Skip to content

README.md

Latest commit

 

History

History
40 lines (29 loc) · 1.51 KB

README.md

File metadata and controls

40 lines (29 loc) · 1.51 KB
Raw

Tellama open SDK

Dependency-free clients for the authenticated Tellama local AI server.

  • python/tellama.py: Python 3.10+ standard-library client
  • javascript/tellama.mjs: JavaScript client using built-in fetch and web streams
  • tests/: deterministic mock server and cross-client contract tests

Both clients provide model discovery, Ollama NDJSON chat streaming, OpenAI SSE chat streaming, timeouts, and structured errors. They never log the API key.

Python

from tellama import TellamaClient

client = TellamaClient("http://PHONE_IP:11434", "tlm_...")
model = client.list_models()[0]["model"]
for chunk in client.ollama_chat(model, [{"role": "user", "content": "Hello"}]):
    print(chunk, end="", flush=True)

JavaScript

import { TellamaClient } from "./tellama.mjs";

const client = new TellamaClient({
  baseUrl: "http://PHONE_IP:11434",
  apiKey: "tlm_...",
});
const [model] = await client.listModels();
for await (const chunk of client.ollamaChat({
  model: model.model,
  messages: [{ role: "user", content: "Hello" }],
})) process.stdout.write(chunk);

Generation controls can be supplied through Python options={"num_predict": 64} or JavaScript options: { num_predict: 64 }. OpenAI-style helpers accept max_tokens/maxTokens and temperature. Streamed server errors are raised as structured TellamaError instances.

Run the contract tests with tests/run.sh. The public SDK and tests are licensed under Apache-2.0; they do not include Tellama Android application source code.