241a5c94ea
## Summary - Bump `@earendil-works/pi-ai` and `@earendil-works/pi-coding-agent` from **0.80.10 → 0.81.1** (exact matched pins). - Update `pnpm-workspace.yaml` overrides so floating `*` consumers (`droid-cli`, `pi-llama-cpp`, runtime plugins) stay on the same ModelRuntime surface. - Refresh pin-guard tests, package-config assertions, and FNXC notes for the new pin. ## What's new in pi 0.81.x - Qwen Token Plan providers - Expanded usage accounting (tools/compaction/branch summaries) - Resilient compaction retries + retry lifecycle events - Full provider-extension registration API - Built-in llama.cpp router management - Provider/catalog fixes (Bedrock env credentials, OpenAI Responses early-stream retry, Codex 272K defaults, extension stream-fallback restore) ## Test plan - [x] `scripts/check-pi-versions-pinned` (4/4) - [x] Typecheck: core, engine, dashboard, cli, pi-claude-cli - [x] `package-config.test.ts` (35) - [x] `provider-registration.test.ts` (14) - [x] `auth-storage-concurrency` + `model-registry-refresh` (15) - [x] `register-model-routes-kimi-k3-supplemental` (1) - [ ] CI gate green - [ ] Spot-check Anthropic OAuth + API key session - [ ] Spot-check openai-codex model picker / supplemental models <!-- This is an auto-generated comment: release notes by coderabbit.ai --> ## Summary by CodeRabbit * **Updates** * Updated the bundled Pi runtime to version 0.81.1. * Added support for newer models and providers, including Qwen Token Plan. * Improved usage accounting and session reliability. * Strengthened compaction retry handling and provider catalog accuracy. * Added support for the expanded maximum thinking level. <!-- end of auto-generated comment: release notes by coderabbit.ai -->
pi-claude-cli
A pi extension that routes LLM calls through the Claude Code CLI as a subprocess. Use your Claude Pro/Max subscription as the LLM backend — no API key, no separate billing.
How it works
The extension registers as a custom pi provider exposing all Claude models. Each request spawns a claude -p subprocess using the stream-json wire protocol, with --resume on follow-up turns to reuse the CLI's session state instead of replaying full history. Claude proposes tool calls, pi executes them natively. Custom pi tools are exposed to Claude via a schema-only MCP server.
Requirements
- Claude Code CLI installed and authenticated (
claudeon PATH) - A Claude Pro or Max subscription
- pi or GSD
Installation
Add to ~/.gsd/agent/settings.json:
{
"packages": ["npm:pi-claude-cli"]
}
Then select a Claude model via /model in the interactive UI. All Claude models appear under the pi-claude-cli provider.
Features
- Streams text, thinking, and tool call tokens in real-time
- Maps tool names and arguments bidirectionally between Claude and pi
- Exposes custom pi tools to Claude via MCP (schema-only, no execution)
- Break-early pattern prevents Claude CLI from auto-executing tools
- Session resume via
--resumeeliminates history replay on follow-up turns - Configurable thinking effort with elevated budgets for Opus models
- Cross-platform subprocess management (Windows, macOS, Linux)
- Inactivity timeout and process registry for cleanup
License
MIT