KimiX

项目详情摘要

Kimi-CLI-X 中文文档 Install `install.py` handles the whole setup: dependency sync, native runtime download, and tool registration. **Note:** This repo supports not only KIMI LLM but also various API keys! Like OpenAI, Anthropic, etc. Default config templates are in `docs/`; use `kimix --config=xx.json` after setup. Why Kimi-CLI-X? Kimi-CLI-X is a deep optimization of the original Kimi-CLI, focusing on **prompt efficiency**, **tool reliability**, and **extensibility**, plus new tools for real-world development. Optimizations 1. **Lean system prompts** — Compressed initial prompts and tool descriptions down to ~2000 tokens while covering nearly all built-in tools. 2. **Hardened permissions & validation** — Properly handles Shell, Glob, and other tool validations to reduce retry loops from failures. 3. **Better subprocess output** — Redirects large outputs to temp files and filters redundant logs for easier backend retrieval. 4. **Simpler concurrency** — Streamlined design for subprocesses, sub-agents, and background tasks. 5. **Programmable prompts** — Allows custom system prompt injection at the upper layer for flexible scenarios. 6. **Explicit conversation management** — Clearer multi-task orchestration and state tracking. 7. **Write-and-validate** — Auto format checks and warnings on strict config files to prevent model-hallucinated errors. 8. **Multi-API support** — Import custom configs compatible with OpenAI, Anthropic, and more. 9. **Verified backends** — Tested against kimi, xai, anthropic, openai_legacy, openai_responses, google_genai/gemini, vertexai/vertex, plus 25+ Hermes-ported providers (deepseek, openrouter, xiaomi, zai, minimax, bedrock, ...). See `kimi-cli/tests/core/test_create_llm.py` and the Supported Providers list. New Capabilities Capability | Descripti