Model Reference

Jeko AI uses Ollama to access and run various LLMs. This section provides details on the supported models, recommended use cases, and available configuration options. Supported Models Commonly used models include: Deepseek-v3.1, Kimi-k2-thinking, GPT-oss, Qwen 3, and Qwen 2.5.

Back to docs

Model Selection Guidelines

Deepseek-v3.1
A model with a focus on strong reasoning and high computational efficiency.
Kimi-k2-thinking
Designed for very long context processing and in-depth document analysis.
Gpt-oss
A versatile open-source model that mimics the capabilities of GPT with great flexibility.
Qwen 3
High-end model with very powerful reasoning, coding, and multimodal capabilities.
Qwen 2.5
A lighter and more efficient version for practical use and scalable deployment.

Daftar Endpoint

POST /api/generate

Endpoint utama untuk menghasilkan jawaban dari model Ollama pilihan pengguna.

{
  "prompt": "string",
  "model": "qwen3-coder:480b-cloud",
  "uploadedImages": ["https://..."],
  "max_tokens": 512
}

POST /api/upload

Mengunggah file (gambar) untuk digunakan sebagai input multimodal.

FormData:
  file: <image-file>

POST /api/login

{
  "password": "string"
}

POST /api/logout

Menghapus sesi dan mengunci aplikasi (Lock mode).

Semua endpoint berjalan sepenuhnya server-side di Next.js Route Handler.

Documentation v0.1