Deep Dives

Multi-Language SDK Comparison

Comprehensive feature matrix comparing ollama-haskell with Python, JavaScript/TypeScript, and Go SDKs.

Feature Matrix

Feature	Haskell (`ollama-haskell`)	Official Python (`ollama-python`)	Official JS/TS (`ollama-js`)	Community Go (`ollama/ollama`)
Strict Type Safety	✅ Compile-time (PVP, Smart Constructors)	⚠️ Type hints (Runtime)	⚠️ TypeScript (Erased at runtime)	✅ Go Structs
Response Streaming	✅ `conduit` (O(1) constant memory)	⚠️ Python Generator	⚠️ Async Iterator	⚠️ Go Channels
Structured Output Derivation	✅ `GHC.Generics` (`ToSchema`)	⚠️ Pydantic BaseModel	⚠️ Zod / JSON Schema	⚠️ Manual JSON Schema
Model Context Protocol (MCP)	✅ Native `mcp-server` Bridge	❌ Manual	❌ Manual	❌ Manual
Thinking / Reasoning Models	✅ Dedicated `Think` ADT	⚠️ Dict parameters	⚠️ Object properties	⚠️ Raw parameters
Transactional Chat Store	✅ STM `InMemoryStore`	❌ None	❌ None	❌ None
Built-in Mock Testing	✅ `Ollama.Testing` (Pure)	❌ None	❌ None	❌ None
Configurable Retry & Backoff	✅ Exponential & Constant ADT	❌ Manual	❌ Manual	❌ Manual
Token Throughput Metrics	✅ Native Calculation Helpers	⚠️ Raw nanoseconds	⚠️ Raw nanoseconds	⚠️ Raw nanoseconds
Environment Auto-Discovery	✅ `clientFromEnv`	✅ Default client	✅ Default client	✅ Default client

Detailed Architectural Comparison

1. Memory Safety & Streaming

ollama-haskell: Uses conduit streams which process chunks incrementally. If a downstream consumer cancels or encounters an exception, connection resources are finalized immediately with ResourceT.
Python / JS SDKs: Often buffer entire JSON payloads into memory when parsing complex tool calls or large context windows, causing memory spikes.

2. Structured Outputs & Type Safety

ollama-haskell: Type declarations (data Weather = ... deriving (Generic, ToSchema)) produce exact JSON schema definitions without requiring external schema generators or runtime reflection.
Python / JS: Rely on heavy third-party dependencies (Pydantic / Zod) that add runtime overhead and potential schema drift.