omp-docs/tools/llama-cpp/index.md

@@ -124,31 +124,8 @@ print(completion.choices[0].message.content)
 
 ---
 
-## Docker
-
-```bash
-docker run -p 8080:8080 -v /models:/models \
-  ghcr.io/ggml-org/llama.cpp:server \
-  -m /models/model.gguf --host 0.0.0.0 --port 8080
-
-# CUDA variant
-docker run -p 8080:8080 --gpus all \
-  ghcr.io/ggml-org/llama.cpp:server-cuda \
-  -m /models/model.gguf --n-gpu-layers 99
-```
-
----
-
-## Environment Variables
-
-All CLI options can be set via `LLAMA_ARG_` prefix:
-
-```bash
-LLAMA_ARG_MODEL=/models/model.gguf LLAMA_ARG_CTX_SIZE=8192 llama-server
-```
-
----
-
 ## Cross-References
 
-- See [Setup Guide](setup.md) for build instructions, GPU offloading, and deployment.
+- See [Setup Guide](setup.md) for build instructions, GPU offloading, the `LLAMA_ARG_*`
+  environment form, the Docker Compose example, and running under systemd. Deployment lives
+  there; this page is the API surface and the GGUF format.

omp-docs/tools/llama-cpp/setup.md

@@ -111,6 +111,14 @@ cmake --build build --config Release -t llama-server
 | `-lora FNAME` | LoRA adapter (repeatable) | — |
 | `-ngl, --gpu-layers N` | Layers to offload to GPU | auto |
 
+### Environment Variables
+
+Every CLI option can be set from the environment instead: prefix the upper-snake form of the
+long flag with `LLAMA_ARG_` — `LLAMA_ARG_MODEL=/models/model.gguf LLAMA_ARG_CTX_SIZE=8192
+llama-server`. This is how the container and systemd units pass configuration (see the Compose
+example below), because a unit file with one flag per line is auditable and a shell command line
+with forty is not.
+
 ### Context & Performance
 
 | Flag | Description | Default |