API Reference

This page documents all public functions, classes, and behaviors in VRBM‑AI.

Global Settings

• HEADER_VISIBLE
• DEFAULT_MODEL
• TEMPERATURE
• ALIVE_TIME

Loaded from vrbm_config if present.

Exceptions

VRBMError

Base class for all vrbm‑AI errors.

InvalidOutputTypeError

Raised when the user requests an unsupported output type.

OllamaConnectionError

Raised when vrbm‑AI cannot reach the Ollama server.

ModelDownloadError

Raised when a model cannot be downloaded.

Functions*

*only generate() is exposed

load_config()

Reads the vrbm_config file and updates global settings.

ollama_chat(model, messages, temp)

Low‑level wrapper around Ollama’s /api/chat. Returns the raw JSON response.

ollama_list()

Returns a list of installed Ollama models.

ollama_pull(model)

Downloads a model if not installed.

ensure_model(model_name)

Checks if a model exists; downloads it if missing.

parse_output(output)

Attempts to parse model output using:

1. ast.literal_eval
2. json.loads
3. raw string fallback

validate_output(value, expected_type)

Ensures the returned value matches the requested type. Returns:

• None if valid
• error message string if invalid

generate(prompt, output_type="unspecified", model=DEFAULT_MODEL, temperature=TEMPERATURE, parse=True)

Main user‑facing function.

Behavior:

1. Ensures model is installed
2. Sends system + user messages
3. Parses output
4. Validates type
5. Retries up to 3 times
6. Returns Python value or None

Parameters:

• prompt: user text
• output_type: Python type name
• model: Ollama model
• temperature: float
• parse: whether to parse output

Returns:

• Python value
• or None on failure

Constants

SYSTEM_PROMPT

Strict instructions sent to Ollama to enforce Python literal output.

VALID_TYPES

Mapping of type names → Python classes.

---

< prev.

home