togolm 0.1.2

Minimal Python client for the TogoLM API

togolm

Minimal Python client for the TogoLM API.

Requires Python 3.9+. Built on httpx.

Install

pip install togolm

Get an API key

from togolm import TogoLM

client = TogoLM()  # no key needed for this call
result = client.register_key(email="[e-mail]", name="Your Name")
api_key = result["api_key"]  # shown once — save it

You can also call client.query(...) without a key — public access is rate-limited (see rate limits).

Usage

from togolm import TogoLM

client = TogoLM(api_key="your_api_key")

result = client.query("Comment créer une entreprise au Togo ?")
print(result["answer"], result["sources"])

Streaming

for event in client.query_stream("Comment créer une entreprise au Togo ?"):
    if event["type"] == "chunk":
        print(event["text"], end="")
    if event["type"] == "sources":
        print("Sources:", event["sources"])

Multi-turn conversations

history = []
for question in ["Comment créer une entreprise au Togo ?", "Et pour les impôts ?"]:
    result = client.query(question, history=history)
    history.append({"role": "user", "content": question})
    history.append({"role": "assistant", "content": result["answer"]})

history is capped at 20 messages server-side; oldest first. max_tokens (default 3000, 50–4096) caps the generated answer length.

Local development

Point at a local API instead of production:

client = TogoLM(base_url="http://localhost:8000/v1")

Error handling

Non-2xx responses raise TogoLMError (status and body attributes):

from togolm import TogoLM, TogoLMError

try:
    client.query("...")
except TogoLMError as err:
    print(err.status, err.body)

Context manager

TogoLM holds an httpx.Client; use it as a context manager to close it automatically:

with TogoLM(api_key="your_api_key") as client:
    result = client.query("...")

All methods

Method	Description
query(question, category=None, language=None, max_tokens=None, history=None)	RAG query, full response
query_stream(question, category=None, language=None, max_tokens=None, history=None)	RAG query, SSE stream (generator)
embed(text)	Generate an embedding vector
search(q)	Full-text search over the corpus
categories()	List available corpus categories
stats()	Public corpus statistics
documents(page=None, page_size=None, category=None, source=None, language=None)	Paginated document list
document(doc_id)	Document detail, including chunks
register_key(email, name, use_case=None)	Request a free API key
me()	Current API key info and usage
close()	Close the underlying HTTP client

Full request/response shapes: API reference.