diff --git a/pkgs/by-name/ol/ollaya/package.nix b/pkgs/by-name/ol/ollaya/package.nix
new file mode 100644
index 000000000000..15afe5bc07c6
--- /dev/null
+++ b/pkgs/by-name/ol/ollaya/package.nix
@@ -0,0 +1,61 @@
+{
+  lib,
+  rustPlatform,
+  fetchFromGitHub,
+  cacert,
+  onnxruntime,
+  llama-cpp,
+  versionCheckHook,
+  nix-update-script,
+}:
+
+rustPlatform.buildRustPackage (finalAttrs: {
+  pname = "ollaya";
+  version = "0.7.5";
+
+  src = fetchFromGitHub {
+    owner = "ollaya-dev";
+    repo = "ollaya";
+    tag = "v${finalAttrs.version}";
+    hash = "sha256-T4uY31ONSPsHdTtE6u89UCKDGIVX/cdyWTt+kEB0/cM=";
+  };
+
+  cargoHash = "sha256-IFFDMauewq3JOerPOEZy9a/HCBvYT+VbKOLXl+TPqfU=";
+
+  buildInputs = [ onnxruntime ];
+
+  __structuredAttrs = true;
+
+  env = {
+    ORT_STRATEGY = "system";
+    ORT_LIB_LOCATION = "${lib.getLib onnxruntime}/lib";
+    ORT_PREFER_DYNAMIC_LINK = "true";
+  };
+
+  # The client tests hit local mock servers, but reqwest refuses to build a client without a trust store.
+  nativeCheckInputs = [ cacert ];
+
+  # GGUF runners dlopen llama.cpp from here and refuse any version but the one pinned in
+  # scripts/llama-cpp.sh, so llama-cpp must match it.
+  postInstall = ''
+    # ort's copy-dylibs duplicates onnxruntime next to the binary.
+    rm $out/lib/libonnxruntime*
+    mkdir -p $out/lib/ollaya
+    ln -s ${lib.getLib llama-cpp}/lib $out/lib/ollaya/llama
+  '';
+
+  doInstallCheck = true;
+  nativeInstallCheckInputs = [ versionCheckHook ];
+
+  passthru.updateScript = nix-update-script { };
+
+  meta = {
+    description = "Run open decision models locally behind a TypeSafe-compatible API";
+    homepage = "https://ollaya.dev";
+    changelog = "https://github.com/ollaya-dev/ollaya/releases/tag/v${finalAttrs.version}";
+    license = lib.licenses.asl20;
+    maintainers = with lib.maintainers; [ happysalada ];
+    mainProgram = "ollaya";
+    platforms = lib.platforms.linux;
+  };
+})