polyharness
Make your AI agent evolve automatically through iterative harness optimization.
Description
Make your AI agent evolve automatically through iterative harness optimization.
Release History
| Version | Changes | Urgency | Date |
|---|---|---|---|
| 0.2.1 | Imported from npm (0.2.1) | Low | 4/21/2026 |
| v0.2.1 | ## v0.2.1 — Shell Hook Auto-Wrap + Harness Guard ### Highlights **`ph shell-hook` — zero-config auto-wrap (3 new commands)** No more typing `ph wrap --auto-evolve` every time. Install a shell hook once and all agent commands are automatically wrapped: ```bash ph shell-hook install # one-time setup → ~/.zshrc claude -p "Fix the auth bug" # auto-becomes: ph wrap --auto-evolve claude -p ... ``` Supports `claude`, `claw`, `codex`, and `opencode`. Fully reversible with `p | Medium | 4/8/2026 |
| v0.2.0 | ## What's New — Online Evolution 🧬 PolyHarness can now **collect traces from your daily agent usage** and **auto-trigger evolution** — your agent improves while you work. ### New Commands | Command | Description | |---------|-------------| | `ph wrap <cmd>` | Transparent command wrapper — records execution traces (agent, exit code, duration, output) | | `ph wrap --auto-evolve` | Auto-triggers evolution when trace count reaches threshold (default: 50) | | `ph traces list / show / stat | Medium | 4/8/2026 |
| v0.1.3 | ## What's New ### Added - `ph new <dir>` — scaffold command that generates starter `harness.py`, `test_cases.json`, and `evaluate.py` - `--template` option for `ph init` — one-command workspace setup from 5 bundled templates (text-classification, math-word-problems, code-generation, rag-qa, api-calling) - `ph upgrade` — check for updates and upgrade PolyHarness via pip - `ph uninstall` — clean removal with confirmation prompt - 7 new tests (128 total) ### Changed - Consolidated `ex | Medium | 4/8/2026 |
| v0.1.2 | ## What's New ### Added - **OpenAI-compatible backend** — `--agent openai` for local models (Ollama, vLLM, LM Studio) and any OpenAI-compatible API - Default workspace changed to `.ph_workspace` for better project isolation ### Fixed - Friendly error message when `evaluate.py` is missing + warning on init - Ruff lint fixes (`E501` long lines, `openai_proposer.py`) ### Docs - Improved README for beginners: harness definition, preparation guide, local model setup, config reference | Medium | 4/8/2026 |
| v0.1.1 | # PolyHarness v0.1.0: Initial Public Release 🚀 We are thrilled to announce the first public release of **PolyHarness** — the open-source performance optimization layer for the AI Agent ecosystem! PolyHarness automatically upgrades existing agents (Claude Code, Claw Code, Codex, etc.) via iterative Meta-Harness search. You run one command — your agent gets smarter. ### 🌟 Key Features * **Automated Self-Evolution**: An iterative search loop that records history, evaluates candidate | Medium | 4/4/2026 |
Dependencies & License Audit
Loading dependencies...
Similar Packages
@bfun-bot/cliBFunBot CLI — deploy tokens, check balances, and manage your AI agent from the terminal1.2.0
@agent-analytics/cliStop juggling dashboards. Let your agent do it. Analytics your AI agent can actually use — track, analyze, experiment, optimize.0.5.21
agent-office-cliRun and manage AI agent sessions locally, with optional relay to agentoffice.top0.1.9