v0.3.0 — Benchmark Harness, Visual Verification, Computer Control

What's New

Benchmark Infrastructure

• selfware bench --suite throughput,multilang --report — one-command benchmark suite
• selfware auto-config --endpoint <url> --save — auto-detect endpoint capabilities
• SWE-bench evaluation pipeline with fuzzy patch application (62% apply rate)
• Multi-language coding benchmark: Rust, Python, JavaScript/TypeScript, Go

Visual Verification System

• Hard gating: blocks progress on visual assertion failures (configurable)
• Stuck-loop detection: screenshot hashing detects repeated screens
• Durable screenshot storage at ~/.selfware/visual_evidence/
• Assertion lifecycle: set → checked each step → cleared on pass

Computer Control (Linux)

• Keyboard: type_text, press_key, key_combo via xdotool
• Mouse: move_to, click, double_click, scroll, drag via xdotool
• Window management: list/focus/active via wmctrl + xdotool

Memory & Storage

• Memory consolidation ("sleep") system — sessions auto-compact to long-term storage
• Encrypted KV store (AES-256-GCM with PBKDF2 key derivation)
• HNSW batch import for vector store

Code Quality

• 54-issue review across 20 perspectives, 40+ resolved
• File decomposition: execution.rs, context_management.rs, config/mod.rs, vector_store.rs split
• 6,285 tests passing, 0 failures
• Release profile: LTO thin, strip, codegen-units=1

Security

• Git tools: path validation via SafetyConfig
• extra_body validation: blocks model/messages/tools override
• CI: secret scanning, benchmark regression detection

Benchmark Results (Qwen3.5-27B, 2x4090)

Metric	Value
Throughput (32x)	290 tok/s
Multi-language	12/12 (100%)
SWE-bench Lite	5/300 resolved
Patch apply rate	68% (fuzzy)

Full Changelog: v0.2.2...v0.3.0

Selfware v0.2.2

Release 0.2.2.

Highlights:

• fixes queued input handling and retry suppression
• improves process tracking, reuse, and port reservation
• adds persistent session/debug logging and debug commands
• aligns local browser/http/vision workflows
• includes broader end-to-end coverage and release packaging cleanup

Full Changelog: v0.2.0...v0.2.2

Selfware v0.2.0

Full Changelog: v0.1.4...v0.2.0

Selfware v0.1.4

Full Changelog: v0.1.3...v0.1.4

Selfware v0.1.3

Full Changelog: v0.1.2...v0.1.3

Selfware v0.1.2

Full Changelog: v0.1.1...v0.1.2

Selfware v0.1.1

What's Changed

📦 Other Changes

• Fix CI blockers and complete TUI enhancements by @architehc in #17
• ci: bump actions/upload-artifact from 4 to 6 by @dependabot[bot] in #1
• ci: bump actions/cache from 4 to 5 by @dependabot[bot] in #2
• ci: bump softprops/action-gh-release from 1 to 2 by @dependabot[bot] in #3
• ci: bump codecov/codecov-action from 4 to 5 by @dependabot[bot] in #4
• ci: bump actions/checkout from 4 to 6 by @dependabot[bot] in #5
• deps: bump colored from 2.2.0 to 3.1.1 by @dependabot[bot] in #6
• deps: bump whoami from 1.6.1 to 2.1.1 by @dependabot[bot] in #7
• deps: bump git2 from 0.19.0 to 0.20.4 by @dependabot[bot] in #8
• deps: bump nix from 0.28.0 to 0.30.1 by @dependabot[bot] in #10
• deps: bump toml from 0.8.23 to 1.0.1+spec-1.1.0 by @dependabot[bot] in #11
• deps: bump thiserror from 1.0.69 to 2.0.18 by @dependabot[bot] in #12
• deps: bump pulldown-cmark from 0.9.6 to 0.13.0 by @dependabot[bot] in #14
• deps: bump rusqlite from 0.31.0 to 0.38.0 by @dependabot[bot] in #15
• deps: bump reqwest from 0.12.28 to 0.13.2 by @dependabot[bot] in #9
• deps: bump reedline from 0.32.0 to 0.45.0 by @dependabot[bot] in #13
• Fix pulldown-cmark 0.13 API changes and TUI improvements by @architehc in #18
• Add developer experience improvements by @architehc in #19
• Refactor feature flags into semantic groups by @architehc in #20
• Fix unit tests for context and git modules by @architehc in #21
• ci: bump actions/checkout from 4 to 6 by @dependabot[bot] in #37
• ci: bump actions/cache from 4 to 5 by @dependabot[bot] in #35
• ci: bump github/codeql-action from 3 to 4 by @dependabot[bot] in #36
• ci: bump actions/download-artifact from 4 to 8 by @dependabot[bot] in #33
• ci: bump codecov/codecov-action from 4 to 5 by @dependabot[bot] in #34
• deps: bump the rust-minor group with 3 updates by @dependabot[bot] in #38
• deps: bump nix from 0.30.1 to 0.31.2 by @dependabot[bot] in #40
• deps: bump reedline from 0.45.0 to 0.46.0 by @dependabot[bot] in #39
• deps: bump thiserror from 1.0.69 to 2.0.18 by @dependabot[bot] in #43
• deps: bump tiktoken-rs from 0.7.0 to 0.9.1 by @dependabot[bot] in #26
• deps: bump criterion from 0.5.1 to 0.8.2 by @dependabot[bot] in #28
• deps: bump dirs from 5.0.1 to 6.0.0 by @dependabot[bot] in #24

New Contributors

• @dependabot[bot] made their first contribution in #1

Full Changelog: v0.1.0...v0.1.1

Selfware v0.1.0

Initial Release

Features

• Core agent framework with PDVR cognitive cycle
• 53 built-in tools for file operations, git, cargo, search
• Safety system with hardened shell command validation
• Path traversal protection with symlink chain detection
• Checkpoint system for task persistence
• Multi-agent collaboration support
• TUI mode with ratatui
• Docker support

Security

• Regex-based shell command validation with obfuscation detection
• Base64-encoded command detection
• Command chaining detection
• Symlink attack prevention

Platforms

• Linux x86_64
• macOS Apple Silicon (ARM64)
• Windows x86_64

Note: Linux ARM64 and macOS Intel builds require additional cross-compilation setup.

🤖 Generated with Claude Code