Metadata-Version: 2.4
Name: gigai
Version: 0.1.10.7
Summary: Contract-first foundations for a local, user-controlled agent runtime
License-Expression: Apache-2.0
Classifier: Development Status :: 2 - Pre-Alpha
Classifier: Operating System :: MacOS
Classifier: Operating System :: POSIX :: Linux
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Requires-Python: >=3.11
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: click>=8.1
Requires-Dist: httpx>=0.27
Requires-Dist: jsonschema>=4.23
Requires-Dist: openpyxl>=3.1
Requires-Dist: pycountry>=24
Requires-Dist: pypdf>=5
Requires-Dist: python-dotenv>=1.0
Requires-Dist: questionary>=2.1.1
Requires-Dist: referencing>=0.36
Requires-Dist: typst<0.16,>=0.15
Provides-Extra: test
Requires-Dist: pydantic>=2.10; extra == "test"
Requires-Dist: pytest>=8.3; extra == "test"
Requires-Dist: pytest-xdist>=3.6; extra == "test"
Dynamic: license-file

# GigAI

GigAI runs AI work on your own computer, under your control. Your settings, your
credentials and your work stay on your machine, and every model call can be inspected.
It works through the model you already use: Codex, Claude Code, Ollama or OpenRouter.

**Scout** is its first Gig, for a job search. It keeps a local store of public job
postings, ranks them against your resume, assesses the ones you approve, asks you the
questions a posting leaves open, remembers your answers and stories, and drafts a
tailored resume you download as a PDF. You use it from the browser, or from your own AI
agent.

![An agent runs gigai scout new: what is new, and the cost, before anything is assessed.](https://raw.githubusercontent.com/karthik446/gigai/main/gigai-docs/public/media/terminal-new.png)

![The Scout Jobs page: stored postings your profiles match, with filter chips, profile tags and Scout's chips.](https://raw.githubusercontent.com/karthik446/gigai/main/gigai-docs/public/media/jobs-light.png)

## Quickstart

1. **Requirements.** macOS or Linux with Python 3.11+, [`uv`](https://docs.astral.sh/uv/getting-started/installation/),
   and one model CLI installed and logged in: Codex (`codex login`) or Claude Code (`claude`, then `/login`).
2. **Install.**

   ```bash
   uv tool install gigai
   ```

3. **Run.**

   ```bash
   gigai scout run      # starts Scout and opens the browser
   ```

   The setup wizard asks for your model, your resume (`.md` or `.txt`) and your
   roles. Then **Update sources** fills the store, the **Jobs** page lists what
   matches, **Assess these** assesses what you approve, and **Tailor resume**
   drafts a resume for one job. Update later with `uv tool upgrade gigai`.

## Use it from your AI agent

```bash
gigai agent-skill          # prints the instructions that teach your agent the loop
gigai agent-permissions    # prints a permissions snippet for you to apply; GigAI applies nothing
gigai scout new            # what is new since your last check
```

1. You ask your agent "what's new?". It runs `gigai scout new` and shows you a grid.
2. Scout asks before it assesses: you see the count and an estimate, and say yes or no.
3. The agent asks you each job's open questions and saves your answers. A longer reply
   becomes a story, if you agree.
4. Jobs you answered are tailored and scored in the background, within daily limits
   (10 jobs per trigger, 40 model calls a day; more waits for your approval).
5. For a PDF the agent gives you an "open in Scout" link. You add your name and contact
   details there, in your browser. The agent never gets them.

[For agents](https://karthik446.github.io/gigai/latest/scout/agents/) has the full workflow, the setup and an example session.

## Privacy

**GigAI never stores your name, email, phone, address or links.** You type them only
when you make a PDF, and GigAI forgets them right after. A resume you add is stored
without its name and contact lines; that removal works on patterns and can't catch
personal details elsewhere in the text, so keep those out. Earlier versions did
store contact lines: a one-time cleanup after you upgrade removes them from the current
files, but older copies can remain in GigAI's local history on your computer.

Everything else you give GigAI (your answers, stories, notes and the body of your
resume) stays on your computer unless you or your agent send it somewhere. GigAI's own
model calls send your resume (without the contact lines) and your answers to the model
you picked. With Ollama the resume stays on this machine.

**Anything GigAI gives your agent is sent to that agent's model provider. Agents get no contact data from GigAI, but an agent with shell access can read local files.**

See [Privacy and security](https://karthik446.github.io/gigai/latest/scout/privacy/) for exactly what is stored, what is sent, and to whom.

## Links

- [Documentation](https://karthik446.github.io/gigai/): quickstart, concepts, CLI and API reference, known limitations
- [What Scout's numbers and labels mean](https://karthik446.github.io/gigai/latest/scout/numbers/): rank, verdict, Scout label, Scout ATS score
- [Roadmap](https://karthik446.github.io/gigai/latest/roadmap/): what is planned, without dates
- [CHANGELOG](https://github.com/karthik446/gigai/blob/main/CHANGELOG.md) and [Releases](https://github.com/karthik446/gigai/releases)
- [CONTRIBUTING](https://github.com/karthik446/gigai/blob/main/CONTRIBUTING.md) and [issues](https://github.com/karthik446/gigai/issues)

GigAI is an alpha; expect rough edges. Apache-2.0, see [LICENSE](https://github.com/karthik446/gigai/blob/main/LICENSE).
