fix: prevent stat inflation in compute_frozen_count, fix token estimate

test: add integration tests for token_headroom mode

feat: add compression_cache stats to /stats endpoint

feat: add token_headroom branch to OpenAI handler

feat: add token_headroom branch to Anthropic handler

feat: configure pipeline based on HEADROOM_MODE at startup

feat: add HEADROOM_MODE config to ProxyConfig

feat: export CompressionCache from cache package

feat: add frozen count, apply_cached, update_from_result to CompressionCache

feat: add CompressionCache with LRU eviction for token headroom mode

fix: add claude-opus-4-6 (1M context) to ANTHROPIC_CONTEXT_LIMITS

Add implementation plan for token headroom mode

Update token headroom mode spec with review fixes

Add design spec for token headroom mode

Add SharedContext for multi-agent, rewrite README, fix proxy cleanup

Updated README and mcp docs

Enhanced MCP server with compress/stats tools, fix proxy torch crash

Fixing tests

Fix proxy crash when torch not installed (kompress lazy imports)

Bump version to 0.4.4

Add wrap commands for Codex/Cursor/Aider with rtk instructions, fix savings metrics

Bump version to 0.4.3

Commit message:

Slim core dependencies: 2.5GB → 195MB install size

fix(wrap): route through proxy, prevent pipe deadlock, bump to 0.4.2

Fixing headroom wrap claude

Fix mypy no-any-return errors in kompress_compressor.py

Format prefix_cache_benchmark.py for ruff 0.15 compatibility

Fix pricing lookup for retired Claude models

Fix proxy backend bugs: env vars, tool forwarding, and provider support

Introducing headroom wrap

Kompress: use argmax for token selection, remove hardcoded 0.5 threshold

Add Kompress: ModernBERT token compressor replacing LLMLingua-2

Updating the README to reflect the right positioning for Headroom

feat: provider-aware prefix cache tracking and combined savings dashboard

fix: dashboard metrics, TTFB tracking, eager LLMLingua loading, and multi-provider consistency

docs: add headroom learn demo GIF to README

Fixing the build AND the perf logs

style: ruff format parser.py, content_router.py, smart_crusher.py

fix(learn): pass explicit model in tests to avoid API key requirement

feat: add `headroom perf` CLI and rewrite `headroom learn` to use LLM analysis

fix: harden edge cases, expand tool exclusions, and fix mypy errors

Merge pull request #40 from smartwatermelon/claude/fix-mcp-startup-and-install-20260306

fix(mcp): fix uninstall symmetry, move subprocess import, add cli-path tests

fix(mcp): use claude mcp add for install and fix startup import

Revert "docs: update README demo GIF to headroom_recording.gif"

docs: update README demo GIF to headroom_recording.gif

feat(router): adaptive compression with Read lifecycle and context-pressure scaling

feat(code): add semantic symbol importance to CodeAwareCompressor

Merge pull request #39 from longtngo/fix/dev-deps-sqlite-vec