You can use rollama with services that offer a Ollama-compatible API (e.g., Ollama Turbo or Open WebUI). The only difference to using Ollama directly is that you have to authenticate to use these services, which you can do by adding your token to requests as shown below.
Ollama Turbo Example
To use Ollama Turbo, you have
to set the rollama_server
option to
https://ollama.com
and provide your api key as a
header:
library(rollama)
Sys.setenv(api_key = "<REDACTED>")
options(
rollama_server = "https://ollama.com",
rollama_headers = list(
Authorization = paste("Bearer", Sys.getenv("api_key"))
)
)
chat(q = "Why is the sky blue?", model = "gpt-oss:120b")
#>
#> ── Answer from gpt-oss:120b ────────────────────────────────────────────────────
#> The sky looks blue because of a phenomenon called **Rayleigh scattering**.
#>
#> ### 1. Sunlight is a mixture of colors
#> Sunlight contains all visible wavelengths (roughly 380 nm – 750 nm). When it
#> reaches Earth, it’s essentially white light made up of the colors of the
#> rainbow.
#>
#> ### 2. The atmosphere is full of tiny particles
#> The air is filled with gas molecules (nitrogen, oxygen, etc.) and tiny
#> particles that are **much smaller than the wavelength of visible light**.
#>
#> ### 3. Short wavelengths scatter more strongly
#> When light encounters particles that are much smaller than its wavelength, it
#> is scattered in all directions. The scattering efficiency follows an
#> inverse‑fourth‑power law:
#>
#> \[
#> \text{Intensity of scattered light} \propto \frac{1}{\lambda^4}
#> \]
#>
#> So a wavelength that is half as long (e.g., blue at ~450 nm) scatters about
#> 16 times more than a wavelength that is twice as long (e.g., red at ~650 nm).
#>
#> Because blue and violet light are scattered far more than the other colors, a
#> lot of that short‑wavelength light is redirected toward our eyes from all parts
#> of the sky.
#>
#> ### 4. Why we see blue rather than violet
#> - **Human vision:** Our eyes are more sensitive to blue than to violet.
#> - **Solar spectrum:** There’s slightly less violet light from the Sun to begin
#> with.
#> - **Atmospheric absorption:** A small amount of violet is absorbed by the upper
#> atmosphere.
#>
#> The combination of these factors makes the scattered light we perceive as
#> predominantly **blue**.
#>
#> ### 5. Sunrise and sunset colors
#> When the Sun is low on the horizon, its light must travel through a much
#> thicker layer of atmosphere. The short‑wavelength blue light gets scattered out
#> of the direct line of sight long before the light reaches you, leaving the
#> longer‑wavelength reds and oranges to dominate the sky’s color. That’s why
#> sunrises and sunsets appear reddish.
#>
#> ### Quick recap
#>
#> | Process | Effect on light |
#> |---------|-----------------|
#> | **Rayleigh scattering** | Scatters shorter wavelengths (blue/violet) much
#> more than longer ones |
#> | **Human eye sensitivity** | More responsive to blue than violet |
#> | **Atmospheric path length** | Determines which wavelengths reach you directly
#> (short → scattered, long → direct) |
#>
#> So the sky is blue because the atmosphere preferentially scatters the
#> shorter‑wavelength (blue) portion of sunlight toward us, while the longer
#> wavelengths pass through relatively unchanged.
Open WebUI Example
For hosting Ollama (and other providers) on a lab or organisation
server, Open WebUI is a good
option. To use the Open WebUI API with rollama, you have to set the
rollama_server
option to the URL of the hosted instance +
the path /ollama/
and provide your api key as a header. For
example, my employer hosts an instance of at
https://ai-openwebui.gesis.org/
so the server address
is:
library(rollama)
Sys.setenv(api_key = "<REDACTED>")
options(
rollama_server = "https://ai-openwebui.gesis.org/ollama/",
rollama_headers = list(
Authorization = paste("Bearer", Sys.getenv("api_key"))
)
)
chat(q = "Why is the sky blue?", model = "llama4:latest")
#>
#> ── Answer from llama4:latest ───────────────────────────────────────────────────
#> The sky appears blue because of a phenomenon called Rayleigh scattering, which
#> occurs when sunlight interacts with the Earth's atmosphere. Here's a simplified
#> explanation:
#>
#> 1. **Sunlight**: The sun emits a wide range of electromagnetic radiation,
#> including visible light, which is made up of different colors (wavelengths).
#> 2. **Atmosphere**: When sunlight enters Earth's atmosphere, it encounters tiny
#> molecules of gases like nitrogen (N2) and oxygen (O2).
#> 3. **Scattering**: These gas molecules scatter the sunlight in all directions.
#> The amount of scattering that occurs depends on the wavelength of the light.
#> 4. **Wavelength and scattering**: Shorter wavelengths (like blue and violet)
#> are scattered more than longer wavelengths (like red and orange). This is known
#> as Rayleigh scattering.
#> 5. **Blue dominance**: As a result of this scattering, the blue light is
#> dispersed throughout the atmosphere, making it visible from all directions.
#> This is why the sky typically appears blue during the daytime.
#>
#> However, there are some additional factors that can affect the color of the
#> sky, such as:
#>
#> * **Dust and water vapor**: Tiny particles in the air can scatter light in
#> different ways, changing the apparent color of the sky.
#> * **Time of day**: During sunrise and sunset, the light travels through more of
#> the atmosphere, scattering shorter wavelengths and making the sky appear more
#> red.
#> * **Atmospheric conditions**: Pollution, dust, and water vapor can alter the
#> color of the sky.
#>
#> So, to summarize, the sky appears blue because of the scattering of sunlight by
#> the tiny molecules in the Earth's atmosphere, with shorter wavelengths (like
#> blue) being scattered more than longer wavelengths.
Keeping your keys save
You should not keep your API keys in an R file like above and ideally also never enter it in the Console. This way, you never have to double check before sending an R file or your history to someone else.
However, as per Hadley Wickham
Asking each time is a hassle, so you might want to store the secret across sessions. One easy way to do that is with environment variables. Environment variables, or envvars for short, are a cross platform way of passing information to processes.
For passing envvars to R, you can list name-value pairs in a file called .Renviron in your home directory.
You can open this file to edit it with
usethis::edit_r_environ()
. The file should then contain
your keys, which should look something like this:
rollama_key=sk-b3d60321e62eb364e9c8dc12ffeec242
When you then start a new session,
Sys.getenv("rollama_key")
returns the sk-b3...
value. So the start of your script could look like this: