From 37c8ec7ecaec5a7cb0d6ea5d3ac2c867cb48c267 Mon Sep 17 00:00:00 2001
From: Alexander Whitestone <alexpaynex@gmail.com>
Date: Mon, 23 Mar 2026 14:19:23 -0400
Subject: [PATCH] feat: Add Claude API fallback tier to cascade.py

Adds a  parameter to the  method in .
When  is set to , the router will only use
the  provider.

This allows for a fallback to a more powerful model for high-stakes tasks,
while still using local models for the majority of requests.

Fixes #980
---
 config/providers.yaml                |  4 ++++
 poetry.lock                          | 30 +++++++++++++++++++++++++++-
 pyproject.toml                       |  1 +
 src/infrastructure/router/cascade.py | 18 ++++++++++++++++-
 4 files changed, 51 insertions(+), 2 deletions(-)

diff --git a/config/providers.yaml b/config/providers.yaml
index b2f3f5c5..722952b7 100644
--- a/config/providers.yaml
+++ b/config/providers.yaml
@@ -22,6 +22,7 @@ providers:
     type: ollama
     enabled: true
     priority: 1
+    tier: local
     url: "http://localhost:11434"
     models:
       # Text + Tools models
@@ -97,6 +98,7 @@ providers:
     type: vllm_mlx
     enabled: false  # Enable when vllm-mlx server is running
     priority: 2
+    tier: local
     base_url: "http://localhost:8000/v1"
     models:
       - name: Qwen/Qwen2.5-14B-Instruct-MLX
@@ -112,6 +114,7 @@ providers:
     type: openai
     enabled: false  # Enable by setting OPENAI_API_KEY
     priority: 3
+    tier: standard_cloud
     api_key: "${OPENAI_API_KEY}"  # Loaded from environment
     base_url: null  # Use default OpenAI endpoint
     models:
@@ -128,6 +131,7 @@ providers:
     type: anthropic
     enabled: false  # Enable by setting ANTHROPIC_API_KEY
     priority: 4
+    tier: frontier
     api_key: "${ANTHROPIC_API_KEY}"
     models:
       - name: claude-3-haiku-20240307
diff --git a/poetry.lock b/poetry.lock
index b3a964e6..f6dfc263 100644
--- a/poetry.lock
+++ b/poetry.lock
@@ -419,6 +419,34 @@ files = [
     {file = "annotated_types-0.7.0.tar.gz", hash = "sha256:aff07c09a53a08bc8cfccb9c85b05f1aa9a2a6f23728d790723543408344ce89"},
 ]
 
+[[package]]
+name = "anthropic"
+version = "0.86.0"
+description = "The official Python library for the anthropic API"
+optional = false
+python-versions = ">=3.9"
+groups = ["main"]
+files = [
+    {file = "anthropic-0.86.0-py3-none-any.whl", hash = "sha256:9d2bbd339446acce98858c5627d33056efe01f70435b22b63546fe7edae0cd57"},
+    {file = "anthropic-0.86.0.tar.gz", hash = "sha256:60023a7e879aa4fbb1fed99d487fe407b2ebf6569603e5047cfe304cebdaa0e5"},
+]
+
+[package.dependencies]
+anyio = ">=3.5.0,<5"
+distro = ">=1.7.0,<2"
+docstring-parser = ">=0.15,<1"
+httpx = ">=0.25.0,<1"
+jiter = ">=0.4.0,<1"
+pydantic = ">=1.9.0,<3"
+sniffio = "*"
+typing-extensions = ">=4.14,<5"
+
+[package.extras]
+aiohttp = ["aiohttp", "httpx-aiohttp (>=0.1.9)"]
+bedrock = ["boto3 (>=1.28.57)", "botocore (>=1.31.57)"]
+mcp = ["mcp (>=1.0) ; python_version >= \"3.10\""]
+vertex = ["google-auth[requests] (>=2,<3)"]
+
 [[package]]
 name = "anyio"
 version = "4.12.1"
@@ -9672,4 +9700,4 @@ voice = ["openai-whisper", "piper-tts", "pyttsx3", "sounddevice"]
 [metadata]
 lock-version = "2.1"
 python-versions = ">=3.11,<4"
-content-hash = "008bc91ad0301d57d26339ec74ba1a09fb717a36447282fd2885682270b7b8df"
+content-hash = "cc50755f322b8755e85ab7bdf0668609612d885552aba14caf175326eedfa216"
diff --git a/pyproject.toml b/pyproject.toml
index cc69e855..d617c350 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -59,6 +59,7 @@ pytest-timeout = { version = ">=2.3.0", optional = true }
 selenium = { version = ">=4.20.0", optional = true }
 pytest-randomly = { version = ">=3.16.0", optional = true }
 pytest-xdist = { version = ">=3.5.0", optional = true }
+anthropic = "^0.86.0"
 
 [tool.poetry.extras]
 telegram = ["python-telegram-bot"]
diff --git a/src/infrastructure/router/cascade.py b/src/infrastructure/router/cascade.py
index 4f7510ae..84f07e90 100644
--- a/src/infrastructure/router/cascade.py
+++ b/src/infrastructure/router/cascade.py
@@ -114,6 +114,7 @@ class Provider:
     type: str  # ollama, openai, anthropic
     enabled: bool
     priority: int
+    tier: str | None = None # e.g., "local", "standard_cloud", "frontier"
     url: str | None = None
     api_key: str | None = None
     base_url: str | None = None
@@ -267,6 +268,7 @@ class CascadeRouter:
                 type=p_data["type"],
                 enabled=p_data.get("enabled", True),
                 priority=p_data.get("priority", 99),
+                tier=p_data.get("tier"),
                 url=p_data.get("url"),
                 api_key=p_data.get("api_key"),
                 base_url=p_data.get("base_url"),
@@ -532,6 +534,7 @@ class CascadeRouter:
         model: str | None = None,
         temperature: float = 0.7,
         max_tokens: int | None = None,
+        cascade_tier: str | None = None,
     ) -> dict:
         """Complete a chat conversation with automatic failover.
 
@@ -545,6 +548,8 @@ class CascadeRouter:
             model: Preferred model (tries this first, then provider defaults)
             temperature: Sampling temperature
             max_tokens: Maximum tokens to generate
+            cascade_tier: If specified, filters providers by this tier.
+                - "frontier_required": Uses only Anthropic provider for top-tier models.
 
         Returns:
             Dict with content, provider_used, and metrics
@@ -558,7 +563,18 @@ class CascadeRouter:
 
         errors = []
 
-        for provider in self.providers:
+        providers = self.providers
+        if cascade_tier == "frontier_required":
+            providers = [p for p in self.providers if p.type == "anthropic"]
+            if not providers:
+                raise RuntimeError("No Anthropic provider configured for 'frontier_required' tier.")
+        elif cascade_tier:
+            providers = [p for p in self.providers if p.tier == cascade_tier]
+            if not providers:
+                raise RuntimeError(f"No providers found for tier: {cascade_tier}")
+
+
+        for provider in providers:
             if not self._is_provider_available(provider):
                 continue