Loomcast CLI reference

Loomcast ships a command-line interface built with Typer and Rich.

Every command is available through the loomcast executable (or python -m loomcast.cli.main).

Command overview

loomcast [COMMAND] [ARGS] [OPTIONS]

1. loomcast status

Shows the operational state (online / offline), the endpoint URL, and the default runtime for every configured backend (Ollama, LM Studio, llama.cpp).

Syntax

loomcast status

Sample output

┌──────────┬────────────────────────┬─────────┬────────┐
│ Runtime  │ Endpoint               │ Default │ Status │
├──────────┼────────────────────────┼─────────┼────────┤
│ OLLAMA   │ http://localhost:11434 │ Yes     │ Online │
│ LMSTUDIO │ http://localhost:1234  │ No      │ Offline│
│ LLAMACPP │ http://localhost:8080  │ No      │ Offline│
└──────────┴────────────────────────┴─────────┴────────┘

2. loomcast list

Lists every model installed and available across the local runtimes.

Syntax

loomcast list [OPTIONS]

Options

  • -b, --backend TEXT: filter by a specific runtime (ollama, lmstudio, llamacpp).

Examples

# List models across every runtime
loomcast list

# List Ollama models only
loomcast list --backend ollama

3. loomcast install

Downloads and installs a model into the target runtime (for runtimes that support hot installation through an API, such as Ollama).

Syntax

loomcast install MODEL [OPTIONS]

Arguments

  • MODEL (required): identifier of the model to install (e.g. qwen2.5:0.5b, llama3).

Options

  • -b, --backend TEXT: target backend (ollama, lmstudio, llamacpp). Defaults to the configured default backend.

Example

loomcast install qwen2.5:0.5b --backend ollama

4. loomcast remove

Removes an installed model from the target runtime (for example through Ollama's DELETE /api/delete API). Prompts for interactive confirmation before deleting, unless --yes is passed.

Syntax

loomcast remove MODEL [OPTIONS]

Arguments

  • MODEL (required): identifier of the model to remove (e.g. qwen2.5:0.5b, llama3).

Options

  • -b, --backend TEXT: target backend (ollama, lmstudio, llamacpp). Defaults to the configured default backend.
  • -y, --yes: skip the interactive confirmation prompt (useful in scripts).

Examples

# Interactive removal
loomcast remove qwen2.5:0.5b

# Removal without confirmation, on an explicit backend
loomcast remove llama3 --backend ollama --yes

5. loomcast update

Updates a model already installed on the target runtime (for example by re-pulling on Ollama and comparing the SHA-256 digest). The message differs depending on whether the model was updated or was already up to date.

Syntax

loomcast update MODEL [OPTIONS]

Arguments

  • MODEL (required): identifier of the model to update (e.g. qwen2.5:0.5b, llama3).

Options

  • -b, --backend TEXT: target backend (ollama, lmstudio, llamacpp). Defaults to the configured default backend.

Example

loomcast update qwen2.5:0.5b --backend ollama

6. loomcast chat

Sends a prompt to a local LLM and displays its response in a formatted panel.

Syntax

loomcast chat MODEL PROMPT [OPTIONS]

Careful: the MODEL argument comes first, followed by the PROMPT text.

Arguments

  • MODEL (required): name of the loaded model.
  • PROMPT (required): the question or request text.

Options

  • -b, --backend TEXT: target backend (ollama, lmstudio, llamacpp).

Examples

# Chat on the default backend
loomcast chat qwen2.5:0.5b "Explain recursion in two sentences."

# Chat with a model on LM Studio
loomcast chat qwen2.5-coder-7b "Write a bash script that lists files." -b lmstudio

7. loomcast compare

Sends one prompt to several runtimes at once and shows their answers side by side, one panel per runtime.

Syntax

loomcast compare PROMPT [OPTIONS]

Arguments

  • PROMPT (required): the prompt sent to every selected runtime.

Options

  • -m, --model TEXT: model to use on every runtime. When omitted, each runtime's first installed model is auto-detected (the three runtimes do not share a model namespace).
  • -b, --backend TEXT: restrict to a specific runtime. Repeatable (-b ollama -b lmstudio). Defaults to all runtimes.

A runtime that is offline, has no usable model, or fails mid-request shows an error panel instead of aborting the whole comparison. Model output is rendered as literal text, so a reply containing markup-like characters is shown verbatim. The exit code is 1 only when no runtime produced a response.

Examples

# Same prompt to every available runtime, auto-detecting each model
loomcast compare "Explain recursion in one sentence."

# Compare two runtimes, forcing the same model name on both
loomcast compare "Write a haiku about the sea." -b ollama -b lmstudio -m llama3

8. loomcast config

Views or updates Loomcast's persistent configuration (config.toml).

Syntax

loomcast config [OPTIONS]

Options

  • --set-default TEXT: set the default backend (ollama, lmstudio, llamacpp).
  • --show / --no-show: show or hide the current configuration table (default: --show).

Examples

# Show the current configuration
loomcast config

# Set LM Studio as the default backend
loomcast config --set-default lmstudio

9. loomcast search

Searches for models in the Loomcast index catalog.

Syntax

loomcast search QUERY

Arguments

  • QUERY (required): search keyword (e.g. code, llama, vision).

Example

loomcast search code

10. loomcast recommend

Finds and recommends models suited to a specific task or tag.

Syntax

loomcast recommend TASK

Arguments

  • TASK (required): task tag or category (e.g. coding, math, translation).

Example

loomcast recommend coding

11. loomcast install-category

Batch-installs every model associated with a given category or tag.

Syntax

loomcast install-category CATEGORY [OPTIONS]

Arguments

  • CATEGORY (required): name of the tag/category to install.

Options

  • -b, --backend TEXT: target backend.
  • -c, --max-concurrent INTEGER: maximum number of parallel downloads (default: 3).
  • -s, --sequential: run the installation sequentially (max-concurrent=1).

Example

loomcast install-category coding --backend ollama --max-concurrent 2

12. loomcast verify

Verifies the cryptographic integrity (SHA-256 digest) of an installed model.

Syntax

loomcast verify MODEL [OPTIONS]

Arguments

  • MODEL (required): name of the model to verify.

Options

  • -b, --backend TEXT: backend on which to verify the model.

Example

loomcast verify qwen2.5:0.5b --backend ollama

13. loomcast doctor

Runs a full diagnostic of the Loomcast environment: TOML configuration, environment variables, local runtime availability, catalog reachability, and SHA-256 integrity of the static catalog.

Syntax

loomcast doctor

Statuses and exit codes

  • ✓ OK: the check passed.
  • ⚠ Warning: non-blocking issue (e.g. an optional runtime offline, or the dynamic catalog falling back to the static one). Exit code 0 if these are the only findings.
  • ✗ Problem: blocking issue (e.g. the default backend offline, a configuration error, or a tampered static catalog). Exit code 1.

Sample output

Loomcast System Diagnostics

✓ Configuration: loaded successfully from /home/user/.config/loomcast/config.toml
✓ Environment variables: none set, using defaults
✗ Runtime OLLAMA (default backend): unavailable. Start the service or change the default backend in the config.
⚠ Runtime LMSTUDIO: unavailable (not the default backend)
⚠ Runtime LLAMACPP: unavailable (not the default backend)
✓ Catalog: dynamic catalog reachable
✓ Static catalog integrity: SHA-256 check passed

1 problem(s) found, 2 warning(s)