feature/Bump-pipelex-drop-gateway · L-261009-405281

.env.example

@@ -1,11 +1,14 @@
-# [OPTIONAL] Free Pipelex Inference API key - Get yours on Discord: https://go.pipelex.com/discord
-# No credit card required, limited time offer
-PIPELEX_GATEWAY_API_KEY=
+# cocode runs its AI workflows on your machine, with your own provider keys.
+#
+# It ships with the OpenAI backend switched on, and the model deck's default tiers are OpenAI models,
+# so OPENAI_API_KEY is all you need to start. A key on its own is not enough for any other provider:
+# its backend also has to be switched on, and the model deck pointed at its models. The README says
+# how, and so does https://docs.pipelex.com/latest/get-started/configure-ai-providers/
 
-# OpenAI: to use models like GPT-4o and GPT-5
+# OpenAI: the provider cocode ships configured for
 OPENAI_API_KEY=
 
-# Azure: to use models like GPT-4o and GPT-5 from OpenAI but on Azure cloud
+# Azure: to use OpenAI models on Azure cloud
 AZURE_API_BASE=
 AZURE_API_VERSION=
 AZURE_API_KEY=
@@ -15,27 +18,51 @@ AWS_ACCESS_KEY_ID=
 AWS_SECRET_ACCESS_KEY=
 AWS_REGION=
 
-# To use Claude models directly from their creator, Anthropic 
+# To use Claude models directly from their creator, Anthropic
 ANTHROPIC_API_KEY=
 
 # To use Mistral models
 MISTRAL_API_KEY=
 
-# To generate images from fal.ai, the service of Forest Labs
+# OpenRouter: one key for models from many providers at once
+OPENROUTER_API_KEY=
+
+# To generate images from fal.ai
 FAL_API_KEY=
 
 # Google AI Studio - Use GOOGLE_API_KEY for direct API access (simpler, rate-limited)
 GOOGLE_API_KEY=
 
 # Google Cloud Platform (GCP) - Use these for production Vertex AI access (requires GCP project setup)
-# We recomment to use GOOGLE_API_KEY OR GCP credentials, not both
+# We recommend using GOOGLE_API_KEY OR GCP credentials, not both
 GCP_PROJECT_ID=
 GCP_LOCATION=
 GCP_CREDENTIALS_FILE_PATH=gcp_credentials.json
 
 # To use Grok from x.ai
 XAI_API_KEY=
-XAI_API_ENDPOINT=https://api.x.ai/v1/
 
 # To get access to many models with one key from blackbox.ai
 BLACKBOX_API_KEY=
+
+# To use open models served by Groq
+GROQ_API_KEY=
+
+# To generate images with models on the Hugging Face Inference API
+HF_TOKEN=
+
+# Portkey: an AI gateway reaching many providers with one key
+PORTKEY_API_KEY=
+
+# To use models hosted on Scaleway
+SCALEWAY_ENDPOINT=
+SCALEWAY_API_KEY=
+
+# To use MiniMax models
+MINIMAX_API_KEY=
+
+# Linkup: the models behind web search and web page extraction
+LINKUP_API_KEY=
+
+# TypeSafe: the models behind judgment steps
+TYPESAFE_API_KEY=

.github/workflows/tests-check.yml

@@ -21,9 +21,8 @@ jobs:
     env:
       VIRTUAL_ENV: ${{ github.workspace }}/.venv
       ENV: dev
-      # Our CI never sends usage statistics: the off switch every Pipelex tool honours. Safe on the pinned pipelex, which
-      # refuses DO_NOT_TRACK only when it boots with inference and the Pipelex Gateway enabled, because `make gha-tests`
-      # always passes --disable-inference.
+      # Our CI never sends usage statistics: the off switch every Pipelex tool honours. The pinned pipelex sends none and
+      # reads DO_NOT_TRACK only to turn its own telemetry off, which it honours at any boot.
       DO_NOT_TRACK: "1"
     steps:
       - uses: actions/checkout@v4

.pipelex/inference/backends.toml

@@ -5,16 +5,23 @@
 # This file configures the inference backends available to Pipelex.
 # Each backend connects to a different AI service provider (OpenAI, Anthropic, Google, etc.).
 #
-# Documentation: https://docs.pipelex.com
+# cocode ships with `openai` switched on, because the model deck's default tiers are OpenAI models:
+# an OPENAI_API_KEY in your .env file is all a run needs. A run refuses to start while an enabled
+# backend's key is missing, so switch a backend on only when you hold its key. The cocode package
+# carries this directory and runs on it from any directory without an inference configuration of
+# its own.
+#
+# To run on another provider without editing this tracked file, write the change in a git-ignored
+# backends_override.toml beside it (for instance `[openai] enabled = false` and
+# `[anthropic] enabled = true`), then point the deck's aliases and the presets that name a model at
+# that provider's models in a deck file of your own, such as deck/x_custom_provider_deck.toml. The
+# README gives both files in full.
+#
+# Documentation: https://docs.pipelex.com/latest/get-started/configure-ai-providers/
 # Support: https://go.pipelex.com/discord
 #
 ####################################################################################################
 
-[pipelex_gateway]
-display_name = "⭐ Pipelex Gateway"
-enabled = true                         # Enable after accepting terms via `pipelex init config`
-api_key = "${PIPELEX_GATEWAY_API_KEY}"
-
 [anthropic]
 enabled = false
 api_key = "${ANTHROPIC_API_KEY}"
@@ -58,6 +65,17 @@ display_name = "Hugging Face"
 enabled = false
 api_key = "${HF_TOKEN}"
 
+[linkup]
+display_name = "Linkup"
+enabled = false
+api_key = "${LINKUP_API_KEY}"
+
+[minimax]
+display_name = "MiniMax"
+enabled = false
+endpoint = "https://api.minimax.io/anthropic"
+api_key = "${MINIMAX_API_KEY}"
+
 [mistral]
 display_name = "Mistral AI"
 enabled = false
@@ -69,9 +87,15 @@ endpoint = "http://localhost:11434/v1"
 
 [openai]
 display_name = "OpenAI"
-enabled = false
+enabled = true
 api_key = "${OPENAI_API_KEY}"
 
+[openrouter]
+display_name = "OpenRouter"
+enabled = false
+endpoint = "https://openrouter.ai/api/v1"
+api_key = "${OPENROUTER_API_KEY}"
+
 [portkey]
 display_name = "Portkey"
 enabled = false
@@ -84,6 +108,11 @@ enabled = false
 endpoint = "${SCALEWAY_ENDPOINT}"
 api_key = "${SCALEWAY_API_KEY}"
 
+[typesafe]
+display_name = "TypeSafe"
+enabled = false
+api_key = "${TYPESAFE_API_KEY}"
+
 [vertexai]
 display_name = "Google Vertex AI"
 enabled = false                                            # This is the only one we disable because setting it up requires internet access just to get credentials so it fails in CI sandboxes
@@ -99,10 +128,3 @@ api_key = "${XAI_API_KEY}"
 
 [internal] # software-only backend, runs internally, without AI
 enabled = true
-
-# Deprecated
-[pipelex_inference]
-display_name = "🛑 Legacy Pipelex Inference"
-enabled = false
-endpoint = "https://inference.pipelex.com/v1"
-api_key = "${PIPELEX_INFERENCE_API_KEY}"

.pipelex/inference/backends/.kit_manifest.json

@@ -0,0 +1,6 @@
+{
+  "files": {
+    "internal.toml": "d99ddf303c4da154d01d88d7f2839c07dc8dd6a98c0446290a75a0f3c4eef0d8"
+  },
+  "kit_version": "0.83.0"
+}

.pipelex/inference/backends/anthropic.toml

@@ -25,6 +25,7 @@ model_type = "llm"
 sdk = "anthropic"
 structure_method = "instructor/anthropic_tools"
 thinking_mode = "manual"
+valued_constraints = { min_thinking_budget = 1024 } # Anthropic refuses a smaller budget_tokens
 
 ################################################################################
 # LANGUAGE MODELS
@@ -100,6 +101,15 @@ outputs = ["text", "structured"]
 max_prompt_images = 100
 costs = { input = 5.0, output = 25.0 }
 
+["claude-4.6-sonnet"]
+model_id = "claude-sonnet-4-6"
+max_tokens = 64000
+inputs = ["text", "images", "pdf"]
+outputs = ["text", "structured"]
+max_prompt_images = 100
+costs = { input = 3.0, output = 15.0 }
+thinking_mode = "adaptive"
+
 ["claude-4.6-opus"]
 model_id = "claude-opus-4-6"
 max_tokens = 64000
@@ -108,3 +118,110 @@ outputs = ["text", "structured"]
 max_prompt_images = 100
 costs = { input = 5.0, output = 25.0 }
 thinking_mode = "adaptive"
+
+# --- Claude 4.7 Series --------------------------------------------------------
+["claude-4.7-opus"]
+model_id = "claude-opus-4-7"
+max_tokens = 128000
+inputs = ["text", "images", "pdf"]
+outputs = ["text", "structured"]
+max_prompt_images = 100
+costs = { input = 5.0, output = 25.0 }
+thinking_mode = "adaptive"
+listed_constraints = ["temperature_unsupported"]
+
+# --- Claude 4.8 Series --------------------------------------------------------
+["claude-4.8-opus"]
+model_id = "claude-opus-4-8"
+max_tokens = 128000
+inputs = ["text", "images", "pdf"]
+outputs = ["text", "structured"]
+max_prompt_images = 100
+costs = { input = 5.0, output = 25.0 }
+thinking_mode = "adaptive"
+listed_constraints = ["temperature_unsupported"]
+
+# --- Claude 5 Series ----------------------------------------------------------
+[claude-5-sonnet]
+model_id = "claude-sonnet-5"
+max_tokens = 64000
+inputs = ["text", "images", "pdf"]
+outputs = ["text", "structured"]
+max_prompt_images = 100
+costs = { input = 2.0, output = 10.0 }
+thinking_mode = "adaptive"
+listed_constraints = ["temperature_unsupported"]
+
+[claude-5-opus]
+model_id = "claude-opus-5"
+max_tokens = 128000
+inputs = ["text", "images", "pdf"]
+outputs = ["text", "structured"]
+max_prompt_images = 100
+costs = { input = 5.0, output = 25.0 }
+thinking_mode = "adaptive"
+listed_constraints = ["temperature_unsupported"]
+
+[claude-5-fable]
+model_id = "claude-fable-5"
+max_tokens = 128000
+inputs = ["text", "images", "pdf"]
+outputs = ["text", "structured"]
+max_prompt_images = 100
+costs = { input = 10.0, output = 50.0 }
+thinking_mode = "adaptive"
+listed_constraints = ["temperature_unsupported"]
+
+# --- Claude 5.1 Series --------------------------------------------------------
+["claude-5.1-fable"]
+model_id = "claude-fable-5-1"
+max_tokens = 128000
+inputs = ["text", "images", "pdf"]
+outputs = ["text", "structured"]
+max_prompt_images = 100
+costs = { input = 10.0, output = 50.0 }
+thinking_mode = "adaptive"
+# Fable 5.1 refuses tool_choice "tool" and "any", which is what the section default
+# instructor/anthropic_tools forces; the reasoning variant leaves tool_choice on auto.
+structure_method = "instructor/anthropic_reasoning_tools"
+listed_constraints = ["temperature_unsupported"]
+
+# --- Claude 5.5 Series --------------------------------------------------------
+["claude-5.5-sonnet"]
+model_id = "claude-sonnet-5-5"
+max_tokens = 128000
+inputs = ["text", "images", "pdf"]
+outputs = ["text", "structured"]
+max_prompt_images = 100
+costs = { input = 2.0, output = 10.0 }
+thinking_mode = "adaptive"
+# Sonnet 5.5 refuses tool_choice "tool" and "any", which is what the section default
+# instructor/anthropic_tools forces; the reasoning variant leaves tool_choice on auto.
+structure_method = "instructor/anthropic_reasoning_tools"
+listed_constraints = ["temperature_unsupported"]
+
+["claude-5.5-opus"]
+model_id = "claude-opus-5-5"
+max_tokens = 128000
+inputs = ["text", "images", "pdf"]
+outputs = ["text", "structured"]
+max_prompt_images = 100
+costs = { input = 4.0, output = 20.0 }
+thinking_mode = "adaptive"
+# Opus 5.5 refuses tool_choice "tool" and "any", like Sonnet 5.5 above.
+structure_method = "instructor/anthropic_reasoning_tools"
+listed_constraints = ["temperature_unsupported"]
+
+["claude-5.5-haiku"]
+model_id = "claude-haiku-5-5"
+max_tokens = 128000
+inputs = ["text", "images", "pdf"]
+outputs = ["text", "structured"]
+max_prompt_images = 100
+# The price of a prompt up to 100,000 tokens; a longer prompt pays $0.50 / $2.50.
+costs = { input = 0.10, output = 0.50 }
+thinking_mode = "adaptive"
+# Unlike Sonnet 5.5 and Opus 5.5, Haiku 5.5 accepts the forced tool_choice that
+# instructor/anthropic_tools sends; it then answers with the tool call and no thinking.
+structure_method = "instructor/anthropic_tools"
+listed_constraints = ["temperature_unsupported"]