Migrate from aevum-llm¶

aevum-llm was deprecated in v0.5.0 and will not receive further updates. This guide shows how to replace it with the purpose-built adapters in aevum-core.

Why it was deprecated¶

aevum-llm was a thin wrapper that bundled provider-specific integration code into a single package. As each integration grew in complexity — Cedar evaluation, snapshot testing, GDPR erasure hooks — maintaining them in a shared package became impractical. Each adapter now lives directly in aevum-core with its own test suite and CI coverage.

See the v0.5.0 section of CHANGELOG.md for the original deprecation notice.

Step 1 — Uninstall aevum-llm¶

pip uninstall aevum-llm

Adapter migration table¶

Use case	New package extra	New import path
Anthropic Claude	`aevum-core[anthropic]`	`aevum.core.adapters.anthropic_adapter.AevumAnthropicAdapter`
LangChain	`aevum-core[langchain]`	`aevum.core.adapters.langchain_callback.AevumLangChainCallback`
LangGraph	`aevum-core[langgraph]`	`aevum.core.adapters.langgraph.AevumCheckpointer`
OpenAI Agents	`aevum-core[openai-agents]`	`aevum.core.adapters.openai_agents.AevumAgentHooks`
CrewAI	`aevum-core[crewai]`	`aevum.core.adapters.crewai.AevumCrewHooks`

Migration by adapter¶

Anthropic¶

pip install "aevum-core[anthropic]"

import anthropic
from aevum.core.adapters.anthropic_adapter import AevumAnthropicAdapter

client = AevumAnthropicAdapter(client=anthropic.Anthropic(), kernel=engine)
message = client.messages.create(
    model="claude-opus-4-7",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello"}],
)

W3C traceparent is injected on every call. tool_use blocks are Cedar-evaluated before your code sees them. See the Anthropic adapter guide for full details.

LangChain¶

pip install "aevum-core[langchain]"

from aevum.core.adapters.langchain_callback import AevumLangChainCallback
from langchain_openai import ChatOpenAI

cb = AevumLangChainCallback(kernel=engine)
llm = ChatOpenAI(model="gpt-4o-mini", callbacks=[cb])

Tool calls are Cedar-evaluated on on_tool_start. LLM invocations are recorded to the sigchain on on_llm_start and on_llm_end. See the LangChain guide for full details including LangGraph StateGraph usage.

LangGraph¶

pip install "aevum-core[langgraph]"

from aevum.core.adapters.langgraph import AevumCheckpointer

checkpointer = AevumCheckpointer.local()
graph = builder.compile(checkpointer=checkpointer)
config = {"configurable": {"thread_id": "alice-session-1"}}
result = graph.invoke(inputs, config)

Every superstep is dual-signed (Ed25519 + ML-DSA-65) and chained. delete_thread("alice-session-1") triggers GDPR Art. 17 crypto-erasure.

OpenAI Agents¶

pip install "aevum-core[openai-agents]"

from aevum.core.adapters.openai_agents import AevumAgentHooks

hooks = AevumAgentHooks(kernel=engine)
# Pass hooks= to your OpenAI Agents SDK Runner

CrewAI¶

pip install "aevum-core[crewai]"

from aevum.core.adapters.crewai import AevumCrewHooks, AevumTaskCallback

hooks = AevumCrewHooks(kernel=engine)

After migration¶

Verify sigchain integrity to confirm the migration did not introduce gaps:

assert engine.verify_sigchain()

If you previously recorded capture gaps in aevum-llm, those events are preserved in the sigchain and remain replayable.