mirror of
https://github.com/headroomlabs-ai/headroom.git
synced 2026-11-22 18:41:07 +08:00
feat/https-compose-caddy
5
Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
|
|
a8905214d9 |
fix(packaging)!: keep PyTorch extras out of all (#4067)
<!-- maintainer-current-main -->
## Maintainer latest-main checkpoint
Published `fb146c0663b30488da0a49cfb33624ee82862074` contains current
upstream main `e4a60ed72f6ef5b28569f8bfbe2163cfda32c948`. 22
dependency/Gemini tests and lock check passed. Earlier live behavior
proofs below retain their commit/platform scope; contributor runtime
logic is unchanged.
The common shard-2 failure was independently confirmed on main's hosted
job 127069149772 and locally (four failures, six passes): the Gemini
memory fixture mocked an unused body reader and provided no request
stream. The maintainer test-only follow-up now constructs a real
FastAPI/ASGI Request carrying JSON bytes and exercises the actual body
reader. All six test-function ASTs/assertions are unchanged, and all 10
parameterized Gemini cases pass. Touched-file Ruff/format and whitespace
checks pass. This fixes the inherited fixture instead of asking
contributors to run CI or weakening the memory/content assertions. Fresh
exact-head hosted qualification is requested. Nothing merged.
<!-- /maintainer-current-main -->
## Description
On Linux, installing `headroom-ai[all]` resolved PyTorch and NVIDIA CUDA
packages even for users who only needed the CLI/proxy profile. The
locked dependency graph included those packages through `ml`, `memory`,
`evals`, and `voice`.
This keeps those PyTorch-backed extras available for explicit
installation while removing them from `[all]`. Image support and the
other CLI/proxy dependencies remain included; dependency metadata cannot
detect installed GPU hardware.
Closes #3339
## Type of Change
- [ ] Bug fix (non-breaking change that fixes an issue)
- [ ] New feature (non-breaking change that adds functionality)
- [x] Breaking change (fix or feature that would cause existing
functionality to change)
- [ ] Documentation update
- [ ] Performance improvement
- [ ] Code refactoring (no functional changes)
## Changes Made
- Remove `ml`, `memory`, `evals`, and `voice` from the `[all]` extra
while keeping each separately installable.
- Retain `image` and the other default CLI/proxy capabilities.
- Update README and installation docs to describe the opt-in extras and
CUDA package implications.
- Add a dependency contract assertion for the `[all]` extra.
## Testing
- [ ] Unit tests pass (`pytest`)
- [ ] Linting passes (`ruff check .`)
- [ ] Type checking passes (`mypy headroom`)
- [ ] New tests added for new functionality
- [x] Manual testing performed
### Test Output
```text
Focused dependency contract tests: 3 passed, 5 deselected.
Python 3.13 locked `uv export` exited 0; the fixed export omits torch, CUDA, NVIDIA, sentence-transformers, and datasets. The baseline export contained them.
`uv lock --check`, scoped Ruff check/format, and `git diff --check` exited 0.
The broader dependency test module had 3 passed and 5 import failures because `opentelemetry` is absent from the local environment.
```
## Real Behavior Proof
- Environment: Linux, Python 3.13, locked uv dependency resolution.
- Exact command / steps: `UV_PYTHON_DOWNLOADS=never uv export --python
3.13 --extra all --no-dev --locked --no-hashes --no-emit-project`.
- Observed result: exit 0; the resolved export no longer contains torch,
CUDA, NVIDIA, sentence-transformers, or datasets, all of which were
present in the baseline export.
- Not tested: a full environment install and GPU runtime behavior.
## Runtime Rollout Safety
- Rollout-managed feature(s): None.
- Minimum rollout channel: Stable package release.
- Stable/default behavior changed: Yes; `[all]` no longer installs the
four PyTorch-backed extras.
- Kill switch / disable path: Not applicable; install the desired extras
explicitly.
- Unsafe override required: No.
- Qualification impact: Users of those capabilities must add the
corresponding extras.
- Rollback path: Restore the extras to `[all]` and regenerate `uv.lock`.
## Review Readiness
- [x] I have performed a self-review
- [x] This PR is ready for human review
## Checklist
- [x] My code follows the project's style guidelines
- [x] I have performed a self-review of my code
- [ ] I have commented my code, particularly in hard-to-understand areas
- [x] I have made corresponding changes to the documentation
- [ ] My changes generate no new warnings
- [x] I have added tests that prove my fix is effective or that my
feature works
- [ ] New and existing unit tests pass locally with my changes
- [x] I did **not** edit `CHANGELOG.md` — it is generated by
release-please from my Conventional Commit PR title (a CI guard enforces
this)
## Screenshots (if applicable)
Not applicable to this dependency-resolution change.
## Additional Notes
The `[all]` install profile intentionally no longer includes
PyTorch-backed optional capabilities; each remains available through its
named extra. The focused regression and locked resolver proof passed,
but the full dependency test module and a full install were not verified
locally.
## Maintainer integration qualification
Published head `3229690b613810d496f015cedd086ca1d7421843` includes
current upstream main `31dcaf41`. The merge preserves main's `lean`
dependency profile and deprecated `sandbox` alias alongside the
contributor's opt-in `[all]` policy. No optional-feature runtime logic
was changed.
- Reconciled the existing macOS dependency test with the declared
policy, retaining the transitive exclusion check and independently
verifying explicit `ml`/`voice` opt-ins on Intel macOS and Linux. All 12
tests across `tests/test_onnx_dependency_contract.py` and
`tests/test_optional_dependencies.py` pass on Windows/Python 3.12.13.
- Evaluation workflow installs now explicitly request `evals`, and the
model-allowed weekly job also requests `ml`, preserving its previous
intended dependencies. The evaluation guide and AI install index
identify those opt-ins. These address the existing Devin workflow/docs
findings.
- `uv lock --check`, touched Python Ruff/format and whitespace checks
pass. All 253 locked package versions are unchanged; only the root
`[all]` dependency metadata/flattened list changes.
- Real resolver commands from this checkout: `uv export --extra all
--no-dev --locked --no-hashes --no-emit-project` and `uv export --extra
all --extra ml --extra voice --extra evals --no-dev --locked --no-hashes
--no-emit-project`. Evaluating their actual requirement markers for
Python 3.12/3.13/3.14 on Linux x86_64, Windows AMD64, macOS arm64 and
macOS x86_64 shows `[all]` excludes torch, sentence-transformers,
datasets and CUDA/NVIDIA packages in all 12 environments. Explicit
opt-ins retain datasets and torch where supported; Intel macOS still
excludes torch.
- Not tested: a fresh full environment install, GPU runtime, live paid
evaluations or a fresh native build. Hosted CI must qualify the
published final head separately.
### Maintainer current-main requalification
Exact head `f1ed1e0019037e7b53ff9ef548843449c3c7a16f`, upstream main
`c01cf07d943d816cfd9d97e6838ed4e38f1ff9a4`.
Current-main integration against upstream
|
||
|
|
b5c52647b4 |
test(deps): fail when a composite extra's uv.lock list drops a nested extra's package (#4598)
## Description
Adds a regression test for a `uv.lock` drift that left `uv sync --extra
lean` without `python-socks`, while `--extra proxy` and the deprecated
`--extra sandbox` alias (which is just `lean`) installed it.
**Scope change:** this PR originally also carried the one-line `uv.lock`
fix. #4614 has since re-locked `main` and picked up the same
`python-socks` line under `lean` ("uv flattening the existing `proxy`
extra that `lean` already includes", per its commit message), so after
rebasing onto `main` (`7e0094d6`) the fix commit was empty and dropped.
What remains is the test that catches this class of drift.
Background: #3357 added `python-socks[asyncio]` to the `proxy` extra
after #3727 had renamed `sandbox` to `lean`. In `uv.lock`, each extra
has its own flattened dependency list, and the `lean` list kept its old
contents. `uv lock --check` accepts a stale list and a plain `uv lock`
leaves it alone, so nothing in CI caught it. The existing
`test_proxy_extra_includes_socks_support_for_websocket_proxy_discovery`
only reads `pyproject.toml`, which is why it stayed green.
## Type of Change
- [ ] Bug fix (non-breaking change that fixes an issue)
- [ ] New feature (non-breaking change that adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to change)
- [ ] Documentation update
- [ ] Performance improvement
- [ ] Code refactoring (no functional changes)
- [x] Test-only change (regression guard)
## Changes Made
- `tests/test_optional_dependencies.py`:
`test_locked_composite_extras_include_their_nested_extras` checks that
every composite extra in `uv.lock` (`lean`, `sandbox`, `all`,
`proxy-prod`, `voice-train`, read from the `headroom-ai[...]`
self-references in `requires-dist`) lists each package its nested extras
list, with at least the same package extras. It reads `uv.lock` directly
with `tomllib`, so it needs no `uv` binary and runs in the normal pytest
shards.
No `uv.lock`, `pyproject.toml` or runtime code changes.
## Testing
- [x] Unit tests pass (`pytest`)
- [x] Linting passes (`ruff check .`)
- [ ] Type checking passes (`mypy headroom`)
- [x] New tests added for new functionality
- [x] Manual testing performed
### Test Output
```text
$ pytest -q tests/test_optional_dependencies.py # this branch, on main
|
||
|
|
79181ca0fe |
fix(proxy): install SOCKS support for websocket proxies (#3357)
## Description Install the SOCKS transport dependency required when `websockets` discovers a system SOCKS proxy. Without it, Codex Responses WebSocket relay logs `connecting through a SOCKS proxy requires python-socks` and cannot establish the upstream WebSocket. ## Type of Change - [x] Bug fix (non-breaking change that fixes an issue) - [ ] New feature (non-breaking change that adds functionality) - [ ] Breaking change (fix or feature that would cause existing functionality to change) - [ ] Documentation update - [ ] Performance improvement - [ ] Code refactoring (no functional changes) ## Changes Made - Add `python-socks[asyncio]>=2.7.1` to the `proxy` extra. - Regenerate `uv.lock`, resolving `python-socks` 3.1.1. Existing package versions are unchanged relative to the refreshed main parent. - Add a package-contract regression test that asserts `headroom-ai[proxy]` includes the SOCKS transport dependency. ## Testing - [x] Focused optional-dependency tests pass - [x] Full branch Ruff lint and formatting checks pass - [ ] Type checking - [x] Package-contract regression test added - [x] Manual SOCKS runtime verification performed ### Test Output ```text $ python -m pytest tests/test_optional_dependencies.py -q 3 passed $ uv lock --check Resolved 252 packages in 5ms $ ruff check . All checks passed! $ ruff format --check . 1715 files already formatted ``` Validation used Python 3.12.11 and Ruff 0.16.8. Comparing package versions against the refreshed main parent found no existing package version changes; only `python-socks` 3.1.1 was added. The full test suite was not rerun locally. The existing hosted `test (2)` failure is `test_category_losing_its_last_pattern_drops_its_heading`, which the maintainer previously reproduced on main. ## Real Behavior Proof - Environment: macOS 26.6.2, Apple Silicon arm64, Python 3.12.11, `websockets` 16.0, locked `python-socks` 3.1.1, live SOCKS proxy at `127.0.0.1:7897`, existing Sub2API OpenAI-compatible provider. - Exact steps: create an isolated environment containing `websockets==16.0` and no `python-socks`. Set `wss_proxy=socks5://127.0.0.1:7897` so `websockets` automatically discovers the SOCKS proxy. Call `websockets.connect(<real upstream /v1/responses WebSocket URL>, additional_headers=<existing bearer>, ssl=<configured TLS context>, open_timeout=30)` without an explicit proxy argument. Repeat with `python-socks[asyncio]==3.1.1` installed, using the same endpoint, proxy, credential, and TLS context. - Installation commands: ```sh uv venv <before-env> --python 3.12 uv pip install --python <before-env>/bin/python websockets==16.0 uv pip install --python <after-env>/bin/python 'python-socks[asyncio]==3.1.1' ``` Captured runtime output on 2026-10-07: ```text # Before Python: 3.12.11 websockets: 16.0 ; python-socks: not installed Discovery: wss_proxy=socks5://127.0.0.1:7897; provider: existing Sub2API endpoint Upstream WebSocket handshake: ImportError connecting through a SOCKS proxy requires python-socks # After Python: 3.12.11 websockets: 16.0 ; python-socks: 3.1.1 Discovery: wss_proxy=socks5://127.0.0.1:7897; provider: existing Sub2API endpoint Upstream WebSocket handshake: InvalidStatus server rejected WebSocket connection: HTTP 426 ``` - Observed result: the missing transport error is gone and the real upstream returns an HTTP handshake response after connecting through SOCKS. The upstream's HTTP 426 remains; this is not evidence of a successful upstream WebSocket session. - Provider/model scope: this is a transport/handshake check, so no model request was sent. The fresh validation above specifically verifies the locked 3.1.1 transport rather than claiming a model completion over WebSocket. - Version clarification: the original manual observation preceded the refreshed lockfile. This new before/after run explicitly tests the currently locked 3.1.1; it does not claim every version allowed by `>=2.7.1` was tested. - Not tested: clean install of a built Headroom distribution, Linux/Windows runtime, successful upstream WebSocket inference, or the full suite. ## Runtime Rollout Safety - Rollout-managed feature(s): none. - Minimum rollout channel: N/A. - Stable/default behavior changed: `headroom-ai[proxy]` and extras that include it install one pure-Python SOCKS transport dependency; no request routing or compression behavior changes without a SOCKS proxy. - Kill switch / disable path: remove SOCKS proxy configuration from the process/system environment. - Unsafe override required: no. - Qualification impact: package extras and lockfile only. - Rollback path: revert this commit or deploy the prior package release. ## Review Readiness - [x] I have performed a self-review - [x] This PR is ready for human review ## Checklist - [x] My code follows the project's style guidelines - [x] I have performed a self-review of my code - [x] I have commented my code, particularly in hard-to-understand areas - [ ] I have made corresponding changes to the documentation - [x] My changes generate no new warnings - [x] I have added tests that prove my fix is effective or that my feature works - [ ] New and existing unit tests pass locally with my changes - [x] I did **not** edit `CHANGELOG.md` — it is generated by release-please from my Conventional Commit PR title (a CI guard enforces this) ## Screenshots (if applicable) N/A. ## Additional Notes ### Dependency review - **Why this package:** `websockets` delegates its optional SOCKS transport to `python-socks`; adding the `asyncio` extra supplies the API its runtime imports. Reimplementing SOCKS in Headroom would duplicate that upstream transport without changing the integration point. - **Maintainer and activity:** upstream is [romis2012/python-socks](https://github.com/romis2012/python-socks), maintained under the `romis2012` account. The repository is not archived. Recent GitHub releases include 2.8.2 (2026-06-23), 3.0.0 (2026-08-11), and 3.1.0/3.1.1 (2026-09-08). - **Security history checked:** an [OSV](https://osv.dev/) query for PyPI `python-socks` 3.1.1 on 2026-10-07 returned no advisory records for that version. This is a bounded database check, not a claim that the package has never had a security issue. - **Install surface:** the locked wheel is `py3-none-any` (pure Python). On Python <3.11, the `asyncio` extra additionally requires `async-timeout>=5.0.1`; on Python >=3.11 it adds no extra dependency. The `trio` and `anyio` extras are not selected. Installation downloads the package from PyPI; using the transport connects to the configured SOCKS proxy and upstream. No separate native build or model download is needed by this dependency. - **Why this version:** the new dependency is needed for SOCKS functionality in `websockets`, not a cosmetic upgrade. The requirement remains `python-socks[asyncio]>=2.7.1`; the refreshed lock resolves 3.1.1, and the current runtime proof tests that exact release. Existing package versions are unchanged relative to the refreshed main parent. The package-contract test verifies the install extra. The real network handshake above separately verifies runtime transport discovery. <!-- devin-review-badge-begin --> --- <a href="https://app.devin.ai/review/headroomlabs-ai/headroom/pull/3357" target="_blank"><picture><source media="(prefers-color-scheme: dark)" srcset="https://static.devin.ai/assets/gh-devin-review-dark.svg?v=4"><img src="https://static.devin.ai/assets/gh-devin-review-light.svg?v=4" alt="Devin Review"></picture></a> <!-- devin-review-badge-end --> ## Maintainer qualification — current main refresh Current head: `0caa86656b9876705bfd80bd590070159e70df27`, containing main `01d9194854233ef6ed04005cf0a5b8c32bf65d6d`. The metadata/lock conflict retains current main's packaging and adds the original `python-socks[asyncio]>=2.7.1` proxy dependency. The regenerated lock preserves all 251 main package identities and all existing non-project package records, adding only `python-socks` **3.1.1**. Current qualification targets that locked release; any earlier 3.0.0 report is historical rather than evidence for the currently locked package. Validation: three optional-dependency tests pass; Ruff, formatting, and changed-file whitespace checks pass. A controlled runtime check on Windows with Python **3.12.13**, `python-socks` **3.1.1**, and `websockets` **16.0** established a WebSocket through a loopback SOCKS5 server and round-tripped `headroom-socks-runtime-probe`. Captured output: `PASS loopback WebSocket through SOCKS5: Python 3.12.13, python-socks 3.1.1, websockets 16.0; no provider/model used`. This is a local transport smoke check, with no provider/model or external provider request. It establishes the selected transport dependency loads and works together with the WebSocket client; it does not claim production-provider qualification. Hosted checks on the refreshed head still require completion, and the existing dependency-provenance review remains available for reconciliation. No PR merge performed. Dependency provenance for the currently locked **3.1.1** release: PyPI and installed wheel metadata identify `https://github.com/romis2012/python-socks`; the upstream repository is not archived and reports activity on **2026-09-08**. The license is **Apache-2.0**. The installed wheel is pure Python (`Root-Is-Purelib: true`, `py3-none-any`). The selected asyncio extra requires `async-timeout>=5.0.1` only on Python below 3.11; Trio and AnyIO dependencies are optional extras and are not selected by this change. A targeted OSV query for PyPI `python-socks` **3.1.1** on **2026-10-09** returned no vulnerability records. Captured package metadata, repository metadata, OSV response, and the controlled transport output support this current-release qualification. Co-authored-by: JerrettDavis <mxjerrett@gmail.com> |
||
|
|
4f3d5ab341 |
fix(install): add orjson to [proxy] extra for LiteLLM provider backends (#2074)
## Description Add `orjson` to the `[proxy]` extra so `uv tool install "headroom-ai[all]"` installs a runtime dependency required by LiteLLM provider backends (e.g. OpenRouter). ## Type of Change - [x] Bug fix (non-breaking change which fixes an issue) - [ ] New feature - [ ] Breaking change - [ ] Documentation update ## Changes Made - Add `orjson>=3.9.14; platform_python_implementation != 'PyPy'` to `[proxy]`. - Regression test in `tests/test_optional_dependencies.py`. - Minimal `uv.lock` update (proxy/all optional-deps + metadata only). ## Motivation Fixes #2056. `headroom-ai[all]` installs `litellm` (core dep) but not `orjson`. LiteLLM provider backends import `orjson` at runtime; LiteLLM only declares it under `litellm[proxy]`, not base deps. Headroom does not depend on `litellm[proxy]` (would pull the full LiteLLM proxy server stack). ## Testing - `uv run --extra dev python -m pytest tests/test_optional_dependencies.py -q` — 2 passed - `uvx --from ruff==0.15.17 ruff check tests/test_optional_dependencies.py` - `uvx --from ruff==0.15.17 ruff format --check tests/test_optional_dependencies.py` ## Real Behavior Proof - **Setup:** Ubuntu, Python 3.12.3 - **Verified:** dependency graph test confirms `orjson` is selected for `[proxy]` and `[all]` extras after this patch. - **Not tested locally:** full `uv tool install` end-to-end (local sdist build requires Rust/C++ toolchain unavailable in this environment). Reporter workaround `uv tool install ... --with orjson` confirms the missing transitive dep diagnosis. ## Review Readiness - [x] I have performed a self-review of my code - [x] This PR is ready for human review ## Notes - Reporter used Python 3.14.4; `litellm` is intentionally skipped on 3.14 (GH #956). This PR fixes the missing `orjson` install path for supported Python versions using `[all]`. Co-authored-by: syf2211 <syf2211@users.noreply.github.com> Co-authored-by: JerrettDavis <mxjerrett@gmail.com> Co-authored-by: Tejas Chopra <chopratejas@gmail.com> |
||
|
|
fd0d29c92d |
fix(packaging): guard torch extras on intel macos (#2011)
## Description Closes #1931 Guard the `ml` and `voice` `torch` optional dependencies on macOS x86_64 so `headroom-ai[all]` remains resolvable on Intel Macs where PyTorch does not publish compatible wheels for this version floor. The lockfile metadata is updated with the same markers. ## Type of Change - [x] Bug fix (non-breaking change which fixes an issue) - [ ] New feature (non-breaking change that adds functionality) - [ ] Breaking change (fix or feature that would cause existing functionality to not work as expected) - [ ] Documentation update - [ ] Refactoring - [ ] Performance improvement - [ ] Test update - [ ] Other ## Changes Made - Added macOS x86_64 environment markers to `torch` in the `ml` and `voice` extras. - Updated `uv.lock` optional dependency metadata to match the guarded extras. - Added a packaging regression test that checks `[all]` keeps `ml` and `voice` while guarding `torch` on macOS x86_64. ## Testing - [x] Unit tests pass (`pytest`) - [x] Linting passes (`ruff check`) - [x] Formatting verified (`ruff format --check`) - [ ] Manual testing performed ### Test Output ```text $ python3 -m pytest tests/test_optional_dependencies.py -q collected 1 item tests/test_optional_dependencies.py . [100%] ============================== 1 passed in 0.26s =============================== $ .venv/bin/ruff check tests/test_optional_dependencies.py All checks passed! $ .venv/bin/ruff format --check tests/test_optional_dependencies.py pyproject.toml 1 file already formatted ``` ## Test verification (RED -> GREEN) RED, with the `torch` markers temporarily removed from `pyproject.toml`: ```text tests/test_optional_dependencies.py F [100%] FAILED tests/test_optional_dependencies.py::test_all_extra_does_not_require_torch_on_macos_x86_64 E assert False ``` GREEN, with this patch applied: ```text tests/test_optional_dependencies.py . [100%] ============================== 1 passed in 0.26s =============================== ``` ## Real Behavior Proof - Environment: Linux, Python 3.12.3, pytest 9.1.1, ruff 0.14.14. - Exact command / steps: Removed the environment markers from `torch`, ran the new packaging test, restored the markers, and reran the test plus targeted ruff checks. - Observed result: The test fails without the macOS x86_64 guard and passes once the `ml` and `voice` `torch` requirements are guarded. - Not tested: Full `uv run pytest`, full-project `uv run ruff check .`, full-project `uv run ruff format --check .`, and `uv run mypy headroom` were not run locally for this targeted packaging change. ## Review Readiness - [x] I have performed a self-review - [x] This PR is ready for human review ## Checklist - [x] My code follows the style guidelines of this project - [x] I have performed a self-review of my code - [x] I have added tests that prove my fix is effective - [x] New and existing targeted tests pass locally with my changes - [x] Any dependent changes have been merged and published in downstream modules ## Screenshots (if applicable) N/A ## Additional Notes No new dependency is added; this only narrows when the existing `torch` optional dependency is selected. |