Coding agent for local LLMs. Ollama & LM Studio。
What it does
- Runs on Ollama or LM Studio on your machine
- Indexes your repo with tree-sitter (symbols, fast lookup)
- Streaming TUI with syntax-colored code and markdown replies
- Headless mode for scripts and CI (
shrimp run)
Install
git clone https://github.com/ddoemonn/shrimp.git
cd shrimp
cargo install --path crates/shrimp-cliUsage
shrimp # open TUI: pick provider, model, chat
shrimp run -P "create hello.py" # one-shot prompt, no TUI
shrimp index # index the current repoProviders: Ollama at localhost:11434 (configurable via OLLAMA_HOST environment variable), LM Studio at localhost:1234 (configurable via LM_STUDIO_HOST environment variable).
| key | action |
|---|---|
| Enter | send message |
| ↑↓ / PgUp/PgDn | scroll transcript |
| ⌘C / Ctrl+Shift+C | copy transcript |
| End | jump to live stream |
| /model /provider | switch model or provider |
| /undo /reindex | undo last edit or rebuild index |
| Esc | quit |
Config
Settings live in .shrimp/config.toml. Every field is optional.
provider = "ollama"
base_url = "http://hub:11434" # custom base URL (optional)
model = "gemma4:12b"
auto_approve = true
max_context_tokens = 8192from https://github.com/ddoemonn/shrimp
No comments:
Post a Comment