OllamaRequest

Defined in: adapters/ollama.ts:55

Ollama API request structure for local models.

Ollama provides a local server for running LLMs. Parameters are nested in an options object, and uses num_predict instead of max_tokens.

See

• OllamaMessage
• OllamaResponse
• https://github.com/ollama/ollama/blob/main/docs/api.md

Example

const request: OllamaRequest = {
  model: 'llama2',
  messages: [{ role: 'user', content: 'Hello!' }],
  options: {
    temperature: 0.7,
    num_predict: 100
  }
};

Properties

messages

messages: OllamaMessage[]

Defined in: adapters/ollama.ts:60

Conversation messages

---

model

model: string

Defined in: adapters/ollama.ts:57

Local model name (e.g., ‘llama2’, ‘mistral’, ‘codellama’)

---

options?

optional options: object

Defined in: adapters/ollama.ts:63

Model parameters nested in options object

num_predict?

optional num_predict: number

Maximum tokens to predict (Ollama uses ‘predict’ not ‘tokens’)

stop?

optional stop: string[]

Stop sequences

temperature?

optional temperature: number

Sampling temperature

top_k?

optional top_k: number

Top-K sampling

top_p?

optional top_p: number

Nucleus sampling

---

stream?

optional stream: boolean

Defined in: adapters/ollama.ts:81

Enable streaming responses