multi-forge 1.0.1__tar.gz → 1.0.2__tar.gz

{multi_forge-1.0.1 → multi_forge-1.0.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.5
 Name: multi-forge
-Version: 1.0.1
+Version: 1.0.2
 Summary: Multi-runtime agent toolkit: proxy routing, cost control, session management, policy enforcement, and workflow orchestration
 Project-URL: Homepage, https://github.com/hapa1i/multi-forge
 Project-URL: Repository, https://github.com/hapa1i/multi-forge

{multi_forge-1.0.1 → multi_forge-1.0.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "multi-forge"
-version = "1.0.1"
+version = "1.0.2"
 description = "Multi-runtime agent toolkit: proxy routing, cost control, session management, policy enforcement, and workflow orchestration"
 readme = "README.md"
 license = "Apache-2.0"

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/cli/claude.py

@@ -129,8 +129,9 @@ def _build_bare_launch_env(
         env_vars["ANTHROPIC_BASE_URL"] = base_url
         if context_limit is not None:
             env_vars["CLAUDE_CODE_AUTO_COMPACT_WINDOW"] = str(context_limit)
-        apply_proxy_context_model_defaults(env_vars, context_limit)
-        if wire_shape := resolve_proxy_wire_shape(proxy_id=proxy_id, template=template):
+        wire_shape = resolve_proxy_wire_shape(proxy_id=proxy_id, template=template)
+        apply_proxy_context_model_defaults(env_vars, context_limit, wire_shape=wire_shape)
+        if wire_shape:
             env_vars[FORGE_PROXY_WIRE_SHAPE_VAR] = wire_shape
         if template:
             env_vars["ACTIVE_TEMPLATE"] = template

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/backends/litellm.yaml

@@ -14,7 +14,7 @@
 #     - gemini/gemini-2.5-pro
 #
 #   Anthropic (requires ANTHROPIC_API_KEY):
-#     - anthropic/claude-haiku-4-5, claude-sonnet-5, claude-opus-5-5, claude-opus-5
+#     - anthropic/claude-haiku-4-5, claude-sonnet-5-5, claude-sonnet-5, claude-opus-5-5, claude-opus-5
 #     - anthropic/claude-fable-5-1, claude-fable-5
 #     - anthropic/claude-opus-4-8, claude-opus-4-6, claude-sonnet-4-6
 #
@@ -26,7 +26,7 @@
 #     - openai/gpt-5.4, gpt-5.4-pro
 #     - openai/gpt-5.5
 #     - openai/gpt-5.6, gpt-5.6-sol, gpt-5.6-terra, gpt-5.6-luna
-#     - openai/gpt-6-astra, gpt-6-sol, gpt-6-luna
+#     - openai/gpt-6-astra, gpt-6.1-sol, gpt-6-sol, gpt-6-luna
 #
 # Start with: forge model backend start litellm --port 4000
 
@@ -140,6 +140,44 @@ model_list:
       model: anthropic/claude-haiku-4-5
       api_key: os.environ/ANTHROPIC_API_KEY
 
+  # LiteLLM 1.102 predates Sonnet 5.5; supply offline pricing and effort capabilities.
+  # https://platform.claude.com/docs/en/models/sonnet-5-5/overview (verified 2026-10-01)
+  - model_name: anthropic/claude-sonnet-5-5
+    litellm_params:
+      model: anthropic/claude-sonnet-5-5
+      api_key: os.environ/ANTHROPIC_API_KEY
+    model_info:
+      base_model: claude-sonnet-5-5
+      mode: chat
+      max_input_tokens: 1000000
+      max_output_tokens: 128000
+      max_tokens: 128000
+      input_cost_per_token: 0.000002
+      output_cost_per_token: 0.000010
+      cache_creation_input_token_cost: 0.0000025
+      cache_creation_input_token_cost_above_1hr: 0.000004
+      cache_read_input_token_cost: 0.0000002
+      supports_vision: true
+      supports_function_calling: true
+      supports_tool_choice: true
+      supports_prompt_caching: true
+      prompt_cache_min_tokens: 512
+      supports_reasoning: true
+      supports_adaptive_thinking: true
+      supports_legacy_thinking: false
+      # Keep disabled visible to upstream validation; dropping it would turn thinking on.
+      thinking_always_on: false
+      supports_output_config: true
+      supports_xhigh_reasoning_effort: true
+      supports_max_reasoning_effort: true
+      default_reasoning_effort: high
+      reasoning_effort_levels: [low, medium, high, xhigh, max]
+      supports_response_schema: true
+      supports_native_structured_output: true
+      supports_assistant_prefill: false
+      provider_specific_entry:
+        us: 1.1
+
   - model_name: anthropic/claude-sonnet-5
     litellm_params:
       model: anthropic/claude-sonnet-5
@@ -340,6 +378,56 @@ model_list:
       model: openai/gpt-6-astra
       api_key: os.environ/OPENAI_API_KEY
 
+  # LiteLLM 1.102 predates GPT-6.1 Sol; cached reads are 5% of input, including the >272k premium.
+  # https://developers.openai.com/api/docs/models/gpt-6.1-sol (verified 2026-10-01)
+  - model_name: openai/gpt-6.1-sol
+    litellm_params:
+      model: openai/gpt-6.1-sol
+      api_key: os.environ/OPENAI_API_KEY
+      use_responses_api: true
+    model_info:
+      base_model: gpt-6.1-sol
+      mode: responses
+      max_input_tokens: 922000
+      max_output_tokens: 128000
+      max_tokens: 128000
+      input_cost_per_token: 0.000002
+      output_cost_per_token: 0.000010
+      cache_read_input_token_cost: 0.0000001
+      cache_creation_input_token_cost: 0.0000025
+      input_cost_per_token_above_272k_tokens: 0.000004
+      output_cost_per_token_above_272k_tokens: 0.000015
+      cache_read_input_token_cost_above_272k_tokens: 0.0000002
+      cache_creation_input_token_cost_above_272k_tokens: 0.000005
+      input_cost_per_token_flex: 0.000001
+      output_cost_per_token_flex: 0.000005
+      cache_read_input_token_cost_flex: 0.00000005
+      cache_creation_input_token_cost_flex: 0.00000125
+      input_cost_per_token_above_272k_tokens_flex: 0.000002
+      output_cost_per_token_above_272k_tokens_flex: 0.0000075
+      cache_read_input_token_cost_above_272k_tokens_flex: 0.0000001
+      cache_creation_input_token_cost_above_272k_tokens_flex: 0.0000025
+      input_cost_per_token_priority: 0.000004
+      output_cost_per_token_priority: 0.000020
+      cache_read_input_token_cost_priority: 0.0000002
+      cache_creation_input_token_cost_priority: 0.000005
+      input_cost_per_token_above_272k_tokens_priority: 0.000008
+      output_cost_per_token_above_272k_tokens_priority: 0.000030
+      cache_read_input_token_cost_above_272k_tokens_priority: 0.0000004
+      cache_creation_input_token_cost_above_272k_tokens_priority: 0.000010
+      supports_vision: true
+      supports_function_calling: true
+      supports_parallel_function_calling: true
+      supports_tool_choice: true
+      supports_prompt_caching: true
+      supports_reasoning: true
+      supports_none_reasoning_effort: false
+      supports_xhigh_reasoning_effort: true
+      supports_max_reasoning_effort: true
+      default_reasoning_effort: medium
+      reasoning_effort_levels: [low, medium, high, xhigh, max]
+      supports_response_schema: true
+
   # LiteLLM 1.102 predates Sol and Luna. Standard, Flex and Fast rates include the >272k-token premium.
   # Flex costs half of standard; Fast (also called priority) costs twice the applicable rate.
   # Remove these overrides when bundled metadata covers them; test_litellm_model_metadata pins that boundary.

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/templates/anthropic-passthrough.yaml

@@ -27,5 +27,5 @@ proxy:
   litellm:
     tiers:
       haiku: claude-haiku-4-5
-      sonnet: claude-sonnet-5
+      sonnet: claude-sonnet-5-5
       opus: claude-opus-5-5

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/templates/codex-responses-local.yaml

@@ -30,10 +30,12 @@ proxy:
       haiku:
         gpt-6-luna: openai/gpt-6-luna
       sonnet:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-luna: openai/gpt-6-luna
         gpt-5.6-sol: openai/gpt-5.6-sol
       opus:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-luna: openai/gpt-6-luna
         gpt-5.6-sol: openai/gpt-5.6-sol

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/templates/litellm-anthropic-local.yaml

@@ -6,9 +6,9 @@
 #
 # Backend (LiteLLM on port 4000) starts automatically.
 #
-# Defaults: sonnet tier -> Sonnet 5, opus tier -> Opus 5.5. Alternatives are
+# Defaults: sonnet tier -> Sonnet 5.5, opus tier -> Opus 5.5. Alternatives are
 # claude-opus-5, claude-fable-5-1, claude-fable-5, claude-opus-4-8, claude-opus-4-6,
-# and claude-sonnet-4-6.
+# claude-sonnet-5, and claude-sonnet-4-6.
 #
 # After adding ANTHROPIC_API_KEY, stop the backend so Forge restarts it with the new environment:
 # forge model backend stop litellm-4000
@@ -22,7 +22,7 @@ proxy:
   litellm:
     tiers:
       haiku: anthropic/claude-haiku-4-5
-      sonnet: anthropic/claude-sonnet-5
+      sonnet: anthropic/claude-sonnet-5-5
       opus: anthropic/claude-opus-5-5
     model_alternatives:
       opus:
@@ -32,6 +32,7 @@ proxy:
         claude-opus-4-8: anthropic/claude-opus-4-8
         claude-opus-4-6: anthropic/claude-opus-4-6
       sonnet:
+        claude-sonnet-5: anthropic/claude-sonnet-5
         claude-sonnet-4-6: anthropic/claude-sonnet-4-6
     prompt_caching: passthrough
     error_hints: true

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/templates/litellm-anthropic.yaml

@@ -1,13 +1,13 @@
 # Template: litellm-anthropic
 # Anthropic Claude via remote/shared LiteLLM
 #
-# Access to Claude Opus 5.5/5, Sonnet 5, Fable 5.1/5, Opus 4.8/4.6, Sonnet 4.6, and Haiku 4.5.
+# Access to Claude Opus 5.5/5, Sonnet 5.5/5, Fable 5.1/5, Opus 4.8/4.6, Sonnet 4.6, and Haiku 4.5.
 # Requires LITELLM_API_KEY and LITELLM_BASE_URL in .env (or environment).
-# NOTE: the shared LiteLLM instance must serve anthropic/claude-opus-5-5 and anthropic/claude-sonnet-5.
+# NOTE: the shared LiteLLM instance must serve anthropic/claude-opus-5-5 and anthropic/claude-sonnet-5-5.
 #
-# Defaults: sonnet tier -> Sonnet 5, opus tier -> Opus 5.5. Alternatives are
+# Defaults: sonnet tier -> Sonnet 5.5, opus tier -> Opus 5.5. Alternatives are
 # claude-opus-5, claude-fable-5-1, claude-fable-5, claude-opus-4-8, claude-opus-4-6,
-# and claude-sonnet-4-6.
+# claude-sonnet-5, and claude-sonnet-4-6.
 # prompt_caching: passthrough respects Anthropic's 4-breakpoint cache_control limit.
 
 proxy:
@@ -19,7 +19,7 @@ proxy:
   litellm:
     tiers:
       haiku: anthropic/claude-haiku-4-5
-      sonnet: anthropic/claude-sonnet-5
+      sonnet: anthropic/claude-sonnet-5-5
       opus: anthropic/claude-opus-5-5
     model_alternatives:
       opus:
@@ -29,6 +29,7 @@ proxy:
         claude-opus-4-8: anthropic/claude-opus-4-8
         claude-opus-4-6: anthropic/claude-opus-4-6
       sonnet:
+        claude-sonnet-5: anthropic/claude-sonnet-5
         claude-sonnet-4-6: anthropic/claude-sonnet-4-6
     prompt_caching: passthrough
     error_hints: true

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/templates/litellm-openai-codex-local.yaml

@@ -24,9 +24,11 @@ proxy:
       haiku:
         gpt-6-luna: openai/gpt-6-luna
       sonnet:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-luna: openai/gpt-6-luna
       opus:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-luna: openai/gpt-6-luna
         gpt-5.6-sol: openai/gpt-5.6-sol

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/templates/litellm-openai-local.yaml

@@ -24,10 +24,12 @@ proxy:
       haiku:
         gpt-6-luna: openai/gpt-6-luna
       sonnet:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-luna: openai/gpt-6-luna
         gpt-5.6-sol: openai/gpt-5.6-sol
       opus:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-luna: openai/gpt-6-luna
         gpt-5.6-sol: openai/gpt-5.6-sol

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/templates/litellm-openai.yaml

@@ -22,10 +22,12 @@ proxy:
       haiku:
         gpt-6-luna: openai/gpt-6-luna
       sonnet:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-luna: openai/gpt-6-luna
         gpt-5.6-sol: openai/gpt-5.6-sol
       opus:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-luna: openai/gpt-6-luna
         gpt-5.6-sol: openai/gpt-5.6-sol

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/templates/openrouter-anthropic.yaml

@@ -1,12 +1,12 @@
 # Template: openrouter-anthropic
 # Anthropic Claude models via OpenRouter (direct, no LiteLLM)
 #
-# Access to Claude Opus 5.5/5, Sonnet 5, Fable 5.1/5, Opus 4.8/4.6, Sonnet 4.6, and Haiku 4.5.
+# Access to Claude Opus 5.5/5, Sonnet 5.5/5, Fable 5.1/5, Opus 4.8/4.6, Sonnet 4.6, and Haiku 4.5.
 # Requires OPENROUTER_API_KEY in .env (or environment).
 #
-# Defaults: sonnet tier -> Sonnet 5, opus tier -> Opus 5.5. Alternatives are
+# Defaults: sonnet tier -> Sonnet 5.5, opus tier -> Opus 5.5. Alternatives are
 # claude-opus-5, claude-fable-5-1, claude-fable-5, claude-opus-4-8, claude-opus-4-6,
-# and claude-sonnet-4-6.
+# claude-sonnet-5, and claude-sonnet-4-6.
 
 proxy:
   family: anthropic
@@ -18,7 +18,7 @@ proxy:
     allow_non_zdr: false
     tiers:
       haiku: anthropic/claude-haiku-4.5
-      sonnet: anthropic/claude-sonnet-5
+      sonnet: anthropic/claude-sonnet-5.5
       opus: anthropic/claude-opus-5.5
     zdr_fallbacks:
       anthropic/claude-fable-5.1: anthropic/claude-opus-5.5
@@ -31,5 +31,6 @@ proxy:
         claude-opus-4-8: anthropic/claude-opus-4.8
         claude-opus-4-6: anthropic/claude-opus-4.6
       sonnet:
+        claude-sonnet-5: anthropic/claude-sonnet-5
         claude-sonnet-4-6: anthropic/claude-sonnet-4.6
     error_hints: true

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/templates/openrouter-openai-codex.yaml

@@ -20,11 +20,13 @@ proxy:
         gpt-6-luna: openai/gpt-6-luna
         gpt-6-luna-pro: openai/gpt-6-luna-pro
       sonnet:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-sol-pro: openai/gpt-6-sol-pro
         gpt-6-luna: openai/gpt-6-luna
         gpt-6-luna-pro: openai/gpt-6-luna-pro
       opus:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-sol-pro: openai/gpt-6-sol-pro
         gpt-6-luna: openai/gpt-6-luna

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/defaults/templates/openrouter-openai.yaml

@@ -1,7 +1,7 @@
 # Template: openrouter-openai
 # OpenAI models via OpenRouter (direct, no LiteLLM)
 #
-# Access to GPT-6 Astra, Sol, Luna, their Pro variants, and GPT-5.4-mini.
+# Access to GPT-6.1 Sol, GPT-6 Astra/Sol/Luna and their Pro variants, 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-6-astra,
@@ -24,6 +24,7 @@ proxy:
         gpt-6-luna: openai/gpt-6-luna
         gpt-6-luna-pro: openai/gpt-6-luna-pro
       sonnet:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-sol-pro: openai/gpt-6-sol-pro
         gpt-6-luna: openai/gpt-6-luna
@@ -31,6 +32,7 @@ proxy:
         gpt-5.6-sol: openai/gpt-5.6-sol
         gpt-6-astra-pro: openai/gpt-6-astra-pro
       opus:
+        gpt-6.1-sol: openai/gpt-6.1-sol
         gpt-6-sol: openai/gpt-6-sol
         gpt-6-sol-pro: openai/gpt-6-sol-pro
         gpt-6-luna: openai/gpt-6-luna

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/config/schema.py

@@ -85,6 +85,7 @@ OPENAI_MODELS = [
     "gpt-5.6-luna",
     "gpt-6-astra",
     "gpt-6-astra-pro",
+    "gpt-6.1-sol",
     "gpt-6-sol",
     "gpt-6-sol-pro",
     "gpt-6-luna",
@@ -1058,10 +1059,14 @@ def _validate_static_tier_override_constraints(tiers: TierModels, overrides: Tie
                 "set that effort or remove the temperature override"
             )
 
-        if spec.thinking_modes == ("adaptive",) and override.thinking_budget_tokens is not None:
+        if (
+            spec.thinking_modes is not None
+            and "enabled" not in spec.thinking_modes
+            and override.thinking_budget_tokens is not None
+        ):
             raise ValueError(
                 f"tier_overrides.{tier}.thinking_budget_tokens is not supported by {canonical_model}; "
-                "this model only supports adaptive thinking"
+                "use reasoning_effort instead of a manual thinking budget"
             )
 
         if (

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/core/data/model_catalog.yaml

@@ -43,11 +43,11 @@ defaults:
     opus: gemini-3.1-pro-preview
   anthropic:
     haiku: claude-haiku-4-5-20251001
-    sonnet: claude-sonnet-5
+    sonnet: claude-sonnet-5-5
     opus: claude-opus-5-5
   openrouter:
     haiku: claude-haiku-4-5-20251001
-    sonnet: claude-sonnet-5
+    sonnet: claude-sonnet-5-5
     opus: claude-opus-5-5
   deepseek:
     haiku: deepseek-v4-flash
@@ -126,6 +126,31 @@ models:
     system_prompt_addendum: system_prompt_addendums/openai.md
     tags: []
 
+  # https://developers.openai.com/api/docs/models/gpt-6.1-sol
+  gpt-6.1-sol:
+    friendly_name: GPT-6.1 Sol
+    context_window_tokens: 1050000
+    max_output_tokens: 128000
+    max_thinking_tokens: null
+    supports_thinking: true
+    supports_images: true
+    supports_verbosity: true
+    verbosity_levels: [low, medium, high]
+    temperature_constraint: fixed
+    temperature: 1.0
+    supports_top_p: false
+    supports_sampling_overrides: false
+    default_timeout_seconds: 180
+    prompt_caching:
+      supports: true
+      mechanism: auto
+    native_thinking_param: reasoning_effort
+    litellm_reasoning_efforts: [low, medium, high, xhigh, max]
+    default_reasoning_effort: medium
+    use_responses_api: true
+    system_prompt_addendum: system_prompt_addendums/openai.md
+    tags: []
+
   gpt-6-sol:
     friendly_name: GPT-6 Sol
     context_window_tokens: 1050000
@@ -1273,6 +1298,7 @@ models:
       max_cache_blocks: 4
     native_thinking_param: output_config.effort
     thinking_modes: [adaptive]
+    supports_forced_tool_choice: false
     litellm_reasoning_efforts: [low, medium, high, xhigh, max]
     default_reasoning_effort: medium
     token_estimate_multiplier: 1.35
@@ -1360,10 +1386,36 @@ models:
     token_estimate_multiplier: 1.35
     tags: [opt-in]
 
-  # --- Claude Sonnet 5 ---
-  # Sonnet 5 is the default sonnet tier (anthropic + openrouter) and shares the
-  # Opus 4.8 request surface: adaptive-only thinking, sampling overrides removed,
-  # native 1M context. Single canonical id (no -1m twin).
+  # Sonnet 5.5 uses between_tools instead of disabled for no up-front thinking.
+  # https://platform.claude.com/docs/en/models/sonnet-5-5/overview
+  claude-sonnet-5-5:
+    friendly_name: Claude Sonnet 5.5
+    context_window_tokens: 1000000
+    max_output_tokens: 128000
+    max_thinking_tokens: 128000
+    supports_thinking: true
+    supports_images: true
+    supports_1m_context: true
+    temperature_constraint: fixed
+    temperature: 1.0
+    supports_top_p: false
+    supports_sampling_overrides: false
+    default_timeout_seconds: 180
+    prompt_caching:
+      supports: true
+      mechanism: cache_control
+      min_tokens: 512
+      max_cache_blocks: 4
+    native_thinking_param: output_config.effort
+    thinking_modes: [adaptive, between_tools]
+    supports_forced_tool_choice: false
+    between_tools_reasoning_efforts: [low, medium, high]
+    litellm_reasoning_efforts: [low, medium, high, xhigh, max]
+    default_reasoning_effort: high
+    token_estimate_multiplier: 1.35
+    tags: []
+
+  # Sonnet 5 remains selectable by its explicit version.
   claude-sonnet-5:
     friendly_name: Claude Sonnet 5
     context_window_tokens: 1000000
@@ -2235,8 +2287,9 @@ aliases:
   astra-pro: gpt-6-astra-pro
   gpt-astra-pro: gpt-6-astra-pro
   openai/gpt-6-astra-pro: gpt-6-astra-pro
-  sol: gpt-6-sol
-  gpt-sol: gpt-6-sol
+  sol: gpt-6.1-sol
+  gpt-sol: gpt-6.1-sol
+  openai/gpt-6.1-sol: gpt-6.1-sol
   openai/gpt-6-sol: gpt-6-sol
   sol-pro: gpt-6-sol-pro
   gpt-sol-pro: gpt-6-sol-pro
@@ -2361,11 +2414,17 @@ aliases:
   anthropic/claude-fable-5: claude-fable-5
   fable-5: claude-fable-5
 
-  # Sonnet 5 (default sonnet tier; released on Anthropic + OpenRouter, native 1M)
+  # Sonnet 5.5 (default sonnet tier; OpenRouter uses a dot-form slug)
+  anthropic/claude-sonnet-5-5: claude-sonnet-5-5
+  anthropic/claude-sonnet-5.5: claude-sonnet-5-5
+  claude-sonnet-5.5: claude-sonnet-5-5
+  sonnet-5.5: claude-sonnet-5-5
+  sonnet-5-5: claude-sonnet-5-5
+  claude-sonnet: claude-sonnet-5-5
+  sonnet: claude-sonnet-5-5
+  # Sonnet 5 remains an explicit pin.
   anthropic/claude-sonnet-5: claude-sonnet-5  # also the OpenRouter slug (no dot)
-  claude-sonnet: claude-sonnet-5  # parity with claude-opus friendly alias
   sonnet-5: claude-sonnet-5
-  sonnet: claude-sonnet-5
 
   # Sonnet 4.6
   anthropic/claude-sonnet-4-6: claude-sonnet-4-6

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/core/data/model_routes.yaml

@@ -74,6 +74,64 @@ models:
         source_id: openrouter
         template: openrouter-openai-codex
         model_ref: openai/gpt-6-astra-pro
+  gpt-6.1-sol:
+    routes:
+      - kind: proxy
+        source_id: openrouter
+        template: openrouter-openai
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: openrouter
+        template: openrouter-openai-codex
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: openrouter
+        template: openrouter-anthropic
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: openrouter
+        template: openrouter-deepseek
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: openrouter
+        template: openrouter-gemini
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: openrouter
+        template: openrouter-gemini-flash
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: openrouter
+        template: openrouter-glm
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: openrouter
+        template: openrouter-kimi
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: openrouter
+        template: openrouter-minimax
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: openrouter
+        template: openrouter-qwen
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: codex-responses-local
+        template: codex-responses-local
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: litellm-remote
+        template: litellm-openai
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: litellm-openai-local
+        template: litellm-openai-codex-local
+        model_ref: openai/gpt-6.1-sol
+      - kind: proxy
+        source_id: litellm-openai-local
+        template: litellm-openai-local
+        model_ref: openai/gpt-6.1-sol
   gpt-6-sol:
     routes:
       - kind: proxy
@@ -840,6 +898,11 @@ models:
       - kind: direct
         runtime: claude_code
         model_ref: claude-fable-5
+  claude-sonnet-5-5:
+    routes:
+      - kind: direct
+        runtime: claude_code
+        model_ref: claude-sonnet-5-5
   claude-sonnet-5:
     routes:
       - kind: direct

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/core/llm/clients/openai_compat.py

@@ -137,18 +137,15 @@ def build_chat_completion_kwargs(
     if "openai" in hyperparams.extra:
         kwargs.update(hyperparams.extra["openai"])
 
-    _apply_conditional_sampling(model, kwargs)
+    _apply_sampling_constraints(model, kwargs)
     return kwargs
 
 
-def _apply_conditional_sampling(model: str, kwargs: dict[str, Any]) -> None:
-    """Enforce effort-dependent sampling after provider extras have been merged."""
+def _apply_sampling_constraints(model: str, kwargs: dict[str, Any]) -> None:
+    """Enforce catalog sampling constraints after provider extras have been merged."""
     if not model_exists(model):
         return
     spec = get_model_spec(model)
-    if not spec.sampling_requires_no_reasoning:
-        return
-
     extra_body = kwargs.get("extra_body")
     reasoning = extra_body.get("reasoning") if isinstance(extra_body, dict) else None
     effort = kwargs.get("reasoning_effort")
@@ -156,7 +153,7 @@ def _apply_conditional_sampling(model: str, kwargs: dict[str, Any]) -> None:
         effort = reasoning.get("effort")
     if spec.supports_sampling_at_effort(effort):
         return
-    for name in ("temperature", "top_p"):
+    for name in ("temperature", "top_p", "top_k"):
         kwargs.pop(name, None)
         if isinstance(extra_body, dict) and name in extra_body:
             # Do not mutate the caller's nested provider extras.

{multi_forge-1.0.1 → multi_forge-1.0.2}/src/forge/core/models/catalog.py

@@ -188,6 +188,7 @@ def _parse_model_spec(model_id: str, data: dict[str, Any]) -> ModelSpec:
             supports_sampling_overrides=bool(data.get("supports_sampling_overrides", True)),
             sampling_requires_no_reasoning=bool(data.get("sampling_requires_no_reasoning", False)),
             supports_1m_context=bool(data.get("supports_1m_context", False)),
+            supports_forced_tool_choice=bool(data.get("supports_forced_tool_choice", True)),
             temperature_constraint=constraint,
             temperature=temperature,
             verbosity_levels=verbosity_levels,
@@ -196,6 +197,9 @@ def _parse_model_spec(model_id: str, data: dict[str, Any]) -> ModelSpec:
             litellm_reasoning_efforts=litellm_reasoning_efforts,
             default_reasoning_effort=data.get("default_reasoning_effort"),
             thinking_modes=thinking_modes,
+            between_tools_reasoning_efforts=_parse_tuple_or_none(
+                model_id, "between_tools_reasoning_efforts", data.get("between_tools_reasoning_efforts")
+            ),
             thinking_levels=thinking_levels,
             default_thinking_level=data.get("default_thinking_level"),
             token_estimate_multiplier=float(data.get("token_estimate_multiplier", 1.0)),