pi-toolkit

All-in-one pi extension toolkit.

Includes:

• Providers: kimi-coding, minimax, xiaomi-mimo, crofai
• Cache optimization for Anthropic-compatible providers
• Claude OAuth adapter for Anthropic OAuth sessions
• Native web search with multiple backends + provider override
• Context management (anchor, view, pivot, recall)

Install

pi install npm:@ersintarhan/pi-toolkit

Local development:

pi -e /Users/ersin/Projects/pi-extensions/pi-toolkit

What this package does

1. Provider registration

Registers these providers:

| Provider | API | Notes | |---|---|---| | kimi-coding | Anthropic-compatible or OpenAI mode | OAuth, API key fallback, image/video upload, Kimi-specific payload mutations | | minimax | Anthropic-compatible | Cache-fixed stream path | | xiaomi-mimo | Anthropic-compatible | Cache-fixed stream path, MiMo quirks handled | | crofai | OpenAI-compatible | Dynamic model list, usage support |

2. Cache optimization

For Anthropic-compatible providers (kimi-coding, minimax, xiaomi-mimo), this package uses a custom cached stream implementation derived from the better-messages-cache approach.

It does three things:

• marks the last assistant tool_use block
• marks the last user block
• repairs partial/invalid streaming JSON during tool call parsing

This keeps cache loss lower across long agentic turns.

3. Claude OAuth adapter

When using /model anthropic/... with Anthropic OAuth:

• strips the Claude Code identity block
• injects the billing header Claude Code expects
• shows footer status like ✓ Claude OAuth ready/active
• docs re-injection is disabled by default

4. Native search

Adds:

• web_search
• web_fetch
• /search

Supported native search backends:

| Backend | Search | Fetch | Auth | |---|---|---|---| | ZAI (GLM) | ✅ | ✅ | ZAI_API_KEY | | Google Gemini | ✅ | ❌ | GEMINI_API_KEY | | OpenAI | ✅ | ❌ | OPENAI_API_KEY | | xAI | ✅ | ❌ | XAI_API_KEY | | Anthropic | ✅ | ❌ | ANTHROPIC_API_KEY | | Claude Code bridge | ✅ | ✅ | Claude CLI auth | | Codex / ChatGPT | ✅ | ❌ | codex login / ~/.codex/auth.json | | Fallback | DuckDuckGo | HTTP fetch | none |

You can pin search to a provider independently from the active model.

Examples:

/search provider zai
/search provider openai
/search provider codex
/search provider auto

Or choose it from the TUI settings panel with /search.

5. Context management

Adds the context tool with:

• view
• recall
• anchor
• pivot

Also includes:

• status line (context=, tool=, anchor=)
• tool-result truncation before the last anchor
• anchor-aware cache support in the integrated context flow

Provider usage

Kimi

/model kimi-coding/kimi-for-coding

Auth options:

• KIMI_API_KEY
• /login kimi-coding OAuth

Optional:

• KIMI_CODE_PROTOCOL=openai
• KIMI_CODE_MODEL=<model>

MiniMax

/model minimax/MiniMax-M2

Requires:

• MINIMAX_API_KEY

Xiaomi MiMo

/model xiaomi-mimo/mimo-v2.5
/model xiaomi-mimo/mimo-v2.5-pro

Requires:

• XIAOMI_TOKEN_PLAN_API_KEY

Optional region override:

export XIAOMI_MIMO_BASE_URL=https://token-plan-ams.xiaomimimo.com/anthropic

CrofAI

/model crofai/openai/gpt-4o

Requires:

• CROFAI_API_KEY

Models are fetched dynamically from:

• https://crof.ai/v1/models

/usage command

Shows quota / plan usage for the active provider.

Supported:

| Provider | Endpoint | |---|---| | kimi-coding | https://api.kimi.com/coding/v1/usages | | minimax | https://api.minimax.io/v1/token_plan/remains | | xiaomi-mimo | https://platform.xiaomimimo.com/api/v1/tokenPlan/usage | | crofai | https://crof.ai/usage_api/ |

Examples:

/usage

/context command

Shows current context-window usage without adding the generated report to future LLM context.

It includes:

• total context usage and model window
• estimated category breakdown
• active tools and slash commands
• extension allocation by source/package
• cache read/write and cost stats

Example:

/context

Search override examples

Use Kimi for coding, but ZAI for search:

/model kimi-coding/kimi-for-coding
/search provider zai

Use any model, but force Codex web search:

/search provider codex

Return to auto-detect:

/search provider auto

Context workflow examples

Anchor after a completed phase:

context(anchor, name="crofai-added", summary="CrofAI provider added with dynamic models and usage support.")

Pivot back to a clean checkpoint:

context(pivot, target="crofai-added", carryover="Search override added later; CrofAI implementation is stable.")

Environment summary

Common env vars:

# Kimi
export KIMI_API_KEY=...
export KIMI_CODE_PROTOCOL=anthropic

# MiniMax
export MINIMAX_API_KEY=...

# Xiaomi MiMo
export XIAOMI_TOKEN_PLAN_API_KEY=...
export XIAOMI_MIMO_BASE_URL=...

# CrofAI
export CROFAI_API_KEY=...

# Search backends
export ZAI_API_KEY=...
export GEMINI_API_KEY=...
export OPENAI_API_KEY=...
export XAI_API_KEY=...
export ANTHROPIC_API_KEY=...

Compatibility notes

• Do not install older overlapping Kimi provider forks at the same time.
• Local pi -e ... development may behave differently from installed npm packages for skill loading.
• Codex search requires codex login first.
• Claude OAuth adapter only activates for the anthropic provider when OAuth is actually in use.

Changelog

0.5.4

• package renamed to @ersintarhan/pi-toolkit
• CrofAI provider added
• Claude OAuth adapter merged
• native search merged
• Codex / ChatGPT search backend added
• search provider override added (including TUI selection)
• pi-auto-context integrated
• cache TTL upgraded to 1h
• status line simplified
• skill metadata added

0.5.3

• search provider override
• Codex backend

0.5.2

• native search merged

0.5.1

• Claude OAuth adapter merged

0.5.0

• CrofAI provider added

License

MIT