zms.ZMSLLMConnector.ZMSLLMConnector

class documentation

class ZMSLLMConnector(ZMSItem.ZMSItem): (source)

Constructor: ZMSLLMConnector(id)

Implements interfaces: Products.zms.IZMSConfigurationProvider.IZMSConfigurationProvider, Products.zms.IZMSLLMConnector.IZMSLLMConnector, Products.zms.IZMSRepositoryProvider.IZMSRepositoryProvider

View In Hierarchy

ZMS LLM Connector – bridges ZMS to AI/LLM provider backends.

Method	`__init__`	Constructor.
Method	`chat`	Send messages to the configured LLM provider.
Method	`chat_with_tools`	Agentic chat loop with ZMS tool calling.
Method	`get_ollama_models`	Fetch the list of locally available models from the configured Ollama server.
Method	`get_provider_info`	Return information about the currently configured LLM provider.
Method	`getAvailableLLMToolsConnectors`	Return available ``_llmtools`` / ``_connector`` ZMSLibrary meta-objects.
Method	`getConfProperty`	Return a configuration property for this connector.
Method	`getEnabledFeatures`	Return a dict of enabled AI feature flags.
Method	`isFeatureEnabled`	Check whether a specific AI feature is enabled.
Method	`manage_changeConfig`	Save connector configuration fields from the Config tab form.
Method	`manage_changeFeatures`	Save feature-flag checkboxes from the Features tab form.
Method	`manage_changeProperties`	Backwards-compatible shim: delegates to manage_changeConfig.
Method	`manage_options`	Return management options delegating to the parent ZMS object.
Method	`manage_sub_options`	Contribute the LLM tab to the ZMS main tab bar.
Method	`provideRepository`	Export connector configuration for repository.
Method	`updateRepository`	Import connector configuration from repository.
Class Variable	`__ac_permissions__`	Undocumented
Class Variable	`__administratorPermissions__`	Undocumented
Class Variable	`manage`	Undocumented
Class Variable	`manage_main`	Undocumented
Class Variable	`manage_options_default_action`	Undocumented
Class Variable	`manage_sub_options__roles__`	Undocumented
Class Variable	`meta_type`	Undocumented
Class Variable	`security`	Undocumented
Class Variable	`zmi_icon`	Undocumented
Instance Variable	`id`	Undocumented
Method	`_as_bool`	Parse common truthy/falsy values from request-like sources.
Method	`_prepare_messages`	Normalize messages and optionally inject HTML-preservation guidance.
Instance Variable	`_config`	Undocumented
Instance Variable	`_p_changed`	Undocumented

def __init__(self, id='llm_connector'): (source) ¶

Constructor.

def chat(self, messages, **kwargs): (source) ¶

Send messages to the configured LLM provider.

Delegates to ``llmapi`` provider factory using this connector as the configuration context so ``getConfProperty()`` resolves correctly.

Parameters
messages	List of ``{"role": ..., "content": ...}`` dicts or a plain string.
**kwargs	Optional overrides: temperature, top_p, max_tokens, etc.
Returns
dict	Response in OpenAI /v1/chat/completions format or error dict.

def chat_with_tools(self, messages, context, max_rounds=5, preserve_html=False): (source) ¶

Agentic chat loop with ZMS tool calling.

Sends ``messages`` plus active tool definitions from ``llmtools.ZMSLLMToolsAdapter`` to the LLM. If the LLM responds with ``tool_calls``, each call is executed through the adapter (custom ``*_llmtools`` profile or built-in fallback), appended as ``tool`` role messages, and the conversation continues until the LLM produces a plain text response or ``max_rounds`` is reached.

Parameters
messages	List of ``{"role": ..., "content": ...}`` dicts or a plain string (auto-wrapped as user message).
context	ZMS acquisition context exposing ``metaobj_manager``.
max_rounds	Maximum tool-calling iterations to prevent infinite loops.
preserve_html	Undocumented
Returns
dict	Dict with keys: ``reply`` (str): Final assistant text response. ``turns`` (list): All conversation turns including tool calls/results, each as ``{"role", "content", "tool_calls"?}`` for UI rendering. ``error`` (str): Set only on unrecoverable errors.