multi-forge 0.2.0__py3-none-any.whl

forge/config/defaults/templates/openrouter-anthropic.yaml

@@ -0,0 +1,23 @@
+# Template: openrouter-anthropic
+# Anthropic Claude models via OpenRouter (direct, no LiteLLM)
+#
+# Access to Claude Opus 4.6/4.7, Sonnet 4.6, and Haiku 4.5.
+# Requires OPENROUTER_API_KEY in .env (or environment).
+#
+# Use --model claude-opus-4-7 to select Opus 4.7 instead of the 4.6 default.
+
+proxy:
+  family: anthropic
+  preferred_provider: openrouter
+  default_port: 8095
+
+  openrouter:
+    tiers:
+      haiku: anthropic/claude-haiku-4.5
+      sonnet: anthropic/claude-sonnet-4.6
+      opus: anthropic/claude-opus-4.6
+    model_alternatives:
+      opus:
+        claude-opus-4-7: anthropic/claude-opus-4.7
+    base_url: "https://openrouter.ai/api/v1"
+    error_hints: true

forge/config/defaults/templates/openrouter-deepseek.yaml

@@ -0,0 +1,26 @@
+# Template: openrouter-deepseek
+# DeepSeek open-source models via OpenRouter (direct, no LiteLLM)
+#
+# Haiku uses DeepSeek V4 Flash for lower-cost, high-throughput work.
+# Sonnet/opus use DeepSeek V4 Pro, with opus requesting the max reasoning tier.
+# Requires OPENROUTER_API_KEY in .env (or environment).
+
+proxy:
+  family: deepseek
+  preferred_provider: openrouter
+  default_port: 8100
+
+  openrouter:
+    tiers:
+      haiku: deepseek/deepseek-v4-flash
+      sonnet: deepseek/deepseek-v4-pro
+      opus: deepseek/deepseek-v4-pro
+    tier_overrides:
+      haiku:
+        reasoning_effort: high
+      sonnet:
+        reasoning_effort: high
+      opus:
+        reasoning_effort: xhigh
+    base_url: "https://openrouter.ai/api/v1"
+    error_hints: true

forge/config/defaults/templates/openrouter-gemini-flash.yaml

@@ -0,0 +1,26 @@
+# Template: openrouter-gemini-flash
+# Gemini 2.5 Flash via OpenRouter (cheap, all tiers use Flash)
+#
+# Requires OPENROUTER_API_KEY in .env (or environment).
+#
+# All tiers use Flash, differentiated by reasoning_effort.
+
+proxy:
+  family: gemini
+  preferred_provider: openrouter
+  default_port: 8099
+
+  openrouter:
+    tiers:
+      haiku: google/gemini-2.5-flash
+      sonnet: google/gemini-2.5-flash
+      opus: google/gemini-2.5-flash
+    tier_overrides:
+      haiku:
+        reasoning_effort: low
+      sonnet:
+        reasoning_effort: medium
+      opus:
+        reasoning_effort: high
+    base_url: "https://openrouter.ai/api/v1"
+    error_hints: true

forge/config/defaults/templates/openrouter-gemini.yaml

@@ -0,0 +1,23 @@
+# Template: openrouter-gemini
+# Google Gemini models via OpenRouter (direct, no LiteLLM)
+#
+# Access to Gemini 3.1 Pro and Flash.
+# Requires OPENROUTER_API_KEY in .env (or environment).
+
+proxy:
+  family: gemini
+  preferred_provider: openrouter
+  default_port: 8097
+
+  openrouter:
+    tiers:
+      haiku: google/gemini-2.5-flash
+      sonnet: google/gemini-3.1-pro-preview
+      opus: google/gemini-3.1-pro-preview
+    tier_overrides:
+      sonnet:
+        reasoning_effort: medium
+      opus:
+        reasoning_effort: high
+    base_url: "https://openrouter.ai/api/v1"
+    error_hints: true

forge/config/defaults/templates/openrouter-glm.yaml

@@ -0,0 +1,23 @@
+# Template: openrouter-glm
+# GLM / Z.ai models via OpenRouter (direct, no LiteLLM)
+#
+# Haiku uses GLM 4.7 Flash. Sonnet/opus use GLM 5.1.
+# Requires OPENROUTER_API_KEY in .env (or environment).
+
+proxy:
+  family: glm
+  preferred_provider: openrouter
+  default_port: 8103
+
+  openrouter:
+    tiers:
+      haiku: z-ai/glm-4.7-flash
+      sonnet: z-ai/glm-5.1
+      opus: z-ai/glm-5.1
+    tier_overrides:
+      sonnet:
+        reasoning_effort: high
+      opus:
+        reasoning_effort: high
+    base_url: "https://openrouter.ai/api/v1"
+    error_hints: true

forge/config/defaults/templates/openrouter-kimi.yaml

@@ -0,0 +1,30 @@
+# Template: openrouter-kimi
+# Kimi models via OpenRouter (direct, no LiteLLM)
+#
+# Haiku uses Gemma 4 31B as a shared low-cost multimodal tier. Sonnet/opus use
+# Kimi K2.6 as the higher-capability Kimi tier. K2.5 is available as a
+# sonnet/opus model_alternative for explicit proxy requests.
+# Requires OPENROUTER_API_KEY in .env (or environment).
+
+proxy:
+  family: kimi
+  preferred_provider: openrouter
+  default_port: 8102
+
+  openrouter:
+    tiers:
+      haiku: google/gemma-4-31b-it
+      sonnet: moonshotai/kimi-k2.6
+      opus: moonshotai/kimi-k2.6
+    model_alternatives:
+      sonnet:
+        kimi-k2.5: moonshotai/kimi-k2.5
+      opus:
+        kimi-k2.5: moonshotai/kimi-k2.5
+    tier_overrides:
+      sonnet:
+        reasoning_effort: high
+      opus:
+        reasoning_effort: high
+    base_url: "https://openrouter.ai/api/v1"
+    error_hints: true

forge/config/defaults/templates/openrouter-minimax.yaml

@@ -0,0 +1,26 @@
+# Template: openrouter-minimax
+# MiniMax models via OpenRouter (direct, no LiteLLM)
+#
+# Haiku uses Gemma 4 31B as a shared low-cost multimodal tier. Sonnet/opus use
+# MiniMax M2.7. M2.5 is available as a sonnet/opus model_alternative for
+# explicit proxy requests.
+# Requires OPENROUTER_API_KEY in .env (or environment).
+
+proxy:
+  family: minimax
+  preferred_provider: openrouter
+  default_port: 8104
+
+  openrouter:
+    tiers:
+      haiku: google/gemma-4-31b-it
+      sonnet: minimax/minimax-m2.7
+      opus: minimax/minimax-m2.7
+    model_alternatives:
+      sonnet:
+        minimax-m2.5: minimax/minimax-m2.5
+      opus:
+        minimax-m2.5: minimax/minimax-m2.5
+    tier_overrides: {}
+    base_url: "https://openrouter.ai/api/v1"
+    error_hints: true

forge/config/defaults/templates/openrouter-openai-codex.yaml

@@ -0,0 +1,23 @@
+# Template: openrouter-openai-codex
+# OpenAI Codex models via OpenRouter (direct, no LiteLLM)
+#
+# Requires OPENROUTER_API_KEY in .env (or environment).
+
+proxy:
+  family: openai
+  preferred_provider: openrouter
+  default_port: 8098
+
+  openrouter:
+    tiers:
+      haiku: openai/gpt-5.1-codex-mini
+      sonnet: openai/gpt-5.3-codex
+      opus: openai/gpt-5.5
+    tier_overrides:
+      sonnet:
+        reasoning_effort: medium
+      opus:
+        reasoning_effort: high
+        verbosity: high
+    base_url: "https://openrouter.ai/api/v1"
+    error_hints: true

forge/config/defaults/templates/openrouter-openai.yaml

@@ -0,0 +1,28 @@
+# Template: openrouter-openai
+# OpenAI models via OpenRouter (direct, no LiteLLM)
+#
+# Access to GPT-5.5 and GPT-5.4-mini.
+# Requires OPENROUTER_API_KEY in .env (or environment).
+#
+# Haiku uses gpt-5.4-mini; sonnet and opus both use gpt-5.5,
+# differentiated by tier_overrides (reasoning_effort and verbosity).
+
+proxy:
+  family: openai
+  preferred_provider: openrouter
+  default_port: 8096
+
+  openrouter:
+    tiers:
+      haiku: openai/gpt-5.4-mini
+      sonnet: openai/gpt-5.5
+      opus: openai/gpt-5.5
+    tier_overrides:
+      sonnet:
+        reasoning_effort: medium
+        verbosity: high
+      opus:
+        reasoning_effort: high
+        verbosity: high
+    base_url: "https://openrouter.ai/api/v1"
+    error_hints: true

forge/config/defaults/templates/openrouter-qwen.yaml

@@ -0,0 +1,25 @@
+# Template: openrouter-qwen
+# Qwen models via OpenRouter (direct, no LiteLLM)
+#
+# Haiku uses Qwen3.6 Flash for fast multimodal work. Sonnet uses Qwen3.6 Plus.
+# Opus uses Qwen3.6 Max Preview for the strongest current Qwen tier.
+# qwen3-coder is available as a model_alternative for explicit proxy requests.
+# Requires OPENROUTER_API_KEY in .env (or environment).
+
+proxy:
+  family: qwen
+  preferred_provider: openrouter
+  default_port: 8101
+
+  openrouter:
+    tiers:
+      haiku: qwen/qwen3.6-flash
+      sonnet: qwen/qwen3.6-plus
+      opus: qwen/qwen3.6-max-preview
+    model_alternatives:
+      sonnet:
+        qwen3-coder: qwen/qwen3-coder
+      opus:
+        qwen3-coder: qwen/qwen3-coder
+    base_url: "https://openrouter.ai/api/v1"
+    error_hints: true