Skip to content

fix: align llm provider fallback models #2501

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
stranske merged 1 commit into from
Jun 22, 2026
Merged

fix: align llm provider fallback models #2501

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 14 additions & 8 deletions templates/consumer-repo/tools/llm_provider.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,8 @@
DEFAULT_MODEL = "codex-mini-latest"
ANTHROPIC_API_KEY_ENV = "CLAUDE_API_STRANSKE"
SHORT_ANALYSIS_CONFIDENCE_CAP = 0.4
DEFAULT_OPENAI_ANALYSIS_MODEL = "gpt-5.4"
DEFAULT_ANTHROPIC_ANALYSIS_MODEL = "claude-sonnet-4-6"


def _configured_langchain_model(provider: str, *, fallback: str) -> str:
Expand Down Expand Up @@ -586,7 +588,7 @@ def _get_client(self):
logger.warning("LangChain client helper not available")
return None

model_name = _configured_langchain_model("openai", fallback="gpt-5.1-codex")
model_name = _configured_langchain_model("openai", fallback=DEFAULT_OPENAI_ANALYSIS_MODEL)
resolved = build_chat_client(provider="openai", model=model_name)
if resolved:
self._model_name = resolved.model
Expand Down Expand Up @@ -626,7 +628,7 @@ def analyze_completion(
model_name=getattr(
self,
"_model_name",
_configured_langchain_model("openai", fallback="gpt-5.1-codex"),
_configured_langchain_model("openai", fallback=DEFAULT_OPENAI_ANALYSIS_MODEL),
),
raw_confidence=result.raw_confidence,
confidence_adjusted=result.confidence_adjusted,
Expand Down Expand Up @@ -657,7 +659,9 @@ def _get_client(self):
logger.warning("LangChain client helper not available")
return None

model_name = _configured_langchain_model("anthropic", fallback="claude-sonnet-4-5-20250929")
model_name = _configured_langchain_model(
"anthropic", fallback=DEFAULT_ANTHROPIC_ANALYSIS_MODEL
)
resolved = build_chat_client(provider="anthropic", model=model_name)
if resolved:
self._model_name = resolved.model
Expand Down Expand Up @@ -707,7 +711,9 @@ def analyze_completion(
model_name=getattr(
self,
"_model_name",
_configured_langchain_model("anthropic", fallback="claude-sonnet-4-5-20250929"),
_configured_langchain_model(
"anthropic", fallback=DEFAULT_ANTHROPIC_ANALYSIS_MODEL
),
),
raw_confidence=result.raw_confidence,
confidence_adjusted=result.confidence_adjusted,
Expand Down Expand Up @@ -959,8 +965,8 @@ def get_llm_provider(force_provider: str | None = None) -> LLMProvider:
Options: "github-models", "openai", "anthropic", "regex-fallback"

Returns a FallbackChainProvider that tries:
1. Anthropic claude-sonnet-4-5 (if CLAUDE_API_STRANSKE set) - Best reasoning
2. OpenAI gpt-5.1-codex (if OPENAI_API_KEY set) - Purpose-built for code analysis
1. Anthropic configured slot model (if CLAUDE_API_STRANSKE set) - Best reasoning
2. OpenAI configured slot model (if OPENAI_API_KEY set) - Code analysis
3. GitHub Models gpt-4.1 (if GITHUB_TOKEN set) - Always available, reliable
4. Regex fallback (always available) - 30% confidence baseline
"""
Expand All @@ -987,8 +993,8 @@ def get_llm_provider(force_provider: str | None = None) -> LLMProvider:
return provider

providers = [
AnthropicProvider(), # Primary: claude-sonnet-4-5 for best reasoning
OpenAIProvider(), # Secondary: gpt-5.1-codex for code-optimized analysis
AnthropicProvider(), # Primary: configured Anthropic slot model
OpenAIProvider(), # Secondary: configured OpenAI slot model
GitHubModelsProvider(), # Tertiary: gpt-4.1 via GITHUB_TOKEN (always available)
RegexFallbackProvider(), # Last resort: 30% confidence pattern matching
]
Expand Down
4 changes: 2 additions & 2 deletions tests/tools/test_llm_provider.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -585,7 +585,7 @@ def test_openai_provider_reports_configured_client_model(self):
):
result = provider.analyze_completion("output", ["task1"])

mock_configured.assert_called_with("openai", fallback="gpt-5.1-codex")
mock_configured.assert_called_with("openai", fallback="gpt-5.4")
mock_build.assert_called_once_with(provider="openai", model="gpt-configured")
assert result.model_name == "gpt-configured"

Expand All @@ -612,7 +612,7 @@ def test_anthropic_provider_reports_configured_client_model(self):
):
result = provider.analyze_completion("output", ["task1"])

mock_configured.assert_called_with("anthropic", fallback="claude-sonnet-4-5-20250929")
mock_configured.assert_called_with("anthropic", fallback="claude-sonnet-4-6")
mock_build.assert_called_once_with(provider="anthropic", model="claude-configured")
assert result.model_name == "claude-configured"

Expand Down
22 changes: 14 additions & 8 deletions tools/llm_provider.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,6 +40,8 @@
DEFAULT_MODEL = "codex-mini-latest"
ANTHROPIC_API_KEY_ENV = "CLAUDE_API_STRANSKE"
SHORT_ANALYSIS_CONFIDENCE_CAP = 0.4
DEFAULT_OPENAI_ANALYSIS_MODEL = "gpt-5.4"
DEFAULT_ANTHROPIC_ANALYSIS_MODEL = "claude-sonnet-4-6"


def _configured_langchain_model(provider: str, *, fallback: str) -> str:
Expand Down Expand Up @@ -598,7 +600,7 @@ def _get_client(self):
logger.warning("LangChain client helper not available")
return None

model_name = _configured_langchain_model("openai", fallback="gpt-5.1-codex")
model_name = _configured_langchain_model("openai", fallback=DEFAULT_OPENAI_ANALYSIS_MODEL)
resolved = build_chat_client(provider="openai", model=model_name)
if resolved:
self._model_name = resolved.model
Expand Down Expand Up @@ -638,7 +640,7 @@ def analyze_completion(
model_name=getattr(
self,
"_model_name",
_configured_langchain_model("openai", fallback="gpt-5.1-codex"),
_configured_langchain_model("openai", fallback=DEFAULT_OPENAI_ANALYSIS_MODEL),
),
raw_confidence=result.raw_confidence,
confidence_adjusted=result.confidence_adjusted,
Expand Down Expand Up @@ -669,7 +671,9 @@ def _get_client(self):
logger.warning("LangChain client helper not available")
return None

model_name = _configured_langchain_model("anthropic", fallback="claude-sonnet-4-5-20250929")
model_name = _configured_langchain_model(
"anthropic", fallback=DEFAULT_ANTHROPIC_ANALYSIS_MODEL
)
resolved = build_chat_client(provider="anthropic", model=model_name)
if resolved:
self._model_name = resolved.model
Expand Down Expand Up @@ -719,7 +723,9 @@ def analyze_completion(
model_name=getattr(
self,
"_model_name",
_configured_langchain_model("anthropic", fallback="claude-sonnet-4-5-20250929"),
_configured_langchain_model(
"anthropic", fallback=DEFAULT_ANTHROPIC_ANALYSIS_MODEL
),
),
raw_confidence=result.raw_confidence,
confidence_adjusted=result.confidence_adjusted,
Expand Down Expand Up @@ -971,8 +977,8 @@ def get_llm_provider(force_provider: str | None = None) -> LLMProvider:
Options: "github-models", "openai", "anthropic", "regex-fallback"

Returns a FallbackChainProvider that tries:
1. Anthropic claude-sonnet-4-5 (if CLAUDE_API_STRANSKE set) - Best reasoning
2. OpenAI gpt-5.1-codex (if OPENAI_API_KEY set) - Purpose-built for code analysis
1. Anthropic configured slot model (if CLAUDE_API_STRANSKE set) - Best reasoning
2. OpenAI configured slot model (if OPENAI_API_KEY set) - Code analysis
3. GitHub Models gpt-4.1 (if GITHUB_TOKEN set) - Always available, reliable
4. Regex fallback (always available) - 30% confidence baseline
"""
Expand All @@ -999,8 +1005,8 @@ def get_llm_provider(force_provider: str | None = None) -> LLMProvider:
return provider

providers = [
AnthropicProvider(), # Primary: claude-sonnet-4-5 for best reasoning
OpenAIProvider(), # Secondary: gpt-5.1-codex for code-optimized analysis
AnthropicProvider(), # Primary: configured Anthropic slot model
OpenAIProvider(), # Secondary: configured OpenAI slot model
GitHubModelsProvider(), # Tertiary: gpt-4.1 via GITHUB_TOKEN (always available)
RegexFallbackProvider(), # Last resort: 30% confidence pattern matching
]
Expand Down
Loading