Use codemap find "SymbolName" to locate any symbol before reading — it gives exact file + line range and is faster than grep across the whole repo.

LLM Provider Connection

Quick start

from lauren_ai import LLMConfig, LLMModule

# Anthropic
cfg = LLMConfig.for_anthropic(model="claude-opus-4-6", api_key="sk-ant-...")

# OpenAI
cfg = LLMConfig.for_openai(model="gpt-4o", api_key="sk-...")

# Ollama (local, no API key needed)
cfg = LLMConfig.for_ollama(model="llama3.2")

# Tests — zero network calls
cfg, mock = LLMConfig.for_testing()

LLMConfig parameters

Direct constructor form

from lauren_ai import LLMConfig

# Explicit constructor (all three providers)
cfg_anthropic = LLMConfig(provider="anthropic", model="claude-opus-4-6", api_key="sk-ant-...")
cfg_openai    = LLMConfig(provider="openai",    model="gpt-4o",          api_key="sk-...")
cfg_ollama    = LLMConfig(provider="ollama",    model="llama3.2")        # no key needed

Wire into Lauren DI

from lauren_ai import LLMModule, LLMConfig
from lauren import module

cfg = LLMConfig.for_anthropic(model="claude-opus-4-6")
LLMProvider = LLMModule.for_root(cfg)

@module(imports=[LLMProvider])
class AppModule: ...

Testing pattern

from lauren_ai import LLMConfig
from lauren_ai._transport import Completion, TokenUsage

cfg, mock = LLMConfig.for_testing()       # provider="anthropic", model="mock-model"

mock.queue_response(
    Completion(
        id="c1",
        model="mock-model",
        content="Hello!",
        tool_calls=[],
        stop_reason="end_turn",
        usage=TokenUsage(input_tokens=10, output_tokens=5),
    )
)
# … run your agent code; mock.calls records every request

Reference files