Replace Qwen2.5-Coder-3B with Phi-4-mini-instruct GGUF

Replace DeepSeek R1 7B with Qwen2.5-Coder-3B

fix: raise context window to model native 128K

fix: expose 64K context for Hermes compatibility

fix: switch Space inference model to 7B GGUF

Increase llama.cpp context to 64K for Hermes compatibility

Fix llama cpp image registry

Use prebuilt llama.cpp CPU server for R1 14B

Fix CPU llama cpp wheel and GGUF filename

Deploy DeepSeek R1 Distill Qwen 14B CPU API

Fix emoji encoding error

Use official Kimi Claw installer

Install Kimi Claw at runtime

Rebuild with Kimi Claw installation

Fix: Install Kimi Claw from tgz package

Fix: Use Node.js 22 for OpenClaw

Add startup script for gateway and bot

Add Node.js and Kimi Claw to container

Fix Kimi WS: use additional_headers

Fix Kimi: use WebSocket instead of REST API

Add Kimi bot platform integration via WebSocket

Add Kimi (Moonshot) AI: kimi-k2.7-code, kimi-k2.6, kimi-k2.5, moonshot-v1-128k

Fix: declare global WS_POLL_TASK and WS_BROADCAST_TASK

Fix: properly stop polling and broadcasting when streams stopped

Fix: WS fallback to MEXC REST poll when Binance blocked

WebSocket real-time streaming: Binance/MEXC live tick data, /stream command

Fix: copy backup.py and restore.py into container

Add /export and /import commands for data migration

Docker compose config

Deployment instructions

Quick deploy script

One-click VPS setup script

Restore script

Backup script

Add backup/restore system: /backup create/list

Add 15 crypto/trading plugins: signals, TA, DeFi, NFT, portfolio, market analysis, whale tracking, tokenomics, on-chain, tax, trading bots

35 plugins: translator, coder, writer, analyst, tutor, email, resume, social, recipe, fitness, travel, business, legal, debate, summarizer, math, philosophy, history, science, motivator, interview, storyteller, jokes, horoscope, music, movie, gaming, parenting, diy, photography, language, investing, mental, writing, presentation, project, startup, design, ai, crypto, code review, devops, database, api, security

Separate /model into Chat Models and Agent Models

Plugin system: 15 plugins, install/remove, custom plugins, auto-detect triggers

Fix: load history before search/crypto context

Natural language commands - no slash needed, just chat naturally

Full agent upgrade: memory, tasks, skills, workspace, API client, self-improvement

Add AI agent: 6 models, /agent /run /fetch /connect github commands

Add beautifulsoup4, lxml for web scraping

Fix crypto: MEXC+CoinGecko fallback, forceful context injection

Fix: use ddgs package

Fix: Binance via CF proxy, ddgs import, crypto detection logging

Auto-detect crypto questions and inject live Binance data as context

Add mplfinance, matplotlib, websockets

Add crypto: /price /chart /alert with Binance API