Plugin Parameters for prompt_bridge

This document describes the configuration parameters used by prompt_bridge. These parameters are typically set in prompt_bridge/config/prompt_bridge.yaml.

Server-Level Parameters

These parameters configure the current available plugin families and their mappings.

Parameter	Description	Example/Default
`frame_id`	Frame ID for published history messages.	`agent`
`cached_transactions`	Number of prompt transactions to keep in history.	10
`prompt_family_names`	List of supported prompt families	`[openai, ollama]`
`prompt_family_plugins`	Mapping from family name to provider class	`openai: prompt::OpenAIProvider`
		`ollama: prompt::OllamaProvider`
`embedding_family_names`	List of supported embedding families	`[openai]`
`embedding_family_plugins`	Mapping from family name to embedding class	`openai: prompt::OpenAIEmbedding`
`tokenizer_family_names`	List of supported tokenizer families	`[openai]`
`tokenizer_family_plugins`	Mapping from family name to tokenizer class	`openai: prompt::OpenAITokenize`

These parameters are related to prompting plugins that provide generic communication with different LLMs

Parameter	Description	Example/Default
`rest.uri`	Endpoint for single prompt requests	`https://api.openai.com/v1/responses`
`rest.chat_uri`	Endpoint for conversation requests	`https://api.openai.com/v1/responses`
`rest.method`	HTTP method to use	`POST`
`rest.auth_type`	Authentication type	`Bearer`
`rest.ssl_verify`	Whether to verify SSL certificates	`true`
`option_keys`	List of option keys available for prompts	`[stream, model]`
`options.stream.value`	Whether to stream responses	`false`
`options.stream.type`	Data type for stream option	`bool`
`options.model.value`	Default model to use	`gpt-5`
`options.model.type`	Data type for model option	`string` in parameters, typically sent as `str` in `ModelOption.type`

These parameters are related to embedding plugins that provide communication with different embedding models

Parameter	Description	Example/Default
`rest.embedding_uri`	The endpoint for embeddings	`https://api.openai.com/v1/embeddings`
`rest.method`	HTTP method to use	`POST`
`rest.auth_type`	Authentication type	`Bearer`
`rest.ssl_verify`	Whether to verify SSL certificates	`true`
`option_keys`	List of option keys available	`[model, dimensions, encoding_format]`
`options.model.value`	Default embedding model	`text-embedding-3-small`
`options.model.type`	Data type for model option	`string`
`options.dimensions.value`	Default embedding dimensions	`1536`
`options.dimensions.type`	Data type for dimensions option	`int`
`options.encoding_format.value`	Default encoding format	`float`
`options.encoding_format.type`	Data type for encoding format option	`string`

These parameters are related to tokenizer plugins that provide an interface for local or remote tokenization.

Parameter	Description	Example/Default
`option_keys`	List of option keys available for tokenizing	`[model]`
`options.model.value`	Default tokenizer model	`O200K_BASE`
`options.model.type`	Data type for model option	`string`

prompt::OpenAIProvider and prompt::OpenAIEmbedding also require the OPENAI_API_KEY environment variable.
prompt::OpenAITokenize does not use REST parameters; it only loads default option values such as model.
prompt::OllamaProvider currently uses prompt parameters only and does not require an API key.

For more details, see the main README and the example configuration in prompt_bridge/config/prompt_bridge.yaml.