Skip to content

Releases: Kayzo/headroom

Release v0.11.1

26 Apr 12:11
ab7b191

Choose a tag to compare

[0.11.1] - 2026-04-26

Other Changes

Release v0.11.0

25 Apr 22:29
342a6a4

Choose a tag to compare

[0.11.0] - 2026-04-25

Features

  • pipeline: Introduce canonical pipeline lifecycle contract (1fd8760)

Bug Fixes

  • memory: resolve Qdrant connection from HEADROOM_QDRANT_* env vars (chopratejas#31) (7c1cff2)
  • memory: collapse and decay error_recovery patterns in MEMORY.md (6dede0c)
  • memory: batch onnx embeddings and sqlite-vec ops (32152f5)
  • init: guide users when no agents are auto-detected + expand e2e (7b99b05)
  • wrap: unwrap codex restores prior config.toml (ab0ee65)
  • learn: show prior patterns block to LLM to prevent dangling cross-references (8d4f21b)
  • proxy: strip cache_control before hashing turn_id (f2c6e90)
  • proxy: emit turn_id linking agent-loop API calls to a single user prompt (2e351c6)
  • subscription: raise default OAuth usage poll interval to 300s (a1d9832)
  • Fix MCP command resolution to use stable CLI entrypoint (247060c)

CI/CD

  • docker: clean up image tags, signatures, and Latest indicator (b2527a0)
  • retry workflow validation dry-runs (c821260)

Chores

  • renormalize line endings to LF (e635bab)
  • deps: bump the npm_and_yarn group across 3 directories with 4 updates (62753e4)
  • deps: bump nltk from 3.9.2 to 3.9.4 in the uv group across 1 directory (9aa6487)
  • sync plugin manifest versions (1b70e5c)

Tests

  • backpressure: harden livez-under-backpressure against CI jitter (429e1d1)
  • expand coverage across helper slices (b80ab7d)

Other Changes

  • Rust stage 2 tokenizer (342a6a4)
  • phase-0: rust workspace scaffolding + parity harness (68c506d)
  • feat+fix(learn): live flush of traffic patterns to CLAUDE.md / MEMORY.md + persist real evidence_count (e4b8720)

Release v0.10.1

22 Apr 19:51
4c9c29c

Choose a tag to compare

[0.10.1] - 2026-04-22

Bug Fixes

  • cache dashboard stats snapshots (4c9c29c)
  • support live copilot oauth runtime (e71ebd5)
  • learn: preserve prior recommendations across runs (40f11d8)
  • proxy: log streaming Anthropic requests so /transformations/feed populates (f5e8bea)
  • unblock CI on main (cli submodule binding + test sys.modules hygiene) (706c5e9)

Release v0.10.0

22 Apr 08:05
5738339

Choose a tag to compare

[0.10.0] - 2026-04-22

Features

  • init: Add durable headroom init command for agent hooks (ed7d494)
  • dashboard: live feed sidebar with message transformations (0d6ba99)
  • bundle ast-grep/difftastic/scc + generic tool_result interceptor framework (21d909a)
  • add Pi/Codex and Cloud Code Assist compatibility routes (e6cdc2f)
  • wire kompress_model through compress pipeline (1039f66)
  • specs: chopratejas#183 scaffold initial application specs. (80cfcd7)
  • canonical HEADROOM_CONFIG_DIR and HEADROOM_WORKSPACE_DIR filesystem contract (cb7bc8c)
  • kompress: add compress_batch with device-aware routing (b119430)
  • AI quota & rate-limit tracking — Anthropic, OpenAI Codex, and GitHub Copilot (54ae7b9)
  • add persistent install lifecycle management (65ee4b1)
  • learn: add CLI-based LLM backends for keyless headroom learn (a245234)
  • cli: add Docker-native install flow and parity docs (9f124c9)
  • add GitHub Copilot CLI wrap support (fb953eb)
  • add proxy liveness and readiness healthchecks (f15ef78)
  • add reproducible devcontainers (5b81395)
  • anonymized telemetry warning across all user-facing paths (cd79d77)
  • openclaw: route configurable gateway providers through headroom (e4f7256)
  • cli: add one-command OpenClaw wrap bootstrap (43f7398)
  • Add Anthropic url overrides via env vars and cli flags (32f97f2)
  • persist proxy savings history (a99d4b9)
  • allow overriding proxy telemetry sdk via HEADROOM_SDK (42ea085)
  • LangGraph compress_tool_messages node for ToolMessage compression (4ae9bd3)
  • add license key validation + phone-home usage reporter (d1fe0d4)
  • add live traffic learning + cross-agent memory writers (--learn flag) (0fd6dfc)
  • make token_headroom the default mode + fix Gemini handler bug (9c5580b)
  • provider-aware prefix cache tracking and combined savings dashboard (4a655fb)
  • add headroom perf CLI and rewrite headroom learn to use LLM analysis (4d14012)
  • router: adaptive compression with Read lifecycle and context-pressure scaling (655df09)
  • code: add semantic symbol importance to CodeAwareCompressor (9c31b22)
  • cli: add IntelligentContext CLI arguments (d57673d)
  • toin: add TOIN + CCR integration to IntelligentContextManager (fef02fa)
  • toin: extend TOIN integration to all ContentRouter compression strategies (4f9a3d1)
  • proxy: add connection pooling, HTTP/2, and multi-worker support (7c6d713)
  • add streaming memory tool support with credential error handling (9825d99)
  • Add CCR architecture, TOIN telemetry, and DevEx improvements (c1feb60)

Bug Fixes

  • support GitHub Copilot OAuth sessions (1f978f9)
  • onnx: reduce retained cpu memory (80920ed)
  • skip commitlint for PR merge commits (329956e)
  • telemetry: add headroom_stack and install_mode identity fields (b789dca)
  • complete fork-friendly release publishing (16d4608)
  • release: resolve openclaw dependency and boost test coverage (7495687)
  • ci: add --allow-same-version to npm version in release workflow (2e20849)
  • proxy: Codex reconnect-storm resilience — bounded pre-upstream + WS session tracking + stage timings (cd11acc)
  • publish docker and package artifacts in release flow (eff1616)
  • support OAuth Bearer token routing for proxy endpoints (956fe40)
  • take highest release bump across unreleased commits (9469bcb)
  • run release versioning without package imports (4651f96)
  • strip accept-encoding from forwarded proxy headers (01fa4aa)
  • restore semantic release versioning (e2ca3ca)
  • repair release and docs pipelines (73f6673)
  • lazy-load Headroom proxy startup for OpenClaw (12adb8e)
  • use MCP-namespaced tool name in Claude Code instructions (47e734c)
  • skip token inflation guard in cache mode (1006bc3)
  • eliminate duplicate proxy logging and guard against token inflation (039399a)
  • initialize parsed_response to prevent UnboundLocalError crashing all streaming responses (f65c09a)
  • eliminate prefix cache busts from frozen count underestimation (22dad13)
  • openclaw: normalize engine messages for discord (eed9792)
  • image verities (ef2d1dc)
  • openclaw: preserve toolCall linkage for streaming tool outputs (f984021)
  • harden openclaw proxy auto-detect and startup (865926a)
  • remove unused imports in test_langgraph.py (ruff F401) (de862d9)
  • split Supabase anon key to avoid GitGuardian false positive (8925be9)
  • restore Authorization header in telemetry beacon (27e026d)
  • remove Bearer auth header for Supabase publishable key (telemetry broken since 0.5.5) (854d9f3)
  • move extract_user_query import out of conditional scope (19d93b2)
  • forward ratelimit headers in streaming responses (f133809)
  • bedrock: add EU/AP region support with graceful fallback (4397607)
  • close TOIN feedback loop — retrievals now flow back for learning (acddeff)
  • memory handler overwrites system prompt when it's a list of content blocks (9897bd1)
  • use cache-aware pricing from LiteLLM for dashboard cost display (017b864)
  • pass through unknown CLI flags to wrapped tools (claude, codex, aider, cursor) (cd258d7)
  • propagate HEADROOM_MODE through wrap CLI and drop cache write premium penalty (3066783)
  • capture Zone 1 + Zone 2 savings in token_headroom stats (aa73c7b)
  • wrap: route through proxy, prevent pipe deadlock, bump to 0.4.2 (384ac34)
  • dashboard metrics, TTFB tracking, eager LLMLingua loading, and multi-provider consistency (08d81f2)
  • learn: pass explicit model in tests to avoid API key requirement (da481a3)
  • harden edge cases, expand tool exclusions, and fix mypy errors (3945c02)
  • mcp: use claude mcp add for install/uninstall, fix startup import (448acb7)
  • ci: add sqlite-vec to dev deps to fix memory tests on CI (d90c47b)
  • docs: fix Docker example and document hnswlib C++ build requirement (62daa89)
  • docs: correct PyPI package name from headroom to headroom-ai (ae661c8)
  • Handle BaseModelOutputWithPooling from SigLIP get_*_features() (d51ea56)
  • Increase MAX_REQUEST_BODY_SIZE from 10MB to 100MB to support image-heavy requests (beddc99)
  • Extend Anthropic format tool protection to IntelligentContextManager (0e04099)
  • Handle Anthropic format tool_use/tool_result as atomic units (adb710b)
  • resolve mypy type errors (58cc207)
  • resolve CI failures (a187d80)
  • improve error handling and add comprehensive test coverage (d329836)
  • correct cost calculation for cache tokens and simplify memory DevEx (787f925)
  • resolve mypy type errors across codebase (74872b2)
  • parse actual token usage from SSE streams instead of byte estimation (8c22a72)
  • account for cached tokens in cost calculation across all providers (6bb35eb)

CI/CD

  • publish Python distributions to GitHub releases (5738339)
  • group dependabot minor/patch updates per ecosystem (062dbfd)
  • implement optimized release pipeline (f95984b)

Chores

  • memory: add EXTERNAL backend extension points (5391761)
  • proxy: add third-party extension point (3373dab)
  • bump openclaw plugin version chopratejas#162 (4a6c546)
  • fix ruff lint issues (import sorting, trailing whitespace) (4baef0e)
  • bump version to 0.5.20 (8bb1b1e)
  • gitignore headroom-saas (lives in separate private repo) (b4d2cb0)
  • bump version to 0.5.2 (b5631cd)
  • bump version to 0.5.1 (62a5183)
  • bump version to 0.5.0 (b435d4b)
  • add mypy to pre-commit hooks (67631fe)

Performance

  • add compress_system_messages and min_tokens_to_compress config (3242efe)

Refactors

  • centralize error/priority pattern detection into error_detection module (c7a76db)

Documentation

  • port Docker-native, filesystem-contract, and persistent-installs pages to Fumadocs (724c298)
  • fix Kompress-base HuggingFace link (e4f3cf7)
  • rewrite README for clarity and highlight Kompress-base, leaderboard, RTK (2593ff2)
  • add OpenClaw plugin section to main README (814701e)
  • add OpenClaw plugin to README, integration guide, index, and TS SDK docs (95f9e7a)
  • add headroom learn demo GIF to README (68a75aa)
  • update README demo GIF to headroom_recording.gif (1b350b4)
  • update documentation for IntelligentContext TOIN + CCR integration (83c0334)

Styles

  • apply ruff format to server.py (ff0788a)
  • fix ruff format in server.py (80bfb04)
  • ruff format parser.py, content_router.py, smart_crusher.py (7fd1f6f)

Other Changes

  • Sync plugins to 0.9.2, pyproject canonical at 0.9.1 [skip ci] (7ed2b0b)
  • Compact /stats-history responses by default (da54d4a)
  • Fix OpenClaw GitHub Packages build (043045f)
  • Add timeout guards to Anthropic pre-upstream work (8ba1ebd)
  • Update documentation links to new domain (31ea3a1)
  • Add support for custom Anthropic API URL (1e80ee3)
  • Update number array tests for schema-preserving compression (d66fcd4)
  • Add --stateless mode and graceful degradation for enterprise deployments (acb8ff3)
  • Add supply chain hardening: SBOM, cosign signing, pinned digests, Dependabot (4641e43)
  • Auto-register codebase-memory-mcp as MCP server with --code-graph (4ee519f)
  • Fix SmartCrusher bugs: schema violation, race condition, thread safety, recursion (14415db)
  • Parallelize headroom learn session scanning with --workers flag (6cc8fd0)
  • Fix CI lint errors and test failures (89061fd)
  • Fix all broken docs links in README after docs migration (chopratejas#161) (51f1fc2)
  • Fix headroom learn failing on project paths with underscores (chopratejas#159) (7c91fe1)
  • Propagate API key to COPILOT_PROVIDER_API_KEY in wrap copilot (7a7b8b6)
  • Fix memory tools returning 'unsupported call' in Chat Completions (744a580)
  • Make KompressCompressor model-configurable: model_id, chunk_words, score_threshold (b501369)
  • Add cross-agent memory interoperability via MCP, sync engine, and atomic facts (82301bb)
  • Bump version to 0.5.25 for PyPI release (e306d6b)
  • Add memory support for Codex via OpenAI Responses API handler (c245129)
  • Fix --memory flag silently dropped when proxy already running (1b5cc9b)
  • Add --memory flag to all wrap subcommands (5f4ff91)
  • Update README with CompressConfig options and document compression example (c197a14)
  • Add CompressConfig: user-facing compression options for c...
Read more