utils/provider/provider_profile library

Classes

BenchmarkedOllamaModel

Benchmarked Ollama model with latency data.

CodexCredentials

Resolved codex credentials.

OllamaModelDescriptor

Ollama model descriptor.

ProfileEnv

Environment variables for a provider profile.

ProfileFile

Persisted profile file structure.

RankedOllamaModel

Ranked Ollama model with score and reasons.

SystemPromptBlock

A block of system prompt text with cache scope.

Enums

CacheScope

Cache scope for system prompt blocks.

ProviderProfile

Supported provider profiles.

RecommendationGoal

Recommendation goal for model selection.

Functions

appendSystemContext({required SystemPrompt systemPrompt, required Map<String, String> context}) → List<String>

Append system context entries to the system prompt.

applyBenchmarkLatency({required List<RankedOllamaModel> models, required Map<String, int?> benchmarkMs, required RecommendationGoal goal}) → List<BenchmarkedOllamaModel>

Apply benchmark latency data to ranked models, re-sort by adjusted score.

buildCodexProfileEnv({String? model, String? baseUrl, String? apiKey, Map<String, String>? processEnv, CodexCredentialResolver? resolveCodexApiCredentials}) → ProfileEnv?

Build a Codex profile environment. Returns null if credentials unavailable.

buildGeminiProfileEnv({String? model, String? baseUrl, String? apiKey, Map<String, String>? processEnv}) → ProfileEnv?

Build a Gemini profile environment. Returns null if no API key found.

buildLaunchEnv({required ProviderProfile profile, required ProfileFile? persisted, required RecommendationGoal goal, Map<String, String>? processEnv, String getOllamaChatBaseUrl(String? baseUrl)?, Future<String> resolveOllamaDefaultModel(RecommendationGoal goal)?, ProviderRequestResolver? resolveProviderRequest, CodexCredentialResolver? resolveCodexApiCredentials, bool isCodexBaseUrl(String url)?}) → Future<Map<String, String>>

Build the full launch environment for a given profile.

buildOllamaProfileEnv({required String model, String? baseUrl, required String getOllamaChatBaseUrl(String? baseUrl)}) → ProfileEnv

Build an Ollama profile environment.

buildOpenAIProfileEnv({required RecommendationGoal goal, String? model, String? baseUrl, String? apiKey, Map<String, String>? processEnv, ProviderRequestResolver? resolveProviderRequest}) → ProfileEnv?

Build an OpenAI profile environment. Returns null if no API key found.

createProfileFile({required ProviderProfile profile, required ProfileEnv env}) → ProfileFile

Create a profile file with current timestamp.

getGoalDefaultOpenAIModel(RecommendationGoal goal) → String

Get the default OpenAI model for a goal.

isViableOllamaChatModel(OllamaModelDescriptor model) → bool

Check if a model is viable for chat (not embedding/reranking).

normalizeRecommendationGoal(String? goal) → RecommendationGoal

Normalize a recommendation goal string to enum.

normalizeToolInput({required String toolName, required Map<String, dynamic> input, String? cwd, String? agentId, Map<String, dynamic> getPlan(String? agentId)?}) → Map<String, dynamic>

Normalize tool input for specific tools before execution.

normalizeToolInputForAPI({required String toolName, required Map<String, dynamic> input}) → Map<String, dynamic>

Strip injected fields before sending to API.

prependUserContext({required List<Map<String, dynamic>> messages, required Map<String, String> context}) → List<Map<String, dynamic>>

Prepend user context as a system-reminder message.

rankOllamaModels({required List<OllamaModelDescriptor> models, required RecommendationGoal goal}) → List<RankedOllamaModel>

Rank Ollama models by composite score for a given goal.

recommendOllamaModel({required List<OllamaModelDescriptor> models, required RecommendationGoal goal}) → RankedOllamaModel?

Recommend the best Ollama model for a given goal.

sanitizeApiKey(String? key) → String?

Sanitize an API key, returning null for empty or placeholder values.

selectAutoProfile({required String? recommendedOllamaModel}) → ProviderProfile

Select auto profile based on recommended Ollama model availability.

selectRecommendedOllamaModel<T extends OllamaModelDescriptor>(List<T> models) → T?

Select the first viable chat model from a list.

splitSysPromptPrefix({required SystemPrompt systemPrompt, bool skipGlobalCacheForSystemPrompt = false, bool useGlobalCacheFeature = false, String? dynamicBoundary, Set<String>? cliSyspromptPrefixes}) → List<SystemPromptBlock>

Split system prompt blocks by content type for API cache control.

Typedefs

CodexCredentialResolver = CodexCredentials? Function({String? codexApiKey, Map<String, String>? processEnv})

Callback type for resolving codex API credentials.

ProviderRequestResolver = String Function({String? baseUrl, required String fallbackModel, String? model})

Callback type for resolving provider transport.

SystemPrompt = List<String>

System prompt is a list of strings.