Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
133 changes: 133 additions & 0 deletions CODE_OF_CONDUCT.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,133 @@
# Contributor Covenant Code of Conduct

## Our Pledge

We as members, contributors, and leaders pledge to make participation in our
community a harassment-free experience for everyone, regardless of age, body
size, visible or invisible disability, ethnicity, sex characteristics, gender
identity and expression, level of experience, education, socio-economic status,
nationality, personal appearance, race, caste, color, religion, or sexual
identity and orientation.

We pledge to act and interact in ways that contribute to an open, welcoming,
diverse, inclusive, and healthy community.

## Our Standards

Examples of behavior that contributes to a positive environment for our
community include:

* Demonstrating empathy and kindness toward other people
* Being respectful of differing opinions, viewpoints, and experiences
* Giving and gracefully accepting constructive feedback
* Accepting responsibility and apologizing to those affected by our mistakes,
and learning from the experience
* Focusing on what is best not just for us as individuals, but for the overall
community

Examples of unacceptable behavior include:

* The use of sexualized language or imagery, and sexual attention or advances of
any kind
* Trolling, insulting or derogatory comments, and personal or political attacks
* Public or private harassment
* Publishing others' private information, such as a physical or email address,
without their explicit permission
* Other conduct which could reasonably be considered inappropriate in a
professional setting

## Enforcement Responsibilities

Community leaders are responsible for clarifying and enforcing our standards of
acceptable behavior and will take appropriate and fair corrective action in
response to any behavior that they deem inappropriate, threatening, offensive,
or harmful.

Community leaders have the right and responsibility to remove, edit, or reject
comments, commits, code, wiki edits, issues, and other contributions that are
not aligned to this Code of Conduct, and will communicate reasons for moderation
decisions when appropriate.

## Scope

This Code of Conduct applies within all community spaces, and also applies when
an individual is officially representing the community in public spaces.
Examples of representing our community include using an official email address,
posting via an official social media account, or acting as an appointed
representative at an online or offline event.

## Enforcement

Instances of abusive, harassing, or otherwise unacceptable behavior may be
reported to the community leaders responsible for enforcement at
**ernest@thedataexperts.us**.

All complaints will be reviewed and investigated promptly and fairly.

All community leaders are obligated to respect the privacy and security of the
reporter of any incident.

## Enforcement Guidelines

Community leaders will follow these Community Impact Guidelines in determining
the consequences for any action they deem in violation of this Code of Conduct:

### 1. Correction

**Community Impact**: Use of inappropriate language or other behavior deemed
unprofessional or unwelcome in the community.

**Consequence**: A private, written warning from community leaders, providing
clarity around the nature of the violation and an explanation of why the
behavior was inappropriate. A public apology may be requested.

### 2. Warning

**Community Impact**: A violation through a single incident or series of
actions.

**Consequence**: A warning with consequences for continued behavior. No
interaction with the people involved, including unsolicited interaction with
those enforcing the Code of Conduct, for a specified period of time. This
includes avoiding interactions in community spaces as well as external channels
like social media. Violating these terms may lead to a temporary or permanent
ban.

### 3. Temporary Ban

**Community Impact**: A serious violation of community standards, including
sustained inappropriate behavior.

**Consequence**: A temporary ban from any sort of interaction or public
communication with the community for a specified period of time. No public or
private interaction with the people involved, including unsolicited interaction
with those enforcing the Code of Conduct, is allowed during this period.
Violating these terms may lead to a permanent ban.

### 4. Permanent Ban

**Community Impact**: Demonstrating a pattern of violation of community
standards, including sustained inappropriate behavior, harassment of an
individual, or aggression toward or disparagement of classes of individuals.

**Consequence**: A permanent ban from any sort of public interaction within the
community.

## Attribution

This Code of Conduct is adapted from the [Contributor Covenant][homepage],
version 2.1, available at
[https://www.contributor-covenant.org/version/2/1/code_of_conduct.html][v2.1].

Community Impact Guidelines were inspired by
[Mozilla's code of conduct enforcement ladder][Mozilla CoC].

For answers to common questions about this code of conduct, see the FAQ at
[https://www.contributor-covenant.org/faq][FAQ]. Translations are available at
[https://www.contributor-covenant.org/translations][translations].

[homepage]: https://www.contributor-covenant.org
[v2.1]: https://www.contributor-covenant.org/version/2/1/code_of_conduct.html
[Mozilla CoC]: https://github.com/mozilla/diversity
[FAQ]: https://www.contributor-covenant.org/faq
[translations]: https://www.contributor-covenant.org/translations
134 changes: 134 additions & 0 deletions CONTRIBUTING.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,134 @@
# Contributing to conclave

Thanks for helping improve conclave — a bring-your-own-keys multi-model council
for the CLI and as a Python library. Contributions are held to a clear bar:
tests green, the **BYO-keys** posture preserved, and the design specs in `docs/`
kept authoritative. This guide gets you set up and explains the rules that keep
the project trustworthy.

By participating you agree to the [Code of Conduct](CODE_OF_CONDUCT.md). Never
report security vulnerabilities as public issues or PRs — follow
[`SECURITY.md`](SECURITY.md).

---

## Dev setup

conclave requires **Python ≥ 3.11**. Use a project-local virtual environment.

```bash
# from a clone of this repo
python -m venv .venv
.venv/bin/pip install -e ".[dev]"

# the CLI is then available as:
.venv/bin/conclave --help
```

`uv` works too if you prefer it (`uv venv .venv && uv pip install --python
.venv/bin/python -e ".[dev]"`).

conclave never needs real API keys to develop or test against — the suite mocks
the httpx transport, so it runs fully offline. You only need real keys (set as
environment variables) to make live calls.

---

## Running tests

```bash
# full suite
.venv/bin/python -m pytest -q

# a single file or test
.venv/bin/python -m pytest tests/test_council.py -q
.venv/bin/python -m pytest tests/test_modes.py::test_debate_dropout -q
```

The suite is **offline and key-free**: `tests/conftest.py` mocks the httpx
transport so no network call or credential is ever made. A change is not done
until `.venv/bin/python -m pytest -q` is fully green. CI runs the same suite on
Python 3.11 / 3.12 / 3.13, plus `ruff` (lint + format) and a coverage floor and
a `gitleaks` secret scan — all five checks must pass to merge.

---

## The BYO-keys contract (read before touching internals)

conclave's core promise is that a user's API keys are **never stored, logged, or
serialized** — they are read from the environment by **variable name**, at call
time, and used only to make the request. If your change touches any of the
following, it must preserve that contract:

- **Key resolution** (`registry.py`, `providers.py`) — keys are looked up by env
var name at call time. Never attach a key value to an object, a log record, a
`ModelAnswer`, or a `CouncilResult`. Key **presence** logic may be exposed;
key **values** never.
- **Error capture** (`providers.py`, `adapters/base.py`) — provider error strings
can contain credentials echoed back by the upstream API. They MUST pass through
`redact()` before landing in `ModelAnswer.error` or any log. A change that adds
a new error path must route it through `redact()`.
- **The `redact()` scrubber** (`adapters/base.py`) — it strips bearer tokens,
`sk-`/`x-api-key` shapes, and known env-var values. Detection must **never
weaken on a real secret** to reduce noise. Add coverage; do not drop it.
- **`CouncilResult` / `ModelAnswer` serialization** (`models.py`) — these are the
stable downstream surface (the mcp-warden dev-time consumer keys on them). No
field may carry a credential, and the shape must stay backward-compatible
unless the change is intentional and version-bumped.

Practical rules:

- **No real secrets in the tree, ever.** Tests use obviously-synthetic,
clearly-fake placeholders. The repo runs a `gitleaks` scan on every push and PR.
- **Adding a provider is one adapter registration.** Follow the existing pattern
in `adapters/` — register one adapter per provider family. An OpenAI-compatible
endpoint needs **no code**: it is configured under the `endpoints:` section of
`~/.conclave/config.yml`. Don't add a bespoke code path for something the
config seam already covers.
- **`call_model` never raises.** The whole provider highway is partial-failure
resilient — a member that errors is captured as a failed `ModelAnswer`, not an
exception that aborts the council. Preserve that.

---

## The PDD is the source of truth

`docs/PRODUCT_DESIGN_DOCUMENT.md` is the **canonical** product and architecture
spec; the 3-core docs (`README.md`, `SYSTEM_CONTEXT_DIAGRAM.md`,
`DOCUMENTATION_INDEX.md`) sit on top of it. When docs disagree, the PDD wins, and
your PR must reconcile them. A behavior or surface change updates the relevant doc
in the same PR.

---

## Proposing a feature

Modes, providers, and output options are welcome. The path that gets a change
merged:

1. **Open an issue first** describing the use case and the proposed surface
(CLI flag / library API / config key). Roadmap items live as labeled issues.
2. **Check the PDD.** If your change shifts scope, non-goals, or the provider bar
(§12), say how — the PDD decision history is authoritative.
3. **Implement** following existing patterns (the provider highway, the mode
orchestration in `modes.py`, the `Council` API in `council.py`).
4. **Test** it: cover the happy path, partial failure, and — for anything in the
key/error path — that no credential can leak.

---

## Pull request expectations

Before you open a PR, confirm:

- [ ] **Tests pass**: `.venv/bin/python -m pytest -q` is green.
- [ ] **BYO-keys preserved.** No key value is stored, logged, or serialized; new
error paths route through `redact()`.
- [ ] **No secrets.** No real credentials anywhere in the tree; fixtures use
obviously-fake placeholders. `gitleaks` is clean.
- [ ] **Lint clean.** `ruff check` and `ruff format --check` pass.
- [ ] **Docs in sync.** README/CLI reference updated if user-facing behavior or
flags changed; the PDD and 3-core docs stay accurate.
- [ ] **Scoped commits** with a clear message describing the *why*.

Thank you for keeping the council honest.
93 changes: 93 additions & 0 deletions SECURITY.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,93 @@
# Security Policy

conclave is a **bring-your-own-keys** multi-model council. It calls foundation-model
APIs with the user's own credentials, read from the environment by **variable name
only**. The most security-sensitive surface in the project is therefore key
handling: a weakness that causes a real API key to be stored, logged, serialized,
or echoed back to the user breaks the core trust promise. We treat reports against
that surface as the highest priority.

## Reporting a vulnerability

**Do not open a public GitHub issue, pull request, or discussion for a security
vulnerability.** Public disclosure before a fix is available puts every user's
credentials at risk.

Report privately through **either** channel (GitHub Security Advisories is
preferred because it keeps the report, the fix, and the CVE in one place):

1. **GitHub Security Advisories** — go to the repository's **Security** tab and
click **"Report a vulnerability"**
(<https://github.com/ernestprovo23/conclave/security/advisories/new>). This
opens a private advisory visible only to you and the maintainers.
2. **Email** — `ernest@thedataexperts.us`. Use a clear subject line such as
`[conclave security]`. If you want to encrypt, say so in a first plaintext
email and we will arrange a key.

### What to include

A good report lets us reproduce and triage fast:

- The conclave version or commit SHA.
- The surface involved (`conclave ask` / `conclave providers`, or the library
entry points `Council.ask` / `debate` / `adversarial`), the mode, and the
relevant flags or config.
- A minimal repro: the `~/.conclave/config.yml` (with key **names**, never
values), the provider/endpoint, and the input that triggers it. Strip or fake
any real credentials first.
- The expected vs. actual behavior and the security impact — e.g. "a real key
appears unredacted in `ModelAnswer.error`", "a key value is written to a log
line", "`CouncilResult` JSON serialization leaks a credential", "the CLI prints
a key value".

Reports that demonstrate a **credential leak** — a real key escaping into any
result field, log, serialized payload, or terminal output — are the highest
priority. The key-handling contract is: keys are read from the environment by
name at call time, never stored on objects, never logged, never serialized; and
provider error strings are scrubbed by `redact()` before they reach a result.

## Supported versions

Security fixes are issued for the latest minor series. Older series are not
patched — upgrade to a supported release.

| Version | Supported |
| ------- | ------------------ |
| 0.3.x | :white_check_mark: |
| < 0.3 | :x: |

> Pre-1.0 note: the public surface is still evolving. The supported series will
> advance with each minor release; only the most recent `0.x` minor receives
> security patches.

## Response window

We aim to:

- **Acknowledge** your report within **3 business days**.
- Provide an **initial assessment** (accepted / needs-info / not-a-vuln, with a
severity estimate) within **7 business days**.
- Ship a fix or a documented mitigation for accepted, validated reports within
**30 days** of acknowledgement for high/critical severity, and on a best-effort
basis for lower severities.

These are targets for a small maintainer team, not contractual SLAs. If a report
stalls, a polite nudge to `ernest@thedataexperts.us` is welcome.

## Disclosure & credit

We follow coordinated disclosure. We will work with you on a disclosure timeline,
publish a GitHub Security Advisory (and request a CVE where warranted) once a fix
is available, and credit you in the advisory unless you ask to remain anonymous.

## Scope notes for this repository

- conclave never persists credentials. If you find a path where a key is written
to disk, a log, a result object, or stdout/stderr, that is in scope.
- Reports about the upstream model providers themselves (OpenAI, Anthropic,
Google, xAI, Perplexity) or about third-party dependencies (httpx, pydantic,
typer) are best filed upstream — but tell us too if conclave's *use* of them is
exploitable.
- conclave is a council aggregator, not a security control. The *content* a model
returns is not adjudicated for safety; that is out of scope. A leak of the
user's own credentials is the security boundary we defend.
Loading