daimon-client 0.2.9

Python client for the daimon AI sidecar

daimon-client

Python client for Daimon — a pluggable AI sidecar runtime.

Installation

pip install daimon-client

Quick start

from daimon_client import Client

with Client(base_url="http://localhost:3500") as c:
    reply = c.chat("claude", "What is a daimon?")
    print(reply)

Async

import asyncio
from daimon_client import AsyncClient

async def main():
    async with AsyncClient(base_url="http://localhost:3500") as c:
        reply = await c.chat("claude", "What is a daimon?")
        print(reply)

asyncio.run(main())

Streaming

with Client() as c:
    for text in c.stream("claude", "Tell me a story."):
        print(text, end="", flush=True)

Multi-turn conversations

Pass a list of messages to carry history yourself:

reply = c.chat("claude", [
    {"role": "user",      "content": "My name is Alice."},
    {"role": "assistant", "content": "Nice to meet you, Alice!"},
    {"role": "user",      "content": "What is my name?"},
])

Sessions

Let the sidecar maintain history server-side with a session_id:

c.chat("claude", "My favourite colour is blue.", session_id="chat-1")
reply = c.chat("claude", "What is my favourite colour?", session_id="chat-1")
# reply contains "blue"

c.clear_session("chat-1")

Inference parameters

All sampling parameters are optional and fall back to the component's configured defaults:

reply = c.chat("gpt4o", "Summarise this.", model="gpt-4o", temperature=0.2, max_tokens=256)

Vector store (memory)

Read and write documents in a configured vector store:

mem = c.memory("my-store")

# Upsert a document (returns the assigned ID)
doc_id = mem.upsert("The Eiffel Tower is 330 metres tall.", id="eiffel", metadata={"source": "wikipedia"})

# Semantic search
results = mem.query("tall Paris structures", top_k=3)
for r in results:
    print(f"{r.score:.3f}  {r.content}")

# Delete
mem.delete("eiffel")

Async vector store

async with AsyncClient() as c:
    mem = c.memory("my-store")
    await mem.upsert("Some content")
    results = await mem.query("my query")

Graph store

Interact with a configured graph database using Cypher:

kg = c.graph("knowledge-graph")

# Add nodes
kg.add_node(id="alice", labels=["Person"], props={"name": "Alice", "age": 30})
kg.add_node(id="bob",   labels=["Person"], props={"name": "Bob"})

# Add a relationship
kg.add_edge("alice", "bob", "KNOWS", props={"since": "2020"})

# Run a Cypher query
rows = kg.cypher(
    "MATCH (a:Person)-[:KNOWS]->(b) RETURN a.name AS from, b.name AS to"
)
print(rows)  # [{"from": "Alice", "to": "Bob"}]

# Delete a node (and all its relationships)
kg.delete_node("alice")

API reference

Client(base_url?, timeout?)

Parameter	Default
base_url	http://127.0.0.1:3500
timeout	120.0 seconds

Use as a context manager (with Client() as c) or call c.close() manually.

c.chat(component, prompt, **kwargs) → str

Returns the full response text.

prompt can be a str or a list of {"role": ..., "content": ...} dicts.

c.stream(component, prompt, **kwargs) → Iterator[str]

Yields text fragments as they arrive.

c.clear_session(session_id) → None

Deletes server-side session history. Safe to call on a non-existent session.

c.memory(store) → MemoryStoreClient

Returns a client scoped to the named vector store.

Method	Description
upsert(content, *, id?, metadata?)	Insert or update a document. Returns the document ID.
query(query, top_k=5)	Semantic search. Returns list[MemoryResult] sorted by descending score.
delete(id)	Delete a document by ID.

c.graph(store) → GraphStoreClient

Returns a client scoped to the named graph store.

Method	Description
add_node(*, id?, labels?, props?)	Add or update a node. Returns the node ID.
add_edge(from_id, to_id, rel_type, *, props?)	Create a directed relationship.
cypher(query, params?)	Run a Cypher query. Returns list[dict].
delete_node(id)	Delete a node and all its relationships.

Keyword arguments for chat / stream

Argument	Description
model	Override the component's default model
system	System prompt shorthand
max_tokens
temperature
top_p
top_k	Anthropic only
stop	List of stop sequences
frequency_penalty
presence_penalty
seed
session_id	Server-side session ID

AsyncClient mirrors Client with async def methods and async for streaming.

Links

• Daimon on GitHub
• Bug reports

License

Apache-2.0