ai_bridge_core library
Core models, interfaces, and utilities for Flutter AI Bridge.
This package provides the foundational abstractions for integrating AI providers into Flutter applications with a unified API.
Classes
- AIAttachment
- An attachment that can be sent with a message.
- AIAudioProvider
- Core interface for Audio capabilities.
- AIBridge
- The main entry point for Flutter AI Bridge.
- AICompletionProvider
- Core interface for AI text completion.
- AIConfig
- Configuration for an AI provider.
- AIDocument
- A chunk of textual data, usually parsed from a larger document.
- AIEmbeddingProvider
- Abstract interface for providers that support text embeddings.
- AIMessage
- A single message in an AI conversation.
- AIProvider
- Full-featured AI provider interface.
- AIRequest
- A request to an AI provider.
- AIResponse
- A unified response from any AI provider.
- AIRouter
- Smart router that decides which AI provider handles each request.
- AIStreamChunk
- A single chunk from a streaming AI response.
- AIStreamingProvider
- Interface for providers that support streaming responses.
- AITool
- Defines a tool or function that the AI can call.
- AIToolCall
- Represents a specific invocation of a tool requested by the AI.
- AIUsage
- Token usage information from an AI response.
- CancellationToken
- A token that allows cancellation of in-flight AI requests.
- CircuitBreaker
- Circuit breaker pattern to prevent cascading failures.
- Conversation
- Represents a conversation with an AI provider.
- ConversationManager
- Manages conversations with AI providers.
- CostOptimizedStrategy
- Selects providers by estimated cost-per-token (cheapest first).
- DocumentChunker
- Splits long text into manageable chunks for vector embeddings.
- FallbackChain
- Executes a request through a chain of providers with automatic fallback.
- InMemoryVectorStore
- A simple, local, list-backed Vector Store.
- LatencyOptimizedStrategy
- Selects providers with the lowest average latency first.
- PrimaryStrategy
- Uses the first available provider (insertion order).
- QualityFirstStrategy
- Selects providers in reverse insertion order (assumed quality order).
- ResponseCache
- Caches AI responses to reduce redundant API calls and costs.
- RetryHandler
- Handles automatic retries for retryable AI errors.
- RoundRobinStrategy
- Round-robin rotation across providers.
- RoutingStrategyHandler
- Abstract interface for routing strategy implementations.
- StorageProvider
- An abstract interface for storing and retrieving conversations.
- StreamHandler
- Manages streaming AI responses with unified handling.
- TokenBudget
- Controls token spending with configurable budgets.
- TokenTracker
- Tracks token usage across requests and sessions.
- ToolExecutionLoop
- Handles the autonomous tool execution loop.
- VectorStore
- Abstract interface for a Vector Database/Store used in RAG.
Enums
- AIAttachmentType
- The type of content in an attachment.
- AICapability
- Capabilities that an AI provider may support.
- AIRole
- Represents a role in an AI conversation.
- CircuitState
- States the circuit breaker can be in.
- RoutingStrategy
- Strategy for routing requests to providers.
Extensions
- RoutingStrategyExtension on RoutingStrategy
- Extension for backward compatibility: convert enum to handler.
Typedefs
- OnChunkCallback = void Function(AIStreamChunk chunk)
- Callback for each received chunk.
- OnCompleteCallback = void Function(String fullText, AIUsage? usage)
- Callback when streaming completes.
- OnStreamErrorCallback = void Function(Object error)
- Callback for stream errors.
Exceptions / Errors
- AIAuthError
- Authentication error — invalid or expired API key.
- AIBudgetExceededError
- Budget/quota exceeded.
- AIContentFilterError
- Content was filtered by safety settings.
- AIError
- Base class for all AI Bridge errors.
- AIModelNotFoundError
- Model not found or not available.
- AINetworkError
- Network error — connection issues, timeouts.
- AIRateLimitError
- Rate limit exceeded — too many requests.
- AIServerError
- Server error from the provider (5xx).
- AITokenOverflowError
- Token limit exceeded — input or output too long.
- AIUnknownError
- Unknown/unclassified error.
- CancelledException
- Exception thrown when an operation is cancelled via CancellationToken.