What happens on the first run
What it is
The first start of loom creates a few files that decide which provider a new project uses. This page lists them, and where they go.
What it is for
- Knowing which files appear, and where.
- Knowing why a new project may point at a local server that you have not started.
How
- Open a terminal in your project folder.
- Run
loom. - Loomcoding creates the missing files below. It never overwrites a file that already exists.
Example
PS> cd <project-folder>
PS> loom
What loom creates
| File | Location | Created when | Content at creation |
|---|---|---|---|
providers.toml | <profile>/.loomcoding/ | It is missing | Three local servers and default = "llama" |
loomcoding.toml | The project folder | It is missing | The model and [provider] of the default server |
.env | The project folder | Never: you create it | Your keys, optional |
Loomcoding probes each local server with one GET <base_url>/models request, with a 1.5-second limit. It never sends a prompt during the probe.
If the system has no home folder, there is no profile and no providers.toml. Then a new loomcoding.toml uses OpenRouter, the model nvidia/nemotron-3.5-super-vl:free, and max_steps = 8.
Settings and defaults
| Name | Default | Requires restart |
|---|---|---|
default in providers.toml | llama | Yes |
model in loomcoding.toml | The model of the default server | Yes after a manual edit; no after /model |
max_steps in loomcoding.toml | 8 in the created file; 30 when the key is missing | Yes after a manual edit |
[provider] kind in loomcoding.toml | The kind of the default server, openai_compat | Yes after a manual edit; no after /settings |
[provider] base_url in loomcoding.toml | The base_url of the default server, http://127.0.0.1:8081/v1 | Yes after a manual edit; no after /settings |
Result
- A new project opens with the default local server selected, even if that server is not running.
- A server that is not running shows
downin/provider. Prompts fail until you start it or switch provider. - To change the provider, see Connect a cloud provider or Run a local model server.
- To start from a clean profile, set
LOOMCODING_HOMEto a new empty folder.