rllama 1.0.1-aarch64-linux-gnu → 1.0.3-aarch64-linux-gnu

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 828251b5eb99195371a26fbb54a9fa7ccce0f96928ba1cd096e72b6548617614
-  data.tar.gz: b6cc86f6b97b330260fc1c7d23b21696cc0157014bd3c1fee6cebbb874b4a269
+  metadata.gz: f19cb41b6197028ad125265d82dc01ca449cc6c48b5179433aefa938f139662b
+  data.tar.gz: 6ae54e060e85ce6abbfc8a22b8e5f6f66180685403516485757147eb3978e542
 SHA512:
-  metadata.gz: de4391b2f2fc9e59be8dd23930aa2992594f3b42664a15cc000f58a5d5b54ec8533ce1aa6e86e4af37080f612d18cb36e268f3ce49bdf7f399bf0eded3579a47
-  data.tar.gz: aade6c526d6c817b05ba82a9c29ccf4d61aa566452dcf4ebca39650447abd02ae933236071bbf08522c5a3aad9d250baafef7eb85fe3edc7b1fd7a41c5d4578a
+  metadata.gz: dd391eaaf38dbf33bb72eb561200c9893fa7645d09948c971763ef8b8a89a1d82328d527c5fa256ae7af42f70b45edff4b2121ba45b65c3b90107aaab91523d5
+  data.tar.gz: 84f9fee026a509fe6e6c149d6d20e72e00bfbdee94e50a999b7dee8ce15d7b56f1dc4015860ce5c2e3aa8e9a643dab89e2826bef1dd1a2c98d24bcf819a27a09

data/README.md

@@ -1,3 +1,5 @@
+<img width="336" height="212.0" alt="Logo" src="https://github.com/user-attachments/assets/e27442fb-22d1-44cf-ba3d-f10b24c13652" />
+
 # Rllama
 
 Ruby bindings for [llama.cpp](https://github.com/ggerganov/llama.cpp) to run open-source language models locally. Run models like GPT-OSS, Qwen 3, Gemma 3, Llama 3, and many others directly in your Ruby application code.
@@ -22,6 +24,38 @@ Or install it yourself as:
 gem install rllama
 ```
 
+## CLI Chat
+
+The `rllama` command-line utility provides an interactive chat interface for conversing with language models. After installing the gem, you can start chatting immediately:
+
+```bash
+rllama
+```
+
+When you run `rllama` without arguments, it will display:
+
+- **Downloaded models**: Any models you've already downloaded to `~/.rllama/models/`
+- **Popular models**: A curated list of popular models available for download, including:
+  - Gemma 3 1B
+  - Llama 3.2 3B
+  - Phi-4
+  - Qwen3 30B
+  - GPT-OSS
+
+Simply enter the number of the model you want to use. If you select a model that hasn't been downloaded yet, it will be automatically downloaded from Hugging Face.
+
+You can also specify a model path or URL directly:
+
+```bash
+rllama path/to/your/model.gguf
+```
+
+```bash
+rllama https://huggingface.co/microsoft/phi-4-gguf/resolve/main/phi-4-Q3_K_S.gguf
+```
+
+Once the model has loaded, you can start chatting.
+
 ## Usage
 
 ### Text Generation
@@ -162,43 +196,8 @@ By default, embedding vectors are normalized. You can disable normalization with
 ```ruby
 # Generate unnormalized embeddings
 embedding = model.embed('Sample text', normalize: false)
-
-# Use custom batch size for processing multiple texts
-embeddings = model.embed(
-  ['roses are red', 'violets are blue', 'sugar is sweet'],
-  normalize: true
-)
-```
-
-## CLI Chat Utility
-
-The `rllama` command-line utility provides an interactive chat interface for conversing with language models. After installing the gem, you can start chatting immediately:
-
-```bash
-rllama
-```
-
-When you run `rllama` without arguments, it will display:
-
-- **Downloaded models**: Any models you've already downloaded to `~/.rllama/models/`
-- **Popular models**: A curated list of popular models available for download, including:
-  - Gemma 3 1B
-  - Llama 3.2 3B
-  - Phi-4
-  - Qwen3 30B
-  - GPT-OSS
-  - And more...
-
-Simply enter the number of the model you want to use. If you select a model that hasn't been downloaded yet, it will be automatically downloaded from Hugging Face.
-
-You can also specify a model path directly:
-
-```bash
-rllama path/to/your/model.gguf
 ```
 
-Once the model loads, you can start chatting.
-
 ## Finding Models
 
 You can download GGUF format models from various sources:

data/lib/rllama/context.rb

@@ -1,10 +1,12 @@
 # frozen_string_literal: true
 
+require 'etc'
+
 module Rllama
   class Context
     attr_reader :messages, :n_ctx, :n_batch, :n_past
 
-    def initialize(model, embeddings: false, n_ctx: nil, n_batch: nil)
+    def initialize(model, embeddings: false, n_ctx: nil, n_batch: nil, n_threads: Etc.nprocessors)
       @model = model
       @n_ctx = n_ctx
       @n_batch = n_batch
@@ -15,6 +17,9 @@ module Rllama
       @ctx_params[:n_ctx] = @n_ctx if @n_ctx
       @ctx_params[:n_batch] = @n_batch if @n_batch
 
+      @ctx_params[:n_threads] = n_threads
+      @ctx_params[:n_threads_batch] = n_threads
+
       if @embeddings
         seq_cap = @model.n_seq_max
 

data/lib/rllama/cpp.rb

@@ -8,7 +8,7 @@ module Rllama
 
     LIB_NAME = 'llama'
 
-    platform =
+    PLATFORM =
       case FFI::Platform::OS
       when 'darwin'
         FFI::Platform::ARCH == 'aarch64' ? 'arm64-darwin' : 'x86_64-darwin'
@@ -28,12 +28,27 @@ module Rllama
         "lib#{LIB_NAME}.so"
       end
 
-    platform_dir = File.join(__dir__, platform)
-    platform_path = File.join(platform_dir, lib_file)
+    PLATFORM_DIR = File.join(__dir__, PLATFORM)
+
+    platform_path = File.join(PLATFORM_DIR, lib_file)
 
     lib_paths = []
+
     lib_paths << platform_path if File.exist?(platform_path)
 
+    ggml_lib_file =
+      case FFI::Platform::OS
+      when 'darwin'
+        'libggml.dylib'
+      when 'windows', 'mingw32'
+        'ggml.dll'
+      else
+        'libggml.so'
+      end
+
+    ggml_platform_path = File.join(PLATFORM_DIR, ggml_lib_file)
+    lib_paths << ggml_platform_path if File.exist?(ggml_platform_path)
+
     lib_paths +=
       case FFI::Platform::OS
       when 'darwin'
@@ -436,6 +451,8 @@ module Rllama
     attach_function :llama_backend_init, [], :void
     attach_function :llama_backend_free, [], :void
     attach_function :llama_numa_init, [:int], :void # ggml_numa_strategy
+    attach_function :ggml_backend_load_all, [], :void
+    attach_function :ggml_backend_load_all_from_path, [:string], :void
 
     # Threadpool
     attach_function :llama_attach_threadpool, %i[llama_context_p ggml_threadpool_t ggml_threadpool_t], :void
@@ -681,10 +698,14 @@ module Rllama
       llama_log_set(@log_callback, nil)
     end
 
-    llama_backend_init
-
     silence_log!
 
+    if File.directory?(PLATFORM_DIR)
+      ggml_backend_load_all_from_path(PLATFORM_DIR)
+    else
+      ggml_backend_load_all
+    end
+
     freeze
   end
 end

data/lib/rllama/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Rllama
-  VERSION = '1.0.1'
+  VERSION = '1.0.3'
 end

metadata

@@ -1,13 +1,13 @@
 --- !ruby/object:Gem::Specification
 name: rllama
 version: !ruby/object:Gem::Version
-  version: 1.0.1
+  version: 1.0.3
 platform: aarch64-linux-gnu
 authors:
 - Pete Matsyburka
 bindir: bin
 cert_chain: []
-date: 2025-10-05 00:00:00.000000000 Z
+date: 2025-10-07 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ffi