ruby_llm-providers-infomaniak 0.1.0

checksums.yaml

@@ -0,0 +1,7 @@
+---
+SHA256:
+  metadata.gz: 264a19f883d98cbfdfb7a49a0e1784d964e1289e224ae42dc46641c1e59e9ba5
+  data.tar.gz: 6cbc48e9ad718684ffb0609d10f2b6a4f88043ef6aeb22d19a99341089a2c27c
+SHA512:
+  metadata.gz: dc84db9b0ff07fb6960b3a03344a1da8667a12758bb11d4e581bca04f67273c14debf118a5006c5b27dd93b6283763a6e3dc738e01279bca9e5ba9906f96dd21
+  data.tar.gz: d1940a9651d6ea0493903893129b86189a12f97ab5181456ab4057c9da44dbe92aa62bb785fd8908d26091f65a6e53bf58baae5130e92dd043c2e652ad95a017

data/CHANGELOG.md

@@ -0,0 +1,7 @@
+# Changelog
+
+## 0.1.0
+
+- First release: `infomaniak` provider for RubyLLM 2.x on Infomaniak AI Tools' OpenAI-compatible API, with
+  product id discovery, streaming, tools, structured output, thinking on/off, images, embeddings and a
+  model catalog built from the account's model list (`rake models`).

data/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Andreas Idogawa
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

data/README.md

@@ -0,0 +1,217 @@
+# ruby_llm-providers-infomaniak
+
+[![Gem Version](https://badge.fury.io/rb/ruby_llm-providers-infomaniak.svg)](https://rubygems.org/gems/ruby_llm-providers-infomaniak)
+[![tests](https://github.com/Largo/ruby_llm-providers-infomaniak/actions/workflows/tests.yml/badge.svg)](https://github.com/Largo/ruby_llm-providers-infomaniak/actions/workflows/tests.yml)
+
+**Open-weight models hosted in Switzerland, through the RubyLLM API you already know.**
+
+This gem adds [Infomaniak AI Tools](https://www.infomaniak.com/en/hosting/ai-tools) as a provider to
+[RubyLLM](https://rubyllm.com). Kimi, Qwen, Gemma, Mistral and Apertus run in Infomaniak's Swiss data
+centres; your Ruby code keeps using `RubyLLM.chat`, tools, schemas, streaming and embeddings as with any
+other provider.
+
+```ruby
+chat = RubyLLM.chat(model: 'moonshotai/Kimi-K2.6', provider: :infomaniak)
+chat.ask('Explain Ruby blocks in one sentence.').content
+# => "A Ruby block is a chunk of code you pass to a method ..."
+```
+
+- Chat, streaming, tool calling, structured output, image input, embeddings
+- Thinking on/off per request, with the model's reasoning on `response.thinking`
+- Only an API token to configure: the AI Tools product is looked up for you
+- A model catalog with context sizes and capabilities, refreshable at runtime
+- Irons out Infomaniak's differences from OpenAI's API, such as Kimi's broken JSON while thinking
+
+## Contents
+
+- [Installation](#installation)
+- [Configuration](#configuration)
+- [Models](#models)
+- [Usage](#usage)
+- [How Infomaniak differs from OpenAI](#how-infomaniak-differs-from-openai)
+- [Development](#development)
+- [Releasing](#releasing)
+
+## Installation
+
+```ruby
+# Gemfile
+gem 'ruby_llm-providers-infomaniak', require: 'ruby_llm/providers/infomaniak'
+```
+
+Requires Ruby 3.2+ and RubyLLM 2.x.
+
+## Configuration
+
+You need an Infomaniak API token with the **AI Tools** scope (Infomaniak Manager > API tokens).
+
+```ruby
+require 'ruby_llm/providers/infomaniak'
+
+RubyLLM.configure do |config|
+  config.infomaniak_api_key = ENV['INFOMANIAK_API_KEY']
+end
+```
+
+That is all. On first use the provider asks `GET /1/ai` which AI Tools product the token belongs to, and
+caches the answer per token.
+
+| Option | Default | |
+|---|---|---|
+| `infomaniak_api_key` | required | API token with the AI Tools scope |
+| `infomaniak_product_id` | looked up | The product to bill. Set it when the token reaches several products (the error lists them) or to skip the lookup |
+| `infomaniak_api_base` | `https://api.infomaniak.com/2/ai/{product_id}/openai/v1` | Override for a proxy or gateway |
+
+## Models
+
+The chat and embedding models Infomaniak served on 2026-09-30, and how each behaved when probed.
+Your product lists its own with `bundle exec rake models`, or live with
+`RubyLLM::Providers::Infomaniak.refresh_models!`.
+
+| Model | Context | Images | Thinking |
+|---|---|---|---|
+| `moonshotai/Kimi-K2.6` (beta) | 256K | yes | on by default |
+| `Qwen/Qwen3.5-397B-A17B-FP8` (beta) | 200K | yes | on by default |
+| `Qwen/Qwen3.5-122B-A10B-FP8` | 200K | yes | on by default |
+| `google/gemma-4-31B-it` | 100K | yes | off, opt in with an effort |
+| `mistralai/Mistral-Small-4-119B-2603` | 256K | yes | off, opt in with an effort |
+| `mistralai/Ministral-3-14B-Instruct-2512` | 100K | yes | none |
+| `swiss-ai/Apertus-v1.5-70B` (beta) | 100K | yes | none |
+
+| Embedding model | Input tokens |
+|---|---|
+| `Qwen/Qwen3-Embedding-8B` | 8192 |
+| `bge_multilingual_gemma2` | 8000 |
+| `mini_lm_l12_v2` | 128 |
+
+Infomaniak also lists image generation (Flux, Photomaker), transcription (Whisper) and rerankers
+(`BAAI/bge-reranker-v2-m3`, `Qwen/Qwen3-Reranker-0.6B`). They live on other endpoints this gem does not
+cover yet.
+
+### The model catalog
+
+Any model id your product serves works as given, catalog or not. The gem ships a catalog, `models.json`,
+so that ruby_llm also knows each model's context size and capabilities. That enables:
+
+- `with_thinking` and `with_thinking(false)`, which need to know how a model switches thinking
+- `RubyLLM.chat(model: 'moonshotai/Kimi-K2.6')` without `provider:`
+- `RubyLLM.models.by_provider(:infomaniak)` without a network call
+
+The catalog is a snapshot from the gem release. To know about models Infomaniak added since, load the
+live list at boot (two API calls):
+
+```ruby
+RubyLLM::Providers::Infomaniak.refresh_models!
+```
+
+`RubyLLM.models.refresh!` does not do this: it skips providers that ship their own catalog.
+
+## Usage
+
+### Chat and streaming
+
+```ruby
+chat = RubyLLM.chat(model: 'moonshotai/Kimi-K2.6', provider: :infomaniak)
+
+chat.with_instructions('Answer like a Swiss train conductor.')
+chat.ask('When does the next train to Bern leave?').content
+
+chat.ask('Tell me a story about a marmot.') { |chunk| print chunk.content }
+```
+
+### Tools
+
+```ruby
+class Weather < RubyLLM::Tool
+  description 'Current weather for a city'
+  parameter :city, description: 'City name'
+
+  def execute(city:) = WeatherService.current(city)
+end
+
+chat.with_tools(Weather).ask('Do I need an umbrella in Lausanne today?').content
+```
+
+### Structured output
+
+```ruby
+schema = {
+  type: 'object',
+  properties: { city: { type: 'string' }, population: { type: 'integer' } },
+  required: %w[city population],
+  additionalProperties: false
+}
+
+chat.with_schema(schema).ask('Largest city in Switzerland?').parsed
+# => {"city" => "Zurich", "population" => 443000}
+```
+
+### Thinking
+
+Kimi and Qwen think before they answer, unless told not to. The reasoning comes back separately from the
+answer:
+
+```ruby
+response = chat.ask('Is 1001 prime?')
+response.thinking&.text   # the model's reasoning
+response.content          # the answer
+
+chat.with_thinking(effort: :none).ask('Quick: 17 * 23?')   # thinking off: faster, cheaper
+RubyLLM.chat(model: 'google/gemma-4-31B-it', provider: :infomaniak)
+       .with_thinking(effort: :high).ask('Plan a 3-day Ticino trip.')   # thinking on
+```
+
+With the model in the catalog, `with_thinking` and `with_thinking(false)` work too.
+
+### Images
+
+```ruby
+chat.ask('What is on this receipt?', with: 'receipt.jpg').content
+```
+
+### Embeddings
+
+```ruby
+RubyLLM.embed('Grüezi mitenand', model: 'Qwen/Qwen3-Embedding-8B', provider: :infomaniak).vectors
+```
+
+## How Infomaniak differs from OpenAI
+
+The gem handles these, so the RubyLLM API behaves as usual:
+
+| | What Infomaniak does | What the gem does |
+|---|---|---|
+| Thinking | `reasoning_effort` is an on/off switch: `none` or `low`/`medium`/`high` | `:none` turns thinking off; `:minimal` is sent as `low`, `:xhigh` and `:max` as `high` |
+| Kimi + schema | While thinking, Kimi answers `{{ ... }` instead of JSON | Kimi schema requests go out with thinking off, unless you ask for thinking, which logs a warning |
+| System prompt | `system` role, not OpenAI's `developer` | Sends `system` |
+| Output limit | `max_completion_tokens` | `with_max_output_tokens` maps to it |
+| Attachments | Images inline; no audio, no PDF | Audio and PDFs raise `UnsupportedAttachmentError` before sending; text files are inlined |
+| End user | `user` field | `with_end_user('id')` is sent as `user` |
+| Prompt caching | Repeated prompt prefixes answer faster, but no cached-token counts come back and `prompt_cache_key` has no visible effect | `with_caching` options are not sent; keep long shared context (instructions, documents) at the start of the conversation to benefit |
+| Errors | `{"error": {"code", "description"}}` on its own endpoints | The description ends up in the `RubyLLM::Error` message |
+
+## Development
+
+```sh
+bundle install
+cp test/.env.example test/.env    # put your API token in test/.env (gitignored)
+```
+
+| Command | |
+|---|---|
+| `bundle exec ruby bin/console` | IRB with `chat`, `models` and a `Weather` tool, configured from `test/.env` |
+| `bundle exec rake test` | Offline tests, all HTTP stubbed with WebMock. What CI runs |
+| `bundle exec rake live` | The same features against the real API, with your token |
+| `bundle exec rake models` | Rebuilds `models.json` from your product's model list |
+
+`test/.env` also takes `INFOMANIAK_MODEL` (the console and live-test model, default Kimi-K2.6),
+`INFOMANIAK_EMBEDDING_MODEL` (adds embeddings to `rake live`) and `RUBYLLM_DEBUG=1` (logs every request).
+
+## Releasing
+
+Tags publish to RubyGems through trusted publishing, without a stored API key. See
+[RELEASING.md](RELEASING.md).
+
+## License
+
+MIT. Not affiliated with Infomaniak.

data/lib/ruby_llm/providers/infomaniak/chat.rb

@@ -0,0 +1,75 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    class Infomaniak < Provider
+      # How Infomaniak's Chat Completions differ from OpenAI's: a plain system
+      # role, max_completion_tokens, reasoning_effort as an on/off switch, no
+      # OpenAI prompt-cache fields, and images as the only binary attachments.
+      module Chat
+        EFFORTS = %w[none low medium high].freeze
+
+        # Models whose schema-constrained output breaks while they think: Kimi
+        # answers "{{ ... }" instead of JSON (seen 2026-09-30).
+        SCHEMA_WITHOUT_THINKING = /kimi/i
+
+        module_function
+
+        def render_payload(messages, tools:, temperature:, model:, stream: false, max_output_tokens: nil,
+                           schema: nil, thinking: nil, citations: false, caching: nil, tool_prefs: nil)
+          payload = super
+          schema_without_thinking(payload, model) if schema && model.id.match?(SCHEMA_WITHOUT_THINKING)
+          payload
+        end
+
+        # Thinking goes off for structured output unless it was asked for.
+        def schema_without_thinking(payload, model)
+          effort = payload[:reasoning_effort]
+          return payload[:reasoning_effort] = 'none' if effort.nil?
+          return if effort == 'none'
+
+          RubyLLM.logger.warn(
+            "#{model.id} returns malformed JSON for a schema while thinking; use with_thinking(effort: :none)"
+          )
+        end
+
+        def format_role(role)
+          role.to_s
+        end
+
+        def max_output_tokens_field(_model)
+          :max_completion_tokens
+        end
+
+        # Infomaniak reads reasoning_effort as a switch: "none" turns thinking
+        # off, any other value turns it on. Efforts outside its four values are
+        # clamped to the nearest one.
+        def resolve_effort(thinking)
+          return 'none' if thinking.respond_to?(:disabled?) && thinking.disabled?
+
+          effort = super
+          return effort if effort.nil? || EFFORTS.include?(effort)
+
+          effort == 'minimal' ? 'low' : 'high'
+        end
+
+        def openai_prompt_caching?
+          false
+        end
+
+        def apply_end_user(payload, identifier)
+          payload.merge(user: identifier)
+        end
+
+        def format_content(content, attachments = [])
+          Protocols::ChatCompletions::Media.format_content(
+            content,
+            attachments,
+            document_attachments: :none,
+            audio_attachments: false
+          )
+        end
+      end
+    end
+  end
+end

data/lib/ruby_llm/providers/infomaniak/models.rb

@@ -0,0 +1,115 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    class Infomaniak < Provider
+      # Model listing. The OpenAI-style list names the ids this product can
+      # call; GET /1/ai/models adds the type, context size and beta flag.
+      # Neither reports vision or thinking, so those come from the model name,
+      # following what the served models did when probed (2026-09-30).
+      module Models
+        EFFORT_OPTION = { type: 'effort', values: Chat::EFFORTS }.freeze
+
+        # No reasoning output even with effort: :high. Mistral Small 4 does think.
+        NO_THINKING = /apertus|ministral|mistral-?3\b|mistral-small-3|llama|granite/i
+        VISION = /
+          \bvl\b|-vl-|vision|pixtral|qwen3\.5|kimi-k2\.[5-9]|gemma-?[34]|
+          ministral-3|mistral-small-4|apertus-v1\.5|llama-4
+        /ix
+        EMBEDDING = /embed|bge|\be5\b|gte|mini_?lm/i
+
+        def list_models
+          details = catalog_details
+          listed_models(@connection.get(models_url)).map do |entry|
+            build_model(entry, details_for(details, entry['id']))
+          end
+        end
+
+        private
+
+        def listed_models(response)
+          data = response.body['data']
+          data.is_a?(Hash) ? [data] : Array(data)
+        end
+
+        # The catalog names a model either by its full id or without the
+        # organisation prefix.
+        def details_for(details, id)
+          key = id.to_s.downcase
+          details[key] || details[key.split('/').last] || {}
+        end
+
+        def catalog_details
+          data = Array(@provider.account_connection.get('1/ai/models').body['data'])
+          data.to_h { |entry| [entry['name'].to_s.downcase, entry] }
+        rescue RubyLLM::Error, Faraday::Error => e
+          RubyLLM.logger.warn("Infomaniak model details unavailable (#{e.message}); listing ids only")
+          {}
+        end
+
+        def build_model(entry, details)
+          id = entry['id']
+          type = model_type(id, details['type'])
+          vision = type == :chat && (id.match?(VISION) || details['description'].to_s.match?(/vision|image/i))
+          thinking = type == :chat && !id.match?(NO_THINKING)
+
+          Model.new(
+            id: id,
+            name: id,
+            provider: @provider.slug,
+            family: id.include?('/') ? id.split('/').first.downcase : nil,
+            created_at: entry['created'] ? Time.at(entry['created']) : nil,
+            context_window: details['max_token_input'],
+            capabilities: capabilities_for(type, vision:, thinking:),
+            modalities: modalities_for(type, vision:),
+            reasoning_options: thinking ? [EFFORT_OPTION] : [],
+            metadata: metadata_for(entry, details)
+          )
+        end
+
+        def model_type(id, type)
+          case type.to_s.downcase
+          when /embed/ then :embedding
+          when /rerank/ then :rerank
+          when /stt|speech|audio|whisper/ then :transcription
+          when /image/ then :image
+          when '' then id.match?(EMBEDDING) ? :embedding : :chat
+          else :chat
+          end
+        end
+
+        def capabilities_for(type, vision:, thinking:)
+          return [] unless type == :chat
+
+          capabilities = %w[streaming function_calling structured_output json_mode]
+          capabilities << 'vision' if vision
+          capabilities << 'reasoning' if thinking
+          capabilities
+        end
+
+        def modalities_for(type, vision:)
+          case type
+          when :embedding then { input: %w[text], output: %w[embeddings] }
+          when :rerank then { input: %w[text], output: %w[rerank] }
+          when :transcription then { input: %w[audio], output: %w[text] }
+          when :image then { input: %w[text], output: %w[image] }
+          else { input: vision ? %w[text image] : %w[text], output: %w[text] }
+          end
+        end
+
+        def metadata_for(entry, details)
+          {
+            owned_by: entry['owned_by'],
+            type: details['type'],
+            description: details['description'],
+            version: details['version'],
+            status: details['info_status'],
+            beta: details.dig('meta', 'is_beta'),
+            coder: details.dig('meta', 'is_coder'),
+            documentation: details['documentation_link']
+          }.compact
+        end
+      end
+    end
+  end
+end

data/lib/ruby_llm/providers/infomaniak/version.rb

@@ -0,0 +1,9 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    class Infomaniak < Provider
+      VERSION = '0.1.0'
+    end
+  end
+end

data/lib/ruby_llm/providers/infomaniak.rb

@@ -0,0 +1,120 @@
+# frozen_string_literal: true
+
+require 'ruby_llm'
+require_relative 'infomaniak/version'
+require_relative 'infomaniak/chat'
+require_relative 'infomaniak/models'
+
+module RubyLLM
+  module Providers
+    # Infomaniak AI Tools: OpenAI-compatible chat completions, embeddings and
+    # model listing under https://api.infomaniak.com/2/ai/{product_id}/openai/v1.
+    class Infomaniak < Provider
+      API_HOST = 'https://api.infomaniak.com'
+
+      # Infomaniak's dialect of the Chat Completions API.
+      class ChatCompletions < Protocols::ChatCompletions
+        include Infomaniak::Chat
+        include Infomaniak::Models
+      end
+
+      protocol :chat_completions, ChatCompletions
+
+      def api_base
+        @config.infomaniak_api_base || "#{API_HOST}/2/ai/#{product_id}/openai/v1"
+      end
+
+      def headers
+        { 'Authorization' => "Bearer #{@config.infomaniak_api_key}" }
+      end
+
+      # The AI Tools product requests are billed to: infomaniak_product_id, or
+      # the token's only product, looked up once per token with GET /1/ai.
+      def product_id
+        configured = @config.infomaniak_product_id.to_s.strip
+        return configured unless configured.empty?
+
+        self.class.product_ids[@config.infomaniak_api_key] ||= discover_product_id
+      end
+
+      # The account-level API, which serves product and model metadata
+      # outside the OpenAI base path.
+      def account_connection # :nodoc:
+        @account_connection ||= Transport::Connection.new(self, @config, api_base: API_HOST)
+      end
+
+      # Infomaniak's own endpoints answer {"result":"error","error":{"code":..,"description":..}}.
+      def parse_error(response)
+        body = parse_error_body(response)
+        error = body['error'] if body.is_a?(Hash)
+        return super unless error.is_a?(Hash) && error['description']
+
+        [error['description'], error['code']].compact.uniq.join(' - ')
+      end
+
+      class << self
+        def configuration_options
+          %i[infomaniak_api_key infomaniak_product_id infomaniak_api_base]
+        end
+
+        def configuration_requirements
+          %i[infomaniak_api_key]
+        end
+
+        # Returns +true+: any model id the product serves is accepted as given.
+        # `rake models` refreshes the bundled catalog with capabilities.
+        def assume_models_exist?
+          true
+        end
+
+        # Loads the live model list into RubyLLM.models in place of the
+        # catalog bundled with the gem, so models Infomaniak added since the
+        # release are known with their capabilities. Returns those models.
+        #
+        #   RubyLLM::Providers::Infomaniak.refresh_models!
+        #
+        def refresh_models!(config = RubyLLM.config)
+          live = new(config).list_models
+          registry = RubyLLM.models
+          others = registry.all_including_unlisted.reject { |model| model.provider == slug }
+          # RubyLLM has no public way to replace one provider's entries, and
+          # RubyLLM.models.refresh! skips providers that ship a catalog.
+          registry.instance_variable_set(:@models, others + live)
+          live
+        end
+
+        def product_ids # :nodoc:
+          @product_ids ||= {}
+        end
+      end
+
+      private
+
+      def discover_product_id
+        products = fetch_products
+        return products.first['product_id'].to_s if products.one?
+
+        raise ConfigurationError, product_error(products)
+      end
+
+      def fetch_products
+        Array(account_connection.get('1/ai').body['data'])
+      rescue RubyLLM::Error, Faraday::Error => e
+        raise ConfigurationError,
+              "Could not look up the Infomaniak AI Tools product (#{e.message}). " \
+              'Set config.infomaniak_product_id, or check that the token has the AI Tools scope.'
+      end
+
+      def product_error(products)
+        return 'No Infomaniak AI Tools product is visible to this token. Set config.infomaniak_product_id.' if products.empty?
+
+        listed = products.map { |p| "#{p['product_id']} (#{p['product_name']}, #{p['account_name']})" }
+        "This token reaches #{products.size} AI Tools products: #{listed.join(', ')}. " \
+          'Set config.infomaniak_product_id to the one to use.'
+      end
+    end
+  end
+end
+
+RubyLLM::Provider.register :infomaniak, RubyLLM::Providers::Infomaniak,
+                           models: File.expand_path('../../../models.json', __dir__)

data/models.json

@@ -0,0 +1,403 @@
+[
+  {
+    "id": "bge_multilingual_gemma2",
+    "name": "bge_multilingual_gemma2",
+    "provider": "infomaniak",
+    "family": null,
+    "created_at": "2024-12-03 23:00:00 UTC",
+    "context_window": 8000,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text"
+      ],
+      "output": [
+        "embeddings"
+      ]
+    },
+    "capabilities": [],
+    "pricing": {},
+    "metadata": {
+      "owned_by": "system",
+      "type": "embedding",
+      "description": "Bge Multilingual Gemma2",
+      "version": "1.0",
+      "status": "coming_soon",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/embeddings"
+    }
+  },
+  {
+    "id": "mini_lm_l12_v2",
+    "name": "mini_lm_l12_v2",
+    "provider": "infomaniak",
+    "family": null,
+    "created_at": "2024-12-03 23:00:00 UTC",
+    "context_window": 128,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text"
+      ],
+      "output": [
+        "embeddings"
+      ]
+    },
+    "capabilities": [],
+    "pricing": {},
+    "metadata": {
+      "owned_by": "system",
+      "type": "embedding",
+      "description": "All MiniLM L12 v2",
+      "version": "2.0",
+      "status": "coming_soon",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/embeddings"
+    }
+  },
+  {
+    "id": "Qwen/Qwen3-Embedding-8B",
+    "name": "Qwen/Qwen3-Embedding-8B",
+    "provider": "infomaniak",
+    "family": "qwen",
+    "created_at": "2025-11-23 23:00:00 UTC",
+    "context_window": 8192,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text"
+      ],
+      "output": [
+        "embeddings"
+      ]
+    },
+    "capabilities": [],
+    "pricing": {},
+    "metadata": {
+      "owned_by": "system",
+      "type": "embedding",
+      "description": "Qwen/Qwen3-Embedding-8B",
+      "status": "coming_soon",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/embeddings"
+    }
+  },
+  {
+    "id": "mistralai/Ministral-3-14B-Instruct-2512",
+    "name": "mistralai/Ministral-3-14B-Instruct-2512",
+    "provider": "infomaniak",
+    "family": "mistralai",
+    "created_at": "2026-02-01 23:00:00 UTC",
+    "context_window": 100000,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text",
+        "image"
+      ],
+      "output": [
+        "text"
+      ]
+    },
+    "capabilities": [
+      "streaming",
+      "function_calling",
+      "structured_output",
+      "json_mode",
+      "vision"
+    ],
+    "pricing": {},
+    "metadata": {
+      "owned_by": "system",
+      "type": "llm",
+      "description": "mistralai/Ministral-3-14B-Instruct-2512",
+      "status": "ready",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions"
+    }
+  },
+  {
+    "id": "Qwen/Qwen3.5-122B-A10B-FP8",
+    "name": "Qwen/Qwen3.5-122B-A10B-FP8",
+    "provider": "infomaniak",
+    "family": "qwen",
+    "created_at": null,
+    "context_window": 200000,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text",
+        "image"
+      ],
+      "output": [
+        "text"
+      ]
+    },
+    "capabilities": [
+      "streaming",
+      "function_calling",
+      "structured_output",
+      "json_mode",
+      "vision",
+      "reasoning"
+    ],
+    "pricing": {},
+    "metadata": {
+      "owned_by": "system",
+      "type": "llm",
+      "description": "Qwen/Qwen3.5-122B-A10B-FP8",
+      "status": "coming_soon",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "reasoning_options": [
+        {
+          "type": "effort",
+          "values": [
+            "none",
+            "low",
+            "medium",
+            "high"
+          ]
+        }
+      ]
+    }
+  },
+  {
+    "id": "google/gemma-4-31B-it",
+    "name": "google/gemma-4-31B-it",
+    "provider": "infomaniak",
+    "family": "google",
+    "created_at": null,
+    "context_window": 100000,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text",
+        "image"
+      ],
+      "output": [
+        "text"
+      ]
+    },
+    "capabilities": [
+      "streaming",
+      "function_calling",
+      "structured_output",
+      "json_mode",
+      "vision",
+      "reasoning"
+    ],
+    "pricing": {},
+    "metadata": {
+      "owned_by": "system",
+      "type": "llm",
+      "description": "google/gemma-4-31B-it",
+      "status": "coming_soon",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "reasoning_options": [
+        {
+          "type": "effort",
+          "values": [
+            "none",
+            "low",
+            "medium",
+            "high"
+          ]
+        }
+      ]
+    }
+  },
+  {
+    "id": "moonshotai/Kimi-K2.6",
+    "name": "moonshotai/Kimi-K2.6",
+    "provider": "infomaniak",
+    "family": "moonshotai",
+    "created_at": null,
+    "context_window": 256000,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text",
+        "image"
+      ],
+      "output": [
+        "text"
+      ]
+    },
+    "capabilities": [
+      "streaming",
+      "function_calling",
+      "structured_output",
+      "json_mode",
+      "vision",
+      "reasoning"
+    ],
+    "pricing": {},
+    "metadata": {
+      "owned_by": "system",
+      "type": "llm",
+      "description": "moonshotai/Kimi-K2.6",
+      "status": "coming_soon",
+      "beta": true,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "reasoning_options": [
+        {
+          "type": "effort",
+          "values": [
+            "none",
+            "low",
+            "medium",
+            "high"
+          ]
+        }
+      ]
+    }
+  },
+  {
+    "id": "mistralai/Mistral-Small-4-119B-2603",
+    "name": "mistralai/Mistral-Small-4-119B-2603",
+    "provider": "infomaniak",
+    "family": "mistralai",
+    "created_at": null,
+    "context_window": 256000,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text",
+        "image"
+      ],
+      "output": [
+        "text"
+      ]
+    },
+    "capabilities": [
+      "streaming",
+      "function_calling",
+      "structured_output",
+      "json_mode",
+      "vision",
+      "reasoning"
+    ],
+    "pricing": {},
+    "metadata": {
+      "owned_by": "system",
+      "type": "llm",
+      "description": "mistralai/Mistral-Small-4-119B-2603",
+      "status": "coming_soon",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "reasoning_options": [
+        {
+          "type": "effort",
+          "values": [
+            "none",
+            "low",
+            "medium",
+            "high"
+          ]
+        }
+      ]
+    }
+  },
+  {
+    "id": "Qwen/Qwen3.5-397B-A17B-FP8",
+    "name": "Qwen/Qwen3.5-397B-A17B-FP8",
+    "provider": "infomaniak",
+    "family": "qwen",
+    "created_at": null,
+    "context_window": 200000,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text",
+        "image"
+      ],
+      "output": [
+        "text"
+      ]
+    },
+    "capabilities": [
+      "streaming",
+      "function_calling",
+      "structured_output",
+      "json_mode",
+      "vision",
+      "reasoning"
+    ],
+    "pricing": {},
+    "metadata": {
+      "owned_by": "system",
+      "type": "llm",
+      "description": "Qwen/Qwen3.5-397B-A17B-FP8",
+      "status": "coming_soon",
+      "beta": true,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "reasoning_options": [
+        {
+          "type": "effort",
+          "values": [
+            "none",
+            "low",
+            "medium",
+            "high"
+          ]
+        }
+      ]
+    }
+  },
+  {
+    "id": "swiss-ai/Apertus-v1.5-70B",
+    "name": "swiss-ai/Apertus-v1.5-70B",
+    "provider": "infomaniak",
+    "family": "swiss-ai",
+    "created_at": null,
+    "context_window": 100000,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text",
+        "image"
+      ],
+      "output": [
+        "text"
+      ]
+    },
+    "capabilities": [
+      "streaming",
+      "function_calling",
+      "structured_output",
+      "json_mode",
+      "vision"
+    ],
+    "pricing": {},
+    "metadata": {
+      "owned_by": "system",
+      "type": "llm",
+      "description": "swiss-ai/Apertus-v1.5-70B",
+      "status": "coming_soon",
+      "beta": true,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions"
+    }
+  }
+]

metadata

@@ -0,0 +1,68 @@
+--- !ruby/object:Gem::Specification
+name: ruby_llm-providers-infomaniak
+version: !ruby/object:Gem::Version
+  version: 0.1.0
+platform: ruby
+authors:
+- Andreas Idogawa
+bindir: bin
+cert_chain: []
+date: 1980-01-02 00:00:00.000000000 Z
+dependencies:
+- !ruby/object:Gem::Dependency
+  name: ruby_llm
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '2.0'
+  type: :runtime
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '2.0'
+description: 'Use the models hosted by Infomaniak AI Tools in Switzerland (Kimi, Qwen,
+  Apertus, Mistral, ...) through RubyLLM: chat, streaming, tools, structured output,
+  thinking, images and embeddings.'
+email:
+- web@idogawa.com
+executables: []
+extensions: []
+extra_rdoc_files: []
+files:
+- CHANGELOG.md
+- LICENSE
+- README.md
+- lib/ruby_llm/providers/infomaniak.rb
+- lib/ruby_llm/providers/infomaniak/chat.rb
+- lib/ruby_llm/providers/infomaniak/models.rb
+- lib/ruby_llm/providers/infomaniak/version.rb
+- models.json
+homepage: https://github.com/Largo/ruby_llm-providers-infomaniak
+licenses:
+- MIT
+metadata:
+  source_code_uri: https://github.com/Largo/ruby_llm-providers-infomaniak
+  changelog_uri: https://github.com/Largo/ruby_llm-providers-infomaniak/blob/main/CHANGELOG.md
+  bug_tracker_uri: https://github.com/Largo/ruby_llm-providers-infomaniak/issues
+  rubygems_mfa_required: 'true'
+rdoc_options: []
+require_paths:
+- lib
+required_ruby_version: !ruby/object:Gem::Requirement
+  requirements:
+  - - ">="
+    - !ruby/object:Gem::Version
+      version: '3.2'
+required_rubygems_version: !ruby/object:Gem::Requirement
+  requirements:
+  - - ">="
+    - !ruby/object:Gem::Version
+      version: '0'
+requirements: []
+rubygems_version: 4.0.20
+specification_version: 4
+summary: RubyLLM provider for Infomaniak AI Tools
+test_files: []