       Internal Message/ToolCall representation
+  session.py        Persistent, named conversation history
+  agent.py           The chat + tool-calling loop
+  providers/          Provider abstraction (unified request/response shape)
+    base.py
+    anthropic.py       Anthropic Messages API
+    openai.py           OpenAI Chat Completions API
+  tools/               Built-in tools the agent can call
+    base.py             Tool + ToolRegistry
+    workspace.py         Path confinement (resolve_within)
+    fs.py                read_file / write_file / edit_file / delete_file / list_dir
+    shell.py             run_shell (opt-in)
+    meta.py              request_more_tools (--progressive-tools)
+  cli.py               argparse-based entry point
+```
+
+Adding a provider means implementing `Provider.complete()` in `quark/providers/` and registering it in `quark/providers/__init__.py`. Adding a tool means subclassing `Tool` in `quark/tools/` and adding it to `default_registry()`.
+
+## Security note
+
+- `read_file`, `write_file`, `edit_file`, `delete_file`, and `list_dir` are confined to `--workspace` (default: the current directory); paths that escape it, including via symlinks, are rejected.
+- `run_shell` executes arbitrary shell commands and is **off by default** — enable it with `--allow-shell` or `QUARK_ALLOW_SHELL=1` only when you trust the prompts/provider you're running against.
+- `write_file` and `delete_file` still mutate the filesystem inside the workspace without asking for confirmation. Only point Quark at trusted providers/prompts, and run it in an environment you're comfortable with an autonomous agent touching.
+
+## Roadmap / not yet implemented
+
+- Streaming responses (print tokens as they arrive instead of waiting for the full reply).
+- Automatic context compaction for long-running sessions that approach the model's context window.
+
+## Development
+
+```bash
+pip install -e ".[dev]"
+pytest
+```
+
+Tests mock the HTTP layer (`httpx.post`), so the suite runs offline with no API keys required.
+
+## Publishing to PyPI
+
+The project builds as `quark-ai` (the CLI command stays `quark` either way — the package name on PyPI doesn't have to match the command it installs). `.github/workflows/publish.yml` publishes to PyPI via [Trusted Publishing](https://docs.pypi.org/trusted-publishers/) (OIDC) whenever a GitHub Release is published — no API token stored in the repo.
+
+One-time setup (done from your own PyPI account, not from CI):
+
+1. Create a PyPI account at [pypi.org](https://pypi.org) if you don't have one, and enable 2FA (PyPI requires it).
+2. Go to [pypi.org/manage/account/publishing](https://pypi.org/manage/account/publishing/) and add a **pending publisher**:
+   PyPI project name `quark-ai`, owner `rmoya81`, repository `Quark`, workflow `publish.yml`, environment `pypi`.
+   This reserves the name immediately — the project is created the moment the first publish succeeds, and only this repo's workflow can publish to it.
+3. In the GitHub repo, create an environment named `pypi` (Settings → Environments), matching the one in the workflow and in the pending publisher config.
+4. Cut a GitHub Release (tag `v0.1.0` or similar) — the workflow builds the sdist/wheel and publishes them automatically.
+
+To claim the name immediately instead of waiting on that setup (e.g. to block squatting right now), a one-off manual publish works too: `pip install build twine && python -m build && twine upload dist/*` with a PyPI API token — `quark-ai`/`quark-cli`/`quarkagent`/`quark-micro-agent`/`quark-llm-agent` are all currently unregistered as of this writing; `quark` and `quark-agent` are already taken by unrelated projects.