Hardens client-selected upstreams, memory identity resolution, downloaded binary integrity, telemetry import, Docker defaults, Neo4j credentials, and archive extraction. Refreshes the branch against current main and preserves newer same-origin and loopback protections.
Adds crates/headroom-core/src/transforms/code_compressor.rs (1,882 lines): the
AST-aware CodeCompressor ported to Rust on tree-sitter, with grammars for
Python, JavaScript, TypeScript, Go, Rust, Java, C and C++.
Parity-only, like #1153. Nothing calls it: the only references outside the
module are the pub mod / pub use declarations in transforms/mod.rs, and
live_zone.rs still routes SourceCode to a no-op. The pyo3 bridge is untouched
and no Python source changes, so the engine is unreachable from the shipped
package. #1155 wires it into live-zone dispatch.
Every grammar is pinned with '=' to the exact version of the corresponding
Python tree-sitter-<lang> PyPI wheel. Same version on crates.io and PyPI means
the same grammar.js, hence the same generated parser.c, hence node-for-node
identical ASTs — the precondition for byte-parity. A canary over 9 samples x 8
languages confirmed identical node-type and line-span trees at these pins;
bumping any pin requires re-running it and re-recording the fixtures.
Ships 30 recorded parity fixtures, a CodeCompressorComparator in
headroom-parity, and scripts/record_code_compressor_fixtures.py.
Verified byte-identical to the recorded Python output:
[code_aware_compressor] total=30 matched=30 skipped=0 diffed=0
Full harness on the merge result: 227 fixtures, 182 matched, 45 skipped
(cache_aligner + ccr stubs), 0 diffed, exit 0 — with kompress at 21/21 under
ONNX Runtime 1.24.4 (see #2591).
Also verified cargo check -p headroom-core --no-default-features passes, so the
static-musl path stays intact.
Adds crates/headroom-core/src/transforms/kompress.rs (672 lines): the Kompress
ML prose compressor ported to Rust, running the ModernBERT tokenizer plus the
kompress-v2-base ONNX model through ort with a cache-only loader that never
touches the network.
Parity-only. Nothing calls it: the only references outside the module are the
pub mod / pub use declarations in transforms/mod.rs. live_zone.rs still carries
TODO(PR-B4), so PlainText dispatch remains a no-op and Python continues to serve
prose compression. The pyo3 bridge is untouched and no Python source changes, so
the new engine is unreachable from the shipped package. #1155 wires it up.
Ships 21 recorded parity fixtures, a KompressComparator in headroom-parity, and
scripts/record_kompress_fixtures.py.
Verified byte-identical to the recorded Python output:
[kompress] total=21 matched=21 skipped=0 diffed=0
That required ONNX Runtime >= 1.24 (see #2591) — below it ort deadlocks instead
of erroring, which is why these fixtures had never been run. In CI the model is
absent from the HF cache, so the comparator errors and the fixtures report
Skipped rather than hanging.
Also gates the module behind the ml feature, matching magika_detector: kompress.rs
uses ort, which is optional = true, so an unconditional pub mod broke
cargo check --no-default-features (the static-musl path). CI does not catch that
class of break because cargo test --workspace only builds default features.
## Description
Ruff currently has three independent versions: `uv.lock` resolves
`0.14.14`, pre-commit runs `0.9.4`, and CI installs `0.15.17`.
Contributors can therefore pass one formatter path and fail another.
Make the exact Ruff pin in `pyproject.toml` the source of truth, align
the lockfile and pre-commit hook to it, and make CI read that pin
through a deterministic consistency verifier instead of carrying another
hardcoded version.
Closes#2398
## 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 causes existing functionality
to change)
- [ ] Documentation update
- [ ] Performance improvement
- [ ] Code refactoring (no functional changes)
## Changes Made
- Pin the existing `[dev]` Ruff dependency to `0.15.17`, the formatter
baseline already used by CI.
- Refresh only Ruff in `uv.lock` with `uv 0.11.29`.
- Align `ruff-pre-commit` to `v0.15.17`.
- Add `scripts/verify-ruff-version.py` and run it from pre-commit and
CI.
- Make CI install the verified version read from `pyproject.toml` rather
than a separate literal.
## Testing
- [ ] Unit tests pass (`pytest`) — not run; no runtime source or test
behavior changed.
- [x] Linting passes (`ruff check .`)
- [x] Type checking passes (`mypy headroom`)
- [x] New deterministic guard proves the configuration fix
- [x] Manual testing performed
### Test Output
```text
# Before: run the verifier with the patched pyproject pin but base-branch
# uv.lock, pre-commit config, and workflow.
Ruff version mismatch detected:
uv.lock uses Ruff 0.14.14, expected 0.15.17
.pre-commit-config.yaml uses Ruff 0.9.4, expected 0.15.17
ci.yml does not run 'python scripts/verify-ruff-version.py --print-version'
ci.yml does not install Ruff from 'steps.ruff-version.outputs.version'
$ python3 scripts/verify-ruff-version.py
Ruff versions aligned at 0.15.17
$ uvx uv@0.11.29 lock --check
Resolved 269 packages
$ uvx uv@0.11.29 tree --locked --package ruff
ruff v0.15.17
$ uvx ruff@0.15.17 check .
All checks passed!
$ uvx ruff@0.15.17 format --check .
1322 files already formatted
$ uvx mypy@1.20.2 headroom --ignore-missing-imports
Success: no issues found in 505 source files
$ uvx --with tomli mypy@1.20.2 scripts/verify-ruff-version.py --ignore-missing-imports
Success: no issues found in 1 source file
$ uvx pre-commit run ruff --all-files
Passed
$ uvx pre-commit run ruff-format --all-files
Passed
$ uvx pre-commit run verify-ruff-version --all-files
Passed
```
## Real Behavior Proof
- Environment: macOS 26.5, Python 3.14.6 locally; Python 3.10 fallback
also exercised; `uv 0.11.29`, Ruff `0.15.17`, mypy `1.20.2`.
- Exact command / steps: reproduced the mismatch using the base branch's
real `uv.lock`, `.pre-commit-config.yaml`, and
`.github/workflows/ci.yml`; then ran the verifier, locked Ruff tree,
full Ruff check/format, mypy, and actual pre-commit hooks after the
patch.
- Observed result: the base state fails with all four drift points
listed; the patched state reports one aligned Ruff version (`0.15.17`)
and every formatter path passes.
- Not tested: runtime proxy behavior and the pytest suite, because the
change is limited to development-tool configuration, lock metadata,
pre-commit, and CI wiring.
## Dependency / Supply-Chain Justification
- Ruff is an existing development-only formatter maintained by Astral;
this PR adds no new package.
- `0.15.17` is required to fix local/CI reproducibility and has already
been the repository's CI formatter baseline since #1295.
- Install surface is limited to the `[dev]` extra, lint CI job, and
pre-commit environment. Production/runtime dependencies are unchanged.
- The `uv.lock` refresh updates only Ruff; no unrelated dependency
upgrades are included.
## 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 the non-obvious consistency checks
- [x] Documentation changes are N/A; contributor commands are unchanged
- [x] My changes generate no new warnings
- [x] The guard fails on the real base-state mismatch and passes after
the fix
- [ ] New and existing unit tests pass locally — not run; no runtime
code changed
- [x] I did not edit `CHANGELOG.md`; release-please will use the
conventional PR title
## Additional Notes
No formatter-driven source changes are included. AI assistance was used
to inspect configuration, implement the verifier, and run validation.
## Description
Adds first-class Grok Build support to Headroom so Grok CLI sessions can
route through the local proxy for context compression and savings
tracking.
This PR introduces `headroom wrap grok-build` / `headroom unwrap
grok-build`, a `grok_build` provider slice, Grok MCP registrar support,
and install/telemetry wiring so Grok traffic is attributed correctly in
the proxy and dashboard.
Review follow-up (`9368c413`): when users already own
`[model.grok-build]` in `~/.grok/config.toml`, wrap rewrites `base_url`
in that table in place instead of appending a duplicate header (invalid
TOML).
## Type of Change
- [x] New feature (non-breaking change that adds functionality)
## Changes Made
- Added `headroom/providers/grok_build/` with runtime helpers,
reversible `~/.grok/config.toml` injection, and install env builders.
- Added `headroom wrap grok-build` and `headroom unwrap grok-build` CLI
commands.
- Added `GrokRegistrar` for Headroom MCP registration in Grok config.
- Wired `grok_build` into install planner/registry, agent savings,
telemetry, and proxy client detection (`grok/` user agent).
- **Review fix:** rewrite `base_url` inside an existing user-owned
`[model.grok-build]` table in place (`# was: …` metadata).
- Added regression tests + docs (`grok-build.mdx`, `proxy.mdx`) and
CHANGELOG entry.
## Testing
- [x] Unit tests pass (`pytest`)
- [x] New tests added for new functionality
- [x] Manual testing performed
### Test Output
```text
$ pytest -q tests/test_provider_grok_build.py tests/test_mcp_registry/test_grok_registrar.py
============================== 12 passed in 1.13s ==============================
```
See **Screenshots** below for terminal captures (pytest, review-fix
in-place rewrite, proxy `/readyz`, unwrap).
## Real Behavior Proof
- Environment: macOS, Python 3.11.12 venv, feat/grok-build @ `9368c413`,
isolated `GROK_HOME` temp dirs, proxy port 8799
- Exact command / steps: see screenshot evidence (wrap/unwrap, in-place
table rewrite, `/readyz`)
- Observed result: see screenshots — 12 tests pass; single
`[model.grok-build]` table after wrap on pre-existing config; proxy
healthy; unwrap restores backup
- Not tested: Live interactive Grok chat with xAI auth through the proxy
## 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
- [x] My changes generate no new warnings
- [x] I have added tests that prove my fix is effective or that my
feature works
- [x] New and existing unit tests pass locally with my changes
- [x] I have updated the CHANGELOG.md if applicable
## Screenshots (if applicable)
Terminal captures from local verification (`9368c413`). Assets hosted on
fork prerelease only — **not** in the source tree.
**1. Pytest — 12 passed (incl. review-fix regression)**
**2. Review fix — in-place `[model.grok-build]` rewrite (single table,
`# was:` metadata)**
**3. Proxy health — `/readyz` healthy on port 8799**
**4. Unwrap — restores pre-wrap backup**
## Additional Notes
Screenshot assets:
https://github.com/aashishtamsya/headroom/releases/tag/pr-1629-evidence
(temporary prerelease; safe to delete after merge).
---------
Co-authored-by: JerrettDavis <mxjerrett@gmail.com>
## Description
Persist bounded, aggregate-only Lifetime dashboard metrics across proxy
restarts and expose them through a new `/stats-lifetime` endpoint. The
change keeps session/runtime stats separate from durable lifetime stats,
gates sensitive dashboard metadata for loopback or explicitly trusted
dashboard clients, and updates the dashboard Lifetime view to consume
the new endpoint.
Closes#2137
## Type of Change
- [ ] Bug fix (non-breaking change that fixes an issue)
- [x] New feature (non-breaking change that adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to change)
- [ ] Documentation update
- [ ] Performance improvement
- [x] Code refactoring (no functional changes)
## Changes Made
- Added bounded persistent lifetime metrics state and wired proxy metric
events into it.
- Added `/stats-lifetime` with sensitive project/persistence details
gated behind dashboard metadata access checks.
- Extended loopback/dashboard metadata access policy for trusted
dashboard client CIDRs without widening admin/debug endpoints.
- Reorganized dashboard session/lifetime presentation around runtime
counters versus durable aggregates.
- Added focused tests for persistent aggregation, persistence, endpoint
registration, loopback gating, trusted dashboard CIDRs, and recent
request ordering.
- Fixed current Ruff/mypy issues in the lifetime metrics normalization
code.
## Testing
- [x] Unit tests pass (`pytest`)
- [x] Linting passes (`ruff check`)
- [x] Type checking passes (`mypy`)
- [x] New tests added for new functionality
- [x] Manual testing performed
### Test Output
```text
uv run --extra proxy --with pytest --with pytest-asyncio pytest tests/test_persistent_metrics.py tests/test_persistent_metrics_integration.py tests/test_persistent_metrics_persistence.py tests/test_proxy_loopback_gating.py tests/test_proxy_stats_recent_requests.py -q
53 passed, 1 warning
uvx ruff==0.15.17 check headroom/proxy/forwarded_headers.py headroom/proxy/loopback_guard.py headroom/proxy/persistent_metrics.py headroom/proxy/savings_tracker.py headroom/proxy/server.py tests/test_forwarded_headers.py tests/test_persistent_metrics.py tests/test_persistent_metrics_integration.py tests/test_persistent_metrics_persistence.py tests/test_proxy_loopback_gating.py tests/test_proxy_project_savings.py tests/test_proxy_stats_recent_requests.py
All checks passed!
uv run --extra proxy --with mypy mypy headroom/proxy/persistent_metrics.py
Success: no issues found in 1 source file
```
## Real Behavior Proof
- Environment: Windows review worktree, Python 3.13.3 via uv.
- Exact command / steps: Ran the focused persistent metrics,
persistence, loopback gating, and recent request tests; ran CI-matching
Ruff on touched files; ran mypy on the new persistent metrics module.
- Observed result: `/stats-lifetime` is registered, non-loopback callers
receive only non-sensitive aggregate data, loopback/trusted dashboard
clients receive the full lifetime payload, admin/debug endpoints remain
loopback-only, and persistent metrics normalize malformed stored state
without type/lint errors.
- Not tested: Full repository pytest suite, full dashboard browser
screenshot pass, or live long-running proxy traffic.
## 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
- [x] 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
- [x] New and existing unit tests pass locally with my changes
- [x] I have updated the CHANGELOG.md if applicable
## Screenshots (if applicable)
N/A
## Additional Notes
No changelog entry is required for this dashboard/internal metrics
iteration. The endpoint intentionally exposes only aggregate lifetime
data to ordinary network callers and strips project/persistence details
unless the caller passes the dashboard metadata access policy.
---------
Co-authored-by: JerrettDavis <mxjerrett@gmail.com>
## Description
Harden the release artifact workflow so maintainers can reproduce npm
and Python release checks locally and so OpenClaw packaging does not
depend on a not-yet-published SDK version. This follow-up also keeps the
OpenClaw loader export contract explicit and updates the PR after the
branch was merged with current `headroomlabs/main`.
## Type of Change
- [x] Bug fix (non-breaking change that fixes an issue)
- [x] New feature (non-breaking change that adds functionality)
- [x] Documentation update
## Changes Made
- Added reusable local release smoke scripts for npm assets, Python
wheel/sdist artifacts, and the combined release gate.
- Switched the release workflow npm asset build to the reusable npm
builder.
- Made OpenClaw release asset building install against the just-built
local SDK tarball before rewriting packed metadata to the release
dependency range.
- Kept the OpenClaw source dependency registry-installable at
`headroom-ai@^0.22.3` while the packed artifact still ships
`^<release-version>`.
- Added `registerHeadroomPlugin` as a named export while preserving the
default `{ register }` OpenClaw loader contract.
- Added Windows development bootstrap docs/script and regression tests
for version sync, npm asset ordering, OpenClaw source installability,
and Python wheel smoke import isolation.
## Testing
- [x] Unit tests pass (`pytest`)
- [x] New tests added for new functionality
- [x] Manual testing performed
### Test Output
```text
uv run --with pytest python -m pytest tests/test_release_workflows.py scripts/tests/test_version_sync.py -q
44 passed, 1 warning
node --check scripts/build_npm_release_assets.mjs
node --check scripts/verify_npm_release_assets.mjs
python -m py_compile scripts/build_python_release_smoke.py scripts/release_smoke_all.py scripts/version-sync.py
python scripts/verify-versions.py
All versions aligned at 0.31.0
npm ci (plugins/openclaw)
added 88 packages, audited 89 packages, found 0 vulnerabilities
python scripts/release_smoke_all.py --out release-assets-local/all-pr1824-postmerge-fixed-20260710-104739
Verified npm release assets for 0.31.0
wheel metadata OK: headroom_ai-0.31.0-cp310-abi3-win_amd64.whl contains headroom/_core.pyd
sdist License-File metadata OK: ['LICENSE', 'NOTICE']
smoke-import OK: version=0.31.0 hello=headroom-core
```
## Real Behavior Proof
- Environment: Windows 11, Python 3.14.3, Node v24.14.0, npm 11.9.0, uv
0.11.15, Rust/Cargo already installed in the local development
environment.
- Exact command / steps: Ran `python scripts/release_smoke_all.py --out
release-assets-local/all-pr1824-postmerge-fixed-20260710-104739` after
syncing this branch with current `headroomlabs/main`.
- Observed result: The command built `headroom-ai-0.31.0.tgz`,
`headroom-openclaw-0.31.0.tgz`,
`headroom_ai-0.31.0-cp310-abi3-win_amd64.whl`, and
`headroom_ai-0.31.0.tar.gz`; npm verifier passed; the wheel installed
into a fresh venv and imported `headroom._core`.
- Not tested: Full GitHub Actions release matrix and publish jobs with
real PyPI/npm/GitHub Packages credentials.
## 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
- [x] My changes generate no new warnings
- [x] I have added tests that prove my fix is effective or that my
feature works
- [x] New and existing unit tests pass locally with my changes
- [ ] I have updated the CHANGELOG.md if applicable
## Screenshots (if applicable)
N/A.
## Additional Notes
The post-merge full smoke initially failed because the OpenClaw source
package depended on `headroom-ai@^0.31.0`, which is not available on
public npm yet. The builder now installs OpenClaw against the just-built
local SDK tarball before build/pack, and then rewrites packed metadata
to `^0.31.0`.
---------
Co-authored-by: JerrettDavis <mxjerrett@gmail.com>
Co-authored-by: Tejas Chopra <chopratejas@gmail.com>
## Description
Repo hygiene for a public OSS project: removes committed `node_modules`,
stray/internal/draft markdown, and commercial-surface references —
keeping every real doc (the published docs site, the wiki guides, and
all component READMEs) intact. Every file was content-audited before
removal, and load-bearing files were verified against the code/CI and
kept.
Net: **1,695 files changed, +23 / −266,409** (the deletions are
dominated by a committed `node_modules` tree).
Closes # (no tracking issue)
## 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)
- [x] Documentation update
- [x] Code refactoring (no functional changes)
## Changes Made
**Removed (verified to have no code/CI dependencies):**
- `examples/vercel-ai-sdk-pr/` — 1,649 committed `node_modules` files
(zero example source); `node_modules/` added to `.gitignore`.
- `docs/spec/` (23 draft "Living Specification" files — orphaned,
`1.0.0-draft`, drifted from the code), `docs/superpowers/` (2 agent
plans), `docs/proposals/` (2 internal/commercial memos).
- 6 orphan `docs/*.md` (auth-modes, bedrock,
claude-code-vertex-headroom, cortex-code, output-token-reduction-guide,
rtk-loop-weighting).
- `PR.md` (committed PR draft), `ENTERPRISE.md`, `.github/FUNDING.yml`.
**Content scrubs:**
- Removed unreleased "Headroom Cloud" / `api.headroom.ai` / `hr_`
references from `configuration.mdx`, `wiki/configuration.md`,
`wiki/typescript-sdk.md`, `sdk/typescript/README.md` (reworded to
neutral, accurate phrasing).
- Dropped a stale "awaiting maintainer before merge" line from
`plugins/headroom-oauth2/SPEC.md`; tidied `.gitignore` comments (kept
the protective `headroom-managed/` ignore rule).
- Fixed the now-dangling links into removed files (README
nav/`output-token-reduction` link, `scripts/README`, `wiki/vertex`).
**Explicitly KEPT (load-bearing — would orphan in-code citations if
removed):**
- `.changelog.md` — consumed by `.github/workflows/release.yml` (read as
the release-notes file).
- `REALIGNMENT/`, `docs/observability.md`, `docs/rtk-architecture.md`,
`wiki/plans/`, `TESTING-copilot-subscription.md` — referenced by the
Rust core / Python / tests as design docs.
## Testing
- [ ] Unit tests pass (`pytest`)
- [x] Linting passes (`ruff check`)
- [ ] Type checking passes (`mypy`)
- [x] New tests added for new functionality
- [x] Manual testing performed
### Test Output
```text
# Docs/markdown + .gitignore only — no Python/Rust source changed, so the
# behavioral test suite is unaffected. Verified the cleanup did not orphan
# references or break the published docs site:
$ git ls-files 'docs/content/docs/*.mdx' | wc -l # published site intact
42
$ # meta.json nav unchanged; no published page removed.
$ grep -rnI "Headroom Cloud|api.headroom.ai|'hr_" $(git ls-files '*.md' '*.mdx')
>>> none
$ # dangling refs to removed files (excl pre-existing P0/P2 spec stubs that
$ # never existed in git): none remaining.
```
## Real Behavior Proof
- Environment: macOS, local git clone of the repo (markdown/.gitignore
changes only — no runtime).
- Exact command / steps: 4 read-only content-audit agents classified
every `.md`/`.mdx` file; each removal candidate was cross-checked
against the codebase (`grep` for citations in `.rs`/`.py`/tests,
workflows, and configs); only files with no dependents were removed; the
tree was re-grepped after removal to confirm no new dangling references;
verified the published docs site page count (`git ls-files
'docs/content/docs/*.mdx' | wc -l` = 42, unchanged).
- Observed result: the 42-page published docs site and all wiki guides
are untouched; no source or workflow references a removed file;
`.changelog.md` (consumed by release.yml) and the code-cited design docs
were detected as dependencies and kept; the committed `node_modules`
tree is removed and `node_modules/` is gitignored so it can't be
re-committed; zero "Headroom Cloud"/`headroom.dev` references remain.
- Not tested: N/A — no executable code changed (only markdown, `.mdx`,
and `.gitignore`), so the behavioral test suite is unaffected.
## 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
- [x] I have made corresponding changes to the documentation
- [x] My changes generate no new warnings
- [ ] I have added tests that prove my fix is effective or that my
feature works
- [x] New and existing unit tests pass locally with my changes
- [ ] I have updated the CHANGELOG.md if applicable
## Additional Notes
- This branch deletes `.github/FUNDING.yml` while PR #1526 edits it —
the two will be sequenced at merge (delete wins).
- A follow-up option (not in this PR): also remove the internal design
docs that are currently cited by the code (`REALIGNMENT/`,
`docs/observability.md`, `docs/rtk-architecture.md`, `wiki/plans/`) —
that requires scrubbing ~15–20 in-code citations so nothing dangles, so
it's deliberately deferred.
- Untracked local working files (`benchmarks/hf_pilot/`,
`tools/copilot-test/`) are intentionally left out of git (not
committed).
## Description
Makes **tokensave**
([github.com/aovestdipaperino/tokensave](https://github.com/aovestdipaperino/tokensave))
the **primary coding-task compressor** that `headroom wrap` installs,
and demotes **Serena** to a **backup**. tokensave is a local semantic
code-graph MCP server (`tokensave serve`): the agent queries it for
symbols, call chains, and impact analysis instead of grepping/reading
whole files — the same role Serena filled, but as a pre-indexed graph.
Serena now only registers when tokensave is unavailable (or when forced
with `--serena`).
Closes #
## Type of Change
- [ ] Bug fix (non-breaking change that fixes an issue)
- [x] 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
- `headroom/graph/tokensave_installer.py` (new): fetch the prebuilt
tokensave release binary for the platform (release-binary only — no
`cargo` compile at wrap time); honors `HEADROOM_BINARIES_OFFLINE`;
returns `None` (→ Serena) when no asset exists (e.g. x86_64 macOS) or
the download fails.
- `mcp_registry`: `build_tokensave_spec()`; registration/disable/migrate
go through the existing `ServerSpec` + ownership-ledger flow, identical
to Serena.
- `cli/wrap.py`: new `_setup_coding_compressor` primary/backup policy;
tokensave setup/disable/migrate/index helpers. New flags
`--no-tokensave` (skip primary) and `--serena` (force backup on);
`--no-serena` now means "never register the backup". Default wrap
removes a previously Headroom-installed Serena entry once tokensave is
primary (user-managed entries preserved). `--code-graph` repointed to
tokensave; the legacy `codebase-memory-mcp` install path is dropped
(unwrap still cleans up legacy entries). `unwrap claude|codex` remove a
ledger-owned tokensave entry.
- Strands `HeadroomBundle`: `enable_tokensave_mcp=True` (primary);
`enable_serena_mcp` now defaults `False` (backup).
- `docs/content/docs/proxy.mdx`: `--code-graph` description updated from
codebase-memory-mcp to tokensave.
- Tests: tokensave installer (incl. error paths),
register/disable/migrate, primary/backup policy, and the
binary-resolution/indexing helpers. A scoped
`tests/test_cli/conftest.py` offline guard keeps the CLI suite hermetic.
## Testing
- [x] Unit tests pass (`pytest`)
- [x] Linting passes (`ruff check .`)
- [x] Type checking passes (`mypy headroom`)
- [x] New tests added for new functionality
- [ ] Manual testing performed
### Test Output
```text
$ uv run pytest -q tests/test_graph_tokensave.py tests/test_cli/test_tokensave_setup.py tests/test_cli/test_tokensave_helpers.py
41 passed
$ uv run pytest -q tests/test_cli/ tests/test_graph.py tests/test_graph_tokensave.py
421 passed # full CLI + graph suites, incl. all pre-existing Serena/unwrap/registry tests
$ uv run pytest -q tests/test_mcp_registry/ tests/test_proxy_healthchecks.py
passed
$ uv run ruff format --check headroom/ tests/ # 822 files already formatted
$ uv run ruff check <changed files> # All checks passed!
$ uv run mypy headroom/graph/tokensave_installer.py headroom/mcp_registry/install.py
Success: no issues found in 2 source files
# Coverage on new module
headroom/graph/tokensave_installer.py 99%
```
## Real Behavior Proof
- Environment: macOS (darwin arm64), Python 3.14, `uv` dev env;
tokensave 7.0.2 binary present on PATH and exercised against this repo's
`.tokensave/` graph during development. The installer pins release
**v7.0.2** (SHA-256-verified) across macOS arm64, Linux aarch64/x86_64,
and Windows x86_64/aarch64.
- Exact command / steps: `headroom wrap claude` registers `tokensave
serve` as the primary MCP code-graph server and indexes the project;
with the binary removed from PATH and `HEADROOM_BINARIES_OFFLINE=1`, the
same command falls back to registering Serena. Behavior is pinned by the
unit tests (binary-present → tokensave registered + Serena entry
removed; binary-absent → Serena fallback; `--serena` forces backup on;
`--no-serena` suppresses it; `--no-tokensave` disables primary).
- Observed result: tokensave registered as primary on the binary-present
path; Serena registered on the unavailable path; unwrap removes only
ledger-owned entries.
- Not tested: live end-to-end agent session inside Claude Code / Codex
against a real provider API; Windows/Linux release-asset download
(covered by unit tests with mocked archives, not a live fetch).
## 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
- [x] 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
- [x] New and existing unit tests pass locally with my changes
- [ ] I have updated the CHANGELOG.md if applicable
## Additional Notes
- CHANGELOG is left untouched: this repo generates it via release-please
from Conventional Commits, so a manual edit is N/A.
- `strands/bundle.py` shows 0% patch coverage because that module
hard-imports the optional `strands` SDK, which CI does not install (the
pre-existing `_make_serena_client` was likewise uncovered) — not a
regression.
- A `test (3)` shard failure on `headroom.memory.bridge` is a
pre-existing offline-CI flake (cannot reach huggingface.co); it touches
no file in this PR and the scoped offline guard only applies under
`tests/test_cli/`.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
## Description
Adds the first levers that reduce the tokens the model **writes back**
(output), complementing Headroom's existing input compression. Output
costs 5× input on Opus-class models and is full of waste (ceremony,
restated code, deep "thinking" on routine steps). Two phases in one
self-contained PR off `main`: the request-side output shaper, then
per-user verbosity learning plus an honest counterfactual savings
estimator and dashboard surfacing.
## Type of Change
- [ ] Bug fix (non-breaking change that fixes an issue)
- [x] New feature (non-breaking change that adds functionality)
- [ ] Breaking change (fix or feature that would cause existing
functionality to change)
- [x] Documentation update
- [ ] Performance improvement
- [ ] Code refactoring (no functional changes)
## Changes Made
- **Output shaper** (`output_shaper.py`, opt-in
`HEADROOM_OUTPUT_SHAPER=1`): cache-safe verbosity steering appended to
the system-prompt tail (5 levels); effort routing that lowers
`output_config.effort` on mechanical tool-result continuations; legacy
`thinking.budget_tokens` clamp. Never injects effort where absent, never
toggles `thinking.type`.
- **`headroom learn --verbosity`**: mines Claude Code transcripts for
behavioral signals (interrupts, length-adaptive fast-skips, echo ratio),
recommends a verbosity level (heuristic + optional `--llm-judge`), and
seeds the savings baseline.
- **Counterfactual estimator** (`output_savings.py`): per-stratum
synthetic-control (estimated) + A/B holdout (measured) with a propagated
95% CI; conversation-stable arm assignment for A/B validity and
prefix-cache safety.
- **AIMD verbosity controller** (`verbosity_controller.py`):
additive-increase / fast-back-off state machine; live signal emission
gated off by default.
- **Wiring + surfaces**: shaper resolves the learned level; recording
rides the existing `transforms_applied` channel through the outcome
funnel (no `RequestOutcome` changes); `headroom output-savings` CLI;
dashboard "Output Tokens Saved" card.
- **Docs**: simple-words user guide + design doc with the counterfactual
methodology.
## Testing
- [x] Unit tests pass (`pytest`)
- [x] Linting passes (`ruff check .`)
- [x] Type checking passes (`mypy headroom`)
- [x] New tests added for new functionality
- [x] Manual testing performed
### Test Output
```text
$ pytest tests/test_output_savings.py tests/test_output_savings_cli.py \
tests/test_verbosity_learn.py tests/test_verbosity_controller.py \
tests/test_output_shaper.py -q
94 passed in 0.54s
$ pytest tests/test_request_outcome.py tests/test_handler_outcome_tag_invariant.py \
tests/test_proxy_dashboard_stats_cache.py -q
44 passed
$ ruff format --check .
831 files already formatted
$ mypy headroom --ignore-missing-imports
Success: no issues found in 361 source files
```
## Real Behavior Proof
- Environment: macOS, Python 3.12 (`.venv`), `anthropic` 0.76, live API
model `claude-opus-4-8`.
- Exact command / steps: `HEADROOM_OUTPUT_SHAPER=1`; `headroom learn
--verbosity --apply` (seeds level + baseline); `python
scripts/eval_output_shaper.py A` (live before/after); simulate holdout
traffic then `headroom output-savings`.
- Observed result: code-review ask — baseline 1,750 output tokens → L2
1,354 (−22.7%) → L3 599 (−65.8%), same bugs found. `learn --verbosity`
on 24 real sessions → 11% interrupt / 26% fast-skip → L3 (high
confidence). Measured A/B path → 31.7% reduction (95% CI 27.7%–35.7%).
94 new tests + 44 existing outcome/dashboard tests green; ruff + mypy
clean.
- Not tested: live streaming-path recording exercised only via unit
tests (the `transforms_applied` funnel is shared across paths); runtime
AIMD signal emission is gated off by default and not exercised live.
## 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
- [x] 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
- [x] New and existing unit tests pass locally with my changes
- [ ] I have updated the CHANGELOG.md if applicable
## Additional Notes
Output savings are counterfactual (we never observe what the model
*would* have written), so the estimator separates **estimated** (vs a
learned baseline) from **measured** (A/B holdout via
`HEADROOM_OUTPUT_HOLDOUT`) and always reports a confidence band — never
a single made-up number. CHANGELOG left unchecked (release-please
manages it). Runtime AIMD self-tuning is intentionally a TODO
(controller built/tested; live signal emission gated behind
`HEADROOM_VERBOSITY_AUTOTUNE`).
`CI / lint (pull_request)` failed because `ruff format --check` detected
formatting drift in the new PR governance script and its tests. This PR
aligns those files with repository formatting rules so the lint job can
pass.
- **Root cause**
- `ruff format --check .` reported two files as non-canonical:
- `scripts/pr-governance.py`
- `scripts/tests/test_pr_governance.py`
- **Change set**
- Applied `ruff` formatting to only the two flagged files.
- No behavioral or logic changes; edits are line-wrap/format
normalization only.
- **Representative update**
```python
parser.add_argument(
"--event", type=Path, required=True, help="Path to the GitHub event
payload JSON."
)
```
---------
Co-authored-by: JerrettDavis <mxjerrett@gmail.com>
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Co-authored-by: copilot-swe-agent[bot] <198982749+Copilot@users.noreply.github.com>
## Summary
Replaces `chopratejas/kompress-base` with
**`chopratejas/kompress-v2-base`** as the default Kompress
text-compression model (the fallback for content not handled by
structured compressors), using a new **weight-only int8 ONNX** artifact
that is fp32-equivalent at 2.2x less memory.
## Why
v2 is the same dual-head ModernBERT (token classifier + span CNN),
LoRA-trained on the v2 dataset. The HF repo originally shipped PyTorch
weights only — pointing Headroom at it naively would have forced the
heavy `[ml]` (torch) path on every `[proxy]` install. We exported ONNX
artifacts reproducing the v1 loader contract (single `final_scores`
output) and published them to the HF repo.
## Eval (labeled dataset_v2 test split, n=500, threshold 0.5)
| artifact | size | f1 | must_keep_recall | keep_rate | fp32 agreement |
|---|---|---|---|---|---|
| fp32 (reference) | 601 MB | 0.9128 | 0.9770 | 0.8100 | 100% |
| **int8-wo (default)** | **261 MB** | **0.9130** | **0.9765** |
**0.8097** | **99.6%** |
| fp16 | 301 MB | 0.9129 | 0.9771 | 0.8099 | 99.9% |
| int4-wo | 230 MB | 0.9102 | 0.9825 | 0.8354 ⚠️ | 94.5% |
| int8-dynamic | 271 MB | 0.9041 | 0.9868 | 0.8857 ⚠️ | 90.5% |
Weight-only int8 (MatMulNBits) keeps activations fp32, avoiding the
upward score bias that makes dynamic int8 keep ~7% more tokens (≈40%
less compression savings). Quantized candidates were generated and
eval-gated by a Modal job in the kompress repo
(`modal_jobs/export_onnx_v2.py`) against the labeled test split.
## Changes
- Default model id → `chopratejas/kompress-v2-base`
- ONNX artifact resolution tries candidates in order (**int8-wo → fp32 →
v1 int8**), falling through on download miss **or session-load failure**
— onnxruntime builds without the MatMulNBits 8-bit kernel fall back to
fp32 instead of losing Kompress
- `HEADROOM_KOMPRESS_ONNX_FILENAME` env pins an exact artifact
- `scripts/export_kompress_v2_onnx.py`: reproducible fp32 export (loads
the merged v2 checkpoint, traces the `final_scores` contract, verifies
vs PyTorch)
- `.gitignore`: local `onnx/` artifacts dir; allowlist the export script
## Testing
- End-to-end: fresh `KompressCompressor().preload()` resolves int8-wo
from HF, loads on onnxruntime 1.23 (CPU, no torch), compresses with
error/traceback content preserved
- fp32 export verified lossless vs PyTorch (max |Δscore| = 0.0, 100%
keep agreement)
- ruff check + format clean, mypy clean, 63 targeted tests pass
## Summary
- add a containerized differential network capture harness for Claude
Code direct vs Claude Code routed through Headroom
- capture both Headroom client-side traffic and Headroom upstream
traffic with sanitized mitmproxy JSONL output
- add `headroom capture network-diff` to compare captures and produce
Markdown/JSON reports, including Anthropic tool-count/tool-byte deltas
for deferred-tool investigations
- add an on-demand GitHub Actions workflow for the harness; it only runs
via `workflow_dispatch`, with live Claude Code/Anthropic capture gated
on `ANTHROPIC_API_KEY`
- document the workflow and ignore generated capture artifacts
## Validation
- `C:\git\headroom\.venv\Scripts\python.exe -m pytest
tests/test_network_diff_capture.py`
- `ruff check headroom/capture headroom/cli/capture.py
tests/test_network_diff_capture.py`
- `ruff format --check headroom/capture headroom/cli/capture.py
tests/test_network_diff_capture.py`
- `C:\git\headroom\.venv\Scripts\python.exe -m mypy
headroom/capture/network_diff.py headroom/cli/capture.py`
- `docker compose -f
docker/differential-network-capture/docker-compose.yml --profile run
config`
- `docker compose -f
docker/differential-network-capture/docker-compose.yml --profile run
build claude-direct`
- `docker run --rm -e CLAUDE_COMMAND="claude --version"
headroom-network-diff-claude-direct:latest`
- parsed `.github/workflows/network-diff-capture.yml` with PyYAML and
confirmed manual-only trigger
Live Claude API capture was not run locally because `ANTHROPIC_API_KEY`
is not set in this environment. The workflow can run it manually in
GitHub Actions when that secret is present; otherwise it emits a visible
skip warning and uploads a skipped artifact.
## Notes
- Full pre-commit mypy still fails on unrelated Windows `fcntl`
attributes in `headroom/subscription/tracker.py`; the feature commit
skipped only that hook after narrow mypy passed for the new modules.
- `tests/test_release_workflows.py` has two Windows-local failures
because it shells out to a missing Unix/Rust command; unrelated workflow
checks in that file passed before those failures.
- Motivated by
https://github.com/chopratejas/headroom/issues/746#issuecomment-4651276818
/ Issue #746.
- Add `Depends(_require_loopback)` to `/debug/memory` endpoint (was missing
while /debug/tasks, /debug/ws-sessions, /debug/warmup all had it)
- Guard `raise last_error` when last_error is None (retry_max_attempts=0 path
raised TypeError); add ProxyConfig.__post_init__ validation rejecting
retry_max_attempts < 1 when retry_enabled=True
- Make initialize_context_tool_session_baseline async; offload subprocess via
asyncio.to_thread so the blocking rtk/lean-ctx subprocess does not stall the
event loop; update call sites in server.py
- Take snapshot list() of SemanticCache._cache.values() before iterating in
get_memory_stats() to avoid dict-size-changed RuntimeError under async load
- Change memory_neo4j_password default from 'password' to '' and emit a
logger.warning at startup when backend=qdrant-neo4j and password is empty
- Replace hardcoded NEO4J_AUTH=neo4j/password in docker-compose.yml with
${NEO4J_AUTH:-neo4j/devpassword}; add .env.example with CHANGEME placeholder
- Format tests/test_provider_proxy_routes.py (pre-existing ruff format drift)
CI ran the Codex compression scheduler stress test on python 3.10/3.11/
3.12/3.13 and all four versions failed with:
ModuleNotFoundError: No module named 'scripts.replay_codex_ws_load'
The test imports ``boot_proxy``, ``warmup``, ``replay_session``, ``Frame``,
and ``Scenario`` from the replay tool to drive 30 concurrent compression
calls and assert no p99 contention tail (the very regression this PR
fixes). The tool was untracked because ``scripts/*`` is gitignored by
allowlist — local-only tools work fine for measurement but CI cannot
import them.
Add the replay tool to the allowlist so it ships. Same pattern as
``scripts/smoke_issue_327.py`` and other single-purpose scripts already
on the allowlist. The tool is genuinely useful beyond this PR: it lets
any contributor reproduce the Codex slowness baseline numbers and
measure their own fix against the same workload shape.
Local re-run with the tool committed: 3 passed, 1 skipped — identical
to the pre-fix branch run.
Issue #355: published headroom_ai-*-manylinux_2_28_*.whl fails to import on Ubuntu 22.04, Debian 11/12, Conda envs with libc < 2.38: 'ImportError: undefined symbol: __isoc23_strtoll'.
Root cause: ORT prebuilt artifacts (downloaded via fastembed's ort-download-binaries-rustls-tls feature) are compiled with gcc-14.2.1 on a glibc-2.38+ host and reference __isoc23_strtoll. Our manylinux build host has glibc 2.38 so the link succeeds; end users with older glibc don't.
Two-part fix: (1) crates/headroom-py/glibc_compat.c provides weak-alias definitions for __isoc23_strtol/strtoll/strtoul/strtoull delegating to the older strtol* family, compiled by build.rs on Linux/glibc only. The dynamic linker prefers glibc's strong symbol when present (>= 2.38) and falls back to ours when not (< 2.38). (2) scripts/audit_wheel_glibc_symbols.py is a release.yml gate that runs objdump -T on every Linux wheel and rejects any UND symbol whose required glibc version exceeds the wheel's manylinux floor.
Validated: the audit correctly rejects the actually-broken v0.20.26 wheel with a precise diagnostic. The shim itself is a tiny static link with zero runtime cost.
Regression tests in tests/test_release_workflows.py pin the shim's load-bearing pieces (.c file, build.rs trigger, [build-dependencies] cc dep) and the audit invocation in release.yml. Future drift fails at PR time, not in the next release.
This is the same bug class as PR #371 (rustls-everywhere). Each instance gets fixed; the audit gate now catches the *class* — any future static linkage that introduces a post-floor symbol gets blocked before publish.
Adds an opt-in compression interceptor that buffers Anthropic
/v1/messages requests, runs IntelligentContextManager over the
messages array, and forwards the (possibly trimmed) body upstream.
All other paths, methods, and content-types stay on the original
streaming passthrough — so existing operators see zero change.
Behaviour gates ALL must be true to buffer + compress:
- --compression flag (or HEADROOM_PROXY_COMPRESSION=1)
- method == POST
- path == /v1/messages
- Content-Type: application/json
- ICM constructed successfully at startup
Falls through to streaming on any failure: parse, missing fields,
unknown model, body-too-large. Compression must never break a
request — that's the safety contract.
Model context windows come from a vendored LiteLLM snapshot at
crates/headroom-proxy/data/model_prices_and_context_window.json
parsed once into an OnceLock<HashMap>. Refresh via
scripts/refresh_model_limits.sh. Rationale documented inline:
hardcoded tables silently rot; LiteLLM is the canonical source
the entire LLM-tooling ecosystem relies on.
New tests:
- 16 unit tests across compression::{anthropic, icm, model_limits}
- 5 integration tests: off-passthrough, on-short-passthrough,
on-oversized-trim, on-non-json-skip, on-non-llm-path-skip
Verification:
- cargo test --workspace -> 884 passed, 0 failed
- cargo clippy --workspace -- -D warnings -> clean
- cargo fmt --check -> clean
The Anthropic token-mode handler walked past prefix_tracker.frozen_message_count
whenever an upcoming tool_result's content-hash matched comp_cache._stable_hashes
or should_defer_compression returned True. That conflated content equality with
positional cache membership.
Anthropic's prefix cache is POSITIONAL: bytes 0..K cached, anything past K is
fresh. _stable_hashes is content-keyed and grows unbounded. In long Claude Code
sessions where tool_result content rhymes across turns (repeated system prompts,
repeated file reads, repeated tool descriptions), the walker advanced
frozen_message_count to len(messages) on every turn and the pipeline produced
transforms_applied=[] on 73% of requests in user SvenMeyer's reported session
(headroom-stats-2026-05-01.json: 74 of 101 eligible requests "prefix_frozen") —
even after the prior fix in 44944fb. The 15 requests that did compress averaged
21%, proving compression itself works when reached.
Fix: delete the walker. The freeze boundary is now
frozen_message_count = min(
prefix_tracker.frozen_message_count, # positional ground truth
comp_cache.compute_frozen_count(messages), # local cache lower bound
)
compute_frozen_count's use of _stable_hashes can only LOWER the freeze via the
min clamp, never raise it past prefix_tracker's value. For any position in the
gap [compute_frozen_count, prefix_tracker.frozen_count], recompressing produces
byte-stable output (compression is deterministic on input content), so
Anthropic's prefix cache stays valid.
Cross-handler verification:
* OpenAI handler (proxy/handlers/openai.py:358-382) does not have this walker
— uses only compute_frozen_count. Codex routes through OpenAI handler. Both
unaffected.
* Streaming and non-streaming both invoke anthropic_pipeline.apply() before the
upstream call. One fix covers both paths.
* Cache mode (is_cache_mode) takes the _extract_cache_stable_delta path and is
independent of the walker. Unaffected.
Tests: six new regression tests lock down the post-fix invariants — clamp to
min(prefix_tracker, compute_frozen_count); fresh tool_result whose hash matches
old _stable_hashes entry is not frozen; frozen prefix byte-stable across the
pipeline; 10-turn session produces non-empty compression suffix every turn;
streaming and non-streaming compute identical frozen_message_count; OpenAI
handler never calls the walker functions. Plus scripts/smoke_issue_327.py
(gated by RUN_LIVE_API=1) drives a 10-turn conversation against
api.anthropic.com in both shapes (string + list-of-blocks) and both modes
(streaming + non-streaming).
ci-precheck clean. 191 tests pass.
Follow-ups (separate PRs):
* Fix _cache perpetually empty (anthropic.py result.messages != working_messages
comparison rarely fires in token mode).
* Cap _stable_hashes with bounded LRU + 1h TTL — hygiene only after the freeze
gate is removed.
* List-shape tool_result content gates at content_router.py:1975 and
intelligent_context.py:657 (cluster A from the audit).
Five gates broke on the 2026-04-27 push of the smart_crusher branch.
Each is fixed below; the second half adds a `make ci-precheck` target
(plus an installable git pre-push hook) so the same dance never happens
again.
Failures fixed:
1. cargo fmt — 22 files had formatting drift introduced over the
stage 3c.1 work. `cargo fmt --all` reformatted them; no semantic
changes. `cargo test --workspace` still green (388 + supporting).
2. wheels job (macOS x86_64) — `fastembed -> ort -> ort-sys` does not
publish prebuilt ONNX Runtime binaries for `x86_64-apple-darwin`.
Removed that target from `.github/workflows/rust.yml`'s wheels
matrix. Apple Silicon (`aarch64-apple-darwin`) covers macOS
distribution; Intel macOS users can build from source. The matrix
now has 2 targets: linux x86_64 + macOS aarch64.
3. test-extras (relevance.py) — `tests/test_relevance.py::TestSmartCrusherIntegration`
constructs a `SmartCrusher`, which hard-imports `headroom._core`
since the python implementation was retired in stage 3c.1b. The
test-extras job didn't build the rust extension. Added the same
`maturin build + symlink` block the main `test` job uses.
4. smoke-test (eval.yml) — same root cause:
`compression_only.evaluate_ccr_lossless` instantiates a SmartCrusher.
Same fix: build the rust extension before the smoke test runs.
5. commitlint — three rules tripped:
- `subject-case` rejects PascalCase identifiers in subjects, but
the project deliberately names classes (SmartCrusher, HfTokenizer,
ContentRouter, DiffCompressor) in commit subjects. Disabled.
- `footer-leading-blank` is a warning that the wagoid action turns
into a CI failure; lines like `Module: foo.rs` in our bodies
match the conventional footer pattern and trip it. Disabled.
- `type-enum` doesn't include `parity`, but the project ships
parity-test infrastructure as its own concern (separate from
`test:`); added `parity` to the allowed types.
Pre-push verification — the prevention half:
`make ci-precheck` runs all of the above CI gates locally:
- `ci-precheck-rust`: cargo fmt --check + clippy + test --workspace.
- `ci-precheck-python`: builds the rust extension via maturin, then
runs the smart_crusher-affected python test files (185 tests across
test_transforms/, test_relevance*, test_ccr, test_acceptance,
test_critical_fixes, test_quality_retention).
- `ci-precheck-commitlint`: `npx commitlint --from origin/main --to
HEAD` against the same config CI uses. Skipped silently if npx is
not on PATH (install Node 18+ to enable).
`make install-git-hooks` (or `scripts/install-git-hooks.sh`) installs
a git pre-push hook that runs `make ci-precheck` automatically.
Bypass with `--no-verify` only when truly needed.
When new CI gates land in `.github/workflows/`, mirror them into a
`make ci-precheck-*` target. The Makefile is the local mirror of the
CI configuration; keeping them in sync is a load-bearing invariant.
Verification: `make ci-precheck` runs green on this commit.
Replace the embedding scorer stub with a real fastembed-rs
implementation. Same library + same model as the Python side will
use after the next commit, giving byte-equal embeddings on identical
inputs.
Cargo.toml: fastembed = "5". Default features pull in `ort` (ONNX
Runtime) with auto-download of the runtime binary at build time
(~21s additional first-build); model weights (BAAI/bge-small-en-v1.5,
~30 MB int8-quantized ONNX) auto-download from HuggingFace Hub on
first use.
embedding.rs:
- EmbeddingScorer wraps Option<Mutex<TextEmbedding>>. Mutex required
because TextEmbedding::embed needs &mut self (single-threaded ONNX
session); concurrent callers serialize on the lock, fine for the
SmartCrusher hot path where inference dominates lock contention.
- EmbeddingScorer::try_new() — explicit construction with HF Hub
download. Returns Result; surface errors to callers.
- EmbeddingScorer::try_new_with_model(EmbeddingModel) — bring your
own model from fastembed's catalog.
- EmbeddingScorer::default() — STUB only (model=None,
is_available()=false). Mirrors Python's "sentence-transformers
not installed" branch byte-for-byte. To get a real scorer, call
try_new() and pass via HybridScorer::with_scorers().
Why default() is a stub: with auto-load Default, model availability
would depend on whether HF Hub cache has the file — non-deterministic
in tests. Explicit try_new() keeps Default cheap and predictable.
cosine_similarity:
- f32 vec inputs (fastembed returns Vec<Vec<f32>>).
- Clamped to [0, 1] (mirrors Python _cosine_similarity — only
positive similarity matters for relevance).
- Defensive: zero vectors / mismatched dims → 0.0.
score / score_batch:
- Empty input / unavailable model → empty score with explanatory
reason.
- Batch encodes items + context in one model call (Python parity:
amortizes model dispatch).
- Inference failures degrade gracefully with empty scores rather
than panicking.
Tests:
- 5 cosine-similarity unit tests (offline).
- 3 unavailable-scorer tests (model=None path).
- 3 model-backed integration tests gated on RUN_FASTEMBED_TESTS=1
(semantic-match-outranks-unrelated, batch-shape, model-loads).
- All 388 headroom-core tests pass without RUN_FASTEMBED_TESTS;
with it set, the gated 3 also pass.
Net: 388 unit tests, clippy clean. HybridScorer's BM25-fallback path
remains correct (default embedding scorer reports unavailable).
Stage 3c.1 next: switch Python's relevance/embedding.py to fastembed
PyPI package + record parity fixtures with real embeddings on both
sides.
The python `DiffCompressor` is replaced by a thin pyo3-backed shim that
delegates to `headroom._core.DiffCompressor`. There is no python
implementation and no env-var fallback — the wheel is a hard import.
Why now: opt-in defaults don't drive python retirement. Byte-equal
parity was already proven across 27 fixtures (stage 3a); keeping a
shadow python impl behind a flag is a permanent maintenance cost with
no operational benefit. Stage 3b deletes ~700 lines of python parser /
scorer / formatter code; the rust crate has its own coverage.
Surface preserved:
- `headroom.transforms.diff_compressor.DiffCompressor` — same class
name, same `__init__`, same `compress(content, context)` shape.
Returns python `DiffCompressionResult` dataclasses so call sites that
destructure with `asdict()` work unchanged.
- `DiffCompressorConfig` and `DiffCompressionResult` dataclasses kept.
- Sidecar `compress_with_stats(...)` exposes the rust-only
`DiffCompressorStats` (per-file hunk drops, context lines trimmed,
file_mode normalizations) for observability.
Removed:
- Python parser / scorer / formatter (~700 lines).
- Internal `DiffHunk` / `DiffFile` parser dataclasses (rust crate has
parallel coverage).
- 12 tests that probed `_parse_diff` / `_score_hunks` or the deleted
parser dataclasses. The 29 public-API tests in `test_diff_compressor.py`
remain and now exercise the rust backend through the same import path.
- `HEADROOM_DIFF_COMPRESSOR_BACKEND` env var — no longer meaningful.
Build:
- `scripts/build_rust_extension.sh` runs `maturin develop` and symlinks
the built `.so` into `headroom/` so `import headroom._core` resolves
past the in-tree package shadowing the maturin overlay.
- `.gitignore` excludes the symlinks and allowlists the build script.
Tests:
- 544 transforms pass; 33 rust-vs-fixture parity tests guard the pyo3
bridge against regressions (`tests/test_transforms/test_diff_compressor_rust_parity.py`).
- Mypy clean.
What this does, in plain terms:
Headroom's proxy now ships with three CLI tools (ast-grep, difftastic,
scc) that it can use to shrink tool_result payloads before they reach
the model. The goal is simple: when Claude Code (or Codex, Aider, etc.)
asks the model to reason about a big file or diff, we swap the verbose
output for a compact, same-meaning version. Fewer tokens per turn, same
answers, lower bill.
Today a single interceptor is wired: ast-grep on Read. When an agent
reads a large code file, the proxy replaces the file body with an
outline of its top-level functions/classes plus docstrings. In live
tests that cut prompt tokens 74–76% on both OpenAI and Anthropic,
same answer either way.
How it works:
- `pip install headroom-ai` now installs ast-grep via a PyPI wheel
(core dep). difftastic and scc are fetched once at proxy startup
from pinned upstream GitHub releases and cached per-user.
- A generic registry (`headroom/proxy/interceptors/`) lets us add more
tool-aware rewrites in one file each: declare `matches()` and
`transform()`, call `register()`, done. No proxy or metrics plumbing
per tool.
- Safety rails built in: pass-through when a Read specifies a line
range; second Read of the same file in a conversation returns full
content (progressive disclosure); any failing interceptor logs and
skips, never crashes a request.
Opt-in for now:
- Off by default while this ships. Turn on with
`headroom proxy --intercept-tool-results` or
`HEADROOM_INTERCEPT_ENABLED=1`, so we can measure before flipping
defaults.
What users see after turning it on:
- First `headroom wrap claude` boot is ~5s longer (binaries fetched).
Every subsequent run is cache-only.
- Existing `transforms_applied` field in metrics gets entries like
`interceptor:ast-grep`, so savings show up in current dashboards
and HTML reports with no UI change.
Other housekeeping in this PR:
- uv.lock moved to .gitignore — regenerated locally per environment.
- 35 unit + integration tests, ruff + mypy clean.
- Dead-code audit done: removed `binaries.run()`, `needs_filesystem`
plumbing, unused `_kind` tuple elements, unused `tool_output`
parameter, and the never-set HEADROOM_SKIP_TOOLS_BOOTSTRAP env.
Align with SpecKit's canonical docs/ structure. Update .gitignore
comment to reflect new location.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The release workflow now uses a loop-free algorithm:
- pyproject.toml is the canonical source of truth (never committed by workflow)
- Git tags use v{canonical}.{height} format (e.g. v0.5.25.3)
- npm publishes use 3-part semver bumped from canonical
- No commit step eliminates infinite release loops
- paths-ignore reduces unnecessary workflow triggers
Also:
- Add .releaseetadata to .gitignore
- Separate npm_version output for semver-compatible npm publishing
- create-release no longer blocks on publish jobs
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add system-native install scripts and host wrappers for running Headroom from Docker while keeping wrapped tools on the host. Document the Docker-native path, add a complete CLI reference with help output and parity details, and add support for root help/version aliases and proxy env-based binding behavior.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
- Remove bandit_result.txt, pip_audit_result.txt, reqs.txt, ruff_result.txt
from repo (security risk: bandit output lists exact vuln locations) + gitignore
- Fix unbound original_tokens in batch handler except block (crash on first
batch request failure)
- Downgrade temporary cache debug log from INFO to DEBUG (fires on every
streaming request, polluting production logs)
- Remove duplicate _extract_anthropic_cache_ttl_metrics from AnthropicHandlerMixin
(StreamingMixin version wins via MRO, duplicate was dead code)
- New `compress()` function: HTTP client calling POST /v1/compress on the proxy
- HeadroomClient: reusable client with retry, fallback, auth support
- Vercel AI SDK adapter: headroomMiddleware() for wrapLanguageModel()
- OpenAI SDK adapter: withHeadroom() Proxy wrapper
- Anthropic SDK adapter: withHeadroom() Proxy wrapper
- Format converters: Vercel AI SDK ↔ OpenAI message format round-trip
- POST /v1/compress proxy endpoint: compression without LLM call
- 90 TypeScript tests (84 unit + 6 integration) + 9 Python tests
- Zero runtime dependencies, all framework peers optional
- Updated README, proxy docs, integration guide, and 6 other doc pages
- New docs/typescript-sdk.md with full SDK documentation
- Removed docs/superpowers/ from tracking (.gitignore)
LLMLingua was the original ML text compressor (BERT-based). Kompress
(ModernBERT, trained on 330K structured tool outputs) replaced it with
better compression quality and simpler architecture.
Removed across 35 files:
- Deleted headroom/transforms/llmlingua_compressor.py
- Deleted tests/test_transforms/test_llmlingua_compressor.py
- Deleted tests/test_proxy_llmlingua.py
- Removed all enable_llmlingua config, _get_llmlingua methods,
LLMLingua fallback paths, LLMLINGUA strategy enum values
- Removed CLI flags, model configs, compression handler references
- Simplified ContentRouter: Kompress is primary and only text compressor
Add comprehensive macOS deployment support for running headroom proxy as a
persistent background service using LaunchAgent. This enables automatic startup,
crash recovery, and proper lifecycle management for local development environments.
Files added:
- examples/deployment/macos-launchagent/com.headroom.proxy.plist.template
- examples/deployment/macos-launchagent/install.sh (shellcheck-clean)
- examples/deployment/macos-launchagent/uninstall.sh (shellcheck-clean)
- examples/deployment/macos-launchagent/shell-integration.sh (bash + zsh)
- examples/deployment/macos-launchagent/README.md
- docs/macos-deployment.md
Key features:
- Configurable port via HEADROOM_PROXY_PORT environment variable (default: 8787)
- Automated installation and uninstallation scripts
- Shell integration supporting both bash and zsh
- Comprehensive documentation with troubleshooting guide
- All shell scripts are shellcheck-clean (zero errors, warnings, or info messages)
Files modified:
- .gitignore: Added CLAUDE.md to prevent committing local config
- docs/README.md: Added Deployment & Operations section with navigation entry
AI review: Pending (will be run by pre-commit hook)
Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>
