ruby_llm-providers-infomaniak 0.1.0 → 0.2.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 264a19f883d98cbfdfb7a49a0e1784d964e1289e224ae42dc46641c1e59e9ba5
-  data.tar.gz: 6cbc48e9ad718684ffb0609d10f2b6a4f88043ef6aeb22d19a99341089a2c27c
+  metadata.gz: 8a7d8b0d086efaa9c17802b1a0fad9b129627e42014334458219956e553ea49d
+  data.tar.gz: 874abaf93d029ac7c471d9c0c1a7a7c1f78ff603159b67b60bf91f213111a321
 SHA512:
-  metadata.gz: dc84db9b0ff07fb6960b3a03344a1da8667a12758bb11d4e581bca04f67273c14debf118a5006c5b27dd93b6283763a6e3dc738e01279bca9e5ba9906f96dd21
-  data.tar.gz: d1940a9651d6ea0493903893129b86189a12f97ab5181456ab4057c9da44dbe92aa62bb785fd8908d26091f65a6e53bf58baae5130e92dd043c2e652ad95a017
+  metadata.gz: 991cb79aae76ef6e1b7a8df3b78468c855d7f9f6b7bcaeaf3c86a8d386f529794b95beacb15320162f492df8df997b78d83a13390b5d51a90fde9c6d7a2c002e
+  data.tar.gz: d70228140d4b2b2eb504f71d2637817355dc18ccee36f9a6eb4d1da736c7788cf7d03e27ce8ad529c4f4c00247bf97263c2637fd277d676575237a9f1d5e7f53

data/CHANGELOG.md

@@ -1,5 +1,17 @@
 # Changelog
 
+## 0.2.0
+
+- Reranking (`RubyLLM.rerank`) with `BAAI/bge-reranker-v2-m3` and `Qwen/Qwen3-Reranker-0.6B`, on the
+  Cohere-compatible endpoint.
+- Image generation (`RubyLLM.paint`) with Flux; the image type is read from the returned bytes (JPEG).
+- Transcription (`RubyLLM.transcribe`) with Whisper, polling Infomaniak's asynchronous results
+  (`infomaniak_poll_interval`).
+- The model catalog now also lists the rerankers, Flux and Whisper.
+- Pricing: the catalog carries Infomaniak's CHF list prices, so `response.cost` works for chat,
+  embeddings and rerank (in CHF). Flux and Whisper per-minute prices are in the model metadata.
+- Author name: Andi Idogawa.
+
 ## 0.1.0
 
 - First release: `infomaniak` provider for RubyLLM 2.x on Infomaniak AI Tools' OpenAI-compatible API, with

data/LICENSE

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2026 Andreas Idogawa
+Copyright (c) 2026 Andi 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

data/README.md

@@ -17,6 +17,7 @@ chat.ask('Explain Ruby blocks in one sentence.').content
 ```
 
 - Chat, streaming, tool calling, structured output, image input, embeddings
+- Reranking, image generation (Flux) and transcription (Whisper)
 - 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
@@ -60,7 +61,8 @@ caches the answer per token.
 |---|---|---|
 | `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 |
+| `infomaniak_api_base` | `https://api.infomaniak.com/2/ai/{product_id}/openai/v1` | Override for a proxy or gateway. Rerank, image and transcription routes are derived from it |
+| `infomaniak_poll_interval` | `2` | Seconds between checks while a transcription runs (bounded by `request_timeout`) |
 
 ## Models
 
@@ -68,25 +70,30 @@ The chat and embedding models Infomaniak served on 2026-09-30, and how each beha
 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 |
+| Model | Context | Images | Thinking | CHF per 1M tokens, in / out |
+|---|---|---|---|---|
+| `moonshotai/Kimi-K2.6` (beta) | 256K | yes | on by default | 0.60 / 3.00 |
+| `Qwen/Qwen3.5-397B-A17B-FP8` (beta) | 200K | yes | on by default | 0.80 / 3.60 |
+| `Qwen/Qwen3.5-122B-A10B-FP8` | 200K | yes | on by default | 0.40 / 3.20 |
+| `google/gemma-4-31B-it` | 100K | yes | off, opt in with an effort | 0.20 / 0.40 |
+| `mistralai/Mistral-Small-4-119B-2603` | 256K | yes | off, opt in with an effort | 0.20 / 0.75 |
+| `mistralai/Ministral-3-14B-Instruct-2512` | 100K | yes | none | 0.30 / 0.40 |
+| `swiss-ai/Apertus-v1.5-70B` (beta) | 100K | yes | none | 0.70 / 2.50 |
+
+| Embedding model | Input tokens | CHF per 1M tokens |
+|---|---|---|
+| `Qwen/Qwen3-Embedding-8B` | 8192 | 0.07 |
+| `bge_multilingual_gemma2` | 8000 | 0.065 |
+| `mini_lm_l12_v2` | 128 | free |
+
+| Other models | For | CHF |
+|---|---|---|
+| `BAAI/bge-reranker-v2-m3` | `RubyLLM.rerank` | 0.01 per 1M tokens |
+| `Qwen/Qwen3-Reranker-0.6B` | `RubyLLM.rerank` | 0.009 per 1M tokens |
+| `flux` | `RubyLLM.paint`, returns JPEG | 0.30 per minute of compute |
+| `whisper` | `RubyLLM.transcribe` | 0.006 per audio minute |
 
-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.
+Photomaker, which needs reference photos on its own route, is not covered.
 
 ### The model catalog
 
@@ -175,6 +182,53 @@ chat.ask('What is on this receipt?', with: 'receipt.jpg').content
 RubyLLM.embed('Grüezi mitenand', model: 'Qwen/Qwen3-Embedding-8B', provider: :infomaniak).vectors
 ```
 
+### Reranking
+
+Sort retrieved passages by relevance before handing them to a chat model:
+
+```ruby
+rerank = RubyLLM.rerank('Which metal is the densest?', passages,
+                        model: 'BAAI/bge-reranker-v2-m3', provider: :infomaniak, top_n: 3)
+rerank.results.map { |result| [result.score, result.document] }
+```
+
+### Image generation
+
+```ruby
+image = RubyLLM.paint('A red Swiss train crossing a stone viaduct, watercolor',
+                      model: 'flux', provider: :infomaniak, size: '1024x1024') # or 1024x1792, 1792x1024
+image.save('train.jpg')
+```
+
+Prompts work best in English, and are limited to 77 tokens. Infomaniak's extra options pass through
+`provider_options:`, e.g. `provider_options: { style: 'photographic' }`.
+
+### Transcription
+
+```ruby
+RubyLLM.transcribe('meeting.m4a', model: 'whisper', provider: :infomaniak, language: 'de').text
+```
+
+Infomaniak runs transcriptions as background jobs: the gem uploads the file, then polls until the
+transcript is ready, so the call blocks for about as long as the job takes. Streaming (passing a block)
+is not possible. mp3, mp4, m4a, wav, flac, ogg, opus, aac, wma and webm are accepted.
+
+### Costs
+
+The catalog carries Infomaniak's list prices, so ruby_llm computes costs from the token counts:
+
+```ruby
+response = chat.ask('Summarise this contract.', with: 'contract.txt')
+response.cost.total   # => 0.0021 (CHF)
+chat.cost.total       # the whole conversation
+```
+
+**Costs from this provider are in CHF**, exactly as Infomaniak bills them, although ruby_llm documents its
+prices as USD. Don't add them to costs from other providers without converting. Chat, embedding and
+rerank costs are computed; Flux and Whisper bill per minute, which ruby_llm cannot compute, so their
+prices are only in `model.metadata[:price_per_minute]`. The prices date from 2026-09-30 and live in
+`lib/ruby_llm/providers/infomaniak/pricing.rb`.
+
 ## How Infomaniak differs from OpenAI
 
 The gem handles these, so the RubyLLM API behaves as usual:
@@ -189,6 +243,10 @@ The gem handles these, so the RubyLLM API behaves as usual:
 | 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 |
+| Routes | Chat and embeddings on `/2/.../openai/v1`, rerank on `/2/.../cohere/v2`, images and transcription on `/1/.../openai` | Each call goes to its route |
+| Reranking | Cohere v2 format, usage as `usage.total_tokens` | Read into `rerank.tokens.input` |
+| Images | Only base64 JPEG, no edits | The image type is read from the bytes; `with:` raises `ArgumentError` |
+| Transcription | Asynchronous: upload returns a batch id, results are polled | Polls every `infomaniak_poll_interval` seconds until done, failed or `request_timeout` |
 
 ## Development
 

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

@@ -0,0 +1,95 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    class Infomaniak < Provider
+      # Image generation (Flux) and transcription (Whisper) on Infomaniak's
+      # OpenAI-style v1 routes. Transcription is asynchronous there: the upload
+      # returns a batch id, and the transcript is polled from /results.
+      class Media < Protocol
+        include Protocols::ChatCompletions::Images
+        include Protocols::ChatCompletions::Transcription
+
+        public :render_transcription_options
+
+        IMAGE_SIGNATURES = { "\x89PNG".b => 'image/png', "\xFF\xD8\xFF".b => 'image/jpeg', 'RIFF'.b => 'image/webp' }.freeze
+        Body = Struct.new(:body)
+
+        def images_url(**)
+          @provider.product_url(1, 'openai/images/generations')
+        end
+
+        def validate_paint_inputs!(with:, mask:)
+          raise ArgumentError, 'Infomaniak generates images from a prompt only; it cannot edit images' if editing?(with, mask)
+        end
+
+        # ruby_llm assumes PNG for OpenAI-style images; read the real type from the bytes.
+        def parse_image_responses(response, model:)
+          entries = Array(response.body['data'])
+          raise Error.new('Infomaniak returned no image', response: response) if entries.empty?
+
+          entries.map do |entry|
+            Image.new(data: entry['b64_json'], mime_type: image_mime_type(entry['b64_json']),
+                      revised_prompt: entry['revised_prompt'], model: model, usage: {})
+          end
+        end
+
+        def transcription_url
+          @provider.product_url(1, 'openai/audio/transcriptions')
+        end
+
+        def stream_transcription(*)
+          raise Error, 'Infomaniak transcribes asynchronously and cannot stream; call transcribe without a block'
+        end
+
+        def parse_transcription_response(response, model:)
+          batch_id = unwrap(response.body)['batch_id']
+          raise Error.new('Infomaniak returned no transcription batch id', response: response) unless batch_id
+
+          super(Body.new(transcript(await_batch(batch_id))), model:)
+        end
+
+        private
+
+        def image_mime_type(base64)
+          head = base64.to_s[0, 24].unpack1('m')
+          IMAGE_SIGNATURES.find { |signature, _| head.start_with?(signature) }&.last || 'image/png'
+        end
+
+        def await_batch(batch_id)
+          deadline = monotonic_now + @config.request_timeout
+          loop do
+            result = unwrap(@connection.get(@provider.product_url(1, "results/#{batch_id}")).body)
+            case result['status']
+            when 'success' then return result
+            when 'failed', 'cancelled' then raise Error, "Infomaniak transcription #{result['status']} (batch #{batch_id})"
+            end
+            raise Error, "Infomaniak transcription not done after #{@config.request_timeout}s (batch #{batch_id})" if monotonic_now > deadline
+
+            sleep(@config.infomaniak_poll_interval || 2)
+          end
+        end
+
+        # The finished batch carries the output inline, or only a download URL.
+        def transcript(result)
+          data = result['data'] || @connection.get(result['url']).body
+          return data unless data.is_a?(String)
+
+          parsed = JSON.parse(data)
+          parsed.is_a?(Hash) ? parsed : data
+        rescue JSON::ParserError
+          data
+        end
+
+        # Infomaniak's own routes wrap payloads as {"result": "success", "data": {...}}.
+        def unwrap(body)
+          body.is_a?(Hash) && body.key?('result') && body['data'].is_a?(Hash) ? body['data'] : body
+        end
+
+        def monotonic_now
+          Process.clock_gettime(Process::CLOCK_MONOTONIC)
+        end
+      end
+    end
+  end
+end

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

@@ -18,11 +18,23 @@ module RubyLLM
         /ix
         EMBEDDING = /embed|bge|\be5\b|gte|mini_?lm/i
 
+        # Catalog types served outside the OpenAI endpoint, so missing from its list.
+        OTHER_ENDPOINT_TYPES = %w[reranker image stt].freeze
+        # Photomaker needs its own route with reference photos, which the gem does not cover.
+        UNSUPPORTED = /photo_?maker/i
+
         def list_models
           details = catalog_details
-          listed_models(@connection.get(models_url)).map do |entry|
-            build_model(entry, details_for(details, entry['id']))
+          listed = listed_models(@connection.get(models_url))
+          listed_ids = listed.map { |entry| entry['id'].to_s.downcase }
+          elsewhere = details.values.select do |entry|
+            name = entry['name'].to_s
+            OTHER_ENDPOINT_TYPES.include?(entry['type']) && !name.match?(UNSUPPORTED) &&
+              !listed_ids.include?(name.downcase)
           end
+
+          listed.map { |entry| build_model(entry, details_for(details, entry['id'])) } +
+            elsewhere.map { |entry| build_model({ 'id' => entry['name'] }, entry) }
         end
 
         private
@@ -63,7 +75,8 @@ module RubyLLM
             capabilities: capabilities_for(type, vision:, thinking:),
             modalities: modalities_for(type, vision:),
             reasoning_options: thinking ? [EFFORT_OPTION] : [],
-            metadata: metadata_for(entry, details)
+            pricing: Pricing.pricing_for(id),
+            metadata: metadata_for(entry, details).merge(Pricing.metadata_for(id))
           )
         end
 

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

@@ -0,0 +1,53 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    class Infomaniak < Provider
+      # Infomaniak's list prices in CHF, from
+      # https://www.infomaniak.com/en/hosting/ai-services/prices. The API does
+      # not report prices, so they are kept here and written into the catalog
+      # by `rake models`. They go in as CHF: ruby_llm calls its prices USD,
+      # but response.cost for this provider is in CHF, matching the invoice.
+      module Pricing
+        AS_OF = '2026-09-30'
+        CURRENCY = 'CHF'
+
+        # Per million tokens: [input, output]. Rerankers and embeddings bill input only.
+        PER_MILLION_TOKENS = {
+          'Qwen/Qwen3.5-122B-A10B-FP8' => [0.40, 3.20],
+          'Qwen/Qwen3.5-397B-A17B-FP8' => [0.80, 3.60],
+          'mistralai/Ministral-3-14B-Instruct-2512' => [0.30, 0.40],
+          'mistralai/Mistral-Small-4-119B-2603' => [0.20, 0.75],
+          'google/gemma-4-31B-it' => [0.20, 0.40],
+          'moonshotai/Kimi-K2.6' => [0.60, 3.00],
+          'swiss-ai/Apertus-v1.5-70B' => [0.70, 2.50],
+          'BAAI/bge-reranker-v2-m3' => [0.01],
+          'Qwen/Qwen3-Reranker-0.6B' => [0.009],
+          'bge_multilingual_gemma2' => [0.065],
+          'Qwen/Qwen3-Embedding-8B' => [0.07],
+          'mini_lm_l12_v2' => [0]
+        }.freeze
+
+        # Billed per minute (audio length for Whisper, compute time for Flux),
+        # which ruby_llm's per-token pricing cannot express. Metadata only.
+        PER_MINUTE = { 'whisper' => 0.006, 'flux' => 0.30 }.freeze
+
+        module_function
+
+        def pricing_for(id)
+          input, output = PER_MILLION_TOKENS[id]
+          return {} unless input
+
+          { text_tokens: { standard: { input_per_million: input, output_per_million: output }.compact } }
+        end
+
+        def metadata_for(id)
+          per_minute = PER_MINUTE[id]
+          return {} unless per_minute || PER_MILLION_TOKENS.key?(id)
+
+          { currency: CURRENCY, prices_as_of: AS_OF, price_per_minute: per_minute }.compact
+        end
+      end
+    end
+  end
+end

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

@@ -0,0 +1,27 @@
+# frozen_string_literal: true
+
+module RubyLLM
+  module Providers
+    class Infomaniak < Provider
+      # Reranking on Infomaniak's Cohere v2 compatible endpoint, which sits
+      # next to the OpenAI one and reports usage OpenAI-style instead of in meta.
+      class CohereRerank < Protocol
+        include Protocols::Cohere::Rerank
+
+        def rerank_url
+          @provider.product_url(2, 'cohere/v2/rerank')
+        end
+
+        def parse_rerank_response(response, model:, documents: [])
+          data = response.body
+          RubyLLM::Rerank.new(
+            results: parse_rerank_results(data, documents),
+            model: data['model'] || model,
+            raw: data,
+            input_tokens: data.dig('usage', 'total_tokens')
+          )
+        end
+      end
+    end
+  end
+end

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

@@ -3,7 +3,7 @@
 module RubyLLM
   module Providers
     class Infomaniak < Provider
-      VERSION = '0.1.0'
+      VERSION = '0.2.0'
     end
   end
 end

data/lib/ruby_llm/providers/infomaniak.rb

@@ -3,12 +3,17 @@
 require 'ruby_llm'
 require_relative 'infomaniak/version'
 require_relative 'infomaniak/chat'
+require_relative 'infomaniak/pricing'
 require_relative 'infomaniak/models'
+require_relative 'infomaniak/rerank'
+require_relative 'infomaniak/media'
 
 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.
+    # model listing under https://api.infomaniak.com/2/ai/{product_id}/openai/v1,
+    # reranking under /2/ai/{product_id}/cohere/v2, and image generation and
+    # transcription under /1/ai/{product_id}/openai.
     class Infomaniak < Provider
       API_HOST = 'https://api.infomaniak.com'
 
@@ -19,6 +24,23 @@ module RubyLLM
       end
 
       protocol :chat_completions, ChatCompletions
+      protocol :rerank, CohereRerank
+      protocol :media, Media
+
+      def protocol_for(model, operation: nil, **)
+        case operation
+        when :rerank then protocols[:rerank]
+        when :paint, :transcribe then protocols[:media]
+        else super
+        end
+      end
+
+      # The URL of +path+ under the product on API +version+ (1 or 2), derived
+      # from api_base so an override moves every route along.
+      def product_url(version, path)
+        base = api_base.delete_suffix('/').delete_suffix('/openai/v1')
+        "#{base.sub(%r{/2/ai/}, "/#{version}/ai/")}/#{path}"
+      end
 
       def api_base
         @config.infomaniak_api_base || "#{API_HOST}/2/ai/#{product_id}/openai/v1"
@@ -54,7 +76,7 @@ module RubyLLM
 
       class << self
         def configuration_options
-          %i[infomaniak_api_key infomaniak_product_id infomaniak_api_base]
+          %i[infomaniak_api_key infomaniak_product_id infomaniak_api_base infomaniak_poll_interval]
         end
 
         def configuration_requirements

data/models.json

@@ -17,7 +17,13 @@
       ]
     },
     "capabilities": [],
-    "pricing": {},
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.065
+        }
+      }
+    },
     "metadata": {
       "owned_by": "system",
       "type": "embedding",
@@ -26,7 +32,9 @@
       "status": "coming_soon",
       "beta": false,
       "coder": false,
-      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/embeddings"
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/embeddings",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30"
     }
   },
   {
@@ -47,7 +55,13 @@
       ]
     },
     "capabilities": [],
-    "pricing": {},
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0
+        }
+      }
+    },
     "metadata": {
       "owned_by": "system",
       "type": "embedding",
@@ -56,7 +70,9 @@
       "status": "coming_soon",
       "beta": false,
       "coder": false,
-      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/embeddings"
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/embeddings",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30"
     }
   },
   {
@@ -77,7 +93,13 @@
       ]
     },
     "capabilities": [],
-    "pricing": {},
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.07
+        }
+      }
+    },
     "metadata": {
       "owned_by": "system",
       "type": "embedding",
@@ -85,7 +107,9 @@
       "status": "coming_soon",
       "beta": false,
       "coder": false,
-      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/embeddings"
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/embeddings",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30"
     }
   },
   {
@@ -113,7 +137,14 @@
       "json_mode",
       "vision"
     ],
-    "pricing": {},
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.3,
+          "output_per_million": 0.4
+        }
+      }
+    },
     "metadata": {
       "owned_by": "system",
       "type": "llm",
@@ -121,7 +152,9 @@
       "status": "ready",
       "beta": false,
       "coder": false,
-      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions"
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30"
     }
   },
   {
@@ -150,7 +183,14 @@
       "vision",
       "reasoning"
     ],
-    "pricing": {},
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.4,
+          "output_per_million": 3.2
+        }
+      }
+    },
     "metadata": {
       "owned_by": "system",
       "type": "llm",
@@ -159,6 +199,8 @@
       "beta": false,
       "coder": false,
       "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30",
       "reasoning_options": [
         {
           "type": "effort",
@@ -198,7 +240,14 @@
       "vision",
       "reasoning"
     ],
-    "pricing": {},
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.2,
+          "output_per_million": 0.4
+        }
+      }
+    },
     "metadata": {
       "owned_by": "system",
       "type": "llm",
@@ -207,6 +256,8 @@
       "beta": false,
       "coder": false,
       "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30",
       "reasoning_options": [
         {
           "type": "effort",
@@ -246,7 +297,14 @@
       "vision",
       "reasoning"
     ],
-    "pricing": {},
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.6,
+          "output_per_million": 3.0
+        }
+      }
+    },
     "metadata": {
       "owned_by": "system",
       "type": "llm",
@@ -255,6 +313,8 @@
       "beta": true,
       "coder": false,
       "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30",
       "reasoning_options": [
         {
           "type": "effort",
@@ -294,7 +354,14 @@
       "vision",
       "reasoning"
     ],
-    "pricing": {},
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.2,
+          "output_per_million": 0.75
+        }
+      }
+    },
     "metadata": {
       "owned_by": "system",
       "type": "llm",
@@ -303,6 +370,8 @@
       "beta": false,
       "coder": false,
       "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30",
       "reasoning_options": [
         {
           "type": "effort",
@@ -342,7 +411,14 @@
       "vision",
       "reasoning"
     ],
-    "pricing": {},
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.8,
+          "output_per_million": 3.6
+        }
+      }
+    },
     "metadata": {
       "owned_by": "system",
       "type": "llm",
@@ -351,6 +427,8 @@
       "beta": true,
       "coder": false,
       "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30",
       "reasoning_options": [
         {
           "type": "effort",
@@ -389,7 +467,14 @@
       "json_mode",
       "vision"
     ],
-    "pricing": {},
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.7,
+          "output_per_million": 2.5
+        }
+      }
+    },
     "metadata": {
       "owned_by": "system",
       "type": "llm",
@@ -397,7 +482,145 @@
       "status": "coming_soon",
       "beta": true,
       "coder": false,
-      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions"
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/openai/v1/chat/completions",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30"
+    }
+  },
+  {
+    "id": "whisper",
+    "name": "whisper",
+    "provider": "infomaniak",
+    "family": null,
+    "created_at": null,
+    "context_window": null,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "audio"
+      ],
+      "output": [
+        "text"
+      ]
+    },
+    "capabilities": [],
+    "pricing": {},
+    "metadata": {
+      "type": "stt",
+      "description": "Whisper V3",
+      "version": "3.0",
+      "status": "ready",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/1/ai/%7Bproduct_id%7D/openai/audio/transcriptions",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30",
+      "price_per_minute": 0.006
+    }
+  },
+  {
+    "id": "flux",
+    "name": "flux",
+    "provider": "infomaniak",
+    "family": null,
+    "created_at": null,
+    "context_window": null,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text"
+      ],
+      "output": [
+        "image"
+      ]
+    },
+    "capabilities": [],
+    "pricing": {},
+    "metadata": {
+      "type": "image",
+      "description": "Flux schnell",
+      "version": "0.1",
+      "status": "ready",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/1/ai/%7Bproduct_id%7D/openai/images/generations",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30",
+      "price_per_minute": 0.3
+    }
+  },
+  {
+    "id": "BAAI/bge-reranker-v2-m3",
+    "name": "BAAI/bge-reranker-v2-m3",
+    "provider": "infomaniak",
+    "family": "baai",
+    "created_at": null,
+    "context_window": null,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text"
+      ],
+      "output": [
+        "rerank"
+      ]
+    },
+    "capabilities": [],
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.01
+        }
+      }
+    },
+    "metadata": {
+      "type": "reranker",
+      "description": "BAAI/bge-reranker-v2-m3",
+      "status": "coming_soon",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/cohere/v2/rerank",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30"
+    }
+  },
+  {
+    "id": "Qwen/Qwen3-Reranker-0.6B",
+    "name": "Qwen/Qwen3-Reranker-0.6B",
+    "provider": "infomaniak",
+    "family": "qwen",
+    "created_at": null,
+    "context_window": null,
+    "max_output_tokens": null,
+    "knowledge_cutoff": null,
+    "modalities": {
+      "input": [
+        "text"
+      ],
+      "output": [
+        "rerank"
+      ]
+    },
+    "capabilities": [],
+    "pricing": {
+      "text_tokens": {
+        "standard": {
+          "input_per_million": 0.009
+        }
+      }
+    },
+    "metadata": {
+      "type": "reranker",
+      "description": "Qwen/Qwen3-Reranker-0.6B",
+      "status": "coming_soon",
+      "beta": false,
+      "coder": false,
+      "documentation": "https://developer.infomaniak.com/docs/api/post/2/ai/%7Bproduct_id%7D/cohere/v2/rerank",
+      "currency": "CHF",
+      "prices_as_of": "2026-09-30"
     }
   }
 ]

metadata

@@ -1,10 +1,10 @@
 --- !ruby/object:Gem::Specification
 name: ruby_llm-providers-infomaniak
 version: !ruby/object:Gem::Version
-  version: 0.1.0
+  version: 0.2.0
 platform: ruby
 authors:
-- Andreas Idogawa
+- Andi Idogawa
 bindir: bin
 cert_chain: []
 date: 1980-01-02 00:00:00.000000000 Z
@@ -37,7 +37,10 @@ files:
 - README.md
 - lib/ruby_llm/providers/infomaniak.rb
 - lib/ruby_llm/providers/infomaniak/chat.rb
+- lib/ruby_llm/providers/infomaniak/media.rb
 - lib/ruby_llm/providers/infomaniak/models.rb
+- lib/ruby_llm/providers/infomaniak/pricing.rb
+- lib/ruby_llm/providers/infomaniak/rerank.rb
 - lib/ruby_llm/providers/infomaniak/version.rb
 - models.json
 homepage: https://github.com/Largo/ruby_llm-providers-infomaniak