https://github.com/retospect/acatome-lambic
Generic MCP-aware LLM shell — ollama/litellm, parallel tool dispatch, streaming TUI
https://github.com/retospect/acatome-lambic
agent llm ollama python tool-calling
Last synced: about 2 months ago
JSON representation
Generic MCP-aware LLM shell — ollama/litellm, parallel tool dispatch, streaming TUI
- Host: GitHub
- URL: https://github.com/retospect/acatome-lambic
- Owner: retospect
- License: mit
- Created: 2026-03-10T19:19:55.000Z (3 months ago)
- Default Branch: main
- Last Pushed: 2026-04-01T22:21:16.000Z (2 months ago)
- Last Synced: 2026-04-01T22:25:22.588Z (2 months ago)
- Topics: agent, llm, ollama, python, tool-calling
- Language: Python
- Size: 167 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- Changelog: CHANGELOG.md
- License: LICENSE
Awesome Lists containing this project
README
# acatome-lambic
MCP-aware LLM shell with provider switching.
Connects to MCP servers via stdio, talks to LLMs (ollama, OpenAI, Anthropic via litellm),
and provides a terminal chat interface with tool calling.
## Usage
```python
from acatome_lambic.tui.app import Shell
from acatome_lambic.core.config import LlmConfig, McpServer, ShellConfig
config = ShellConfig(
llm=LlmConfig(provider="ollama", model="qwen3.5:9b"),
servers=[
McpServer(name="precis", cmd=["uv", "run", "precis"]),
],
system_prompt="You are a research assistant.",
)
Shell(config).run()
```
## Commands
- `/model ` — switch LLM
- `/think on|off` — toggle reasoning mode (default: on)
- `/tools` — list tools with on/off status
- `/tools off ` — disable tools matching pattern
- `/tools on ` — enable tools matching pattern
- `/expand ` — show full (untruncated) tool result
- `/status` — show session info
- `/clear` — clear message history
- `/help` — show command help
- `/quit` — exit
Applications can register custom commands (`custom_commands`) and
LLM-routed message commands (`message_commands`) via `ShellConfig`.
All registered commands appear in tab autocomplete.